Theses for the.” Degree 0f Ph D; MlCHIGAN STATE UNIVERSITY;5:51;, A“; LIBRART A Michigan State University . This is to certify that the thesis entitled ECONOMIC DEVELOPMENT, POLITICAL LABOR MOVEMENTS, AND PUBLIC POLICY: A QUANTITATIVE COMPARISON OF TIME-SERIES INDICATORS FOR NORWAY AND SWEDEN presented by Charles David Klingman has been ‘accepted towards fulfillment of the requirements for Ph.D. dpgn¥,hl Political Science fi/«J/f A». Major professor \ May 29, 1973 Date 0-7 639 ..___.. _.._-—_ i i ! fir lllllllllzlllfllllllllllljllllWW L ABSTRACT ECONOMIC DEVELOPMENT, POLITICAL LABOR MOVEMENTS, AND PUBLIC POLICY: A QUANTITATIVE COMPARISON OF TIME-SERIES INDICATORS FOR NORWAY AND SWEDEN By Charles David Klingman This dissertation compares the patterns of political development fin‘Norway and Sweden from T875 to T965 by means of correlation and regression analysis of aggregate time—series data. Historical accounts mnlprevious analyses of industrialization, political labor movements, mnlsocial legislation in Norway and Sweden were reorganized within theiflamework of the mobilization model of political development. This Adelded fourteen specific hypotheses which anticipated certain smfilarities and differences between Norway and Sweden in the patterns ofintercorrelation among indicators of the following concepts: social nnbilization, economic wealth, political mobilization, democratization, government penetration, government expenditures by sector (health, enucation, and welfare), and the impacts of those expenditures, namely personnel services and objective-security conditions, in each corre— sponding societal sector. The data base consisted of aggregate national statistics on Norway and Sweden ultimately derived from official sources (the Central lhneau of Statistics in each country) for years in which regularly Charles David Kl ingman scheduled, lower-house parliamentary elections occurred (every third and fourth year) from T875 to 1965. The criterion for acceptance of both bivariate and multivariate relationships was significant corre— lation and lack of significant autocorrelation. The hypotheses were evaluated by comparing the two countries on the number of acceptable relationships among the indicators for each hypothesized set of related concepts. In an effort to reduce the high level of auto— correlation, this procedure was also used for a shorter time-span, the period since the disruption of the loose union between the two countries in T905. Only three of the original fourteen hypotheses were confirmed. First, the indicators of social mobilization were more substantially associated with the indicators of government penetration for Norway than for Sweden. Second, the indicators of political mobilization were more Substantially associated with the indicators of democrati- zation for Norway than for Sweden. Third, there was no significant difference between the two countries in the strength of association between the indicators of government penetration and government expenditures. The latter two confirmations applied only to the post-1905 time period. 30 few of the hypotheses were confirmed because of severe auto- correlation, which probably resulted from curvilinearity in the regression parameters over time as well as from exogenous variables. Furthermore, the patterns of interrelationship among the indicators suggested that per capita monetary measures intercorrelate more Charles David Kl ingman strongly with each other than do social and political variables measured on a proportional basis. Future research should seek to reduce the autocorrelation not only by incorporating exogenous variables, but also by using every-year data analyzed in successive brief time periods in an attempt to locate the time-points at which the regression parameters change. Re-examination of the acceptable relationships outside the framework of the original hypotheses indicated that in both countries social and economic variables had a far more substantial impact on indicators of public policy than did political variables. Further— more, the only significant differences between the patterns of development for Norway and Sweden were the closer association between the rise of the socialist movement and democratization in Norway than in Sweden. and the earlier expansion of the Civil Service in Norway than in Sweden. Both of these differences were probably due to stronger resistance of the Swedish political elites to leftist demands for democratic reform and expansion of government social programs. ECONOMIC DEVELOPMENT, POLITICAL LABOR MOVEMENTS, AND PUBLIC POLICY: A QUANTITATIVE COMPARISON OF TIME-SERIES INDICATORS FOR NORWAY AND SWEDEN By Charles David Klingman A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1973 Copyright © by CHARLES DAVID KLINGMAN 1973 ACKNOWLEDGMENTS Even though the responsibility for this dissertation rests solely with its author, several people have contributed signifi- cantly to its completion. Recognition must go first of all to my ammfittee, Professors Timothy M. Hennessey, Paul R. Abramson, and (Heo H. Cherryholmes, for their guidance. As adviser, Professor Hennessey was particularly helpful in the interpretation stages of the project. Appreciation is also due the following organizations at Michigan State University: the Department of Political Science, the Computer Institute for Social Science Research, and the Social Science Multidisciplinary Program for various forms of support; and the Computer Laboratory and STAT Systems Group for computer service. A few words cannot adequately express my gratitude to Professor Guy Peters of Emory University for his advice, encouragement, and friendship throughout my graduate career (not to mention his data on Sweden). Similarly, I owe more to my parents than could ever be expressed here. I only hope their faith in me can be even further justified in the years ahead. Finally, my wife, Charlotte, who typed the various drafts, and son, Jeffrey, who occasionally had to accept proofreadings as bedtime stories, have endured much the past few years. This dissertation would not have been completed without them. iv TABLE OF CONTENTS CHAPTER ONE. INTRODUCTION ................... I. Review of Literature ................. A. The Development of Political Development ..... 3.. Comparative Research Design ............ II. The Setting: Norway and Sweden ............ A. General Similarities ............... B. General Differences ................ CHAPTER (TWO. A DEVELOPMENTAL HISTORY OF NORWAY ........ CHAPTER THREE. A DEVELOPMENTAL HISTORY OF SWEDEN ....... CHAPTER FOUR. THEORY AND METHODOLOGY ............. 1. ,Theoretical Background ,- . .............. A. The Bull-Galenson Hypothesis . . . . ....... B. The Mobilization Model of Public Policy . Development .................. II. The Data ............... . ........ ' A. The Observations ................. B. The Variables ................... III. The Hypotheses .................... IV. The Methods ...................... CHAPTER FIVE. ANALYSIS OF THE TOTAL TIME PERIOD (T875 TO'l965) ........... ' ........ I. . ReSUItS oooooooo o o o o t ooooooooooo . A. Linear, Bivariate Relationships . . . ....... B. Linear, Multiple Relationships .......... II. Acceptable Relationships ............... CHAPTER SIX. ANALYSIS OF THE POST-1905 TIME PERIOD ...... I. Results . . . . . . . . . . . . . ........... A. Linear, Bivariate Relationships .......... B. Linear, Multiple Relationships .......... II. Acceptable Relationships ....... _ ........ III. Summary of the Analysis ................ lO4 108 III 138 156 165 I70 CHAPTER SEVEN. INTERPRETATION ................. I. Explaining the Failure of the Hypotheses ....... A. Comparison with Previous Works .......... B. Methodological and Historical Interpretations . . . II. A Substantive Interpretation ............. III. Suggestions for Future Research ............ BIBLIOGRAPHY .......................... APPENDIX ............................ vi Page l73 l73 l76 l85 l95 205 208 217 Table 5-1. 5-2. 5-3. 5-4. 5-10. 5-11. 5-12. 5-13. LIST OF TABLES Social Mobilization and Political Mobilization, Total Period .................... . Social Mobilization and Economic Health, Total Period ............ . ........... Political Mobilization and Democratization, Total Period ..................... Social Mobilization and Democratization, Total Period ........................ Democratization and Government Penetration, Total Period ..................... Social Mobilization and Government Penetration, Total Period ..................... Political Mobilization and Government Penetration, TOtal Period ..................... Government Penetration and Government Expenditures, Total Period ..................... Democratization and Government Expenditures, -Total Period. . . . . ................. Economic Wealth and Government Expenditures, Total Period ..................... vii Table 5-14. 5-15. 6-1. 6-2. 6-3. 6-4. 6-10. 6-11. 6-12. 6-13. 6-14. 6-15. Page Economic Wealth and Objective Security, Total Period . . . 129 Acceptable Multiple Regression Equations, Total Period ........................ 132 Social Mobilization and Political Mobilization, Post-1905 Period ................... 139 Social Mobilization and Economic Health, Post- 1905 Period . ...; ................... 142 Political Mobilization and Democratization, Post-1905 Period ........ - ........... 144 Social Mobilization and Democratization, Post 1905 Period ...................... 146 Democratization and Government Penetration, Post-1905 Period . . . .- ............... 147 Social Mobilization and Government Penetration, POSt-1905 Period ..... , .............. 148 Political Mobilization and Government Penetration, Post-1905 Period ................... 149 Government Penetration and Government Expenditures, Poet-1905 Period ....... .. . . ‘ ......... 150 Democratization and Government Expenditures, Post-1905 Period ........... 7 ........ 151 Economic Wealth and Government Expenditures, Post-1905 Period ................... 152 Government Expenditures and Personnel Services, Post-1905 Period ................... 153 Government Expenditures and Objective Security, Post-1905 Period ................... 154 Personnel Serices and Objective Security, Post- 1905 Period ...................... 155 Economic Wealth and Objective Security, Post-1905 Period . . ...................... Acceptable Multiple Regression Equations, Post- 1905 Period ...................... 158 viii Table 5-14. 5-15. 6-1. 6-5. 6-6. 6-7. 6-8. 6-9. 6-10. Page Economic Wealth and Objective Security, Total Period . . . 129 Acceptable Multiple Regression Equations, Total Period ........................ 132 Social Mobilization and Political Mobilization, Post-1905 Period ................... 139 Social Mobilization and Economic Wealth, Post- 1905 Period . .... ................... 142 Political Mobilization and Democratization, Post-1905 Period ................... 144 Social Mobilization and Democratization, Post 1905 Period . .. .................... 146 Democratization and Government Penetration, Post-1905 Period . . . .- ............... 147 Social Mobilization and Government Penetration, POSt-1905 Period ................... 148 Political Mobilization and Government Penetration, Post-1905 Period ................... 149 Government Penetration and Government Expenditures, PoSt-1905 Period ....... g ............ 150 Democratization and Government Expenditures, Post-1905 Period ........... , ........ 151 Economic Wealth and Government Expenditures, Post-1905 Period ................... 152 Government Expenditures and Personnel Services, Post-1905 Period ................... 153 Government Expenditures and Objective Security, Post-1905 Period ................... 154 Personnel Serices and Objective Security, Post- 1905 Period ...................... 155 Economic Wealth and Objective Security, Post-1905 Period ........................ 156 Acceptable Multiple Regression Equations, Post- 158 1905 Period ...................... viii LIST OF FIGURES Figure Page 2-1. Important Periods in the Development of Norway, 1814-1965 ...................... 46 3-1. Important Periods in the Development of Sweden, 1809-1962 ...................... 68 4-1. Diagram of the General Mobilization Model ....... 90 5-1. Acceptable Relationships, Total Period ......... 134 64. Acceptable Relationships, Post-1905 Period ....... 166 7-1. Patterns of Residuals for Selected Relationships, Post-1905 Period Only ................ 183 7-2. Confirmed Models of Development for Norway and Sweden, Based on ACCeptable Relationships Among Concept Indicators, POst-1905 Period Only ...... 198 7-3. Common and Different Conceptual Links in Confirmed MOdels of Development for Norway and Sweden, with the Country for Which the Relationship Was More Substantial Indicated, Postw1905 Period Only . . . . 198 A-l. Univariate Distributions of Enfranchisement and Left Vote ...................... 217 A-2. Univariate Distributions of Representation and . Voter Turnout .................... 218 A-3. Univariate Distributions of Labor Conflict and Urban Population .................. 219 A-4. Univariate Distributions of Civil Service and ............. . . . . . 220 Union Membership A-S. Univariate Distributions of Revenue and Investment . . . 221 A-6. Univariate Distributions of Gross National Product and Non-Defense Expenditures ............ 222 ix CHAPTER ONE INTRODUCTION 1. Review of Literature A. The DeveTOpment of Political Development For years empirical theorists have considered the field of polHfical development to be one of the most backward in political mfience. One cannot build theory on amorphous foundations, and the very meaning of the term "development" has fluctuated ever shme the spread of its usage in the 1950's, when the primary con- cern seemed to be with strengthening the newly independent nations against the perceived threat of international communism. Western mflitical science literature at least implicitly has viewed the end- Tnoduct of the development process as being the establishment of Amstern-style democratic institutions, processes, and attitudes. Thwever, it has camouflaged this view under a variety of pseudonyms amifms emphasized different aspects of democratization. 1. Mass Participation One segment of the literature has emphasized mass participation in the political process, with its requisite attitudes, values, and institutions, as being the major criterion of development. Perhaps the best-known of the earlier writers in the field, Daniel Lerner (1958), attempted to demonstrate through a limited cross-sectional correlation analysis that political develOpment consisted of a sequence of five major variables. That is, urbanization, literacy, and mass media participation fostered the growth of empathetic attitudes, laying the trusting foundation necessary for organized mass political participation, the crucial aspect of political development. Banfield (1958) and Wylie (1964), in their "participant observer" studies of villages in southern Italy and France, respec- tively, also emphasized the role of culturally induced lack of trust of others and alienation from organized politics in perpetuating backwardness. Pye (1962), in his study of Burma, also blamed the cultural lack of trust and the identity crises produced by the socialization process for the friction between politicians and administrators and the lack of democratic political participation. His other works (1965a; 1965b; 1966) also focussed on democratic participation as being the major criterion. of political development. Almond, writing with several different co—authors (1960; 1963; 1966), utilized the political culture approach to emphasize the importance of the development of attitudes and values supporting democratic political participation. He also employed the structural-functional approach to highlight the role of institutions fostering such participation. Deutsch (1953; 1961) popularized the term "social mobilization“ in referring to the process which breaks down old social, economic, and psychological commitments and replaces them with new patterns of socialization, attitudes, and behavior, thus laying the groundwork for political mobilization or participation. Geertz (1963) also emphasized the process of developing participant loyalties to the new states which complement old diverse loyalties or ”primordial sentiments." Finally, Marshall (1965) and Bendix (1968) focussed on the process of incorporation of the lower classes and minority groups into society through the attainmentof economic, social, and finally political rights of citizenship. 2. Elite Institutions Another segment of the literature, (however, has emphasized the development of strong societal institutions deemed legitimate by the people and mediating between mass participation and the elite political processes responsible for the actual operation of the system. This is not to say that the aforementioned "mass participa- tion“ writers ignored institutionalization entirely. But their concern did focus largely on the role of institutions in fostering democratic political participation. Other authors, however, focussed on the control of mass participation by societal institutions and the overriding importance of elite political processes. Lipset (1959) and Olson (1963) pointed to the dangers of zealous mass participation, polarization, and rapid change in social structures and norms for the stability of the political system. Kornhauser (1959) emphasized the need for strong intermediate groups or secondary organizations to mediate between masses available for mobilization and elites accessible to influence. Nordlinger (1968) pointed to the need for the development of national identity and strong institutions before the moderately-paced advent of mass participation. Apter (1965) discussed the suitability of various types of authority and value structures for handling the pressures produced by economic modernization. Kautsky (1962) also discussed the impact of industrialization on agrarian societies in terms of the shifting influence of particular interest groups in the political process. Eckstein (1966) based his "theory of stable democracy" on the degree of congruence among authority patterns in the various institutions and organizations of society. Various authors in the field of comparative public administration emphasized the importance of strong yet adaptable bureaucracies in transforming I socio-economic pressures into policies in developed as well as developing states (Crozier 1964; Riggs 1964; LaPalombara 1967; Raphaeli 1967). Much of the literature in both "schools" emphasized the importance of a strong political party system, whether in foster- ing or controlling mass participation (LaPalombara and Weiner 1966). 3. Resolving the Conceptual Confusion Obviously the early literature harbored considerable confusion as to whether the major criterion of political development should be democratic mass-participation or strong elite institutions. The solution invOlved taking a more comprehensive view of the relationship between the political system and its environment. Huntington (1968) clarified the idea that development involves both participation and institutionalization, although his emphasis was still on the latter. The argument seemed to be that social mobilization inevitably results frmneconomic modernization and that mass political participation results from the gap between economic promise and performance, regard- less of whether a society has strong institutions fostering such participation. Rather, political development involves the growth of institutions capable of dealing with the demands created by social mobilization. But Huntington went a step further and considered the consequences of the failure of political institutions to control mass participation, namely violence and corruption. Thus the field of political development has begun to shift its focus from the inputs of the political system to its outputs or performance in solving the problems of the society which it serves. Throughout much of the literature of both developmental 'schools“ there was some discussion of political development as being the increasing capacity of the political system to adapt to the changing demands of its socio-economic environment. But the real focus remained on the inputs from that environment to the I I political system, mainly political participation, whether the emphasis was on encouraging or controlling it. More recently Alker l (1968) and Almond (1969) have called for the evaluation of the capabilities and performance of political systems, and Karl de Schweinitz (1970) has called for the formulation of a measure of the "Gross Political Product" of nations. Others such as Pennock (1966) and Mitchell and Mitchell (1969) have also drawn parallels between the political system and the economic system and have called for economics-style concepts and methods of measurement and evaluation in the study of politics. 4. Politics and Economic Development Obviously the economy is one of the most important aspects of society on which the political system has an impact and on which it partly depends for its performance. Most of the literature recognized a strong relationship between political development, whatever its . definition, and economic development, usually defined as increasing wealth and industrialization. Many economists and some political scientists were more concerned with the role of political institu- . 'fions in fostering economic growth than with the effect of economic development on politics. Organski (1965) focussed on economics in Ms characterization of the stages of political development. McClelland (1961) pointed to the impact of social norms and institu- tions on the development of achievement motivation and entrepreneur- ship, vital to economic growth. Hirschman (1958) argued that governments in economically underdeveloped nations should induce 'Wnbalanced growth,” helping “tertiary” industries in the hope that these would spur growth in the more traditional primary sectors, such as agriculture. Rostow (1960) argued that governments in I nations with some degree of economic sophistication should intervene in all sectors of their economies in an effort to push them to the ”take-off“ stage of sustained growth. He also attempted to evaluate the suitability of various types of political organization for this task. However, Holt and Turner (1966) utilized an historical com- parison of France and China with England and Japan to illustrate their thesis that governments in economically underdeveloped nations must notintervene in the private economy prior to "take-off,“ but rather must encourage the steady concentration of private wealth for investment and industrialization. Adelman and Morris (1967) employed a factor analysis of subjectively-coded data on 74 economically underdeveloped countries and concluded that political factors were unimportant relative to social and economic ones at the lowest and middle level Of development, and didn't become important until the highest level of development had been reached. 5. Public Policy Analysis Controversy has been raging in the new field of "comparative American state politics,“ thoroughly reviewed elsewhere (Hennessey 1969; Salisbury 1968), over whether socio—economic (environmental) variables or political development (system) variables explain the most variance in public policy (outcome) variables. Examples of the competing independent variables have been level of industriali- zation and urbanization versus level of inter-party competition and legislative malapportionment, while the dependent variable has usually been measured by various expenditure levels and distribution patterns. This approach has begun to appear in studies of cross-national com- parisons as well. Cutright (1965) used cross—sectional analysis to determine the antecedents of government social security programs in 76 nations, and later (1967a; 1967b) to assess the impact of govern— ment activity on inequality and income redistribution in some 40 nations. Flanigan and Fogelman (1971) used "longitudinal“ as well as cross-sectional analysis of data on 44 countries from 1800 to 1960 to explore the patterns of urbanization and agricultural employment associated with the authors‘ index of democratization. Peters (1970) examined the impact of government expenditures on societal conditions in the areas of health, education, and welfare, and explored the influence of social and political mobilization, economic wealth, government penetration, and indices of democratization on those expenditures, using correlation and regression analysis of time- series data for every fifth year since 1850.for Britain, France, and Sweden. Using these data and similar techniques in a more (recent work (1972), he addressed the now—traditional argument by assessing the relative importance of a socio-economic variable, per capita GNP, and a political variable, number of civil servants per capita, in determining sectoral expenditure patterns. Such quantitative, com- parative analyses of the determinants and consequences of public policies, combined with the principles of budgetary analysis (Wildavsky 1968), should eventually develop into comparative evalua— tion of the performance of political systems. B. Comparative Research Design 1. Methodological Sophistication As the field of political development has begun to resolve some of the conceptual confusion plaguing it, the methodology which it has employed has become more SOphisticated. Early empirical works relied largely on attitude surveys or simple correlation and factor analyses of cross-sectional aggregate data. More recent works have employed more complicated regression analysis of time—series data and analysis of variance as well as correlation analysis of cross-sectional data. Much of the controversy in the field has become methodological in nature. The best-known example began when Cutright (1963) sought to test more thoroughly the hypotheses of Lipset (1959) and Lerner (1958) by developing a cumulative index of democratization representing the 21-year period from 1940 to 1960 for 77 countries, criticizing Lipset for using a dichotomy rather than such an index in his analysis. McCrone and Cnudde (1967) criticized Cutright's use of mere correla— timianalysis and analyzed his data with the Simon—Blalock method of causal modelling (Blalock 1964), the limitations of which have been pointed out by Forbes and Tufte (1968). In another response to Cutright, Neubauer (1967) constructed his Own “Index of Democratic Political Performance” based on cross-sectional data on only 23 democratic and relatively developed countries, so that the two entirely different indices could not be expected to yield comparable results. One need only review the more recent studies in comparative politics to perceive the increasing sophistication displayed by political scientists in handling complex methods of statistical analysis. It is no accident that theory and methodology have increased in sophistication simultaneously. Measurement is scientific theory in action for a specific purpose (Stinchcomb 1968: 43), and the quality of a measure greatly depends on the clarity of the concept which it is designed to measure. Furthermore, the scientific validation of hypotheses involves not only logical concept clarification, but also the use of measures which truly represent those concepts. ”A theory lO 'uTbe useful must be specific enough to be disproved" (Stinchcomb 1968:6D by empirical testing as well as by logic. Thus the increas- imjnmthodological sophistication of the discipline of political science as a whole has probably induced the increasing theoretical sophistication of its various sub-fields, and will continue to do so. 2. Multi-level Analysis The field of comparative political science has also begun to expkne more complex analytic perspectives as well as statistical nethods. Although it is obvious that comparison is the basis of all science (Campbell and Stanley 1966: 6), indeed of all human knowledge, mflitical scientists remain uncertain about what should be compared. lbst of the aforementioned researchers confined their comparisons to efifimr the level of individuals or the level of aggregate units, usually the nation-state. Recent treatises of empirical theory in comparative political science, reviewed by LOewenberg (1971), have emphasized the need for "multi-level analysis," or the use of variables from at least two different levels Of aggregation. Allardt (1969) argued that multi-level analysis increases the fruitfulness or hfibrmative value of comparative research because it decreases the likelihood that one's variables are measuring the same phenomena. He cited Riley's (1964) classification of multi-level analysis into structural analysis, involving group-level dependent variables and individual-level independent variables, versus contextual analysis, involving individual-level dependent variables and group-level independent variables. He also pointed out that Riley should have utility the di f similar too man, explana Hriabl me M if caus. isolate. ale TH. “1059 u. liriabl. Vll‘iabh ll (fistinguished between group-level variables that are merely aggregated individual characteristics and those that are truly "global" or group-level characteristics. Obviously there are many levels Of groups, with the nation-state perhaps being at the highest level, and some theorists have focussed on middle-level units of analysis. Dogan and Rokkan (1969) urged the comparison of regional as well as individual and national variations across nations, and Lorwin (1968) wentihrther in urging the study of regional variations in small countries. Perhaps the most significant of the recent treatises of multi— level comparative analysis, because of its thoroughness and logical rigor, was that of Przeworski and Teune (1970). They discounted the utility of the "most similar systems" design, which involves comparing the differences in aggregate-level characteristics among substantially similar systems. Their argument maintained that there will always be too many significant differences to permit causal interpretations or explanations among the variables (p. 34). That is, any dependent variable will be "overdetermined," i.e. interchangeably explained by more than one independent variable, violating one of their criteria of causality, the other being that "the system of variables is isolated--the explanatory pattern does not change when new variables are added" (p. 23). These two criteria were quite different from those usually accepted by most social scientists, namely: the two variables must be strongly and consistently associated; the dependent variable cannot precede the independent variable in time; and all 12 alternative causes of the dependent variable must be ruled out or controlled (Selltiz 1959: 83-88). In fact, Przeworski and Teune's criteria seem to have been selected to bolster their preference for the "most different systems“ design, which involves validating within— system relationships among variables across maximally differing systems, substituting "systematic" variables for the "proper names" of those systems as explanations only when those within-system relationships become dissimilar. The authors seemed to be thinking primarily of individual-level relationships as measured by survey research, for they devoted much of the second half of the book to the serious problem of equivalence of within—system measurement across different-systems. This-problem obviously affects aggregate measurement as well, even llhard" monetary measures Of economic concepts, because of biased estimates and dif- ferences in definition and specific techniques of measurement (Deane 1968; Ohlin 1968). Frey (1970) and Hymes (1970) thoroughly treated the serious practical problems Of cross-national survey research, such as accurate sampling frames and the training of honest inter- viewers, but especially the loss of equivalence through linguistic translation. Przeworski and Teune's solution for the problem of equivalence was to'use system-specific measures that yieldisimilar empirical results in terms of the intercorrelations among variables. Again they were never really clear whether they meant correlations among measures of the same or of different concepts. Surely they meant the former, for if they meant the latter their logic would be 13 inutological because their measures would then certainly show the vfithin-system relationships to be similar across different systems, again bolstering their preference for the “most different systems“ design by reducing "systemic interference" in their measures. And although by this they seemed to be contaminating their within-system measures with system-level variance rather than keeping that variance in the form of separate variables, they did talk of comparing within- system correlations with pooled cross-system correlations in order to validate their measures. 3. Inferential Fallacies Przeworski and Teune were also never really clear about what constitutes system-level and individual-level variables. They often seemed to consider even aggregated individual characteristics, such as total voter turnout, as individual-level or within-system variables, reserving the term "systemic" only for truly global system-level characteristics, although one can also compare within- syshmirelationships among system-level time—series variables. The real question concerning levels of analysis revolves around the unit of analysis over which the variance in a variable occurs, i.e. whether that unit is the individual, a type Of group or organization, a specific level of geographic region, or the nation-state. Observations on a sample of at least about 15 of that type of unit are necessary if that variable is to have sufficient variance to work with. But another consideration is the level of analysis at which inferences are to be made, and social scientists have often been 14 guilty of making inferences beyond the information provided by their data. That is, generalizations about one level Of unit of analysis have often been based on variance derived from some other level of unit of analysis. Initially the problem cited most frequently involved inferring the behavior of individuals from data on geographic aggre- gates, known as the "ecological fallacy" (Robinson 1950). (Subse- quently scientists have attempted to develOp methods of adjusting ecological or aggregate-level correlations to permit inferences about individuals (Shively 1969). But in the field of political development the problem has more frequently been the reverse: Survey data on individual attitudes and reported behavior have been used to make inferences about the "behavior" of geographic aggregates, usually nation-states. The best-known example was the five-nation study by lUnmnd and Verba (1963). The making of these kinds of inferences has been called the "individualistic fallacy." The problem in both types of fallacies is the failure to consider the structure of those aggregates, and the fact that some of their members are simply more influential in determining aggregate “behavior" (Scheuch 1966). A molecule is more than just a cluster of atoms, and the "aggregate behavior" of a molecule depends on how its atoms are arranged as well as on the number and types of atoms it contains (Scheuch 1969). Alker (1969) has provided a very thorough and general typology of cross-level and other inferential fallacies. One of the major limitations of the recent treatises on compara- tive research design is the apparent assumption that the ultimate can be reside 5C1€11C1 dictinq implem the av: Other 1 15 dependent-variable unit of analysis must necessarily be the individual human being. Most of the discussions of multi-level analysis, principally Przeworski and Teune (1970), have emphasized tmainmact of system-level variables on individual-level relation- ships. Valkonnen (1969) has even provided a very precise and thorough nethod for conducting such individual-oriented multi-level analysis twing regression models. However, it is not entirely clear why the dependent variables of comparative political science must always or even usually reside at the individual level of analysis. In fact, it can be argued that the truly significant concerns of political science reside at the level of aggregate systems. That is, if political science is ever to be useful to society it must be capable of pre— dicting what will happen when one policy rather than another is implemented. At the risk of overstating the case, knowing what makes the average citizen participate in politics or vote one way or the other may be inherently interesting, and knowing what makes him participate violently may be important. Furthermore, knowing, through the simple aggregation of such individual behavior, which political party will win an election (Campbell, et al., 1966) or under what conditions civil strife will occur (Gurr 1970) may also be interesting or even important. But what makes a political system perform ade- 9uately in producing policies designed to solve the problems confront- ing the society it serves is crucial. Being able to predict which party will be in power or when violence will occur may help determine two of the many important inputs into the policy-making process. But in th them hi hi models subsys Were it at “i9 sh until 5 Very ill hence: 16 in that actual process individual preferences, especially those of the masses, cannot be simply aggregated to predict policy outcomes. Besides the practical difficulties involved in doing cross— national survey research, especially on the masses, such research is of doubtful utility for the truly significant concerns of political science, namely the determinants and consequences of public policy. The discipline of economics has gotten by very well on l'macro— analysis" at the system level, and political science may indeed do Unasame.. Of course, macro-economics has enjoyed strong support from a thorough understanding of individually rational economic behavior through "micro-analysis," and political science will also need such a thorough understanding of individually rational political behavior. Some efforts along these lines have already been undertaken, notably Downs (1957), Buchanan and Tullock (1962), Olson (1965), and Curry and Wade (1968). Even better would be the development of the reverse of Valkonnen's (1969) methodology, namely a system Of regression models with system-level dependent variables and some individual and subsystem-level independent variables, including indicators of system structure. And any individual-level variables should measure the attitudes and behavior of elites rather than masses, because they are simply more important in the determination of public policy. But until such a methodology is developed, political scientists will do very well to concentrate on analyzing the determinants and conse- quences of public policy using aggregate-level variables. 17 4. The Time Dimension Most of the aforementioned treatises on comparative research design make at least passing mention of the need to consider the time dimension in comparative research, especially in the field of [xflitical development. After all, development is a process which takes place over a rather long period of time, and using cross— sectional data on a large number of polities supposedly arrayed on some presumed developmental dimension cannot capture the actual process of developmental change (Harsanyi 1960; Huntington 1971). But most amfirfical studies incorporating the time dimension of develOpment have only assessed the impact of history in a methodologically non- quantitative fashion, for example Eisenstadt (1963), Moore (1966), and Holt and Turner (1966). Some fruitful attempts have been made at analyzing successive cross-sections using correlation analysis (Converse 1969) and factor analysis (McRae and Meldrum 1969). But Flanigan and Fogelman (1971) tried analyzing the same set of data using both successive cross-sections and time-series, and concluded that only time-series analysis can Specify the conditions under which some dependent variable such as democratization will increase or decrease in a given political entity. What is ultimately needed is a methodology for what Harsanyi (1960) called "comparative dynamics," or what Thrupp (1970) and others called ”diachronic” analysis, or what Flanigan and Fogelman (1971) called "longitudinal” analysis. Przeworski and Teune‘s (1970) method (H comparing maximally different systems, which resulted from their 18 focus on cross-sectional analysis and individual—level dependent variables, would leave uncontrolled too many alternative explanations of any given variable. The best method of analyzing aggregate-level relationships over time is to compare maximally similar systems (Thrupp 1970), in an effort to approximate the principle of control underlying experimental research design. Time-series analysis over- comes many of the practical limitations Of the comparative method identified by Lijphart (1971), primarily the lack of a sufficient number of cases for analysis, which restricts the use of statistical and other non-experimental control. Just as the best hope for the theory of political development lies in using economics-style con- cepts to concentrate on the determinants and consequences of public policy, the best hope for the methodology of comparing aggregate— level relationships over time lies in econometrics. This involves primarily correlation and regression analysis of time—series variables and special techniques to overcome the peculiar problems of auto- correlation within the variables and time—lags among them (Kmenta 1971). This dissertation will attempt to use some of those methods on such data for two highly similar systems, Norway and Sweden- II. The Setting: Norway and Sweden The previous section concluded that system-level developmental processes can best be studied using correlatidn and regression analysis of aggregate time-series data on similar systems. SUch data must therefore be available on thoSe systems for their periods of development. At the international level very few systems satisfy 19 both criteria: similarity, and completion of development recently enough to have sufficient national statistics available on their periods of development. Three of the five Scandinavian countries satisfy those criteria. "Denmark, Norway and Sweden have not only gone through much the same developments culturally, economically and secially: these countries have also reached strikingly similar political solutions to the hnoblems posed by these developments," which occurred in the space of approximately the past 150 years (Rokkan and Valen 1960: 104). Adequate official statistics are available for about two-thirds of that period, when the most significant development occurred.‘ An added advantage is that Scandinavian statistical sources have recent English translations applicable to earlier years. This dissertation will focus on Norway and Sweden because of their particularly inter— esting pattern of similarities and differences as well as for practical reasons of manageability. A. General Similarities In addition to their proximity Norway and Sweden share geographic characteristics which have influenced their social development. The harsh northern climate, moderated by the Gulf Stream, and the rugged terrain and distribution of resources have curtailed the scale of the population and its primary sector, limiting farming and urban areas to the fertile plains and valleys of the southern regions Of the countries, and scattering forestry, fishing and mining far from the centers of commerce. A lack of large coal deposits has forced 20 the countries to rely on their wood and water resources to furnish power and to locate their industries in remote regions. Through the early 1900's the rugged terrain made the sea and inland water- ways the primary means of transportation and communication for much of the countries' populations. Shipping and foreign trade have developed to export their wood and mineral products, to import agri— cultural products as well as coal and fuel oils as further sources of industrial power, and to exchange manufactured goods (Lauwreys 1958: 9-26; Lindgren 1959: 18-19). Despite the dispersion of natural and human resources the Scandinavian cultures display remarkable homogeneity derived from the common source of Viking communalism. Except for the Finns and Lapps the peoples Of Scandinavia share the same racial characteris— tics. Because their geographic characteristics discouraged immigra- tion, Norway and Sweden have resisted racial “contamination“ more effectively than the other Scandinavian countries. They have inherited from the ancient Vikings not only their ethnicity but also similar languages and collectivity-oriented cultural norms emphasizing cooperation, egalitarianism, and individual liberty (Lauwreys 1958: 15-36). These characteristics have limited the number and intensity of their cultural cleavages and controlled their cross-cutting social and economic cleavages (Eckstein 1966: 111-131). Based on this common cultural heritage the Nordic countries joined together under Danish leadership in the Union of Kalmar in fl 1197, but split in and Sweden-Finland. these unions treater of Finnish and Norm unions had hereditar representation (nob' lutheran state churn although the topogra ing difficult (Andre 1952: 110). In 180! soon joined with Ru: . who was allied with During the defeat 01 Hurray to Sweden as Norwegians resisted demce and establish pressured them into Sweden and her allie under her new Consti domestic Constitutir dating from the end all aspects of an in government and a mil Stockholm handled a1 both countries, and 21 1397, but split in 1523 into two separate unions, Denmark-Norway and Sweden-Finland. In essence the first-named country in each of these unions treated the second as a province, fostering the growth of Finnish and Norwegian nationalism over the centuries. Both unions had hereditary monarchies, limited parliaments with Estate representation (nobility, clergy, burghers and peasants), Official Lutheran state churches, and feudal socio-economic structures, although the topography of Sweden and Norway made large-scale farm- ing difficult (Andren 1964: 13-14; Lauwreys 1958: 36-47; Galenson 1952: 110). In 1809 Sweden lost Finland in a war with Russia, but soon joined with Russia and England in an alliance against Napolean, who was allied with Denmark-Norway despite Norwegian opposition. During the defeat of Napolean in 1814 the Treaty of Kiel transferred Norway to Sweden as compensation for the loss of Finland. The Norwegians resisted the move and attempted to declare their indepen- dence and establish their own monarchy; but a two—week war quickly pressured them into accepting a loose union under the Swedish crown. Sweden and her allies permitted Norway to keep her domestic sovereignty under her new Constitution of 1814. Sweden had established her own domestic Constitution in 1809 as a reaction against royal absolutism dating from the end of the Era of Liberty in 1772. Norway retained all aspects of an independent nation, including citizenship, a government and a military force, without Swedish interference. But Stockholm handled all foreign diplomatic and commercial affairs for both countries, and Norwegians were subjects of the Swedish Crown. é hmegian nationali almost going to war maintained close re 1963: 12-33). Despite certai for dovnestic sovere essentially similar development from th tries they develope more smoothly. The highly industrializ but with far less d industrializing nat economic and politi such as the organiz smoothly despite th processes of the he change frovn constit Universal suffrage; method; a multi-par Farmers, Socialists each country); cabi administration and effective interest ststern. Cooperativ 22 Norwegian nationalism grew over the next 91 years, and after twice almost going to war the two nations dissolved the union in 1905, but maintained close relations thereafter (Lindgren 1959: 3-18; Storing 1963: 12-33). Despite certain key differences facilitated by the provision for domestic sovereignty in the union, the two countries showed essentially similar patterns of economic, social and political development from the time of their union. Relative to other coun- tries they developed not only later and more rapidly, but also more smoothly. Their basically poor, agricultural economies became lfighly industrialized and affluent over a shorter period of time but with far Tess disruption of social Tire than in most other industrializing nations. The changes in life-style and in socio— ecOnomic and political cleavages associated with industrialization, such as the organization of interest groups, occurred relatively smoothly despite their rapidity. The political institutions and processes of the two countries also underwent rapid but fairly stable change from constitutional monarchy to parliamentary democracy: Universal suffrage; proportional representation by the Sainte-Lague method; a multi-party system (basically Conservatives, Liberals, Farmers, Socialists and Communists, but with varying strength in each country); cabinet government with functionally organized administration and semi-autonomous local governmental agencies; and effective interest organizations, now characterize the political system. Cooperative and public enterprise and extensive public-sector fl planning and income growth in the priva of these developrnen B. General Differ The geographic and historical simi might be viewed as contaminating vari a ferences between th not quite complete, each of those dimen Geographically than Norway. Her sh extensive, with him tnlovway's three pu mineral resources, 1 nostimportantly, i‘ boasts an abundance have been exploited Processes such as a imity to the open 5‘ abundant rainfall , and foreign trade ( Demographi cal 1 has encouraged the 23 planning and income redistribution developed along with vigorous growth in the private sector (Lauwreys 1958: 21-47; 107-125). All of these developments will be detailed in the next two chapters. B. General Differences The geographic, demographic, socio-economic, cultural, political and historical similarities between Norway and Sweden sketched above might be viewed as quasi-experimental controls on extraneous or contaminating variables in any attempt to explain certain key dif- ferences between the two countries. Unfortunately these controls are not quite complete, for differences as well as similarities exist on each of those dimensions. Geographically, Sweden generally possesses more natural resources than Norway. Her southern farmlands and northern forests are more extensive, with nine percent of her land under cultivation, compared to Norway's three percent (Lauwreys 1958: 22). She also holds more mfineral resources, particulary copper, manganese, zinc, lead and, most importantly,iron that yields quality steel. However, Norway boasts an abundance of waterfalls, only about one-fourth of which have been exploited for hydroelectric power, useful in electrochemical processes such as aluminum production. ‘Also, Norway's closer prox— imity to the open sea has fostered not only a milder climate and more abundant rainfall, but also a greater reliance on fishing, shipping and foreign trade (Lauwreys 1958: 9-15; Lindgren 1959: 18-19). Demographically, Sweden's greater store of natural resources has encouraged the development of greater human resources as well. fl Her population has editor to 3.8 mill per square mile), a 25 percent in 1965) of natural and huma industrialization. in agriculture in 1 15 percent of her i percent for Norway, 1958: 22-23). Swed greater wealth: Sw $16.8 billion compa 1031166 per capita Culturally, No largely to past Dan socioeconomic and domcracies, center differences. But N mimg from the use 0 1814. Isolated fro learning, rationali in remote and rural like the other Scan tithe Vikings. Af w *See Section II in 24 Her population has been over twice as large as Norway's (about 7.6 million to 3.8 million in l965), higher in density (about 40 to 30 per square mile), and more highly urbanized (about 45 percent to 25 percent in 1965).* Socio-economically, Sweden's greaterstore of natural and human resources has facilitated a higher level of industrialization. About nine percent of her population was engaged in agriculture in l965 compared with Norway's sixteen percent. About 75 percent of her farms are under 25 acres, compared with over 90 percent for Norway, and her farm productivity is higher (Lauwreys 1958: 22-23). Sweden's resources and industrialization have yielded greater wealth: Sweden's Gross National Product in 1965 was about $16.8 billion compared with Norway's $6.6 billion, or about $2200 to $l766 per capita.* Culturally, Norway manifests more cleavages than Sweden, due largely to past Danish domination. Both countries share the usual socio-economic and political cleavages associated with industrial democracies, centering around occupational, residential, and income differences. But Norway bears an added complex of cleavages stemw nfing from the use of Danish as the official language in Norway before l8l4. Isolated from the urban centers of officialdom, commerce, learning, rationalized religion, and high culture, Norwegians living in remote and rural areas spoke the old Norwegian dialect derived, like the other Scandinavian languages, from Old Norse, the language of the Vikings. After independence from Denmark was achieved in _ *See Section II in Chapter Four. on some groups 1““ language, Bokmal or I bra Norwegian intel‘ the country, CONPHE‘ old Norwegian dialer: languages are off i do has aggravated the no upper and lower clas: the educated and unet sects (Lauwreys 1958 lMs multi-face' political realm in a deomy overtly rel' lumeh Party (Rokka the political setting indigenous nobility : lNh he Riksdag, he tlocountries,dating me to 1772, the Era hhlhtsand Caps), Ml;Rustou l955: lg. Stallinavian country democracy in 1884, wl he fact that Norway Slide I . "5 ‘3 currentgl 25 HH4-some groups launched efforts to “Norwegianize” the official language, Bokmal or Riksmal. In the 1840's Ivar Aasen, supported inra Norwegian intellectual organization, toured the rural areas of the country, compiled a standardized dictionary and grammar of the old Norwegian dialect, Landsmal, and renamed it Nynorsk. Today both languages are officially accepted, but the "language controversy“ lms aggravated the normal tensions between urban and rural areas, the upper and lower classes, the agricultural and industrial sectors, the educated and uneducated, and.fundamentalist and liberal religious sects (Lauwreys 1958: 18-21; Eckstein 1966: 44-47; Storing 1963: 7). This multi-faceted cleavage has manifested itself in the political realm in a complex manner, including the develOpment of the only overtly religious party in Scandinavia, the Christian People's Party (Rokkan and Valen 1960: 106). Another difference in the political setting is that Norway, unlike Sweden, has not had an indigenous nobility since the union with Denmark. Sweden's parlia- ment, the Riksdag, had existed long before the union between the two countries, dating from 1435 and dominating the government from 1718 to 1772, the Era of Liberty, during which parliamentary factions (the Hats and Caps), resembled political parties (Andren 1964: 139— 141; Rustow 1955: 11-12). Norway, on the other hand, was the first Scandinavian country to achieve the principle of true parliamentary democracy in 1884, while Sweden waited until 1917 (Lauwreys 1958: 36). The fact that Norway's parliament, the Storting, is unicameral whereas Sweden's is currently bicameral and has always been multicameral, helped the Swedish n block donesti c pol it tradition of separat self-government in N especially the rural independence from Sun Sweden's Riksdag has constitutional funct connittees to corres nent departments, a loruegian distrust o lhndren 1964: 185). hdninistrativel executive authority councillors, in actu nunisters is elected cabinet's policy dec choice of a cabinet does participate in onadninistrative de county agencies whic decisions. Despite stoll exercise a gre health, education. a d three of local auto 26 helped the Swedish nobility and their conservative supporters to block domestic political change. This feature, along with the tradition of separation of powers and the establishment of local self—government in Norway in 1837, also helped the Norwegians, especially the rural peasant classes, to establish some degree of independence from Swedish domination (Andren 1964: 117-120). Sweden's Riksdag has organized its committee system according to constitutional function; Norway's Storting has organized its committees to correspond with the functions of the various govern— ment departments, a characteristic indicative of the deeply—ingrained Norwegian distrust of the once foreign—dominated civil service (Andren 1964: 185). Administratively, although both countries have vested formal executive authority in the King and his formally appointed 1 councillors, in actuality the cabinet of functionally specific ministers is elected by parliament, which usually ratifies the cabinet's policy decisions. The King always accepts parliament‘s choice of a cabinet and the latter's policy decisions, although he does participate in those decisions. In Norway each minister heads an administrative department containing directorates and local county agencies which share the detailed administration of cabinet decisions. Despite the trend toward centralization, local agencies still exercise a great deal of autonomy, especially in the areas of health, education, and welfare. But Sweden displays an even greater degree of local autonomy and a ”dual system” of national administration, with semi autonomous ninistries. These b appeal and usualll’ C also representatives to the actions of th anployee organizatio on gouernnent-appoi n and policy solutions legislation. Sweden openness for the pub because of interest- oilaw can dismiss c dismissal cannot be civil servants can b hobudsnan also wield llainst civil servan' lilies more on legal Sweden's constitution requires all adminis (usually numbers of ' that tho document has her so concerning nat' loin ' ' Clple,’ along wi' 11011 organization an Swedish citizen with 27 wiUosemi-autonomous administrative boards supplementing the nfinistries. These boards often act as courts of administrative appeal and usually contain not only government representatives but also representatives of the major private interest groups affected by the actions of that particular board, such as employer and employee organizations. Interest groups also gain representation cnogovernment-appointed commissions investigating social problems and policy solutions, and their advice is usually sought on proposed legislation. Sweden's "dual system" provides not only greater openness for the public but also greater security for civil servants because of interest-group support. In both countries only a court of law can dismiss civil servants after a trial and conviction; and dismissal cannot be based on political reasons, even though Sweden's civil servants can be members of parliament. But Sweden's legal (mmudsman also wields greater authority to initiate legal action against civil servants upon appeal by a citizen, whereas Norway relies more on legal courts for administrative appeal. Finally, Sweden's constitution contains a Freedom of the Press Act which requires all administrative agencies to permit any private citizen (usually members of the press) to examine any public document unless that document has been exempted by special legislation (usually only those concerning national defense). This extraordinary "publicity principle,“ along with the legal Ombudsman and effective interest- group organization and administrative representation, provides the Swedish citizen with unique non-political avenues of appeal against decisions of policy 189-192; 204-206). The general sin Sweden sketched abov economic, social, an described separately chronologically for 28 decisions of policy and administration which affect him (Andren 1964: 189—192; 204-206). The general similarities and differences between Norway and Sweden sketched above serve as the setting for the history of the economic, social, and political development of each country, described separately in the next two chapters, but integrated chronologically for subsequent comparison with time-series data. A | Norway's social conflict between cen' tion(Rokkan 1967: 31 area demonstrated age named Hans Nielsen Ha uovenent against the church (Lafferty 197‘ nationalism over the of Norway to Sweden c section and its anti- the creation of the 1 developed into pro-So king tried to strengl Nessure in 1821 and The Norwegian Cc 101a wide suffrage c 40 percent of men ove Widest "estate" sys None in favor of ti CHAPTER TWO A DEVELOPMENTAL HISTORY OF NORWAY Norway's social history revolves around the multidimensional conflict between center and periphery stemming from Danish domina— tfion (Rokkan 1967: 389). As early as 1765 peasants in the Bergen area demonstrated against Danish tax policy. Around 1800 a merchant named Hans Nielsen Hauge led a national fundamentalist revival mmvement against the secularized and urban-oriented official state church (Lafferty 1971: 117). Despite the growth of Norwegian nationalism over the centuries of Danish domination, the transfer of Norway to Sweden engendered strong pro-Danish sentiment. This faction and its anti-Danish rival resembled political parties during the creation of the Norwegian Constitution in 1814. This soon ckweloped into pro-Swedish and anti-Swedish factions as the Swedish lfing tried to strengthen his dominance of the Union by military pnessure in 1821 and by legislative proposals from 1824 through 1884. The Norwegian Constitution, very liberal for its time, provided for a wide suffrage compared with other European nations, with over 40 percent of men over age 25 eligible to vote. But it employed the indirect "estate“ system of representation, weighted as much as five 'uoone in favor of the “urban“ Estates, the burghers and the officials. 29 Although a “Peasant that of the rural P6 proportions were far portional suffrable d efforts at mobi 1 ill" culminated in 1833 N that continued to er succeeded in establi hen during the next urban and rural lowe temperance and reli g of the Union (Rokkan 119). Most of the his idticulture dominated nineteenth century. andkmerica in the do geoIiilhic limitation industries were smal‘ 9””115- In Sweden as ”'b'fOtce consisted themselves solely by lendence hampered the diodes. During the c increased and the 3'01 30 Although a "Peasant Clause" limited urban representation to one-half that of the rural Peasant Estate, at that time the actual demographic proportions were far less. Despite a continuing decline in the pro- portional suffrage due to increasing numbers among the disenfranchised, efforts at mobilizing a rural opposition to urban and Swedish dominance culminated in 1833 with a protest movement led by Ole Gabriel Ueland that continued to erode the disparity in national representation and succeeded in establishing local self-government councils in 1837. Then during the next two decades a movement swelled among both the urban and rural lower classes for a new language, cultural revival, temperance and religious fundamentalism, and against Swedish domination of the Union (Rokkan 1967: 368-374, 379, 386-387; Storing 1963: 117— 119). Most of the historical literature on Scandinavia emphasizes that agriculture dominated economic life until the last two decades of the nineteenth century. Scandinavia lagged behind her European neighbors and America in the development of industry, largely because of the geographic limitations mentioned previously. The few existing industries were small in scale, craft-oriented and organized into guilds. In Sweden and especially Norway much of the early industrial wmrk-force consisted of part-time peasants who could not sustain themselves solely by small-scale farming and fishing. Their inde- pendence hampered the ability of the guilds to monOpolize their trades. During the early 1800's the ratio of journeymen to masters increased and the journeyman often organized to negotiate, and occasionally to 5t” conditions (Galenson In 1848 Marcus withamixture of sc events on the Conti r tions" composed P81” journeymen, achievir agitation for legisl and for extension of some strike activity oilhrane and other forerunner of Norway Philanthropic worker lectual organizatior cooperative purchasi Meanwhile, the teachers and other p 5"”th under the 1 Created in 1859, was edoctoral means of " of the "Peasant Clau h the 1860's: only 10 percent bothered Joined forces with t m 1865 to pressure 31 cmcasionally to strike for, better wages, benefits and working conditions (Galenson 1952: 107-111). In 1848 Marcus Thrane, an unemployed Norwegian school teacher wfith a mixture of socialist and religious beliefs and influenced by events on the Continent, organized a network of "workers' associa- tions“ composed primarily of rural peasants and laborers as well as journeymen, achieving a membership of over 20,000 by 1851. Their agitation for legislative protection of landless agricultural labor and for extension of the suffrage included mass demonstrations and some strike activity, resulting in the arrest and later emigration of Thrane and other leaders of the waning movement, considered the fOrerunner of Norway's leftist labor and political organizations. Philanthropic workers' societies sponsored by religious and intel— lectual organizations continued to provide insurance funds and cooperative purchasing during the next two decades (Galenson 1949: 7). Meanwhile, the urban liberal opposition, representing lawyers, teachers and other professional groups, was organizing in the Storting under the leadership of Johan Sverdrup. His "Reform Club," created in 1859, was legislatively ineffective due to the lack of electoral means of "party" discipline. Despite strict enforcement of the "Peasant Clause," political participation reached its low ebb in the 1860's: only 30 percent of adult males could vote and only 10 percent bothered to do so. But in 1869 the reform representatives joined forces with the "Friends of the Peasant" movement, organized in 1865 to pressure rural voters to participate and electors to select liberal national 1‘9 soon developed '"to Party (actually V9” and rural areas, 611' periphery 81131"St C' Norway's 08515 but quickly fading ' emigration to the i the Liberals pressei legislatively they : the Viceroy with a l and attacked the con Stunting and the K11 the king's Council ' veto. In accordanm kissed in 1880 for ' every three years, ‘ amendment to be val‘ t“Neath eleven mm Conservatives had a' StrOldest in the ea: already beginning to TheElection of 1881 it turned out twice 1 “Vote by Purchasir 32 liberal national representatives. This urban-rural reform alliance soon developed into Norway's first true political party, the Liberal Party (actually Venstre, the Left), strongest in the western regions and rural areas, and embodying all aspects of the opposition of periphery against center (Rokkan 1967: 374-375, 387, 391). Norway's first trade unions formed in 1872, instigating strikes but quickly fading due to a recession, geographic dispersion, and emigration 'u) the United States (Galenson 1949: 8). Meanwhile, the Liberals pressed for voter registration and participation. Legislatively they succeeded in forcing the Swedish King to replace the Viceroy with a Minister of State for Norway (Storing 1963: 29—30), and attacked the constitutional separation of powers between the Storting and the King, repeatedly proposing an amendment to involve the King's Council in Storting deliberations, despite his repeated veto. In accordance with the Constitution, after the amendment passed in 1880 for the third consecutive Storting session, held every three years, the Liberal-dominated Storting declared the amendment to be valid law without the King's consent and proceeded to impeach eleven members of his Cabinet. By then the pro-Swedish Conservatives had also organized a political party (Hoyre, the Right), strongest in the eastern regions and growing urban areas, which were already beginning to be underrepresented by the “Peasant Clause." The election of 1882 became Norway‘s first truly partisan election. It turned out twice as many voters as in 1879, many of whom qualified to vote by purchasing worthless strips of real estate, and returned arm-thirds fiber“ struggle after a .Y‘ preserve the Union Cabinet, firmlli 95' government in Nomi partially extended new criterion of m‘ entranchising econt ignoring the scattc 1981: 315-383). During the 182 reappeared, primari oijob-seekers from enloved the support to enact full suffv working day caused and the Moderates. soon absorbed into alternated control he formation of if if reform social ism created another spl Dirty t0 Support. acComplnilntl the de independent strikes 33 a two-thirds Liberal majority, sufficient to win the impeachment struggle after a year. To prevent civil war in Norway and to preserve the Union, the King in 1884 asked the Liberals to form a Cabinet, firmly establishing the principle of parliamentary cabinet government in Norway. In that same year the new government partially extended the suffrage by approving legislation adding a new criterion of minimum income to existing voting requirements, enfranchising economically mobilized urban workers but largely ignoring the scattered and unmobilized rural proletariat (Rokkan 1967: 375-383). During the 1880's seventy local and two national trade unions reappeared, primarily to protect urban workers against the influx of job-seekers from the countryside. At first the Liberal Party enjoyed the support of many of these organizations, but its failure to enact full suffrage and such pro-labor legislation as a ten—hour working day caused the party to split into two factions, the Pures and the Moderates. The latter, concentrated in the Southwest, were soon absorbed into the Conservative Party, and the Left and Right alternated control of the government for the next five decades. The formation of the Norwegian Labor Party in 1887, with a program of reform socialism under the leadership of Christian Knudsen, created another split in the labor union ranks over which political party to support. A surge of economic growth from 1887 to 1889 accompanying the development of foreign trade spawned several independent strikes, and the government's use of imported strikebreakers and further split the 9 The Pure facti Norwegian indellende on foreign affairs separate trade rela lemony) based prim and Norway's mercha (tindgren1959: 38- in 1891 the Pures p aseparate consular vetoed. This strug issues spurred mobi highest level ever in the election of the consular issue tariff law was perm N looser resist an. extension of the su inslicence law in 18! "mm“ Poor~re1ie' over 25 not on Debt But the "GNU enfrau heavily at first N leg ‘ acned unrepresent 34 strikebreakers and police suppression strengthened the radicals and further split the union movement (Galenson 1949: 8). The Pure faction of the Liberal Party also favored greater Norwegian independence from the Swedish King in determining her own foreign affairs. The two countries had been developing separate trade relationships (Norway with Britain and Sweden with Germany) based primarily on their expanding lumber industries, and Norway's merchant marine was becoming far superior to Sweden's (Lindgren 1959: 38-42). Upon winning a clear electoral victory in 1891 the Pures passed legislation in the Storting establishing a separate consular service for Norway, which the King promptly vetoed. This struggle for national identity and other partisan issUes spurred mobilization of political participation to its highest level ever (over 90 percent turnout of eligible voters) in the election of 1894, returning another mandate for the Pures. The consular issue remained deadlocked, and the countries' common tariff law was permitted to lapse in 1897. But the Storting could no longer resist the pressure for social legislation and further extension of the suffrage. It enacted an industrial accident insurance law in 1894, care for neglected children in 1896, and a national poor-relief law in 1900. In 1898 it enfranchised all men over 25 not on public assistance or in bankruptcy proceedings. But the newly enfranchised lower-class voters did not participate heavily at first, reducing the total turnout. The Labor Party remained unrepresented in the Storting and thus supported the programs of the PU 384-386). From 1889 to from about 3,000, I toms and cities. involved primarily to finance the num unions of locals i | geographic dispers trialization, as 1111 workers. This dua‘ labor ranks, espeC‘ Scandinavian labor organizations to is ”hidden Federatic than the Federatior third of Norway's v large national unit ‘ l honscentral stri central councils. Association, formec 1904, and the Feder hookers. But becau St1099. the Federat iiiiliation of its 35 programs of the Pure Liberals (Storing 1963: 30-31; Rokkan 1967: 384-386). From 1889 to 1899 the number of organized workers had grown from about 3,000, mostly in Oslo, to about 20,000, mostly in the towns and cities. ‘Efforts at inter-union labor organization had involved primarily city central councils wielding permanent funds to finance the numerous strikes of that period. But national unions of locals in one craft or industry also developed despite geographic dispersion and despite Norway's relatively late indus- trialization, as measured by the proportion of truly industrial workers. This dual organization created further friction in the labor ranks, especially during strikes; but in 1897 an inter- Scandinavian labor congress held in Stockholm urged the various organizations to join together in national federations. The Norwegian Federation of Labor emerged in 1899, more centralized than the Federations in Sweden and Denmark but with less than a third of Norway's workers affiliated because of the refusal of several large national unions to accept direct assessments for the Federa- tion's central strike fund and automatic affiliation with the city central councils. In the face of lockouts by a strong Employer's Association, formed in 1900, these unions returned to the fold in 1904, and the Federation finally represented a majority of Norway's workers. But because support for the Pure Liberals was still strong, the Federation also initially failed to secure automatic affiliation of its locals with the Labor Party, although it did establish shared 61 organizations and l in the ill-seat Stt labor representatio from urban areas he industry (Rokkan 15 indicates that More retained some eleme rather than complel The issue of 1 election of 1903, i together temporaril leader, Christian N service for Norway tions broke down, a 88111 creating a 5 b1 the Swedish King killing to form a g This the Storting a mi) Popular refer to i‘eSune control . agreed to abide by 1100 favored disso “0'80 and after an. a ' . nonvvtatgon to ash 36 establish shared executive representation between the two organizations and helped the party finally secure representation in the ll7-seat Storting in 1903 (Galenson 1949: 8-16). All four Labor representatives came from the northern periphery rather than from urban areas because of the dispersed nature of Norway's early industry (Rokkan 1967: 394; Valen and Katz 1964: 22-24). This indicates that Norway's emerging socio-economic class cleavage retained some elements of the fundamental center-periphery cleavage, rather than completely cross-cutting it (Rokkan 1966: 73-90). The issue of the union with Sweden still dominated the election of 1903, in which the Liberals and Conservatives joined together temporarily in a coalition government under the Liberal leader, Christian Michelsen, pledged to secure a separate consular service for Norway through negotiation. But in 1904 the negotia— tions broke down, and in 1905 the coalition government again passed a bill creating a separate consular service. Again it was vetoed ln/the Swedish King, who could not find a Norwegian party leader willing to form a government after the Michelsen cabinet resigned. Thus the Storting declared the Union dissolved, called for Norway's first popular referendum to ratify its decision, and asked Michelsen to resume control. The Swedish parliament decided to avert war and agreed to abide by the verdict of the referendum, which overwhelm- ingly favored dissolution. The Swedish King abdicated as King of Norway and after another plebiscite in 1905 a Danish prince accepted an invitation to assume the Norwegian Throne (Storing 1963: 32-34). Even during ‘ hinternal socia' lotion providing 1 loans for rural in laws, so called be country's natural pour relief in 190 meet compensation . for some industri e benefits to includ enchant-marine fl sive system of hea use of forest land finally, on the W system of represen‘ based on single-m plurality run-offs suffrage to includ basis as men in 19 Most of the l 1905 also marks th indicated by heavy lover, electrochem expansion of the r the purely industr 37 Even during the Union crisis Norway had turned its attention to internal social problems. In 1903 the Storting enacted legis- lation providing for subdivision of large rural estates and building loans for rural improvement. This was the first of the Concession Laws, so called because they established public dominion over the country's natural resources. The Storting also provided further poor relief in 1904, care of foster children in 1905, and unemploy— ment compensation, employment agencies, and a lO-hour working day for some industries in 1906. It extended industrial accident benefits to include fishermen in 1908 and seamen in the vital merchant-marine fleet in 1911. In 1909 it instituted a comprehen— sive system of health insurance and regulation of the acquisition and use of forest land (Nordskog 1935: 110-115; Storing 1963: 181). Finally, on the political front it replaced the indirect "estate" system of representation in 1905 with a system of direct elections based on single-member constituencies with majority decisions and plurality run-offs (Valen and Katz 1964: 19). It also extended the suffrage to include women on an income basis in 1907, and on the same basis as men in 1913 (Rokkan 1967: 385-386). Most of the literature on Norway asserts that the period around 1905 also marks the beginning of rapid industrialization, as indicated by heavy investment in the development of hydroelectric power, electrochemical plants, paper mills and metal refineries; by expansion of the road and railroad networks; and by a doubling of the purely industrial work-force between 1905 and 1920 (Galenson 1952: 107). Much areas of the rura industrial constrl unskilled farmers, opportunity for gr smelled the ranks federation of Labo 60,800 in 1912, an increased in that while its share of 32 percent in 1915 carved two new spl' lurker Democrats iv 1909(Na1en and Kat The shock of the rigors of indu generated a great owners. Affiliate vorkers came under led by Martin Tra influenced by the if french syndicali in Norway and Tran collective bargaini unsuccessful 1y t0 9 38 1952: 107). Much of the industrial plant was located in isolated areas of the rural periphery, drawing much of the labor force for industrial construction and operation from mostly young and unskilled farmers, farm laborers, and fishermen attracted by the opportunity for greater income and a new way of life. They swelled the ranks of the labor union movement and the Labor Party: Federation of Labor membership increased from 25,300 in 1906 to 60,800 in 1912, and the Labor Party's Storting delegation increased in that period from 10 to 23 (Galenson 1949: 59—61), while its share of the vote doubled from 16 percent in 1906 to 32 percent in 1915. Industrialization and increased suffrage also carved two new splinter parties from the Liberals: the leftist Worker Democrats in 1905 and the rightist National Liberals in 1909 (Valen and Katz 1964: 24—25). The shock of the transition from traditional environments to the rigors of industrial work and frontier boom-town living generated a great deal of resentment against management and owners. Affiliated primarily with the Laborers' Union, the new workers came under the influence of a semi—syndicalist movement led by Martin Tranmael, who had been to the United States and was influenced by the International Workers of the World movement and by French syndicalism. But parliamentarianism was firmly established in Norway and Tranmael accepted the need for electoral activity and collective bargaining. He and his followers first attempted unsuccessfully to gain control of the Federation of Labor, controlled by the tradition 1949: 16-23). Although Nov not escape its at index tripled frc responded with or Concession Laws t limestone deposit mines, and other ing industrial wo tion, and institu children. In 191: rent, and another instituting some r eight-hour working front it replaced system of multi-m method of proport the development 0 Party, composed o to small farmers. Party had by now but also cocnnerci 19-20, 26-27). The economic Nrticularly hard. 39 by the traditional craft unions, at its congress in 1910 (Galenson 1949: 16-23). Although Norway remained neutral during World War I, she did not escape its economic impact: production declined and the price index tripled from 1914 to 1920. During that period the Storting responded with another wave of social legislation, extending the Concession Laws to cover the acquisition and use of peat-bog areas, limestone deposits, mountain tracts, cultivated land, and waterfalls, mines, and other real property. In 1915 it passed a new law regard— ing industrial working conditions, extended unemployment compensa- tion, and instituted public care and education for handicapped children. In 1918 it passed a law regulating housing conditions and rent, and another regulating non-industrial working conditions and instituting some minimum wages. Finally, in 1919 it established the eight-hour working day (Nordskog 1935: 100-115). On the political front it replaced the single—member electoral system in 1920 with a system of multi-member districts allocating seats by the d'Hondt method of proportional representation. This further encouraged the development of small parties, and in that same year the Agrarian Party, composed of big farmers opposed to radical policies favorable to small farmers, split from the Liberal Party. The Conservative Party had by now become the spokesman for not only civil servants but also commercial and industrial interests (Valen and Katz 1964: 19-20, 26-27). The economic conditions of that period hit the working class particularly hard. Tranmael's movement gained control of several local units of t activity increas lation establish coaupulsory arbitn ageneral protes it after a week. and the Federatic heartedly in 1911 to force further the Russian Revol that succeeded in and securing its labor unions. In socialist manifes International and calling for revol congress approved councils and reo However, tha severe economic r in the first year percent and membe 142,600 to 96,000 supported Tranmae bargaining positi agreements advant local units of the Federation and the Labor Party, and strike activity increased. In 1915 the Liberal government passed legis- lation establishing a Labor Court and mediation boards for compulsory arbitration of labor disputes, and Tranmael called for a general protest strike, but the Federation leadership terminated it after a week. Mass demonstrations against inflation occurred and the Federation urged government action, which came half— heartedly in 1917, and again Tranmael's call for a general strike to force further action met Federation opposition. Influenced by the Russian Revolution, Tranmael's movement created worker councils that succeeded in 1918 in taking over control of the Labor Party and securing its membership by collective affiliation of individual labor unions. In 1919 the Federation and the Party issued a joint socialist manifesto and the Party severed its ties with the Second International and joined the Communist International, with Tranmael calling for revolution and dictatorship. In 1920 the Federation congress approved Tranmael's program of establishing central trade councils and reorganizing local craft unions along industrial lines. However, that program was never fully implemented, because a severe economic recession began in late 1920 and lasted until 1933. In the first year unemployment increased from 2.3 percent to 17.6 percent and membership in the Federation of Labor dropped from 142,600 to 96,000, mostly among the new unskilled laborers who supported Tranmael. These conditions adversely affected labor's bargaining position by making long-term collective bargaining agreements advantageous, rather than the weapons of syndicalism. igeneral strike even voted in fa in 1922 (Galenso The Party fr demands for strio collective affil' party, “democrati expulsion of dis: parliamentary act of the Party had Party with eight and in late 1923 lbscow ultimatum tunintern. The p the immunist Par leaving the origi of 1924 the Labor nunists and eight Party still consi soon it resumed i iavocrats in 1927 waists won only lost ground there 10er liar II. Be Party formed the 41 A general strike in 1921 failed, and the Labor Party representatives even voted in favor of further compulsory arbitration legislation in 1922 (Galenson 1949: 23-27). The Party found it difficult to live with the Comintern's' demands for strictly controlled individual membership rather than collective affiliation of labor unions, subordination of unions to party, "democratic centralism" rather than a federative structure, expulsion of dissidents, armed revolution, and rejection of parliamentary action, previously proven successful. The right wing of the Party had broken off in 1921 to form the Social Democratic Party with eight out of a total of 150 Storting representatives, and in late 1923 the Labor Party itself refused to accept the I Moscow ultimatum for reorganization and was expelled from the Comintern. The pro-Moscow element immediately broke off and formed the Communist Party with thirteen Storting representatives, leaving the original Labor Party with fifteen. But in the election of 1924 the Labor Party won 24 representatives to six for the Com— munists and eight for the Social Democrats. At first the Labor Party still considered itself a revolutionary communist party; but soon it resumed its reformist nature and reunited with the Social Democrats in 1927, winning 59 Starting representatives. The Com- munists won only three representatives in that election and steadily lost ground thereafter except for a brief resurgence following World War II. Being the largest party in the Starting, the Labor Party formed the first socialist government in Norway. But it lasted only one month i banking and fisc During this significant soci. 1923 it passed a persons over age include real prop another Concessir missions to regu' rellllating restra it established a 118-120). In 1926 the the prohibition c referendum in 19] massive mobilizat reh.‘llous Values and the Agrarians to accomplish mm; the creation of t lrimaniiy at the ‘ organizations New 1933 Of the Nathl 42 only one month in 1928 because of a parliamentary battle over banking and fiscal policy (Galenson 1949: 61-69). .During this trying period for the political labor movement significant social legislation still got through the Storting. In 1923 it passed a law establishing old-age pensions for needy persons over age 70. In 1925 it extended the Concession Laws to include real property in fishing areas, and in 1928 it enacted another ConcesSion Law establishing a system of district land com- missions to regulate the sale of land. In 1926 it enacted a law regulating restraint of competition and price-fixing, and in l928 it established a state grain monopoly (Nordskog 1935: 100—115, 118-120). In l926 the voters approved a special referendum abolishing the prohibition of alcohol that had been instituted by a previous referendum in 1919. The election of 1930 thus saw the first massive mobilization of women in "defense of traditional moral and religious values against the threats of secularism and Socialism” (Rokkan 1967: 398-399). The Labor Party lost 12 Storting seats and the Agrarians formed their first government, but were unable to accomplish much because of their minority position, thus causing the creation of the new Christian People's Party in l933, again primarily at the expense of the declining Liberals. Dut fascist organizations were also emerging, culminating in the creation in 1933 of the National Socialist Party, led by Vidkun Quisling. In the face of these threatening developments the now thoroughly domesticated Lat across on a pla absolute majorit port of the Agra acceded to power class and center prices and incre social welfare v facilities and u pull out of the 1949: 69-70). However, th recovery, forcin london, where th was ruled by an appointed Prime Proved invaluabl Germans had been leconstructi on. the North Atlant' Stance (Storing ' The election With in absolute embarked on a pm social legislatic 43 domesticated Labor Party launced an all-out effort in the election of 1933 on a platform of economic relief and fell just short of an absolute majority in the Starting. But it eventually won the sup- port of the Agrarians by promising a favorable farm policy and acceded to power in 1935, again affirming the overlap between the class and center-periphery cleavages. The subsequent control of prices and increased government spending, especially in the area of social welfare via public works and the extension of credit facilities and unemployment and pension benefits, helped Norway pull out of the recession (Valen and Katz 1964: 29-30; Galenson 1949: 69-70). However, the German invasion of 1939 cut short this period_of recovery, forcing the King and most political leaders to flee to London, where they established a government-in-exile, while Norway was ruled by an occupation government superior even to the appointed Prime Minister Quisling. The Norwegian merchant marine proved invaluable to the Allies during the War, and after the Germans had been driven out in 1944 the Marshall Plan greatly aided reconstruction. Norway joined not only the United Nations but also the North Atlantic Treaty Organization, thus dr0pping her neutralist stance (Storing 1963: 35-38). The election of 1945 returned the Labor government to power with an absolute majority in the Storting, and it immediately embarked on a program of increased government spending, further social legislation and extensive economic planning. Production, income. enployme in the post-liar I the earlier precr nationalizing pro voluntary prograi massive public hr general disabilii rally, after 1945 three, and in 195 portional represe to the Sainte-Lag small parties (R0 increased its par Splinter party, t ”0”th the balan. entry into the Eur into confusion, or 11 Agrarians (no, and Christians 5p] longer count on u h” 1963 an adminis vote of no COHflde major opDOSltion p M” the sullllort socialist than the 44 income, employment and prices soared as well as government spending in the post-Mar era. Most of the social legislation merely modified the earlier precedent-setting laws, for example, extending benefits, nationalizing programs previously administered locally, or making voluntary programs compulsory. But the government also introduced massive public housing in 1945, family allowances in 1946, and general disability insurance in 1960 (Storing 1963: 181). Politi- cally, after 1945 elections were held every four years rather than three, and in 1952 the "Peasant Clause" was abolished and the pro- portional representation system was changed from the d'Hondt method to the Sainte-Lague method, further encouraging the development of small parties (Rokkan 1966: 88). But the Labor Party steadily I increased its parliamentary majority until 1961 when a new leftist splinter party, the Socialist Pe0p1e's Party, gained two seats holding the balance of power in the Storting. The issue of Norwegian entry into the European Common Market threw political alignments into confusion, with Labor and Conservatives in favor, Socialists and Agrarians (now called the Center Party) opposed, and Liberals and Christians split. The minority Labor government could no longer count on the support of any of the opposition parties, and in 1963 an administrative scandal over a mining disaster brought a vote of no confidence and a coalition government among the four major opposition parties. It lasted only four weeks because Labor re-won the support of the Socialists by shrewdly acting even more socialist than they. But in the election of 1965 Labor lost six seats and the op government, perh between left and policies did not gram, indicating politicization in Figure 2-1 : of Norway from 1! 45 seats and the opposition parties again formed a durable coalition government, perhaps signaling a new era of alternation in power between left and right. Nevertheless, the new government's policies did not differ significantly from the previous Labor pro- gram, indicating a process of "de-ideologization" if not de- politicization in post-War Norway (Rokkan 1967: 402-403). Figure 2-1 summarizes the important periods in the development of Norway from 1814 to 1965. Fhme2¥L Impc 196E Inde Swed Swed Legi Uelaz domh Loca‘ Thrai "Refc tive: “Frie Merge event Brief repla Final Parli cabin Conse elect turnm Liberi EXtEn: Re-emr leere former First 46 Figure 2-1. Important Periods in the Development of Norway, 1814- 1965. Dates Events 1814 Independence from Denmark. Constitution. Union with Sweden. 1821 Swedish military pressure to dominate Union. 1824-1844 Legislative pressure to strengthen Swedish dominance. 1833 Ueland rural movement against urban and Swedish dominance. 1837 Local self-gOvernment. 1848-1852 Thrane movement of "workers' associations." 1859 "Reform Club" of urban liberal Storting representa- tives. 1865 "Friends of the Peasant" movement. 1869 Merger of reform representatives and rural opposition, eventually forming the Liberal Party. 1872 Brief emergence of first labor unions. Viceroy replaced by Minister of State. 1880 Final Storting passage of amendment establishing parliamentary principle. Impeachment of King's cabinet begun. 1882 Conservative Party organized.. First truly partisan election: Two-thirds Liberal majority, doubled turnout. 1884 Liberal cabinet: Parliamentary principle established. Extension of suffrage by minimum income requnrement. 1880-1887 Re-emergence of some labor unions. 1887 Liberals split into Pures and Moderates. Labor Party formed. 1887-1889 First surge of economic growth. Strike activity. figure 2-1 (conti .——-————_— M has 1889-1899 Grow actii WM Eleca serV' hM Anotl 1894-1900 F lrsi hh Unive h” Feder iooo [mp], hM Labor 1903-1911 gem, 1903-1905 roan Swede WM Union insti 1905.19” SECOni Rapid Rise 1 ”hiM3 Women activi social 1 W9 Labor Peters 1 ho Prepay formed 47 Figure 2-1 (continued). Dates Events 1889-1899 Growth of unions. Inter-union organization. Strike activity. 1891 Election victory for Pure Liberals. Separate consular service first passed in Storting. 1894 Another victory for Pures, highest turnout ever. 1894-1900 First true social legislation. 1898 Universal manhood suffrage. 1899 Federation of Labor formed. 1900 Employers' Association formed. 1903 Labor Party first represented in Storting. 1903-1911 Second wave of social legislation. 1903-1905 Coalition government during crisis over Union with Sweden. 1905 Union with Sweden dissolved. Direct elections instituted. 1905-1913 Second surge of economic growth and union membership. Rapid industrialization. Rapid growth of Labor Party. Rise of Tranmael radicals in both unions and party. 1909-1913 Nomen enfranchised. 1914-1920 Rise in inflation, decline in production. Strike activity. Compulsory arbitration. Third wave of social legislation. 1917-1920 Tranmael radicals take over Labor Party and Federation. h919 Labor Party joins Communist International. Prohibition referendum passes. 1920 Proportional representation instituted. Agrarian Party formed. Figure 2-1 (conti Dues M 1004933 Seve 1004927 Decl 101 Gene righ 102 Furt 103 Labor breai 004927 Four' hhh Proh' 101 Socia ' large 004940 Unior First 1%0 Elect tion ”33 Chris (QUis in el ”33.1940 ECOnO 1%5 Labor Incre legis 1%0 Germa: 104 Libero 1 9% Labor Plan. 48 Figure 2-1 (continued). W W Dates Events 1920-1933 Severe economic recession. High unemployment. 1920-1927 Decline in union membership. 1921 General strike. Social Democratic Party breaks off to right. 1922 Further compulsory arbitration legislation. 1923 Labor Party expelled from Comintern. Communist Party breaks off to left. 1923-1927 Fourth wave of social legislation. l926 Prohibition abolished by referendum. 1927 Social Democrats re-join Labor Party, which becomes ‘ largest party in Storting. 1927-1940 Unions resume growth. 1928 First brief Labor government. 1930 Election setback for Labor due to first big mobiliza- tion of women voters. . 1933 Christian People's Party and National Socialist _ (Quisling) Party formed. Labor almost wans majority in election. 1933-1940 Economic recovery. 1935 Labor forms government with support of Agrarians. , Increase in government activity. Flfth wave of socnal legislation. 1940 German invasion. Government in exile. 1944 Liberation. 1945 Labor wins absolute majority in election. Marshall Plan. Figure 2-1 (conti M _——-—-——-— 005 M 1945-1965 Econ gove Elec 102 "Pea tion 00 Soci. major power 004963 Conti real' )%3 Brief scanc 1%5 Durat \ 49 Figure 2-1 (continued). Dates Events 1945-1965 Economic reconstruction. Further increase in government activity. Final wave of social legislation. Elections now every four years. 1952 "Peasant Clause" abolished. Lague method of propor- tional representation. 1961 Socialist People's Party formed. Labor loses majority in election. Socialists hold balance of power with two seats. 1961-1963 Controversy over European Common Market. Political realignment. 1963 Brief non-Labor coalition after administrative scandal. 1965 Durable non-Labor coalition with Laborite policies. Sweden's hi! and evolutionary the homogeneity ‘ European great-114 centuries, and a1 formidable 9409re history of social Sweden’s Parl i ame dominated the 90V 1112, during ivth and th 1 eCaps rese cries in England absolutism lastin losing Finland in In that year restored a limi to the French Marsha Swedish Throne. ' defeat Napolean, r CHAPTER THREE A DEVELOPMENTAL HISTORY OF SWEDEN Sweden's history is distinctive for its seemingly uneventful and evolutionary nature, a characteristic closely associated with the homogeneity of the nation's culture. Despite involvement in European great-power politics in the seventeenth and eighteenth centuries, and aided by her isolated position, resources, and formidable geography, Sweden has enjoyed a relatively continuous history of social development, uninterrupted by foreign domination. Sweden's Parliament, the Riksdag, has existed since 1435 and even dominated the government during the Era of Liberty from 1718 to 1772, during which two parliamentary factions known as the Hats and the Caps resembled political parties, similar to the Whigs and Tories in England. But a coup in 1772 brought a new period of royal absolutism lasting until 1809, when the King lost his throne for losing Finland in a war with Russia. I In that year the Riksdag wrote a new constitution that restored a limited monarchy sharing power with the Riksdag, and the French Marshall Bernadotte was invited in 1810 to assume the Swedish Throne. Then Sweden joined with Russia and England to defeat Napolean, obtaining Norway from Denmark in 1814 as 50 compensation for the King to consc impeached by the politically respc equal power to pa over finances, ti the partnership d lheRiksdag was h every five years into four separat Clergy, Burghers, the King (Board 1 Like Norway, 10 other European “though somewhat still few to numb. 0(linized into gu mmphtiht becaug 110"» In additior suitable to large. Small and incapabl less hihorers, th the ratio of 101m tourism," from be to care adequateU 51 compensation for the loss of Finland. The new constitution required the King to consult with his appointed Council, which could be impeached by the Riksdag, but which was neither chosen by, nor politically responsible to, that body. Although the Riksdag shared equal power to pass laws with the King and had exclusive control over finances, the King and his administrative officials dominated the partnership during the first half of the nineteenth century. The Riksdag was handicapped by the fact that until 1844 it met only every five years (afterwards every three), and that it was divided into four separate chambers representing the Four Estates (Nobility, Clergy, Burghers, and Peasants), making it difficult to unite against the King (Board 1970: 19-28). Like Norway, Sweden remained economically backward relative to other European nations throughout most of the nineteenth century. Although somewhat more developed than in Norway, industries were still few in number, small in scale, geographically dispersed, and organized into guilds that were unable to monopolize their trades completely because of constraints on communications and transporta— tion. In addition, although Sweden's topography is somewhat more suitable to large-scale farming than Norway's, most farms were still small and incapable of completely sustaining the peasants and land— less laborers, who thus worked part—time in small industries. Also, the ratio of journeymen to masters increased, preventing most journeymen from becoming masters and rendering old masters unable to care adequately for their journeymen and their dependents. The tuild mon0110l ies workers' trade a (Galenson 1952‘ thus hasan 15° 5h the Riksdag 14$S was administered Pressure 1‘0 the 1840's, comi burghers, but al. the administrati‘ supported reform plan until 1866, de beer, got the to meet annual 1y; house, indirectly to wealth, became especially the 01 but by a very lin became the libera fact that represe towns. In the ye 30 percent of mal interest was rath parties and signi “1111116 bothered 52 guild monopolies were abolished in 1846, replaced by philanthropic workers' trade associations, not unions in the modern sense (Galenson 1952: 108-110; Carlson 1969: 14-16). Care for the needy thus began to shift more heavily to the government, and in 1847 the Riksdag passed the first Poor Relief Act, which, as in Norway, was administered locally (Rosenthal 1967: 9). Pressure for reform of the Riksdag itself also swelled during the 1840's, coming not only from the peasants and entrepreneurial burghers, but also from landholding noblemen envious of the power of the administrative nobility. After 1848 even the King and officials supported reform, but the various factions could not agree upon a plan until 1866, when the Minister of State for Justice, Baron Louis de Geer, got the Riksdag to reorganize itself into two chambers and to meet annually, despite opposition from the clergy. The upper house, indirectly elected in local assemblies apportioned according to wealth, became the conservative stronghold of the nobility, especially the officials; whereas the lower house, directly elected but by a very limited franchise based on land ownership or income, became the liberal stronghold of primarily the farmers, despite the fact that representation was weighted four to one in favor of the towns. In the years after the bicameral reform of 1866 only about 20 percent of males over age 21 could vote, and because political interest was rather low during that period due to a lack of political parties and significant events, only about 20 to 25 percent of those eligible bothered to vote. These unfavorable electoral conditions led to the format the Ruralist Part restraint and chi military conscrip organized the shc liberal Party spl in 1811 over the unsettled for dec only in the Riksc (Rustow 1955: 14. The De Geer private capital is )"dUSthli which n hon 1867 to 1888 and the replaceme firms ”"1 by prof “”59"th conditi th91860's and i8 arecession begin societies, Consum d"Alter. 141 In 18' complete victory . 53 led to the formation in 1867 of Sweden's first political party, A the Ruralist Party, whose platform advocated primarily fiscal restraint and changes in the inequitablemethods of taxation and military conscription. In that same year the De Geer government organized the short-lived Ministerial Party. In 1868 the New Liberal Party split from the Ministerial Party, only to disband in 1871 over the conscription and taxation issue, which remained unsettled for decades to come. These early parties were organized only in the Riksdag and did not succeed in mobilizing the electorate (Rustow 1955: 14-35; Board 1970: 28). The De Geer government had been encouraging the development of private capitalism and foreign investment, especially in the lumber industry, which was being fed by the industrialization of Europe. From 1867 to 1888 the number of purely industrial workers doubled, and the replacement of the traditional guild organizations with firms run by profit-oriented managers for absentee owners meant worsening conditions for the workers. In addition, the Riksdag in 1871 restricted the scope of the Poor Relief Act of 1847. During the 1860's and 1870's several craft unions began to organize, but a recession beginning in 1875 retarded their development. Suffrage societies, consumers' cooperatives, Social Democratic clubs, temperance societies and Nonconformist religious organizations also developed. In 1879 Nonconformist workers instigated Sweden's first major strike in the sawmill area around Sundsvall, resulting in complete victory for the employers but creating a great deal of resehhle"t among first labof'me' craft 111110"S and central 411d 5mm from Europe V" ' 1880's, and "19C“ adhere to liberd' tion also began ' unions, designed political activii emerged in 1885 h Rustow 1955: 45-‘ Numerous P91 coupled with over spawned a wave of and late 1880's. tariffs throughou Sweden. Organize and in that same belda special e1 scheduled electio far, about half 0 aclear mandate b liver house, part than their entire 54 resentment among Sweden's workers. In 1881 the Riksdag passed its first labor-oriented legislation, a child-labor law. Many new local craft unions and political labor organizations, located mainly in central and southern regions and influenced greatly by ideas imported from Europe via the Danish labor movement, Sprang up in the early 1880's, and much bickering occurred over whether the movement should adhere to liberal or socialist philosophies. Inter-union organiza- tion also began in the 1880's: The first city central council of unions, designed primarily to administer strike funds and engage in political activity, formed in 1883 and the first national union emerged in 1886 (Galenson 1952: 112—114; Carlson 1969: 16—20, 30; Rustow 1955: 45-48: Rosenthal 1967: 9; Verney 1957: 106). Numerous periods of economic recession from 1875 to 1895, coupled with overpopulation among the landless rural proletariat, spawned a wave of emigration that reached its peak in the early and late 1880's. Falling prices were also accompanied by rising tariffs throughout Europe, and pressure for retaliation mounted in Sweden. Organizations for and against tariffs developed in 1887, and in that same year the lower house of the Riksdag dissolved and held a special election on the issue in addition to the regularly scheduled election. Voter turnout reached its highest level thus far, about half of those eligible, and the anti-tariff forces won a clear mandate but lost the issue to the conservative, proétariff upper house, partly because of an electoral technicality that cost them their entire Stockholm delegation. The dispute caused a complete politic luraliSts 5th free-l"ade 01d R aprotectiOnist i 180 the We“ Gustaf BOStrom’ mise solutiOn to relieved the burr military, and ha) rapidly 00")"9 ' urban seatS (RUS‘ Deshlte the first state-SUPP‘ that same year tl followed in 1890 the Riksdag inst' insurance 0001")2 Suffragists 1'01"1 reform petitionS excepta show of political labor 11 with the legal 911 since 1864, helpe Socialists, relat luralist Party re complete political realignment in the Riksdag in 1888: The Ruralists split into a protectionist New Ruralist Party and a free-trade 01d Ruralist Party, while the conservatives split into a protectionist Majority Party and a free-trade Center Party. In 1892 the protectionist alliance, under the leadership of Erik ‘ Gustaf Bostrom, increased industrial tariffs, instituted a compro- mise solution to the old conscription and taxation issue that relieved the burdens on the farmers while strengthening the military, and halted the increasing overrepresentation of the rapidly growing urban areas by setting a fixed ratio of rural and urban seats (Rustow 1955: 35-42). Despite the tariff dispute the Riksdag in 1889 enacted its first state-supported industrial accident insurance system. In that same year the Social Democratic Labor Party was founded, followed in 1890 by the Universal Suffrage Association. In 1891 the Riksdag instituted government regulation of private health insurance organizations. In 1893 and 1896 the Socialists and Suffragists jointly sponsored "people's parliaments" that presented reform petitions to the King and Riksdag without immediate results except a show of strength. This simultaneous development of the political labor movement and the liberal suffrage movement, along with the legal guarantee of freedom of assembly and association since 1864, helped moderate the philosophy and actions of the Socialists, relative to other European labor parties. In 1895 the Ruralist Party reunited, although it retained conservative ties and never regain same 1194'" the Ce with disaffected Party. With sup sent their first 1955: 48-58; Pet Around 1895 expansion i ncrea and by the develi lhe number of 1111' to 1902, and bet1 from 15,000 to 6| industrialization and comunicatior Because these def. entrenched, the s organized along ‘ councils decl inec Scandinavian labc federation of Lal authority to fine federation requi r th th 9 Social Democr e netalvorkers, 1900 the requiren 56 and never regained the complete loyalty of rural Sweden. In that same year the Center Party disbanded and its liberal wing joined wfith disaffected 01d Ruralists to create the reformist People's Party. With support from these liberals the Socialists in 1896 sent their first delegate to the Riksdag, Hjalmar Branting (Rustow 1955: 48-58; Peters 1970: 304-306). Around 1895 prices in Sweden began rising as industrial expansion increased, aided by technical assistance from England and by the development of the technology of hydroelectric power. The number of purely industrial workers doubled again from 1888 to 1902, and between 1895 and 1900 labor union membership soared from 15,000 to 66,000 as uncoordinated strike activity increased. Industrialization also brought improvements in the transportation and communication links between geographically dispersed industries. Because these developments began before craft unions became firmly entrenched, the greatest growth occurred in the national unions organized along industrial lines, and the role of the city central councils declined. In 1898, following a resolution of the inter- Scandinavian labor conference, several unions founded the Swedish Federation of Labor, a decentralized organization with limited authority to finance lockouts but not strikes. Initially the Federation required each member union to affiliate collectively with the Social Democratic Party, causing several large unions, especially the metalworkers, to refrain from joining the Federation, and in 1900 the requirement was dropped. Despite the objections of the labor movement 1 with stri kebreak passed a new chi hmmmm19 workers. For t1 general strike 1 awnnemmo industries were Carlson 1969: 21 in the elec 1900 on the basi 0f the Pe0p1e's halt) in the low lhe issue of pro tires hopeful of ihined the issue 041 of the tart Ruralists form e d Mnmumu ”1905 the Libel with 13 Seats, nu Peacefully the C] the Libepaps form an," bill that 57 labor movement the Riksdag in 1898 made it illegal to interfere with stri.kebreakers,often imported from Denmark. But in 1900 it passed a new child-labor law that also regulated women's working hours, and in 1901 it required employers to compensate injured workers. For this reason and in response to'a socialist-led general strike lasting three days, the Employers' Association and a separate employers organization for the metal and building industries were formed in 1902 (Galenson 1952: 114-116, 134; Carlson 1969: 21-26, 30). In the election of 1902 the new Liberal Party, founded in 1900 on the basis of the growing electoral and petition success of the People's Party, surpassed the Ruralists as the largest party in the lower house of the Riksdag with 106 of 230 seats. The issue of proportional representation, supported by conserva— tives hopeful of party-splintering, and thus opposed by Liberals, joined the issue of universal suffrage in supplanting the supre— nmcy of the tariff controversy. The conservatives and the Ruralists formed a broad "General Voters' Alliance" in 1904 in an effort to stem the rising tide of liberalism. But in the election of 1905 the Liberals, still with 106 seats, and the Socialists, with 13 seats, won a bare majority in the lower house of the Riksdag. After a brief coalition government was formed to settle peacefully the crisis over dissolution of the Union with Norway, the Liberals formed a new cabinet and introduced an electoral reform bill that passed the lower chamber but not the upper chamber. Karl 1 to dissolve the refused and the In 1907 the con: acompromise ale to all men over and had paid the member district: method for both indirect electi< chambers, but be amendments, rece has required. 1 their delegation asPllt of a N31 flee-trade Moder 1905: gained 31, servatives were conservative cat dence. The refc electorate and r (Rustow 1955: SE Following a pl‘lllCiple of col work industry, t 58 chamber. Karl Staaff, the Liberal Prime Minister, asked the King to dissolve the Riksdag and call new elections, but the King refused and the Staaff cabinet resigned, replaced by conservatives. In l907 the conservative Prime Minister, Arvid Lindman, introduced a compromise electoral reform bill that granted universal suffrage to all men over age 24 who had performed their military service and had paid their taxes, and which introduced a system of multi- member districts and proportional representation by the d'Hondt method for both the direct elections of the lower house and the indirect elections of the upper house. The bill passed both chambers, but because some of the changes involved constitutional amendments, reconfirmation after the next lower-chamber election was required. In the election of l908 the Socialists increased their delegation from l3 to 34 while the conservatives, despite a split of a National Progressive Party from the Ruralists and a free-trade Moderate Party from the conservative alliance after l905, gained six and the Liberals lost six. The Liberals and con- servatives were now equal in strength, but the King retained the conservative cabinet despite continued expressions of no confi- dence. The reform bill passed again in 1909, doubling the electorate and necessitating mass electoral party organization (Rustow l955: 58-78; Verney l957: T74). Following a major strike by the metalworkers in l905 the )rinciple of collective bargaining became accepted by both mmloyer and employee organizations, not only those in the metal- mrk industry, but also the Federation of Labor and the Employers' Association. arbitration ma recession beg‘ and curtailed the last major ageneral str‘ some areas. I several union: weakened, and United States destroy the m for almost a a rellllations fl and supervisia the smdrcan: the Federation Struction, f0] Conditions p0, advocacy of d‘ action, the S, bargain COllea strengthening organl‘Zdtion r till l967: 7)‘ The Libel plied to becm 59 Association. In l906 the government established non-compulsory arbitration machinery available for labor disputes. An economic recession beginning in l908 reduced wages, boosted unemployment, and curtailed union membership for the first time, resulting in the last major attempt to solve a labor dispute by open conflict, a general strike and lockout in l909 lasting up to six months in some areas. It was a major defeat for the union movement, as several unions disbanded, Federation leadership and control weakened, and many blacklisted members emigrated to Norway or the United States. But the employers did not seize the opportunity to destroy the movement completely, although it did not fully recover for almost a decade. In 1909 and l910 the Riksdag even passed new regulations for factory working conditions and a law subsidizing and supervising private health insurance organizations. In l9l0 the syndicalist Swedish Workers' Central Organization seceded from the Federation of Labor, appealing primarily to workers in con- struction, forestry, and mining, where wages were lowest and conditions poorest. But primarily because of their impractical advocacy of direct strike action rather than cooperative political action, the syndicalists attracted few members and were forced to bargain collectively. Another result of the l909 disaster was a strengthening of the movement for industrial rather than craft organization (Galenson l952: ll8-l54; Carlson l969: 27-30; Rosen— thal l967: 7). The Liberals and Socialists were organizationally best pre- pared to become mass parties as necessitated by the onset of universal suff of lAll, despi Alliance, the tives' 64 seat at W seats. tives dropped formed his sec- in joint sessia passed with co l955: 72-80). accident insur. the first Natir System of Old-a AS veil as pr'ir In 1912 tl create two M house and the A Campalined undc the need fer d' Agrarian Party The Staaff gove with the King c Senate in the asked the Conse f Atce the issue T—__i 60 universal suffrage and proportional representation. In the election of 1911, despite the efforts of the conservatives' General Voters‘ Alliance, the Socialists continued to gain, equalling the conserva- tives' 64 seats in the lower chamber, while the Liberals remained at 101 seats. Even in the upper chamber the non-Ruralist conserva- tives dropped from 133 to 86 seats, and the Liberal leader Staaff formed his second cabinet, commanding a Liberal-Socialist majority in joint sessions of the Riksdag, although some measures were passed with conservative rather than Socialist support (Rustow 1955: 72-80). In 1912 the Riksdag thoroughly revised the industrial accident insurance system (Carlson 1969: 30), and in 1913 it passed the first National Pension Act, establishing a social security system of old-age pensions and housing allowances financed by taxes as well as private contributions (Rosenthal 1967: 9). In 1912 the Ruralist and conservative forces combined to create two joint parties, the Rural andUrban Party in the lower house and the Nationalist Party in the upper house, but still campaigned under the General Votersl Alliance. Farmers soon felt the need for distinct Riksdag representation and formed the Agrarian Party in 1913 and a National Farmers' Association in 1915. The Staaff government resigned early in 1914 after a disagreement with the King over whether to extend the length of military service in the face of the threat of war in Europe. The King then asked the conservative Hjalmar Hammerskjold to form a cabinet and force the issue, but the Riksdag balked and the King dissolved the lover chamber. the Liberals 1c conservatives 2 the outbreak of acquiesced to t strengthened de Liberals lost 1 status in less Riksdag (Rustow Patriotic wartime economr' after losing 27 farmerSl groups AWN dissident: the elections, bl” 300" relucta by the new Libel Parliamentary p] from the left e, i0 further Elect removal of the l the Riksdag, alt indirect eiectic for New fOur 3, elections We 3 r____i 61 lower chamber. In the special dissolution election that follOwed the Liberals lost 31 seats while the Socialists gained 11 and the conservatives 22, still falling far short of a majority. With the outbreak of World War I in August of 1914 the Liberals acquiesced to the King's policy of neutrality combined with a strengthened defense. In the regular November elections the Liberals lost 14 more seats to the Socialists, thus reducing its status in less than a year from largest to smallest party in the Riksdag (Rustow 1955: 72-83). Patriotic support of the conservative government waned as wartime economic conditions worsened. The government resigned after losing 27 seats in the election of 1917, including 14 to the farmers' groups and 11 to the Left Socialists, a group of radical young dissidents who had seceded from the Social Democrats before the elections. The King attempted to form a coalition government, but soon reluctantly appointed a Liberal-Socialist cabinet headed by the new Liberal leader, Nils Eden, thus establishing the parliamentary principle in Sweden. In the face of intense pressure rom the left even the upper-house conservatives submitted in 1918 0 further electoral reforms, including suffrage for women and emoval of the remaining suffrage restrictions for both houses of he Riksdag, although the upper house retained its system of ’ndirect elections. Also, lower house elections were scheduled or every four years rather than every three, and upper house lections were scheduled regularly in between the lower house elections. Mos constitutional (Rustow l955: 7 Poor Relief Act working week, a to combat epidea 31). For almost single party be ten minority go' the Liberals be' than. Two other contained not or nentary princip? coalition with e held under the l ”fused picture Slight SEtback 1. economic depress Adenine in uni labor lOl/enema a 1921 a maiom‘ ty Internatgma] an minority to spin factionar di Spu t 62 lections. Most of these reforms again required ratification of onstitutional amendments and some did not take effect until 1921 Rustow 1955: 75-85). In 1918 the Riksdag arso passed a liberalized oor Relief Act, still administered locally, and enacted a 48-hour orking week, and in 1919 it passed a comprehensive healthprogram to combat epidemic diseases (Peters 1970c 305, 307; Carlson 1969: H). For almost two decades after the parliamentary reform no fingle party held a majority in both houses of the Riksdag, and en minority governments were attempted by all but the Agrarians, me Liberals being most successful because of their centrist posi- jon. Two other cabinets appointed by the King during that period ontained not one member of the Riksdag, contrary to the parlia— entary principle, because the major parties refused to form a oalition with each other. A special election in 1921, the first eld under the liberalized suffrage, failed to clear up the onfused picture, although the Socialists rebounded from their light setback in the regular election of 1920. Although a severe conomic depression struck in late 1920, Sweden did not experience decline in union membership or as serious a split in the political abor movement as did Norway. Prior to the special election of 921 a majority of the Left Socialists voted to join the Communist ternational and renamed itself the Communist Party, causing the inority to split off and retain the Left Socialist title. Such ctional disputes rendered the leftists ineffectual, and the Federation of Social Democra into a single rejoined the S split into pro defeat of a re liberal partie Socialists, ex. continued to gr three Sociali s Public unemplog the Riksdag est ””500 arbitra however, did pa Act of 1924, ar include mental insurance Organ WA: 305). The Great DOSl'Wal‘ depres CdUSEd an inCY‘e Mmmmm election of 193 ing the hardest. Various attGMpt; 63 Federation of Labor continued to work closely with the original Social Democrats. In 1921 the two farmers' organizations merged into a single Agrarian Party, and in 1923 the Left Socialists rejoined the Social Democrats. In that same year the Liberals split into prohibitionist and urban liberal parties following the defeat of a referendum to institute prohibition in 1922, and both liberal parties continued to decline in strength while the Socialists, except for a slight setback in 1928, and Agrarians continued to gain. During this period of "minority parliamentarism“ three Socialist governments fell because they attempted to introduce public unemployment benefits (Rustow 1955: 85-101), and in 1928 the Riksdag established a Labor Court with powers bordering on com— pulsory arbitration (Carlson 1969: 33). Other social legislation, however, did pass, such as the locally administered Child Welfare Act of 1924, and an extension of the health program in 1929 to include mental illness, and of assistance to the private health insurance organizations in 1931 (Rosenthal 1967: 7, 10; Peters 1970: 305). The Great Depression hit Sweden even harder than had the post-war depression, so that rising unemployment and falling wages caused an increase in strike activity. (In 1931 five striking workers were killed by inexperienced police at Adalen, and the election of 1932 brought major gains to the two parties represent~ ing the hardest-hit groups, the Socialists and the Agrarians. Various attempts at a majority coalition were made, but the King fhaHy chose a Hansson. In 1E mmhma leaders of the mmmmn MmHmm MMMHM whle the two 1 hatsame year formed the Soci mrwtive Party hm,Mthap WWHMH hhmtruction in, "495, but the 1 Mmmm, 1“ lowernmentl1 thBend of 1934 The Agrarie mmhmm nilitary appropr WMMHM: iefhgt parties and the Agrarian three ”lining in 64 finally chose a near-majority Socialist cabinet led by Per Albin Hansson. In 1933 Hansson succeeded in passing an anti-depression program known as the "Swedish New Deal" by going beneath the leaders of the center parties to secure rank-and-file support. The Riksdag also expanded the Child Welfare Act in 1934 and the National Pension Act in 1935. The Agrarians actually shifted to the left of the Liberals, and in 1934 replaced their leadership, while the two liberal parties reunited under new leadership. In that same year a group of dissident Communists and Social Democrats formed the Socialist Party, and the youth organization of the Con- servative Party split off to form the insignificant Nationalist Party, with a philosophy closer to the equally insignificant National Socialist organizations. Also, a strike occurred in the construction industry over an attempt by the employers to lower wages, but the Federation of Labor eventually pressured the unions into accepting a bad settlement in the collective interest of aiding the government's anti-depression policy, which seemed to succeed by the end of 1934 (Rustow 1955: 101-107; Carlson 1969: 29~38). The Agrarians soon began voting with the opposition once again, and the Hansson government finally fell in 1936 in a dispute over military appropriations and pension increases. The first Agrarian cabinet was thus formed, but the election of 1936 gave the three leftist parties a bare majority in the lower house of the Riksdag, and the Agrarian "Vacation Government“ had to resign after only three months in office. But the Socialists still faced a bourgeois majority in th government wit previously est and workers fr 1939 the four ment headed by 110). In 1938 a government emp Employers' Ass compulsory arb settle dispute; orionizations c ment imposed a Federation of | 01ionization at But unlike the c0"llhtition tn the new 100 am Carlson 1959: E DESPite 11 during the War elections We Communists trip Socialists by c 65 majority in the upper house, and thus formed a strong coalition government with the Agrarians that succeeded in expanding previously established social programs beneficial to both farmers and workers from 1937 to 1939. When World War II broke out in 1939 the four major Swedish parties joined in a coalition govern— ment headed by Hansson and declared neutrality (Rustow r955} 107- 110). In 1938 a new organization of white-collar workers and government employees emerged, and the Federation of Labor and the Employers' Association, seeking to avoid further government-imposed compulsory arbitration, entered into a Basic Agreement designed to settle disputes peacefully. During the War the two labor market organizations even agreed to regulate wages, although the govern- ment imposed a wage ceiling and commodity rationing. In 1941 the Federation of Labor amended its constitution to centralize its organization and to strengthen its authority over member unions. But unlike the Employers' Association the Federation still faced competition from the syndicalist Workers' Central Organization and the new TCO and other white-collar unions (Galenson 1952: ll9-144; Carl son 1969: 38-42). 1 Despite limitations on political as well as economic activity during the War in order to avoid provoking the belligerents, elections were held on schedule. In the election of 1944 the Communists tripled their share of the vote at the expense of the Socialists by criticizing the government's economic restrictions in 191 put in resour their but 11 manage 66 l concessions to the Germans, causing the Socialists to move ’tward and to give up any notions of continuing the national clition. In 1945 Hansson formed a purely Socialist cabinet and rpped up the pre-war program of social legislation and economic mnsion while maintaining Sweden's neutral foreign policy. In -6 the Riksdag further expanded the National Pension Act, and 1947 it authorized compulsory health insurance, but it was not rinto effect until 1955 because of a shortage of medical ources. In the election of 1948 the Liberals more than doubled ir seats in the Riksdag at the expense of all other parties, the Socialists still held a majority in the upper house and aged to muster a leftist majority in the lower house (Rustow 5: 110-115). As Socialist support continued to dwindle the Agrarians, now ling themselves the Center Party, in 1951 again agreed to form oalition government with the Socialists. In 1952 the d'Hondt hod of proportional representation was replaced by the Sainte— Je method, and in 1956 the Law on Social Help replaced the “ Relief Act of 1918. The Socialist-Agrarian government igned in 1957 in a dispute over proposals for revising the sion system, and since no feasible alternative coalitions Id be found a special lower house election was held in 1958, rlting in a leftist majority and a purely Socialist cabinet. :e then the division between leftist and non-leftist forces has lined in near-even balance, and the Socialists have clung 67 :ariously to power (Stjernquist 1966s 124-126). In 1959 the sdag established inflation-proof supplementary pensions, and in 2 the National Insurance Act consolidated all existing health social security programs (Rosenthal 1967: 6-8). Figure 3-1 summarizes the important periods in the development iweden from 1809 to 1962. are 3-1. 68 Important Periods in the Development of Sweden, 1809- 1962. ates Events 5-1847 5-1867 3-1871 5-1895 3-1886 7-1888 -l892 1-1896 End of royal absolutism. New constitution. Union with Norway. Guild monopolies abolished. Philanthropic workers' associations emerge. First Poor Relief Act. Bicameral reform of Riksdag. Ruralist Party and Ministerial Party formed. Industrialization begins. New Liberal Party splits from Ministerial Party. Craft unions begin to form. Poor Relief Act restricted. Period of numerous economic recessions. First major labor strike. First child-labor law. Regional and national inter-union organizations formed. Anti-tariff forces win special election, but denied part of Riksdag delegation. Party realignment 1n Riksdag: Ruralists and conservatives split into protectionist and free-trade parties. Social Democratic Party founded. First state- supported industrial accident 1nsurance. Universal Suffrage Association formed. Riksdag regulates private health insurance, raises tariffs, solves conscription and taxation issue, reduces overrepresentation of c1t1es. "People's Parliaments" sponsored by Socialists and Suffragists. Ruralists reunite. Liberal Peoples' Party formed. 69 *e 3-1 (continued). 1* Events “es -1900 Rising prices. Increased industrial expansion. Rising labor union membership. First Socialist representative sent to Riksdag. Swedish Federation of Labor formed. Riksdag outlaws interference with strikebreakers. -1901 Liberal Party founded. New child-labor law. Work- men's compensation. General strike. Employers' Association formed. Conservative General Voters'.Alliance formed. Union with Norway dissolved. First Liberal cabinet. Metalworkers' strike. Collective bargaining accepted. Non-compulsory government arbitration made available. Compromise extension of suffrage with proportional representation passes Riksdag. Liberals equal conservatives' strength in lower chamber of Riksdag. Economic recession. First set- back in union membership growth. Electoral reform takes effect. General strike and lockout, defeat for labor. New regulation of factory conditions. Subsidization and regulation of private health insurance. Syndicalists secede from Federation of Labor. Socialists equal conservative strength in lower chamber of Riksdag. Second.Liberal cabinet. Ruralists and conservatives join forces. Industrial accident insurance revrsed. National Pension Act: Social security system. Agrarian Party founded. 70 re 3-1 (continued). l— tes Events ~1919 -l921 -1932 -l935 Wartime inflation begins. Liberals reduced from largest to smallest party in Riksdag. Conservatives lose heavily in election. First Liberal-Socialist cabinet: Parliamentary principle established. Left Socialists secede from Socialists. Riksdag passes universal suffrage for both sexes and schedules elections every four years. Liberalized Poor Relief Act. 48-hour work—week. Comprehensive disease-control program. . . Electoral reform takes effect. Beginning of severe economic recession. Minority parliamentarism: »All parties try governing. Socialists and Agrarians gain at polls. Communist Party splits off from Left Socialists. Defeat of prohibition referendum. Left Socialists rejoin Socialists. Liberals split into prohibitionist and urban liberal parties. Child Welfare Act. Labor Court established. Mental illness included in health program. Assistance to private health insurance extended. Strikers killed at Adalen. . Great Depression. Socialist cabinet following near- majority electoral victory. Anti—depression program passed. Child Welfare Act expanded. Liberals reunite. Construction strike. National Pension Act expanded. Agrarian cabinet. Socialist election victory. Socialist-Agrarian cabinet. 71 ure 3-1 (continued). I.___ I_‘i Events ates 7-1939 Expansion of previous social legislation. 8 Basic Agreement between labor market organizations. White-collar unions organized separate from Federation of Labor. 9-1945 Coalition government of major parties. Wartime economic restrictions. 1 Federation of Labor centralized. 5 Socialist cabinet following Communist electoral resurgence. Economic expansion begins. 6-1947 National Pension Act expanded. Compulsory health insurance passed. 1 Socialist-Agrarian cabinet. 2 Sainte-Lague proportional representation method. 6 Law on Social Help replaces Poor Relief Act. 3-1965 Socialist cabinet following special election. 9 Supplementary pensions. IV National Insurance Act consolidates health and welfare programs. Swe imp ewo lon soc' aftc a or issc 0hr : CHAPTER FOUR THEORY AND METHODOLOGY The descriptive and historical literature on Norway and weden, summarized in the previous chapters, leaves the general ‘ mpression that Sweden's development progressed more smoothly than id Norway's. Developmental processes seem to have begun earlier, volved more slowly and with less social disruption, and taken onger to complete in Sweden than in Norway. Issues surrounding ocial change seem to have been settled more often by compromise fter long periods of public debate than by heated competition in crisis atmosphere; and the social divisions associated with those ssues seem to have been less profound in Sweden than in Norway. This chapter will organize presumed differences between Norway rd Sweden into a theoretical framework composed of interrelated rpotheses, and will specify indicators for measuring the concepts rvolved, the nature and sources of the data for those indicators, and 1e methods for analyzing those data and evaluating those hypotheses. Theoretical Background The Bull-Galenson Hypothesis The literature comparing Norway and Sweden heavily emphasizes re specific set of differences: Sweden's earlier and more gradual 72 no car or cor amc 1111 (11 quot Demm fere indu 73 ndustrialization, and Norway's more radical political labor ovement. Furthermore, the literature seems to assume a direct ausal relationship between these two differences. This assumption riginated in 1922 in a comparison of all three Scandinavian :ountries by the Norwegian historian Edvard Bull, Sr. (Bull 1922), md was popularized in English by the American labor economist hlter Galenson (1949; 1952). It has since been cited by Lipset 1963: 54) and Kornhauser (1959: 153), among others, as evidence f the ill effects of too-rapid economic growth on social structures nd processes, as well as by such analysts of Scandinavia as Rokkan 1960: 107-110; 1967: 395). Only recently has the assumption been uestioned (Rokkan 1970: 136) and tested empirically as the "Bull- alenson hypothesis" by the American political scientist, William afferty (1971). ' The basic premise underlying this hypothesis varies slightly om the Bull version to the Galenson version. Bull's formulation, oted in Lafferty (1971: 21), emphasizes Norway's differences from nmark and Sweden, whereas Galenson's emphasizes Denmark's dif- rences from Sweden and Norway. Bull asserts that Norway's. dustrial development progressed much faster than did the other 0 countries', and that the sudden, disruptive recruitment of a w industrial labor class from a traditional peasant society, and e isolated location of industries close to hydroelectric power ants, made the Norwegian workers more susceptible to radical eas than were the workers in Sweden and Denmark, where the dustrial labor classes had developed more slowly. to C01 C01 185 11101 wea lldl' ear has mos rad' enpr' on t 1011" majn hetw the 1"AU: of $1 74 Galenson, on the other hand, emphasizes Denmark's earlier and 'e gradual industrialization and stronger guild tradition compared both Sweden and Norway, which had to await the development of ~technology of hydroelectric power before real industrialization ld begin, and whose guilds were unable to monopolize their trades pletely because of geographic barriers. But Galenson also erts that Norway's industrial development was even later and e rapid and isolated than Sweden's; her guild tradition even ker; and her labor movement even more radical. Lafferty sum- izes the hypothesis as representing two sets of continua, with three countries ordered similarly on each: Denmark with the liest and most gradual industrialization, and the most guild— ad and least radical labor movement; Norway with the latest and t rapid industrialization, and the least guild-based and most ical labor movement; and Sweden in between. 1 Patterns of Economic Development Lafferty (1971) points out that Galenson's only systematic rical support for the hypothesis consists of cross-section data :he non-agricultural work-force and urban population of each rtry at widely-spaced intervals. Galenson places Denmark's r period of economic growth between 1880 and 1900, Sweden's een 1895 and 1914, and Norway's between 1905 and 1920. Lafferty rts that Galenson's only support for the statement that Norway's strialization was more rapid than Sweden's was the construction everal electrochemical plants and the doubling of the industrial 75 work-force between 1905 and 1920. Lafferty also cites one other attempt at empirical verification of the Bull-Galenson hypothesis, that of Asbjorn Bjornset, a Norwegian historian, whose findings generally support the hypothesis but minimize the differences between Norway and Sweden. Lafferty points out, however, that Bjornset's data on the "strictly" industrial work-force, i.e. workers covered by industrial accident insurance, are unreliable because of changes in the legal definitions of eligibility for benefits; because of missing data; and because Bjornset used an incremental index rather than percentage of total work-force, thus distorting the true picture. Lafferty then proceeds to derive extremely specific hypotheses from just the industrialization portion of the Bull-Galenson hypothesis, involving not only the timing and rate of industriali- zation but also its "intensity" and “regularity," then evaluates hese hypotheses by examining a variety of macro-economic indicators ver time. The first involves decennial figures on the distribu— ion of the labor force in the agricultural, industrial, and service ectors from 1870 to 1930. These data indicate that Sweden rather han Norway was the least industrialized Scandinavian country in 870 and also industrialized the fastest. Denmark had already begun eveloping by 1870 and continued to do so gradually and evenly hroughout the entire period, with the greatest growth in the service ector. Norway's agricultural sector was almost as low as Denmark's n 1870 due to her larger service sector, primarily in shipping. 76 Norway's industrial sector was also slightly larger than den's in 1870, and both her service and industrial sectors grew adily and evenly throughout the entire period. Sweden grew most idly in the industrial sector from 1870 to 1900, in the service tor from 1900 to 1920, then in the industrial sector again from 0 to 1930. These findings thus contradict the hypothesis by icating that Norway and Sweden should switch positions.on the ustrialization continuum, and that Norway's pattern of develop- rt was more similar to Denmark's than to Sweden's Lafferty then examines rates of change in gross and net restic product per capita for decade averages from 1870-1878 ~ough 1929-1938, and finds that Sweden's was highest and Norway's rest until l909-1918, after which the rates for all three Intries were similar. The patterns for Sweden and Denmark seem have been more similar to each other than either was to Norway's, rtradicting not only the Bull-Galenson hypothesized continuum for es of industrialization, but also the findings of the labor- ce data, although Denmark's overall rate of growth was slightly ser to Norway's than to Sweden's. Examination of growth rates ween arbitrarily selected and overlapping five-year averages, ever, confirms the hypothesized timing of industrialization: mark and Sweden both accelerated from 1871—1875 through 1901- 5 while Norway remained stagnant; then Denmark tapered off from 6-1900 through 1916-1920 while Norway began and Sweden continued elerating; then Sweden began and Denmark continued to taper 77 pff from 1911-1915 through 1931-1935 while Norway continued accelerating. In order to determine more carefully the timing of industrial surges, Lafferty then examines data on absolute and incremental ross and net national investment ratios based on five-year averages. he results indicate that each country had two principle surges, he second larger than the first, and that Norway's were the latest nnd weakest: Denmark had a fairly strong one from 1880 to 1885, a stronger one from 1890 to 1900, and a fairly weak one from 1920 to 1930; Sweden had a fairly strong one from 1895 to 1900 and a stronger rne from 1920 to 1925; and Norway had a weak one from 1890 to 1900, 1 strong one from 1905 to 1915, and a stronger one from 1925 to 1930. Lafferty then examines the relationship of investment ratios nth the labor-force sector shifts, concluding that Norway's ‘ndustrial surges were not strong enough to cause large shifts; and fith per capita product, concluding that Denmark showed a strong multiplier effect" of product upon investment during the time eriod, whereas Norway showed a strong "production effect" of invest- ent upon product, while Sweden showed a mixed effect. Lafferty hen correlates the economic indicators across countries, concluding hat Norway and Sweden are more similar to Denmark than they are to ach other, and finally delves into an esoteric examination of the elationships among the economic indicators within countries. Lafferty summarizes all of these findings by concluding that enmark had already industrialized and achieved "sustained growth" 78 r 1900, and then became a I'mature" economy; that Sweden possessed fficient resources and had begun industrialization before 1900, t had to wait for her service sector to catch up before a second rge of industrializatidn could carry her into sustained growth; d that Norway lacked adequate resources and thus had only a mild dustrial surge by 1900, but her labor force was properly portioned for steady acceleration thereafter. Although Lafferty es not state it clearly, the general conclusion must be that the dustrialization portion of the Bull-Galenson hypothesis is correct ncerning sequence but incorrect concerning rate (and its confus- gly related concepts of intensity and regularity) for Norway and veden. Labor Response to Economic Development Following his thorough examination of the differences in rtterns of industrialization among the Scandinavian countries, rfferty turns his attention to the various manifestations of the esponse of the political labor movements in those countries to :onomic development. The first indicator of that response he insiders is voter support for all leftist parties, as measured by arcentage of the total vote, although it is not entirely clear V rich "moderate socialist" splinter parties he includes. He com- rres the within-country correlations of this variable with not 11y the previous economic indicators but also with indicators of 1 "economic man" intervening variable, price and wage indices, rd finally with an indicator of a "sociological man“ intervening 79 variable, the suicide rate. He then compares the within-country correlations of these same variables with per capita membership in the national Federation of Labor, and with the ratio of man-days lost due to strikes and lockouts to maximum potential working time of members of the national Federation of Labor. All of these cor- relations are based on five-year averages from 1900 to 1935, an insufficient data base for thorough correlation and regression analysis, and the non-economic variables are not compared cross- nationally. Lafferty concludes that in general the results support the Bull-Galenson hypothesis, in that the correlations confirm the expected pattern of labor response associated with each country's particular stage of economic development during that period. lorway generally showed the strongest positive correlations not nly between the labor-response indicators and the indicators of er particular pattern of economic development, but also between eft vote and union membership; whereas Sweden and Denmark generally how appropriately mixed results, with Denmark displaying the most egative and weakest positive correlations. Lafferty seems to mply that the correlations exhibit this pattern because the ountries happened to be at different stages of economic develop- ent while their political labor movements developed coterminously nd differed in "radicalness" due to other factors than economic velopment alone. i I , I He seems to draw a similar general conclusion from a thorough 'storical comparison of the political contexts surrounding the 80 political labor movements in each country, including a detailed examination of "sub-system" factors in Norway involving individual delegates and groups at the Labor Party convention in 1918. Using Lipset and Rokkan's (1967) framework of thresholds of legitimation, incorporation, representation, and executive power, Lafferty determines that although the thresholds in all three countries were "softer" than in most other countries, Norway's thresholds were more difficult to cross and thus were crossed later than in Sweden and Denmark, although Sweden's barriers were somewhat "harder" than Denmark's. To reach this conclusion Lafferty relies not only on dates of introduction of such characteristics as universal suffrage, direct elections, proportional representation, and socialist voting, representation, and cabinet participation, but also on a comparison of the "voter-mandate ratio,“ the ratio of the proportion of parliamentary seats to the proportion of total vote, for leftist parties. Although Lafferty does not state them succinctly, the chief causes of Norway's more radical labor response uncovered by his analysis seem to be the following: (1) the coincidental occurrence of the first major industrial surge, albeit the mildest of the Scandinavian surges in terms of labor shifts into the industrial sector, and the unionization movement, resulting in (2) a more rapid influx of many young workers into the Federation of Laborthan in Denmark and Sweden, a far more radicalizing factor than either Galenson's doubling of the number of insured industrial workers 81 (probably due largely to changes in the official definition of ”industrial"), or the increase in the total work-force, including women and business employees; (3) the early achievement of the parliamentary principle and universal suffrage, partly due to the lack of an indigenous nobility and thus of an upper chamber in the parliament, resulting in the absence of a fight for electoral reform that would have necessitated the degree of cooperation between socialists and liberals found in Denmark and Sweden, thus isolating the Norwegian socialists and excluding them from cabinet participation until 1928; (4) the underrepresentation of the socialists after 1905 due to the single-member run-off system of direct elections; and (5) the radical leadership of Martin Tranmael. Overall, Lafferty's work seems to confirm the Bull-Galenson hypothesis only partially: Apparently Norway did industrialize later than Denmark and Sweden, but not more rapidly, and her politi- cal labor movement was apparently more radical, although he never explicitly compares the indicators of labor response cross-nationally. And apparently that radicalness was "more internal-party oriented than national-system oriented" (Lafferty 1971: 325), and was not caused by later industrialization directly, but rather by several coincidental social and political factors as well as economic ones. B. The Mobilization Model of Public Policy Development 1. Social Mobilization and Political Mobilization The Bull-Galenson hypothesis, as refined by Lafferty, fits into a more general model of developmental processes popularized by 82 Deutsch (1953; 1961) and discussed throughout much of the literature of comparative politics, the "mobilization model." Peters (1970) extended this model of the developing political system to include the development of public policy and tested it using correlation and regression analysis of quintennial data on Britain, France, and Sweden from 1850 to 1965. His formulation of the model involves changing societal conditions that generate inputs composed of increasing demands for public services and of resources available for meeting those demands; democratic institutions which penetrate the society to perceive the demands, obtain the needed resources, and provide service machinery; and outputs of public expenditures which have impacts on societal conditions, or “objective security." The input side of the model involves social mobilization, socio-economic resources, and political mobilization. Social mobilization involves industrialization and urbanization, which move people from traditional agricultural environments to more modern and rationalistic ones, disrupting accustomed authority and security patterns; make people more dependent on employers and the money economy, and more susceptible to organized influence; and aften create intolerable working conditions, unemployment, and population pressures in urban areas and in young and old age groups. ”hese social conditions generate demands for specific public services, such as factory legislation, unemployment benefits, mass transit, rublic education, and old-age pensions. Industrialization also ’acilitates the development of greater socio-economic resources, 83 namely higher levels of economic wealth, education, and technology, all necessary for not only the private production of complex goods and services, but also the formulation and delivery of public services capable of solving complex societal problems. But increasing resources reciprocally foster further industrialization by providing surpluses for investment, and produce further demands for public services, especially public education and technological research. The dislocations associated with these fundamental social changes, plus the development of modern values and the social and political awareness associated with education, combine to generate political mobilization of the demands of the working class. Political mobili- zation involves labor unionization, voting for leftist parties, voter turnout, and rate of enfranchisement, all indicating the strength of working-class demands. This formulation requires further modification of the Bull— Salenson hypothesis in order to specify the distinction between the sub-concepts of social mobilization and economic wealth within the general concept of economic development. Lafferty's indicators of economic development include Gross Domestic Product, a measure of realth in the mobilization model; labor-force sector proportions, a reasure of industrialization in the mobilization model; plus Gross mmestic Investment, a measure of industrialization not used in the obilization model. Although Lafferty's analysis of Scandinavia oes not suggest a strong relationship between social and political obilization, it does seem to warrant specifying industrialization 84 rather than wealth as the economic factor most likely to motivate political mobilization. It also seems to suggest separating unionization from political mobilization and perhaps placing it between social and political mobilization, although unionization and left voting seem to have grown simultaneously in Scandinavia. In addition, Lafferty does not really consider urbanization as a possible explanation of labor response, probably because of the dispersed, non-urban character of industry in Norway and Sweden. Peters, on the other hand, does not include labor disputes as an indicator of political mobilization, perhaps because the "radical- ness" of the demands of political labor movements was deemed unimportant in the general model of mobilization, or perhaps because strike data was not completely available for all three of his test countries. He also does not include a price index in his analysis. But a test of the general model for Norway and Sweden should include these variables, although industrialization and urbanization are known to be almost interchangeable, mutually causal indicators in most settings, and although labor disputes must be considered contemporaneous with other indicators of politi- cal mobilization and not caused by any of them. 2. Political Institutions and Public Policy The general mobilization model of political development deals next with the translation of demands and resources into public policy, involving the development of democratic institutions to hear the mobilized demands; government penetration of society to 85 btain needed resources and provide service mechanisms; government xpenditures designed to alleviate the offending societal condi- ions; and the impacts of those expenditures on objective security. 5 measures of democratic institutionalization Peters (1970) tilizes an index developed by Cutright (1963) and another developed y Flanigan and Fogelman (1971), both of which award points to bservations exhibiting defined democratic characteristics. Govern- ent penetration is measured by the size of the civil service and he amount of governmental revenue. Public employees are needed 0 perceive mobilized demands, make decisions that produce policies 0 meet those demands, and execute the detailed administration of :hose policies. While the size of the civil service indicates the legree of involvement of the government in the labor force, the rmount of public revenue indicates the degree of government involve- ent in the nation's economy and the amount of resources available or public policy expenditures. Both variables could be considered 0 represent the willingness of political institutions to act. Because of the functional specificity of public demands and he policies required to meet them, the expenditures themselves are easured not only in totals but also by sectors: defense, health, ducation, and such social services as pensions and relief. The ealth and education sectors also require intervening measures of he number of personnel available to deliver to the public the rvices provided by the expenditures, namely doctors and teachers. inally, the impact of these expenditures and services on societal 86 conditions or objective security are measured in each sector: for health, infant deaths, total deaths, and life expectancy; for education, the number of pupils being educated; and for welfare, the number of relief recipients and pensioners. Peters tests the hypothesized relationships within the obilization model using correlation and regression analysis of ata on Britain, France, and Sweden for every fifth year from 1865 to 1965. In addition, using similar methods and analysis of ariance, he compares the explanatory power of the model against everal simpler models that attempt to explain the development of xpenditures and impacts, such as the occurrence of great events, changing elite ideology, and the general increase in public action over time. On the input side of the mobilization model he finds support for the Deutsch (1961) formulation of mobilization thresh— )lds rather than a linear relationship between social and political mobilization. By splitting the total time period, comparing the ‘esults for the different portions, and considering the stages of social mobilization which each country was experiencing during those reriods, he suggests the existence of not only a threshold level of social mobilization which must be attained before it will “spill over" into political mobilization, but also a second threshold level cfter which there will again be no relationship between the two 'ariables. The relationship between political mobilization and government enetration is substantiated for different indicators in each 87 country due to historical reasons, such as Sweden's.intense financial investment and de-emphasis of bureaucratization in her social pro- grams, compared with the political entrenchment and belated resource utilization of France's unpopular bureaucracy. However, Peters uncovers a stronger and more consistent relationship between social mobilization and government penetration, indicating independent effects of social and political mobilization on penetration. In fact, Sweden exhibits no political mobilization effect when social mobilization is partialled out. This phenomenon and the relation- ship of social mobilization to revenue extraction in France indicate the development of what Peters calls the "cybernetic system," in which needs arising from changing social conditions are anticipated technologically and assuaged directly by ruling elites, thus reduc- ing the importance of political mobilization in the later stages of development. The results involving the indices of democratization are incon- sistent, due largely to their lack of variance in the latter portions of the time period. The Flanigan-Fogelman Index seems to work the best, and Peters concludes that, like penetration, it is independently affected by both social and political mobilization, thus strengthening his interpretation of the cybernetic system. He also tries using the rate of enfranchisement as an indicator of democratization rather than political mobilization because of its greater variance and because it has as much construct validity as the two indices in terms of the changing nature of elite institutions in response to 88 social and political demands. However, these results are also inconsistent, and it remains unclear whether enfranchisement should be considered a measure of mobilization or of democratization, although the weight of the evidence seems to favor the latter. Another useful indicator of democratization exhibiting sufficient variability might be Lafferty's "voter-mandate ratio," the ratio of the proportion of parliamentary seats to the proportion of the total vote, for left-wing parties. This indicator, like enfranchisement, measures the changing nature of elite institutions in response to social and political demands, although both measures deal only with institutions for representation and ignore such other processes as selection of the executive. Contrary to the findings of the analysts of expenditures in the American states, both sets of indicators of governmental struc- , ture, penetration and democratization, have significant effects upon expenditures, although the effects of democratization are weaker and seem to feed through government penetration. In addition, Gross National Product must be included in the multiple regressions on expenditures involving government structure in order to eliminate autocorrelation. This suggests that both political and economic variables are important in determining levels of public expenditures, in that spending requires both willingness and capability. The impacts of these expenditures on "policy outputs," or the levels of service personnel and objective security, vary for the different sectors, countries, and time periods, again due to 89 historical factors. Health expenditures show greater impacts than expected, with those in Britain and Sweden indicating an increasing trend vis-a-vis the impacts of economic resources and technology, whereas in France the government became invdlved in health care earlier and has been declining in importance relative to resources. The opposite proves true in the education sector, although complete data is not available for all three countries on the number of teachers. The impact of educational expenditures on the number of pupils is not as strong as expected, and seems to be declining in Britain and Sweden and increasing in France relative to the role of the resource measure for this sector, adult educational attainment. Finally, the impact of welfare expenditures on relief and pension recipients is much weaker than expected and seems to have become more functionally differentiated over time. The pension system is more directly dependent on social mobilization, via the breakdown of the extended family, than on government expenditures, and is thus deemed more legitimate than the relief system, which is becom- ing increasingly dependent on public support. Peters concludes that in general the alternative models do not explain expenditures and impacts as well as does the mobilization model. The "General Amelioration" model, based on the notion of an incremental increase in public action over time, can be rejected because of significant autocorrelation in the admittedly strong relationships of policy variables with time, especially in Sweden, probably due to exponential increases in the later time periods. 90 The "Ideology of the Elites" model, based on the presumed importance of ideas in the formulation of policy, does not work well except for the output measures in France, particularly education, probably due to the greater structural centralization and variability in elite ideology in that country. The "Great Events" model, based on the presumed lasting effects of wars and depressions on policy, works well in explaining expenditures except in Sweden, which has been I relatively isolated from such events, but did not work for policy outputs. The modified general mobilization model of political develop- ment is summarized diagrammatically in Figure 4-l. Political-———————-+Democrati- Personnel Mobilization zation Services \\\\\\\\X Social Government Government Objective obilization ,\\\\\Pene::a:ion/ Expenditures“—‘T Security ‘ Resources // Figure 4-l. Diagram of the General Mobilization Model. I. The Data The Observations Comparing Norway and Sweden within the conceptual framework of he modified general mobilization model of political development 91 requires certain modifications in the data base and techniques of analysis used by Peters, mostly due to constraints imposed by the sources of data for Norway. The first constraint is that, without travel funds, virtually all of the Norwegian data has to be obtained in the United States from publications of the Norwegian Central Bureau of Statistics, either directly or indirectly: Even secondary sources cite some such publication. This means that data for most of the variables are not available before 1876, a few years after the establishment of the Bureau, whereas data on Sweden is available in the U.S. back to 1865. Using an interval of every five years from l880 to l965 would yield only l8 observations for analysis, probably an insufficient sample size even for analyses of this type. However, using a short interval such as every year over long periods tends to produce severe utocorrelation due to the lingering effects of extraneous factors omprising the disturbance term in regression equations (Kmenta 97l: 269—270). Some interval in between is obviously indicated, nd since some of the data involve election statistics, a reasonable hoice is the election years, initially every three years and then very four years, yielding 27 observations. A related problem is hat data on Norway is not available for the period of World War II, hen the l939 and l942 elections were not held because of the German ccupation. But fortunately, Sweden switched over to quadrennial lections beginning in l920, whereas Norway waited until after the ar, beginning in 1945. Consequently Norway's two "lost" elections 92 are regained via the greater frequency of elections during the l920's and l930's. This procedure generates some obvious objections. First, is it reasonable to presume that Norway's post—War development is an approximate continuation of her pre-war development? In other words, was the War period in Norway such a deviation from the normal pattern of development that inclusion of those observations would have dis- torted the comparison between Norway and Sweden, since the latter was relatively isolated from the impact of the War? The answer appears to be yes: Most of the Norwegian series do not exhibit severe discontinuities between the l936 and l945 observations; rather, the latter usually seems to pick up where the former left off. In addition, linear interpolation, an equally objectionable procedure, would have been required for many variables if the war years had een included. Second, doesn't the lack of synchronization of the observations istort the comparison between the two countries? The answer here '5 yes, slightly. There are three dimensions to this problem. irst, and least serious, is the one-year difference between the bservations, even when the elections in both countries occurred at qual intervals. That is, Norway's elections occurred in l876, 879, etc., and l945, l949, etc., whereas Sweden's occurred in l875, 878, etc., and l944, l948, etc. Second, the unequal spacing of he observations during the l920's and l930's certainly distorts he comparison to some extent. Both of these problems revolve around 93 the influence of events on the variables and the relationships among them. Specifically, simultaneous events whose effects last less than a year might be missed by one of the countries; but such events are probably insignificant anyway. More importantly, from l924 Norway's observations lag further and further behind Sweden's until Norway's l936 is taken as the equivalent of Sweden's l940. However, all this really means is that Sweden has one war-time observation to Norway's none, while Norway has one more Depression-era observa- tion than Sweden; and one could argue that the debilitating effects of these two observations are equivalent and thus that the differ- ence is self-cancelling, if not inherently insignificant. Finally, the wider spacing of the observations in the later time periods nakes the series appear more exponential than they really are. As long as this is true of all series and both countries the comparison "emains undistorted. The major problem again lies in the latter half if the l930's, in which an equal function might appear slightly flgher in Sweden. This distortion is probably no worse than the roblems associated with the alternatives of either an unequal or ow number of observations, or linear interpolation for many of orway's variables during the War years. Thus the data base for this dissertation consists of aggregate :atistics on Norway and Sweden for years in which regularly heduled, lower-house parliamentary elections occurred, from l875 l965. All of the Sweden data except labor statistics were tained from data on every year from l865 to l965 gathered from 94 official sources by Peters (l970; l972).* The Norway data were obtained from official publications of the Norwegian Central Bureau of Statistics, listed at the end of the bibliography; and the Swedish labor statistics were obtained from official publications of the Swedish Central Bureau of Statistics and of the Swedish Federation of Labor (LO), also listed at the end of the bibliography. These data are the most reliable of their type available and provide an ccurate basis for comparison of the general patterns of development in Norway and Sweden. The Variables The variables included in the data base for this dissertation Nlll be operationally defined and discussed under the concepts and sub-concepts which they indicate, listed in order of their approxi- nate position in the general mobilization model of political ievelopment. Social Mobilization Population Measures, reported and estimated by official census ublications, include the total population, the youth population ages 0 to l4), the elderly population (over age 65), the pre-school opulation (ages 0 to 4), the school-age population (youth minus re-school), and the working-age population (total minus youth and lderly). These measures not only indicate population pressures At this point I would like to thank Professor Peters for kindly ermitting me to use his data on Sweden in this dissertation. 95 likely to produce demands for public action, but also serve as bases of standardization for other measures. Industrialization is measured by the declining percentage of the population employed in agriculture, as recorded and linearly interpolated from official census figures. This measure is generally considered by non-economists to be the most valid measure of indus- trialization because it indicates the changing life-style of service 5 well as industrial workers. However, agricultural employment is monotonically decreasing function and thus does not reflect indus- rial surges. Thus Lafferty's (l97l) indicator of Gross Domestic Investment, as reported by official sources, will be used as a supplementary measure. Urbanization is measured by the population living in cities of )ver 20,000, as recorded and linearly interpolated from official figures based on the census and other scattered estimates, divided 3y the total population. This is another indicator of the changes 'n life-style of the population likely to generate demands for public ction. Unemployment is measured slightly differently for each country, ut still indicates a social condition likely to generate demands or public action. For Sweden it is the total number of unemployed ersons divided by the working-age population, based on official nd unofficial estimates. For Norway it is the percentage unemployed mong members of the Federation of Labor, as reported by that rganization back to 1903, thus rendering this variable usable only or later periods. 96 Price Level is also measured slightly differently for each country, but still indicates conditions of rising and falling economic conditions likely to generate demands for public action. For Sweden it is the l9l3-based consumer's cost-of—living index, whereas in Norway it is the 196l—based production price index, both based on official sources. The consumer's cost-of—living index for Norway was available only back to l90l, and its deviations from the production price index are slight. 2. Resources Economic Health is measured by the Gross Domestic Product, as reported by official sources, deflated by the price index and divided by the total population. This will be the only measure of the resources or capability of the system,as the two other measures used by Peters, the number of people with at least an elementary education (lagged 20 years) and the number of patents issued, which are both highly associated with economic resources anyway, cannot be found for Norway. Attempts are being made to develop an indicator of technological capability based on energy consumption, but presently these figures are incomplete. 3. Political Mobilization Unionization is measured by the number of members in the Federa- tion of Labor in each country, as reported by that organization, divided by the working-age population. This measure is a better indicator of the strength of working-class demands than total labor 97 union membership, which is not available for Norway, because the latter is absolute rather than standardized by population, and because membership in the major central organization indicates the unity of the political labor movement while still reflecting total membership. Labor Conflict is measured by the number of man-days lost due to strikes and lockouts for members of the Federation of Labor, as reported by that organization, divided by total possible working man-days, computed by multiplying the total membership in the Federation of Labor by 300. Data for this variable are available only back to 1903 for both countries, and may thus be used only for later periods. Also, because this measure exhibits such wild annual fluctuations that important strike activity might easily be missed by triennial observations, the annual figures were averaged over each inter-election period and recorded for the election year terminating that period. In any case, it is an indicator of the strength and perhaps the "radicalness“ of working-class demands. Left Voting is measured by the proportion of the total popular vote going to all left-wing parties, defined as the Labor Party in Norway and the Social Democratic Party in Sweden, plus all splinter parties ultimately derived from them or standing to the left of them. This does not include “worker's" splinter parties derived from the Liberal Party. Norwegian electoral figures after l903 are derived from official sources, and before that date are estimated from Rokkan (l966: 86). This measure is another indicator of the strength of working-class demands. 98 Turnout is measured by the total vote divided by the total number of people eligible to vote, as reported by election statistics, and is an indicator of electoral mobilization, particularly of the working class. 4. Democratization Enfranchisement is measured by the total number of people eligible to vote, as reported by official election statistics, ivided by the working-age population. This measure indicates the structural response of the governing elites to working-class demands for some degree of political participation. Representation is measured by the proportion of the total number of parliamentary seats going to all left-wing parties, defined above, divided by the proportion of the total popular vote going to such parties. This measure also indicates the structural response of the governing elites to working-class demands for an )pening of the system. Its data are drawn from official sources 5 reported by Rokkan (l966: 85; l967: 403) for Norway, and by erney (l957: 246, l96, 90) and Stjernquist (l966: 405) for Sweden. The two indices of democratization used by Peters will not be sed in this dissertation because of their lack of variance between 5 well as within Norway and Sweden, and because of serious questions oncerning their status as interval-level measures and their bjectivity and reliability, considering the fact that they are ubjectively coded. 5. Government Penetration Economic Penetration is measured by the total amount of government revenue, as reported by official sources, divided by the population size, rather than by thetotal Gross Domestic Product, as Peters did, because of the desirability of maintaining similar bases of standardization for as many of the variables as possible. This measure is an indicator of the degree of government involvement in the nation's economy. Social Penetration is measured by the number of civil servants, as recorded and linearly interpolated from official census figures, divided by the working-age population. This measure is an indicator of government involvement in the nation's social system. 6. Expenditures Government expenditures, as reported by official sources, are deflated by the price index and divided by the population size, and include total, non-defense,.health, education, and welfare. The latter category is reported by Norway as “social purposes,“ and the separate, specific welfare categories used by Peters are not reported for the entire time period. For Sweden this category is computed by summing the separate welfare figures obtained from Peters. 7. Personnel Services Health Personnel is measured by the number of doctors, as reported by official sources, divided by the population size, and 100 is an indicator of the effort to provide health care for the public. The number of hospital beds is not available for the entire time period in Norway. Educational Personnel is measured by the total number of teachers in public elementary and secondary schools, as recorded and linearly interpolated from official sources, divided by the population size, and is an indicator of the effort to provide public education. Indicators of welfare personnel, such as the number of social workers, are not completely available. 8. ObjectiveSeCUrity The following indicators of social conditions in the sectors of health, education, and welfare are all reported by official sources: Health_is measured by the number cf infant deaths per l000 live births, and by the total number of deaths per lOOO population. Education is measured by the number of pupils in public elemen— tary schools, divided by the school-age population. Welfare is measured by the number of persons receiving public elief payments, divided by the total population size. The number of ersons receiving public old-age pensions is not available for the ntire time period for Norway. II. The Hypotheses The following is a list of the hypotheses arising from the dis- ussion of the comparative histories of Norway and Sweden, the Bull- alenson hypothesis, the general mObilization model, and the ndicators of the concepts involved. They will be couched in the lOl onceptual terminology of the mobilization model, and will be eneralized to all indicators of those concepts. l) The indicators of social mobilization will be more strongly associated with the indicators of political mobilization for Norway than for Sweden. This hypothesis is based on Lafferty's finding that Norway's ndustrial surges coincided more closely with the rise of the olitical labor movement than did Sweden's. 2) The indicators of social mobilization will be more strongly associated with economic wealth for Norway than for Sweden. This hypothesis is based on Lafferty's finding that Norway's pro- huflzbeganand continued accelerating during her industrial surges, mereas Sweden's product accelerated in between her industrial surges. 3) The indicators of political mobilization will be more strongly associated with the indicators of democratization for Norway than for Sweden. This is based on the assertion that the greatest growth of the olitical labor movement occurred at about the same time as electoral eform in Norway, whereas electoral reform occurred later than the ncrease in working-class demands in Sweden. 4) The indicators of social mobilization will be more strongly associated with the indicators of democrati- zation for Norway than for Sweden. Norwa '5 major industrial surges seem to have occurred at about y e same time as her periods of electoral reform, whereas Sweden's rges seem to have occurred before and after electoral reform. 5) The indicators of democratization will be more strongly associated with the indicators of govern- ment penetration for Norway than for Sweden. 102 This hypothesis rests on the assumption that government activity increases with the passage of social legislation, and on the fact that major social legislation seems to have begun in the late l800's and early l900's in both countries, about the same time as electoral reform in Norway, but before it in Sweden. 6) The indicators of social mobilization will be more strongly associated with the indicators of government penetration for Norway than for Sweden. This hypothesis rests on the same assumption as hypothesis (5), and on the fact that Norway's social legislation can be grouped into waves which seem to coincide with her industrial surges, whereas Sweden's social legislation seems to have begun more smoothly and in between her industrial surges. - 7) There will be no differencebetween Norway and Sweden in the strength of association between the indicators of political mobilization and government penetration. This hypothesis, although resting on the same assumption as hypotheses (5) and (6), arises from the fact that social legislation seems to have developed simultaneous with the political labor move— ment in both Norway and Sweden, although much of it seems to have come during labor setbacks. I 8) There will be no difference between Norway and Sweden in the strength of association between the indicators of government penetration and government expenditures. The assumption concerning increased government activity asso- ciated with the passage of social legislation should apply to expenditures as well as to penetration, so that both sets of indi- cators should reflect that increase simultaneously. 103 9) The indicators of democratization will be more strongly associated with government expenditures for Norway than for Sweden. This follows logically from hypotheses (5) and (8), in that Norway's electoral reform occurred about the same time as the beginning of her social legislation, which should be reflected in increased expenditures. l0) There will be no difference between Norway and Sweden in the strength of association between the indicators of economic wealth and government expenditures. This is again based on the finding that per capita product accelerated at about the same time that social legislation got under way in both countries. The remaining four hypotheses deal with the relationships between government expenditures and the impacts of those expendi- tures on society. There is no information in the literature previously reviewed that would suggest any differences between Norway and Sweden in that area. Although Norway's levels of objective security and per capita product were cited in Chapter One as seeming to have been lower than Sweden's, government expen— itures probably have been lower, also. Direct cross-national omparisons of monetary indicators are made difficult by the dif- erence in currency values, and exchange rates are not even vailable for the entire time period. At any rate, there is no eason to anticipate any differences in the relationships between xpenditures and their impacts, although such differences might ell emerge from the examination of those relationships. Thus, 104 :he following four hypotheses all state an expectation of no lifference between the two countries. ll) There will be no difference between Norway and Sweden in the strength of association between government expenditures and the indicators of personnel services. l2) There will be no difference between Norway and Sweden in the strength of association between government expenditures and the indicators of objective security. l3) There will be no difference between Norway and Sweden in the strength of association between the indicators of personnel services and objective security. 14) There will be no difference between Norway and Sweden in the strength of association between the indicators of economic wealth and objective security. IV. The Methods The methods for evaluating the above hypotheses consist )rimarily of simple linear correlation and regression analysis. Time-series data composed of aggregate statistics such as the ones employed here meet most of the basic assumptions for that type of nalysis (Kmenta l97l: l97-304). Thus the first step is to compute he simple correlation coefficients for the bivariate relationships mong the specified indicators, and determine their statistical ignificance. In this regard the simple statistical significance f the coefficients themselves is not very helpful, because corre- ation coefficients for data of this type are usually quite high. hus, while correlation coefficients based on such data are sta- istically significant at the 0.0005 level when as low as 0.62, most re much higher than that, often attaining 0.95 and higher. This 105 is due to the similarity of time-series functions, especially for developmental variables, which tend to increase or decrease monotonically over time. Such variables may or may not be func- tionally related, and this high incidence of spurious correlation renders causal interpretation difficult and demands adequate theories to guide the analysis. But one other statistical technique is useful in determining the significance of serial correlation and regression coefficients. Such coefficients are often inflated by a phenomenon known as auto- correlation or autoregression, which arises from systematic associa- tion among the residual errors in predicting the dependent variable from the independent variable. These errors, which comprise the iisturbance term in regression equations, are supposed to be random and unassociated, each limited to that one observation, and together 1aving a mean of zero. However, when they are not, such as when :hey are systematically negative in one portion of the time period 1nd systematically positive in the other, a systematic disturbance, mually an exogenous variable whose effects linger over several bservations, is indicated. In this case the most effective method f eliminating autocorrelation is to bring the variable causing it nto a multiple regression equation, if that combination of indepen- ent variables can be justified theoretically. Another method of reducing autocorrelation is to split the ital time period into smaller portions, although this reduces the mber of observations and thus can distort the estimation of the 106 true regression parameters. In addition, autocorrelation can also indicate curvilinear relationships among the variables, in which case it is apprOpriate to attempt theoretically justified linear transformations of the variables. Finally, there are several specific procedures designed to estimate regression parameters from autocorrelated data (Kmenta l97l: 282-294). Since some of the labor variables can be used only from l905-l906 to l964-l965, and since the mobilization model frequently specifies multiple relationships, the first two of the aforementioned methods of solving the problem of autocorrelation will definitely be employed. Another problem, multicollinearity, arises in connection with the method of multiple regression. When the independent variables in a multiple regression equation are highly related to each other the estimation of the true regresssion parameters becomes distorted, reflected by an increase in the standard error of estimate (Kmenta l97l: 380-39l). Variables considered likely candidates for inclusion in an equation on the basis of theory, strong correlation with the dependent variable, and an appropriate pattern of residuals must be screened on the basis of the strength of their relationship with the independent variable(s) already included in the equation. Even if the simple or multiple correlation and regression coefficients prove to be statistically significant and free of autocorrelation and multicollinearity, there remains the problem >f evaluating the differences in those coefficients between Norway 107 1nd Sweden. Standard statistical tests of significance for the lifference between independent correlation coefficients require lifferences that are unrealistically large for time-series data with only 27 observations in each sample. That is, because the :orrelations for both countries will be rather large, very few of the cross-national differences would be statistically significant, aven if the countries were not so similar in "background" char- icteristics. Thusfthe differences will have to be judged rather subjectively in light of the previously reviewed literature. Finally, it will frequently be necessary to bolster the inter- aretation of differences in coefficients by visually examining Inivariate plots of superimposed individual variables. CHAPTER FIVE ANALYSIS OF THE TOTAL TIME PERIOD (l875 To l965) The previous chapter presented hypothesized differences between Norway and Sweden in terms of relationships among indicators of the conceptual components of the mobilization model of political develop- ment. It also introduced the appropriate methods required to test those hypotheses. The first step involves examining the differences in the linear, bivariate relationships between the indicators of the hypothesized pairs of concepts for Norway and Sweden, and assessing the extent of autocorrelation in those relationships. For this urpose we employed a linear least-squares correlation and regres- ion analysis program called LS, written in FORTRAN for the CDC 3600 omputer at Michigan State University by the MSU Agricultural Experi- ent Station's AES STAT Programming Section, and currently maintained. y the MSU STAT GrOUp, a division of the Computer Laboratory. The 5 program calculates numerous descriptive and inferential statistics or multiple and partial correlation and-regression analysis. It ffers an option to print the residual errors in predicting the ependent variable from the independent variable, accompanied by everal statistics, including the Durbin-Watson "d," a test of 109 significance for the extent of autocorrelation among those residuals. The Durbin-Watson Statistic is approximately normally dis- tributed around a "perfect" zero-autocorrelation value of 2.0, with the critical values defining the regions of rejection for a two- tailed test of the null hypothesis of no autocorrelation falling at varying distances above and below the perfect value, depending upon the number of observations in the sample, the number of independent variables in the regression equation, and the level of significance chosen. The regions below the perfect value indicate significant positive autocorrelation, and the regions above it indicate significant negative autocorrelation. For a one-tailed test only the regions on one side of the perfect value are used. In addition, for any given sample size, number of independent vari- ables, and the level of significance, an inconclusive region exists between the region of rejection and the region of acceptance, both below and above the perfect value. That is, in the case of positive autocorrelation the statistic must attain a value below the lower limit in order to be considered significant or above the upper limit in order to be considered not significant. If it falls in between the upper and lower limits, then the test is considered to be inconclusive (Kmenta l97l: 294-297). For our purposes, with 27 observations for the total time period of l875 to l965, only one independent variable, and a significance level of 0.05, the upper limit of the Durbin-Watson Statistic for a 110 one-tailed test is 1.47 and the lower limit is l.32. A relationship will be deemed acceptable if the Durbin-Watson Statistic is not definitely significant; that is, if it does not fall into the region of rejection below the lower limit. This criterion seems rather ‘ lenient in terms of both the significance level and the limit chosen. But time-series data tend to harbor considerable autocorrelation, so that lenience is necessary if any acceptable relationships are to be found. 0n the other hand, because correlation coefficients for such data tend to be rather high,airelatively severe criterion of sig? nificance is needed in order to screen out unreliable correlations. Our criterion will be a level of significance of 0.0005 for the proportion of variance in the dependent variable explained by the independent variable, according to an F-test performed by the LS program (Hays 1963: 573). This level seems unduly severe, but it corresponds to a critical value for the correlation coefficient of only about 0.62 or higher. The results of this analysis will be presented in an order orresponding to the hypotheses stated in the previous chapter. 0 facilitate direct comparison and thus evaluation of the hypotheses, he statistics for both countries will be presented side-by-side or 'n tandem for each relationship in each table. The Durbin-Watson tatistics will be presented in parentheses beneath the correlation oefficients for each relationship. Thus, a relationship will be deemed acceptable if the proportion f variance in the dependent variable explained by the independent 111 ariable is significant at 0.005 (corresponding to a correlation oefficient of about 0.62) and_if the Durbin-Watson Statistic is 393_ ignificant at 0.05 (greater than or equal to 1.32). The hypotheses ill be evaluated by examining the patterns of correlation and auto- rrelation in the acceptable relationships for the two countries. terpretation of these results will be reserved until after the alysis is completed. Results Linear, Bivariate Relationships Social Mobilization and Political Mobilization The first hypothesis presented in the previous chapter stated Tat the indicators of social mobilization will be more strongly ssociated with the indicators of political mobilization for Norway Tan for Sweden. Table 5-l shows that this is not the case: Of the ifteen correlations between indicators of social and political >bilization, ten are stronger for Sweden and only five are stronger r Norway. Relationships involving either Voter Turnout or Urban pulation consistently favor Sweden, although the correlation tween Gross Domestic Investment and Voter Turnout is not statis- cally significant. The correlations involving Union Membership, cept with Gross Domestic Investment, also consistently favor eden. The correlations involving Left Vote, except with Urban pulation, consistently favor Norway. Finally, of the five corre- tions which favor Norway, only the two involving Gross Domestic vestment are substantially stronger. However, the three 112 correlations favoring Sweden but ggt_involving either Voter Turnout or Urban Population are not substantially stronger, either. Social Mobilization and Political Mobilization, Total Period. Table 5-1 Correlation Coefficients and Durbin-Watson Statistics. Political Mobilization Social Union Left Voter Mobilization Membership Vote Turnout Norway Sweden Norway Sweden Norway Sweden Total 0.956 0.960 0.956 0.920 0.498 0.938 Population (0.365) (0.206) (0.404) (0.247) (0.880) (0.769) Agricultural ~0.935 -0.953 -0.923 -0.890 -0.472 -0.934 Employment (0.269) (0.207) (0.297) (0.203) (0.853) (0.723) Gross 0.950 0.782 0.767 0.534 _ 0.491 0.612 Investment (0.407) (0.191) (0.213 (0.103) (0.835) (0.229) Urban 0.791 0.968 0.794 0.872 0.507 0.912 Population (0.160) (0.288) (0.252) (0.199) (0.832) (0.584) Price 0.897 0.907 0.785 0.754 0.377 0.745 Index (0.513) (0.496) (0.291). (0.279) (0.809) (0.447) But more importantly, all of the relationships are severely auto— correlated, indicating either a missing variable or combination of variables needed to explain political mobilization, or curvilinear association between social and political mobilization. The extent of autocorrelation is higher for Sweden in every case save one, the relationship between Urban Population and Union Membership. HoweVer, the extent of autocorrelation is substantially higher only for rela- tionships involving Total Population, Gross Domestic Investment, and, 113 interestingly, Voter Turnout. At any rate, the first hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of the concepts of social and political mobilization. 2. Social Mobilization and Economic Wealth The second hypothesis presented in the previous chapter stated that the indicators of social mobilization will be more strongly associated with economic wealth for Norway than for Sweden. Again the results indicate that this is not the case: Table 5—2 shows that three of the five correlations are stronger for Sweden. How— ever, only the one involving Urban Population is appreciably stronger, and the two that favor Norway, Total Population and Agricultural Employment, are both appreciably stronger. Table 5—2 Social Mobilization and Economic Wealth, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Gross Social Mobilization National Product Total Agricul- Gross Urban P . .per Popula— tural Domestic Popula- Iréce Capita tion Employment Investment tion n ex Norway 0.942 -0.934 0.986 0.811 0.915 (0.238) (0.201) (0.671) (0.123) (0.691) 0.852 -0.877 0.989 0.905 0.958 Sweden (0.225) (0.266) (0.325) (0.229) (0.687) 114 The results are equally mixed in terms of the extent of autocorrelation present in the relationships. All are severely autocorrelated, more so for Norway in three of the five relation? ships. But of those, only Gross Domestic Investment shows a sub- stantially higher level of autocorrelation, whereas only the relationship for Urban Population is substantially more autocorre- lated for Sweden. Interestingly, this is the same one that shows a substantially higher correlation. Again, the second hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of social mobilization and economic wealth. 3. Political Mobilization and Democratization The third hypothesis presented in the previous chapter states that the indicators of political mobilization will be more strongly associated with the indicators of democratization for Norway than for Sweden. Table 5-3 again shows that this is not necessarily the case: Of the six correlations, three are stronger for Norway and three are stronger for Sweden. Of the three that favor Norway, only one, Union Membership versus Representation, is substantially stronger, and both correlations are relatively low. On the other hand, all three correlations that favor Sweden are substantially stronger. Voter Turnout correlates vastly more strongly with both Enfranchisement and Representation for Sweden than for Norway. Left Vote correlates slightly more strongly with both Enfranchisement and 115 Representation for Norway than for Sweden. Union Membership correlates substantially more strongly with Enfranchisement for Sweden than for Norway, but substantially more strongly with Representation for Norway than for Sweden. Table 5-3 Political Mobilization and Democratization, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Democratization Political Mobiliza- Enfranchisement Representation tion Norway Sweden Norway Sweden Union 0.827 0.930 ' 0.791 0.669 Membership (0.134) (0.654) (0.338) (0.219) Left 0.976 0.941 0.963 0.905 Vote (0.960) (0.487) (1.107) (0.616) Voter 0.374 0.836 0.393 0.889 Turnout (0.178) (0.537) (0.202) (0.477) This table also exhibits the interesting phenomenon of substantially lower autocorrelation for the country having the stronger correlation coefficient. Nevertheless, all of the relationships are heavily autocorrelated, so that none can be accepted, although the relationships involving Left Vote for Norway are by far the least autocorrelated. Again, the third hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between indicators of the concepts of political mobilization and democratization. 116 4. Social Mobilization and Democratization The fourth hypothesis presented in the previous chapter stated that the indicators of social mobilization will be more strongly related to the indicators of democratization for Norway than for Sweden. Table 5-4 shows that this does seem to be true, on the basis of the correlations alone: Of the ten correlations, most of which are rather weak, only three are stronger for Sweden, and only one of those, Urban Population versus Enfranchisement, can be con- sidered substantially stronger. Again, Urban Population consistently favors Sweden, although its correlation with Representation is not substantially stronger. The third correlation that slightly favors Sweden is the Price Index versus Enfranchisement. Of those seven correlations that favor Norway, only those inv01ving either Gross Domestic Investment or Representation can be considered even notice— ably stronger, and only the correlation between those two indicators can be considered substantially stronger. Table 5-4 also displays the interesting feature that all of the relationships involving Enfranchisement are less autocorrelated for Sweden than for Norway, whereas all of the relationships involving Representation are less autocorrelated for Norway than for Sweden. This also means that, for Representation, the stronger relationships are also less autocorrelated, except in the case of Urban Population. 0n the other hand, the pattern for Enfranchisement is mixed, since all of the relationships are less autocorrelated for Sweden and only two of the correlation coefficients are stronger for Sweden. 117 Table 5-4 Social Mobilization and Democratization, Total Period. Correlation Coefficients and Durbin-Watson Statistics. _fi Democratization Social . . Mobilization Enfranchisement Representation Norway Sweden Norway Sweden Total 0. 922 o. 916 0.880 0.805 Population (0.203) (0.464) (0.457) (0.302) Agricultural -O.892 -0.890 -0.838 —0.780 Employment (0.182) (0.409) =(0.385) (0.283) Gross 0.686 0.601 0.636 0.375 Investment (0.105) (0.209) (0.210) (0.148) Urban 0.799 0.887 0.712 0.738 Population (0.187) (0.437) (0.313) (0.250) Price 0.748 0.778 0.640 0.581 Index (0.208) (0.564) (0.319) (0.260) Nevertheless, all of the relationships in the table are severely auto- correlated, rendering them unacceptable. Again, the fourth hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of social mobilization and democratization. 5. Democratization and Government Penetration The fifth hypothesis presented in the previous chapter stated that the indicators of democratization will be more strongly asso- ciated with the indicators of government penetration for Norway than for Sweden. Table 5-5 shows that this does seem to be the case, but 118 only on the basis of the correlations alone: All are substantially stronger for Norway, and those involving Representation are vastly stronger. However, all of the relationships are unacceptable because of severe autocorrelation, so that the fifth hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of democratization and government penetration. Table 5-5 Democratization and Government Penetration, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Government Penetration Democrati- . . . zation Revenue C1Vl1 Serv1ce Norway Sweden Norway Sweden Enfran- 0.700 0.602 0.704 0.584 chisement (0.191) (0.249) (0.157) (0.185) Represen- 0.645 0.371 0.662 0.374 tation (0.228) (0.166) (0.193) (0.128) 6. Social Mobilization and Government Penetration The sixth hypothesis presented in the previous chapter stated that the indicators of social mobilization will be more strongly associated with the indicators of government penetration for Norway than for Sweden. Table 5-6 finally uncovers some acceptable relation- ships, and two of these three are Norwegian. Overall, of the ten correlations, only three favor Sweden, and none of these are sub- stantially stronger. Again, relationships involving Urban Population 119 consistently favor Sweden, although only slightly. The third relationship that very slightly favors Sweden is Gross Domestic Investment versus Revenue. Of the seven correlations that favor Norway, four can be considered substantially stronger, and those involve either Total Population or Agricultural Employment. Table 5-6 Social Mobilization and Government Penetration, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Government Penetration Mobili;giion Revenue Civil Service Norway Sweden Norway Sweden Total 0.902 0.779 0.914 0.788 Population (0.361) (0.235) (0.220) (0.170) Agricultural -o.904 -0.811 -0.918 -0.821 Employment (0.353) (0.272) (0.218) (0.214) Gross o 986 * 0.999 * 0.977 , 0.936 Investment (1.904) (1.847) (1.588) (0.520) Urban 0.798 0.843 0.802 0.855 Population (0.160) (0.236) (0.240) (0.174) Price 0.926 0.913 0.905 0.870 Index (0.799) (0.592) (0.398) (0.471) *Correlation significant at 0.0005 and_Durbin—Watson statistic not_ significant at 0.05. Three relationships are acceptable in terms of both signifi- cance of the correlation coefficient and lack of significance of the Durbin-Watson Statistic for autocorrelation. All three involve Gross Domestic Investment, and only the relationship of that indicator 120 ith Civil Service for Sweden is not acceptable. However, the ifference between the Norwegian and Swedish correlation coefficients or the non-autocorrelated relationship between Gross Domestic. nvestment and Revenue is negligible. But on the basis of the fact hat the relationship between Gross Domestic Investment and Civil rvice is acceptable for Norway but not for Sweden, and the fact at most of the relationships are stronger for Norway, the sixth pothesis will be very tentatively accepted for the total time riod. Political Mobilization and Government Penetration The seventh hypothesis presented in the previous chapter stated Tat there will be no difference between Norway and Sweden in the trength of association between the indicators of political mobiliza- ion and government penetration. Table 5-7 shows that this is not Ie case: The correlation coefficients are substantially and con- stently stronger for Norway in relationships involving either hon Membership or Left Vote. Voter Turnout again favors Sweden, th both indicators of government penetration correlating sub- antially more strongly with that indicator of political mobiliza— on for Sweden than for Norway. However, both correlations for eden are weak, and Revenue versus Turnout is not even significant. But once again all of the relationships are severely autocorre- ted. In three of the four relationships that favor Norway, the tent of autocorrelation is substantially lower, with the exception Left Vote versus Civil Service. For Sweden the extent of 121 Table 5-7 Political Mobilization and Government Penetration, Total Period. Correlation Coefficients and Durbin-Watson Statistics. ‘— Government Penetration Political . . . Mobilization Revenue C1v11 SerVTCe Norway Sweden Norway Sweden Union 0.956 0.785 0 958 0.789 Membership (0.758) (0.297) (0.836) (0.226) Left 0.789 0.531 0.780 0.509 Vote (0.287) (0.190) (0.145) (0.155) Voter 0.497 0.612 0.502 0.670 Turnout (0.292) (0.255) (0.332) (0.198) autocorrelation is substantially lower only in the relationship between Voter Turnout and Civil Service. Because there are no acceptable relationships, due to autocorrelation, the seventh hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of political mobili- zation and government penetration. 8. Government Penetration and Government Expenditures The eighth hypothesis presented in the previous chapter stated that there will be no difference between Norway and Sweden in the strength of association between the indicators of government pene- tration and government expenditures. Table 5—8 shows that this is not the case: Six acceptable relationships emerge, five of them in 'the Norwegian data. In four of those five the correlation coeffi- cients are stronger for Norway, although only slightly: Revenue 122 Table 5-8 Government Penetration and Government Expenditures, Total Period. Correlation Coefficients and Durbin-Watson Statistics. .__,_ Government Penetration Government . . . Expenditures Revenue C1v11 Serv10e Norway Sweden Norway Sweden 0.994 0.962 0.966 0.933 T°tal (1.051) (0.565) (1.428)* (0.666) Non- 0.991 , 0.985 0.969 , 0.956 defense (1.534) (0.731) (1.530) (0.450) 0.940 0.998 0.920 0.945 “ealth (0.734) (2.449)* (0.855) (0.579) . 0.875 0.984 0.853 0 926 Educat'on (0.635) (0.508) (0.724) (0.436) 0.965 0.959 0.941 0.959 “alfa'e (1.855)* (0.745) (1.520)* (0.468) *Correlation significant at 0.0005 afld_0urbin—Watson statistic ngt_ significant at 0.05. versus Non—defense and Welfare Expenditures, and Civil Service versus Total and Non-defense Expenditures. The only exception is the rela- tionship between Civil Service and Welfare Expenditures, which is non-autocorrelated for Norway but slightly stronger for Sweden. Overall, the stronger relationships split evenly between the two countries at five apiece. HoWever, for only the two relationships involving Education Expenditures is the difference substantial, with Sweden's correlations being higher than Norway's in both cases. On the other hand, Sweden shows a higher level of autocorrelation in even the unacceptable relationships and the Education correlations. 123 The lone exception is the relationship between Revenue and Health Expenditures, which is even acceptable on the basis of lack of autocorrelation. But on the basis of the fact that most of the acceptable relationships between the indicators of government pene- tration and government expenditures favor Norway, the eighth hypothesis can be rejected for the total time period. 9. Democratization and Government Expenditures The ninth hypothesis presented in the previous chapter stated that the indicators of democratization will be more strongly asso- ciated with government expenditures for Norway than for Sweden. Table 5-9 shows that this is not the case: The stronger correla- tions are almost evenly divided between the two countries, with six favoring Norway and four favoring Sweden. However, all of the corre— lations are low, and all of the ones that favor Norway are below or nearly below the chosen level of significance, whereas all four of the ones that favor Sweden are well above that value. Representa— tion is consistently stronger for Norway, although the correlations are not significant; whereas for Sweden, Enfranchisement is more strongly related to every class of expenditure except Health. Sweden shows higher autocorrelation in all but two cases, Enfranchisement versus Total and Health Expenditures. But once again, all of the relationships are severely autocorrelated, so that the ninth hypothesis cannot be accepted for the total time period on the basis of simple linear relationships among the indicators of democratization and government expenditures. 124 Table 5~9 Democratization and Government Expenditures, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Democratization. Government . . Expenditures Enfranchisement Representat10n Norway Sweden Norway Sweden Total 0.673 0.735 0.612 0.488 (0.203) (0.269) (0.225) (0.147) Non- 0.695 0.703 0.640 0.459 defense (0.236) (0.216) (0.250) (0.130) 0.686 0.598 0.636 0.366 ”ealth (0.230) (0.255) (0.223) (0.174) . 0.560 0.719 0.550 0.471 Educat'on (0.310) (0.262) (0.308) (0.157) 0.641 0.719 0.590 0,434 welfare (0.241) (0.219) (0.266) (0.122) 10. Economic Wealth and Government Expenditures The tenth hypothesis presented in the previous chapter stated that there will be no difference between Norway and Sweden in the strength of association between the indicators of economic wealth and government expenditures. be true: Table 5-10 shows that this does not seem to Three acceptable relationships emerge, two of which are Swedish; and all of the correlations in the table, although all very strong and not substantially different, consistently favor Sweden. Except for Health Expenditures, all of the Swedish relationships are also consistently less autocorrelated, and the one acceptable rela- tionship for Norway, involving Non-defense Expenditures, has a 125 Durbin-Watson Statistic that is barely above the critical value of 1.32. Thus, the tenth hypothesis cannot be accepted for the total time period on the basis of the simple linear relationships between economic wealth and government expenditures. Table 5-10 Economic Wealth and Government Expenditures, Total Period. Correlation Coefficients and Durbin-Watson Statistics. *— Gross Government Expenditures National Product Non- ' per Capita Total defense Health Education Welfare ‘ 0.968 0.968 , 0.973 Y 0.929 07939 N°rway (1.077) (1.352) (0.931) (0.623) (0.963) Sweden 0.981 0.995 * 0.985 0.992 * 0.977 (1.210) (2.032) (0.806) (1.480) (1.102) *Correlation significant at 0.0005 and Durbin-Watson Statistic ggt_ significant at 0.05. The next four hypotheses focus upon the impacts of government policy; and thus government expenditures, personnel services, and measures of objective security are grouped into sectors. The health sector consists of Health Expenditures, Doctors, and Infant Mortality and the Death Rate. The education sector consists of Education Expenditures, Teachers, and the Pupil Rate. The welfare sector con- sishs of Welfare Expenditures and the Relief Rate: No measure of welfare personnel was available. Only relationships among indicators in the same sector were analyzed. 126 11. Government Expenditures and Personnel Services The eleventh hypothesis presented in the previous chapter stated that there will be no difference between Norway and Sweden in the strength of association between government expenditures and the indicators of personnel services. Table 5-11 indicates that this hypothesis cannot be evaluated with the evidence at hand: The health—sector correlations are very similar, although Norway's is very slightly stronger and less autocorrelated; whereas the education-sector correlation is stronger for Sweden but also some— what more autocorrelated. At any rate, all four relationships are severely autocorrelated; and since there are not enough relation- ships to show a pattern, the eleventh hypothesis cannot be evaluated for the total time period. Table 5-11 Government Expenditures and Personnel Services, Total Period. Correlation Coefficients and Durbin—Watson Statistics. Personnel Services Government Expenditures in Sector Health Education 0.930 0.731 Norway (0.479) (0.203) 0.923 0.846 Sweden (0.340) (0.102) 12. Government Expenditures and Objective Security The twelfth hypothesis presented in the previous chapter stated that there will be no difference between Norway and Sweden in the strength of association between government expenditures and the indicators of objective security. 127 Table 5-12 shows that this is not the case, at least not on the basis of the correlation coeffi- cients alone: 0f the four relatively weak correlations, three are substantially stronger and less autocorrelated for Norway than for Sweden. The welfare-sector correlation is vastly stronger, and its Durbin-Watson Statistic is just barely below the critical value of 1.32. The exception is the education sector, where the relationship is substantially stronger and less autocorrelated for Sweden. How- ever, even the correlation for Sweden is far from being statistically significant, so that the pro-Norway pattern still holds. less, all of the relationships are relatively weak and severely Neverthe- autocorrelated, so that the twelfth hypothesis cannot be accepted for the total time period on the basis of linear, bivariate rela- tionships between the indicators of government expenditures and objective security. Table 5-12 Government Expenditures and Objective Security, Total Period. Correlation Coefficients and Durbin-Watson Statistics. Objective Security Government . Exipne nSde1ctiilorres MoIrntgalrilty DReaatteh PRuaptie1 RS; if: Norway 28:34:) 23:23?) (8:2;g) (7:288) Sweden gg;$$2) E33383) '(83353) (Siiig) __¥ 128 13. Personnel Services and Objective Security The thirteenth hypothesis presented in the previous chapter stated that there will be no difference between Norway and Sweden in the strength of association between the indicators of personnel services and objective security. Table 5-13 shows that this is not the case, at least not on the basis of the correlation coefficients alone: All three are stronger for Norway, although only the education-sector correlation is substantially so. All three are also severely autocorrelated, and Norway's are less autocorrelated than Sweden's, although the education-sector relationship is only slightly so. Thus the thirteenth hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relationships between the indicators of personnel services and objective security. Table 5-13 Personnel Services and Objective Security, Total Period. Correlation Coefficients and Durbin-Watson Stat1st1cs. Objective Security Personnel . 15:22:50: Moi‘rizgil'i‘t y DReaatteh Plijaptle1 “my 28:212.) It???) (8:32?) We" (8:323) (8:333) (8:313) 14. Economic Wealth and Objective Security The fourteenth and final hypothesis presented in the previous chapter stated that there will be no difference between Norway and 129 Sweden in the strength of association between the indicators of economic wealth and objective security. Table 5-14 shows that this is not the case: All four relationships are substantially stronger and less autocorrelated for Norway than for Sweden, although the education-sector correlation is far from being statistically sig- nificant and is less autocorrelated for Sweden than for Norway. In addition, all four relationships are also rather weak and severely autocorrelated, so that the fourteenth hypothesis cannot be accepted for the total time period on the basis of linear, bivariate relation- ships between the indicators of economic wealth and objective security. Table 5-14 1 Economic Wealth and Objective Security, Total Period. Correlation Coefficients and Durbin—Watson Statistics. Gross Objective Security National Product Infant Death Pupil Relief per Capita Mortality Rate Rate Rate -0.874 . -0.795 0.422 -0.698 Norway (0,357) (0.823) (0.629) (0.923) -0.757 -0.709 0.370 -O.232 Sweden (0.203) (0,292) (0.983) (0.382) B. Linear, Multiple Relationships 1. Procedure Most of the relationships just presented were severely autocor- related, rendering them unacceptable. As stated in the previous chapter, the most theoretically palatable method of removing 130 autocorrelation from linear, bivariate relationships is to find the missing variable or variables that are systematically distorting the prediction of the dependent variable from the independent variable. Thus the second step in the search for acceptable relationships on which to base evaluation of the hypotheses involved examining the residual errors in prediction printed by the LS program. The mobilization model guided the identification of independent- variable candidates for inclusion in multiple regression equations. That is, in the attempt to build multiple equations designed to explain dependent-variable concepts for which the mobilization model specifies two or more concepts as independent variables, only the indicators of those concepts were considered for inclusion in those particular equations. In the case of dependent-variable concepts for which only one concept was specified as independent, only indi— cators of that concept were considered for inclusion. The problem with the latter procedure is multicollinearity: Indicators of the same concept should correlate very highly with each other, and thus should not be used together as independent variables in the same equation. But multicollinearity can be considered after the equa- tions have been built according to more immediate criteria. Once the candidates for inclusion were identified according to the mobilization model, the residuals of the relationships between the dependent variable and the independent variables proposed for an equation were compared. If their patterns were even roughly opposite, so that they could be expected to compensate for each other's errors % 131 in predicting the dependent variable, then a multiple regression equation including those variables was tried on the LS program. The equation was again deemed acceptable if the F—test for the pro- portion of variance explained in the dependent variable was signifi- cant at the 0.0005 level and the Durbin-Watson Statistic was not significant at the 0.05 level. The corresponding critical value of the Durbin-Watson Statistic with two independent variables was now 1.24, and the significance level of the proportion of variance explained was again determined directly from the LS program. 2. Results and Implications for the Hypotheses Despite numerous attempted equations, this procedure yielded no acceptable equations for Norway and only one for Sweden: Total Expenditures versus Enfranchisement and Gross National Product per Capita. Table 5-15 presents the statistics for the multiple equation on the first line and the simple and partial statistics for the bivariate relationships between each independent variable and the dependent variable on the second and third lines. It shows that this multiple equation is only a slight improvement over the bivariate relationship between Total Expenditures and Gross National Product per Capita: The correlation is improved by only two one—hundredths; the standard error of estimate is only slightly lower; and the Durbin-Watson Statistic is barely non-significant. The partial correlations, betas, and significance levels indicate that GNP is far more important in the equation than is Enfranchisement. 132 wwwqmo Lea mzu mooo.o mom.o #om.o mw.mw ~mm.o OPN.P Pmo.o mo_.o . omm.o mo.mw wom.o mmm.o mom.o pcmammwgocmewcm mooo.o me.N~ ewm.o 468.0 Pmm._ .mmtzoeeeoexm memm 2 m o m 3 m mcowamscm mpnmquuu< oz > < 3 a o z WHW>m4 mwcmwmz mcowp mcocem cmcmmpqu mcowp more up cmw mpwm -epmcgou cementum meanwce> -mFmeeou -mwpmpm .44: Pm Fewest; ecu :owp mFQEwm new comet: mmPnewem> 16—mgcoo xnwemmcwp newness o_aeppaz -_oo_p_:z mowumwpmpm mepcea mowpmwpepm «Faewm new mPavazz .uowcwm Peach .mcowumscm :owmmwcmmm mpawp_:z mFpramoo< mpnm mFQwH F— 133 These results do not change the evaluation of the ninth and tenth hypotheses: Norway still does not show a stronger relation— ship between democratization and government expenditures; and the one acceptable equation bolsters Sweden's stronger relationship between economic wealth and government expenditures, thus further rejecting the hypothesis of no difference between the two countries for that relationship. II. Acceptable Relationships Figure 5-1 summarizes the acceptable relationships established for the total time period on the basis of both bivariate and multi- variate linear patterns of correlation and autocorrelation. Again, a relationship was deemed acceptable if the F-test for the proportion of variance explained in the dependent variable was significant at the 0.0005 level and the Durbin-Watson Statistic was not significant at the 0.05 level. This means that the acceptable relationships had significant correlation and significant lack of autocorrelation. The major dependent—variable concepts of the mobilization model are listed above their indicators that can be explained by the listed independent variables. The first striking feature of this figure is that Norway has eight acceptable relationships, whereas Sweden has only five. This suggests that although the countries were roughly equal in number of stronger relationships, the Swedish data must have been more autocorrelated. The second interesting feature of this figure is that only two relationships are duplicated: Revenue versus Investment; and 134 Figure 5-1. Acceptable Relationships, Total Period. —r x a v—w v—vv- . a? ——w N O R'W A Y S W E D E N Dependent Independent Dependent Independent Variable Variables Variable Variables fi-t: Political Mobilization None None Economic Wealth None None Democratization None . None Government Penetration Revenue Investment RevenUe Investment Civil Service Investment Government Expenditures Total Civil Service Total Enfranchisement, Non-defense Revenue GNP per Capita Non-defense Civil Service Non-defense GNP per Capita Non-defense GNP per Capita Welfare Revenue Health Revenue Welfare Civil Service Education GNP per Capita Objective Security None None 135 Non-defense Expenditures versus Gross National Product per Capita. This should indicate that there are significant differences between the two countries. The most noticeable difference is the appearance of Civil Service as both an independent and dependent variable in the Norway column, and its absence in the Sweden column. Similarly, Welfare Expenditures appears in the Norway column but not in the Sweden column, whereas Health and Education Expenditures appear in the Sweden column but not in the Norway column. Gross National Product per Capita appears relatively more frequently in the Sweden column than in the Norway column, and Enfranchisement only improves Sweden's relationship between GNP and Total Expenditures enough to make it barely acceptable. But perhaps the most significant feature of Figure 5—1 is that, except for Civil Service and Enfranchisement, all of the acceptable relationships involve economic, monetary measures: None of them involve social or political measures. The Price Index is the only economic measure not involved in an acceptable relationship. Further- more, Civil Service can easily be conceptualized as part of a milieu of measures indicating the relative size of the central government in the nation's economy, including Revenue and Expenditures. And it has already been pointed out that Enfranchisement is useful only for removing just enough of the autocorrelation from Sweden's relation- ship between GNP and Total Expenditures to make it acceptable. This feature may mean one or both of two things. First, economic measures may be more reliable than social and political measures at 136 this level of analysis, perhaps because of the common monetary unit of measurement, so that they correlate more readily without significant autocorrelation. Second, economic measures may be more interrelated than social and political measures because they are part of a separate and more cohesive milieu of phenomena, whereas socio-political phenomena are interrelated in more ambiguous and complex ways. But such inter- pretations must await further analysis. CHAPTER SIX ANALYSIS OF THE POST-1905 TIME PERIOD As mentioned in Chapter Four, the third step in the search for acceptable linear relationships involves splitting the total time period into shorter portions in order to reduce the variance in the systematic disturbance causing the autocorrelation. As with finding additional independent variables, the search for temporal cutting points must be guided by theory as well as by such practical con- siderations as the point at which the autocorrelated residuals change sign, the number of observations included in the chosen time portion, or the measures available during that portion. In this respect we were quite fortunate, because one of the most theoretically justifiable cutting points for both Norway and Sweden happens to coincide with the point at which some new measures become available. Norway and Sweden were joined in a loose union until 1905, about the same time that the national federations of labor had begun to gather statistics on phenomena associated with organized labor. Two such measures, Labor Conflict and Unemployment, were included in our set of data but were heretofore not included in the analysis because they were unavailable before 1902. Furthermore, 137 138 visual examination of plots of the time-series (see the Appendix) indicates that both countries did not really begin developing economically, socially, and politically until the early part of the Twentieth Century. That is, there is very little variance in the series before then. Finally, in many of the bivariate relationships the autocorrelated residuals change sign at some point in the first decade of the Twentieth Century. 1 Thus, 1905 was chosen as the temporal cutting point, and since that left only ten observations in the early portion, analysis was performed only on the 17 observations in the post-1905 period. The lower number of observations changed the critical values to about 0.75 for the correlation coefficient and 1.13 for the Durbin—Watson Statistic, given the same levels of significance as before. How- ever, the significance level for the proportion of variance explained in the dependent variable was again determined directly from the LS program. I. Results A. Linear, Bivariate Relationships 1. Social Mobilization and Political Mobilization Table 6-1 casts further doubt on the first hypothesis: Four acceptable relationships emerge, compared with none for the total time period, and three of those favor Sweden. Overall, the stronger correlations split evenly between the two countries at twelve apiece, and Norway shows eight relationships that are substantially stronger, while Sweden shows seven. The correlations that are not statistically .00.0 00 00000000000.mmm o_0000000 000003-000200.mmm 0000.0 00 00000000000 00000_otco04 2_00.Fv 2000.00 2004.00 2004.00 2000.00 2400000 2004.00 2004.00 x0000 400.0 040.0 000.0 0_0.0 404.0- 400.0- 00000 040.0 00000 2004.00 2000.00 200_.0v 204_.0V 2040._0 2000.00 2000.00 200_.0v 040.0- 00_.0 00_.0 0.0.0 040.0- 040.0 000.0- 000.0- 000000_00000 2000._0 2404.00 2040.00 2400.00 2040.00 2000._v 2000.00 2_F4.0V 0000000000 4 _00.0 000.0 000.0 000.0 404.0- 004.0- 000.0 000.0 00020 2040._V 2000.00 2000.00 2000.00 2000.00 2020.20 2400.00 2000.00 0000000>00 200.0 000.0 004.0 000.0 040.0- 000.0- 000.0 000.0 00020 2000.20 20_0.0v 2000.00 2004.00 2040.00 2040._v. 2000.00 2040.00 000000_0s0 o. 4 000.0- 400.0- 000.0- 040.0- 004.0 __0.0 " 000.0- 200.0- _0t00_0o_t0< a. - 4200N._V. 2__0._0 2000.00 2000.00 2004.00 2000.20 2004.00 2000.00 0000020000 000.0 4 000.0 400.0 0_0.0 0_0.0- 044.0- 000.0 _00.0 20000 :mumzm 202002 cmumzm 203202 :mvmzm 203202 cmvmzm 203g02 pzocgzp muo> pUWchou awgmemnsmz :00p0N0—42o2 cwuo> “mm; Loam; cows: meoom 00000N0_0002 _0o_00000 .mowpmwpmpm compmzncwagzo 2:0 mpcwwowtmmou cowpepmgcou .000200 000_ 0000 .00000N0_0002 _0o000200 000 00000N0_0002 _00000 Flo mpnwh 140 significant are also split evenly at six apiece. Four of the non-significant Norwegian correlations are in the substantially— stronger category, compared with three for Sweden. Most of the non-significant correlations involve either of the two new variables added to the relationship between social mobili- zation and political mobilization for the post-1905 period, Labor Conflict and Unemployment. The only other non—significant corre- lations are between the Price Index and Left Vote, which is stronger for Norway, and Voter Turnout, which is stronger for Sweden; and between Urban Population and Left Vote, which is also stronger for Sweden. All of the correlations for the two new variables are statistically non-significant, although the ones involving Labor Conflict are substantially non-autocorrelated, attributable to that variable's wild fluctuations. (See Figure A-3 in the Appendix.) Interestingly, the relationship between these two indicators is almost non-existent for Sweden and fairly strongly positive, although still far from significant, for Norway. Finally, the relationships between these two indicators and the other indicators of social and political mobilization are all negative except for Unemployment and Left Vote for both countries, and Unemployment and Turnout for Norway. Thus, neither of these new indicators is very useful. As for the consistency of the other indicators, Gross Domestic Investment again consistently favors Norway. Voter Turnout again consistently favors Sweden in every case except Investment, and there the difference between the two countries is slight. Urban l4l Population also consistently favors Sweden in every case except Labor Conflict, and there the difference between the two countries is negligible. Compared with the total time period, the extent of autocorrela- tion is generally reduced; and all of the Swedish relationships involving Union Membership and Left Vote are consistently more auto- correlated, with the single exception of Unemployment versus Left Vote. However, only the six relationships involving Total Population, Agricultural Employment, and Gross Domestic Investment show sub— stantially more autocorrelation. Besides, all eleven of the I relationships are significantly autocorrelated. For Norway, all of the relationships involving Labor Conflict and Voter Turnout are substantially more autocorrelated, but only five of these, Voter Turnout versus every social mobilization indicator except Total Population, are significantly autocorrelated. No clear pattern of association between correlation and auto— correlation emerges, but the pattern of non-significant autocorrela— tion is clear: All four of the acceptable relationships in Table 6-l involve Voter Turnout, and three of these are Swedish. A fourth Swedish relationship, Turnout versus the Price Index, just misses being acceptable because its correlation coefficient is barely non-significant. The only two indicators of social mobili- zation which do not relate acceptably to Turnout are Unemployment, which has been shown to be useless, and Gross Domestic Investment, which has heretofore tended to work better for Norway. Even the 142 relationship that is acceptable for both countries, Turnout versus Total Population, is stronger for Sweden, although not substantially. All of this evidence indicates that the first hypothesis still can- not be accepted: If anything, the results so far indicate a stronger relationship between social and political mobilization for Sweden than for Norway. 2. Social Mobilization and Economic Wealth Table 6-2 indicates that the evidence relevant for the second hypothesis is still ambiguous; but the one acceptable relationship does favor Norway. Table 6-2 Social Mobilization and Economic Wealth, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Social Mobilization Gross fiifi;fifl:g Agricul—' Gross er Total tural Domestic Urban CJJita Popula- Employ- Invest— Popula- Unem- Price p tion ment ment tion ployment Index Norwa 0.973 -0.990 * 0.986 0.9ll -0.255 0.874 y (0.6l3) (l.259) (0.9l7) (0.48l) (0.l24) (0.665) Sweden 0.916 -0.947 0.992 0.946 0.451 0.955 (0.398) (0.653) (0.466) (0.370) (0.270) (0.931) *Correlation significant at 0.0005 ang_Durbin-watson Statistic not_ significant at 0.05. Compared with the total time period, all of the correlations are stronger except Investment for Norway, and all are less autocorrelated except the Price Index for Norway. Overall, the correlations are 143 stronger for Sweden than for Norway in four of the six relationships, but four of the six are also substantially more autocorrelated for Sweden. 0f the six relationships, only two are appreciably stronger for either country: Unemployment and the Price Index. Both favor Sweden and show more autocorrelation for Norway, but the correla- tions for Unemployment in both countries are once again far from being statistically significant. Although the one acceptable rela- tionship shows a correlation that is not substantially stronger for Norway, the second hypothesis will be very tentatively accepted for the post-l905 time period. 3. Political Mobilization and Democratization Table 6-3 lends some support to the third hypothesis: The two acceptable relationships both favor Norway. Compared with the total time period, the correlations are weaker except for Norway's relation- ships between Enfranchisement and Left Vote, Enfranchisement and Turnout, and Representation and Turnout. Overall, the correlations involving Enfranchisement are all stronger for Sweden, except in the case of Voter Turnout. This is ironic considering that most relationships involving Turnout have heretofore favored Sweden. But the latter correlation is barely non-significant, and the difference between the two countries there is very slight. 0n the other hand, all of the correlations involving Representation are stronger for Norway, except in the case of Labor Conflict, whose correlations are highly non—significant anyway. Thus, the only statistically significant correlations are Enfranchisement versus both Union 144 Membership and Left Vote for both countries, but favoring Sweden; and the two acceptable relationships, Representation versus both Left Vote and Voter Turnout for Norway. Table 6-3 Political Mobilization and Democratization, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Democratization Mg:1))g:¥)0n Enfranchisement Representation Norway Sweden Norway Sweden Union 0.768 0.868 ‘ 0.634 0.165 Membership (0.354) (0.751) (0.806) (0.762) Labor —0.l68 -0.529 ~0.121 ~0.l87 Conflict (0.211) (1.026) (0.491) (0.704) Left 0.909 0.921 0.832 * 0.538 Vote (1.097) (0.865) (1.247) (0.955) Voter 0.721 0.714 0.836 * 0.393 Turnout (0.735) (0.884) (1.834) (0.681) *Correlation significant at 0.0005 and_Durbin-Watson Statistic not. significant at 0.05. Compared with the total time period, the extent of autocorrelation is reduced in all of the relationships. The more strongly correlated relationships are also considerably less autocorrelated in every case except Enfranchisement versus Left Vote, which is more autocorrelated for Sweden, and Enfranchisement versus Voter Turnout, which is more autocorrelated for Norway. Although the correlations for the two acceptable relationships are relatively low, they are substantially 145 stronger for Norway than for Sweden. Thus, the third hypothesis will be very tentatively accepted for the post-1905 time period. 4. Social Mobilization and Democratization Table 6-4 lends no further support to the fourth hypothesis. Compared with the total time period, all of the relationships are weaker but less autocorrelated. All of the twelve relationships are still severely autocorrelated, and only three of them are statiscially significant. Furthermore, although eight of them are stronger for Norway, none of these is statistically significant. All six of the correlations involving Representation are substan- tially stronger for Norway except Urban Population, which is only somewhat stronger. Enfranchisement versus both Gross Domestic Investment and the Price Index are also stronger for Norway, although the latter is not substantially so. Thus, Representation and Invest— ment continue to favor Norway consistently. 0f the four correlations that are stronger for Sweden, only Urban Population is substantially so. But all of them except Unemployment are statistically signifi- cant. Interestingly, the relationships involving Representation are not only stronger for Norway than for Sweden but also substantially more autocorrelated, except for Total Population, which is somewhat more autocorrelated for Sweden. In the relationships involving Enfranchisement, the stronger relationships also tend to be some- what less autocorrelated, except for Unemployment and the Price Index. But since all of the relationships are not acceptable because 146 of the extent of autocorrelation, the fourth hypothesis still cannot be accepted. Table 6-4 Social Mobilization and Democratization, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Democratization M0b??§:gl10n Enfranchisement Representation Norway Sweden Norway Sweden Total 0.825 ' 0.829 0.681 0.243 Population (0.426) (0.505) (0.815) (0.772) Agricultural -0.715 -0.776 -0.555 -0.l90 Employment (0.325) (0.452) (0.651) (0.761) Gross 0.662 0.508 0.514 0.048 Investment (0.308) (0.306) (0.611) (0.743) Urban , 0.502 0.771 0.260 0.201 Population (0.248) (0.478) (0.494) (0.760) 0.152 0.172 0.402 0.229 Unemployment (00179) (00175) (0.376) (0.754) Price 0.683 0.632 0.340 0.142 Index (0.396) (0.588) ‘ (0.575) (0.782) 5. Democratization and Government Penetration Table 6-5 lends no further support to the fifth hypothesis. Compared with the total time period, all of the relationships are weaker, and all but Sweden's relationships involving Representation are less autocorrelated. All of the relationships are severly auto- correlated and statistically non-significant. But all are substan- tially stronger and less autocorrelated for Norway, except for 147 Enfranchisement versus Revenue, which is slightly less autocorrelated for Sweden. Thus, the fifth hypothesis still cannot be accepted. T_at2_l£_§-_5 Democratization and Government Penetration, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. ‘2 Government Penetration Democrati- . . . zation Revenue C1v11 Serv1ce Norway Sweden Norway Sweden Enfran- 0.661 0.514 0.661 0.497 chisement (0.268) (0.271) (0.273) (0.193) Represen- 0.491 0.047 0.547 g 0.026 tation (0.301) (0.151) (0.304) (0.108) 6. Social Mobilization and Government Penetration Table 6-6 strengthens the support for the sixth hypothesis: A fourth acceptable relationship, Agricultural Employment versus Revenue, is added to those uncovered in the total time period, and that relationship is Norwegian. Compared with the total time period, all of the relationships are weaker except Norway's relationships involving Gross Domestic Investment and the Price Index, and all of the relationships are less autocorrelated. Overall, all of the cor- relations favor Norway, except those involving either Unemployment or the Price Index, plus the relationship between Urban Population and Civil Service, and the one relationship that is acceptable for both countries, Investment versus Revenue. None of the correlations that favor Norway are substantially stronger, and only the Tab1e 6-6 148 Social Mobilization and Government Penetration, Post-1905 Period. Correlation Coefficients and Durbin-Watson Statistics. Government Penetration Mob111iglion Revenue Civil Service Norway Sweden Norway Sweden Total 0.939 0.869 0.952 0.875 Population (0.687) (0.365) (0.452) (0.207) Agricultural -0.977 * -0.909 -0.981 -0.908 Employment (1.486) (0.546) (0.984) (0.381) Gross 0.982 * 0.999 * 0.973 * 0.945 Investment (1.953) (1.867) (1.815) (0.611) Urban 0.946 0.908 0.888 0.916 Population (0.636) (0.338) (0.876) (0.220) -0.381 -0.473 -0.378 -0.594 Unemmoyment (0.321) (0.348) (0.195) (0.425) Price 0.899 0.922 0.875 0.890 Index (0.865) (0.422) (0.703) (0.815) *Correlation significant at 0.0005 aflg_Durbin—watson Statistic not_ significant at 0.05. statistically non-significant relationships involving Unemployment are substantially stronger for Sweden. The extent of autocorrelation is considerably lower in every stronger relationship except Urban Population versus Civil Service. But on the basis of the two rela- tionships that are both acceptable and stronger for Norway and not for Sweden, Agricultural Employment versus Revenue and Investment versus Civil Service, the sixth hypothesis can be accepted with a little more assurance. 149 7. Political Mobilization and Government Penetration Table 6-7 casts further doubt on the seventh hypothesis: The one acceptable relationship, Union Membership versus Civil Service, is Norwegian; and it is substantially stronger than its Swedish counterpart, indicating a difference between the two countries. Table 6-7 Political Mobilization and Government Penetration, Post—l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Government Penetration Mopbglliiizziactaign Revenue Civil Service Norway Sweden Norway Sweden Union 0.954 0.773 0.965 * 0.797 Membership (0.991) (0.334) (1.442) (0.249) Labor -0.576 -0.345 -0.579 —0.303 Conflict (0.704) (0.406) (0.810) (0.320) Left 0.786 0.454 0.840 0.410 Vote (0.529) (0.226) (0.306) (0.177) Voter 0.719 0.727 0.748 0.775 Turnout (0.553) (0.782) (0.805) (0.750) *Correlation significant at 0.0005 aflg_Durbin-Watson Statistic 02$. significant at 0.05. The differences between the total time period and the post-1905 period in terms of the strength of the correlations is mixed; but all of the post—1905 relationships are less autocorrelated. Over- all, the relationships involving Union Membership, Labor Conflict, or Left Vote are substantially and consistently stronger and less autocorrelated for Norway than for Sweden, although the ones 150 involving Labor Conflict are not statistically significant. Again, Voter Turnout favors Sweden, although the differences are very slight. Furthermore, the relationship for Revenue is just barely non-significant, but substantially less autocorrelated for Sweden than for Norway; and the relationship for Civil Service, although barely significant, is more autocorrelated for Sweden than for Norway. But the one acceptable relationship means that the seventh hypothesis still cannot be accepted. 8. Government Penetration and Government Expenditures Table 6-8 does not change the earlier assessment of the eighth hypothesis: The relationships between indicators of government Table 6-8 Government Penetration and Government Expenditures, Post—l905 Period. Correlation Coefficients and Durbin-watson Statistics. Government Penetration Government . . . Expenditures Revenue C1v11 SerV1ce Norway Sweden Norway Sweden 0.993 0 963 0.957 * 0.943 Total (1,085) (0.684) (1.517) (0.901) Non- 0.987 * 0.987 0.962 * 0.969 defense (1.530) (0.941) (1.678) (0.608) 0.918 0.997 * 0.893 0.955 Health (0.719) (2.446) (0.851) (0.710) . 0.839 0.988 0.813 0.933 Educatlon (0.634) (0.758) (0.725) (0.532) 0.960 0.953 0.933 * 0.973 ”alfare (2.063)* (0.769) (1.721) (0.674) *Correlation significant at 0.0005 agg_Durbin-Watson statistic 09$. significant at 0.05. 151 penetration and government expenditures for the post-1905 period are not very different from the same relationships for the total time period. For all of the relationships involving Revenue and two of the relationships involving Civil Service, the differences in strength of correlation are reduced slightly. Furthermore, the extent of autocorrelation in the relationships is generally reduced. But five of the six acceptable relationships are still Norwegian, and thus the eighth hypothesis still cannot be accepted. 9. Democratization and Government Expenditures Table 6-9 does not change the earlier assessment of the ninth hypothesis: Compared with the total time period, all of the Table 6-9 Democratization and Government Expenditures, Post—l905 Period. Correlation Coefficients and Durbin-Natson Statistics. Democratization Eggzfigggfiggs Enfranchisement Representation Norway Sweden Norway Sweden Total 0.635 0.633 0.452 0.083 (0.274) (0.292) -(0.282) (0.121) Non- 0.648 0.602 0.472 0.078 defense (0.320) (0.236). (0.310) (0.108) 0.621 0.511 0.464 0.041 Hea]th (0.299) (0.275) (0.284) (0.162) . 0.519 0.619 0.372 0.085 Education (0.375) (0.284) (0.345) (0.137) 0.665 0.648 0.512 0.068 welfare (0.359) (0.260) (0.377) (0.102) 152 relationships are weaker and now statistically non—significant, although all but Sweden's relationships involving Representation are also less autocorrelated. But the fact that none of the rela- tionships are acceptable means that the ninth hypothesis still cannot be accepted. 10. Economic Wealth and Government Expenditures Table 6-10 does not change the earlier assessment of the tenth hypothesis: Compared with the total time period, all of the rela- tionships for Sweden except the one involving Health Expenditures are weaker; but all are also less autocorrelated. Table 6-10 Economic Wealth and Government Expenditures, Post-l905 Period. Correlation Coefficients and Durbin—Watson Statistics. Gross Government Expenditures National Product Non— per Capita Total defense Health Education Welfare 0.963 * 0.958 * 0.965 0.917 0.934 ”Orway (1.193) (1.409) (0.924) (0.633) (1.128) 0.976 * 0.994 * 0 988 0.990 * 0.972 SWEde” (1.235) (2.078) (1.124) (1.503) (1.124) *Correlation significant at 0.0005 ang_Durbin-Watson Statistic not_ significant at 0.05. Two acceptable relationships, involving Total Expenditures for both countries, are added to the three remaining from the total time period. The relationships involving Total and Non-defense Expenditures 153 are acceptable for both countries, and those involving Welfare Expenditures just barely miss being acceptable for both countries. But the relationship involving Education Expenditures is acceptable only for Sweden, and the relationship involving Welfare Expenditures just misses being acceptable for Sweden. Furthermore, all of the correlations in the table are stronger for Sweden, although most of them not substantially. Nevertheless, the relationships in general seem to favor Sweden, so that the tenth hypothesis still cannot be accepted. 11. Government Expenditures and Personnel Services Table 6-11 sheds no further light on the eleventh hypothesis: Compared with the total time period, all of the relationships except for the health sector are stronger, and all are less autocorrelated. Although Sweden's relationships are now consistently stronger and less autocorrelated than Norway's, none are acceptable because of severe autocorrelation, so that the eleventh hypothesis still can- not be accepted. Table 6-11 Government Expenditures and Personnel Services, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Government Personnel Services Expenditures in Sector Health Education 0.925 0.792 Norway (0.560) (0.565) 0.941 0.902 Swede" (0.492) (0.369) 154 12. Government Expenditures and Objective Security Table 6-12 does not change the earlier assessment of the twelfth hypothesis: The differences between the total period and the post—1905 period in terms of the strength of correlation are mixed, and the relationships are all less autocorrelated for the post-1905 period. However, only the relationships involving Infant Mortality are statistically significant, and none are acceptable because of significant autocorrelation. Thus, the twelfth hypothe- sis still cannot be accepted. Table 6-12 Government Expenditures and Objective Security, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Objective Security Government Etfif1§;§;;?s Infant Death Pupil Relief Mortality Rate Rate Rate -0.871 -O.66l -0.063 ' -0.741 N°rway (0.388) (0.937) (1.291) (1.298) -O.758 -0.669 0.520 -O.328 SWGde" (0.247) (0.276) (1.003) (0.392) 13. Personnel Services and Objective Security Table 6-13 does not change the earlier assessment of the thirteenth hypothesis: Although the pattern of difference is mixed, there is only one acceptable relationship, and it is Norwegian. The differences between the total period and the post-1905 period in terms of the strength of correlation are also mixed, and the relationships are all less autocorrelated for the post-1905 period. The relationships for 155 the education sector for both countries are not statistically significant; the difference between the countries in the relation- ships involving Infant Mortality are slight; and the acceptable relationship involving the Death Rate for Norway is just barely acceptable and actually shows a weaker correlation than for Sweden. All of this evidence argues for no significant difference between the countries. Thus, the thirteenth hypothesis still cannot be accepted. Table 6-13 Personnel Services and Objective Security, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Objective Security Personnel 'S"egve1ccte°sr Molrntgnitty 03:;- Pé'apg; 23:8;- 21:00:» 21:24- SWEde" (83541) (8 238) (82532) *Correlation significant at 0.0005 aflg_0urbin-Watson Statistic not_ significant at 0.05. ' 14. Economic Wealth and Objective Security Table 6-14 does not change the earlier assessment of the four- teenth hypothesis: Although the pattern of differences is mixed, there are still no acceptable relationships in the table. The dif- ferences between the total period and the post-1905 period in terms of the strength of correlation are also mixed, and the relationships 156 are all less autocorrelated for the post-1905 period. But only the relationships involving Infant Mortality are statistically signifi- cant, and there Norway's is substantially stronger than Sweden's. Furthermore, there is only one relationship, involving the Death Rate for Norway, that comes close to being acceptable. Thus, the fourteenth hypothesis still cannot be accepted. Table 6-14 Economic Wealth and Objective Security, Post-l905 Period. Correlation Coefficients and Durbin-Watson Statistics. Gross Objective Security National Product Infant Death Pupil Relief per Capita Mortality Rate Rate Rate -O.939 -0.734 -0.250 -O.657 N°rway (0 407) (1.182) (1.287) (0.943) -O.829 -O.743 0.445 -O.349 SWGde” (0.277) (0.367) (1.041) (0.403) B. Linear, Multiple Relationships 1. Procedure In order to attempt to remove the autocorrelation from the rela- tionships for the post-1905 period, the procedure previously outlined in Section B of Chapter Five was again employed to try to find the missing variable causing the systematic disturbance in each auto— correlated relationship. Again, the mobilization model guided the identification of independent-variable candidates for inclusion in such multiple regression equations. Then followed a comparison of the autocorrelated pattern of the residuals for the relationship 157 between each proposed independent variable and the particular dependent variable needing explanation. If their patterns were even roughly opposite, so that they could be expected to compensate for each other's errors in predicting the dependent variable, then a multiple regression equation including those variables was tried on the LS program. The equation was again deemed acceptable if the F-test for the proportion of variance explained in the dependent variable was significant at the 0.0005 level and the Durbin-Watson Statistic was not significant at the 0.05 level. The corresponding critical value of the Durbin-Watson Statistic with two independent variables was now 1.02, and the significance level of the proportion of variance explained was again determined directly from the LS program. 2. Results Despite numerous attempted equations, this procedure yielded only two acceptable equations for Norway and nine for Sweden. These equations are statistically summarized in Table 6-15, using the same format as in Table 5-15. The first acceptable relationship for Norway is only a slight improvement over the bivariate relationship between Left Vote and Enfranchisement, and shows the effects of high multi- collinearity: The standard error of estimate for the multiple equation is actually slightly higher than for the simple bivariate relationship between Left Vote and Enfranchisement. This can be ' attributed to the high correlation, 0.915, between the two independent variables in the multiple equation. This correlation is higher than 158 0000.0 000.0- 000.0- 0.0000 040.0- 000.0 0000000000 .00000 0000.0 040.0 000.0 0.004 000.0 000 0 004.0 0000000000 00000 0000.0 0.000 000.0 000.0- 000.0 000000 000.020 0000.0 000.0 000.0 0.0000 040 0 000.0 0000000000 00000 0000.0 040.0 000 0 0.004 000.0 000.0 004.0 0000000000 00000 0000.0 0.00_ 000.0 000.0 000.0 000000 0001020 0000.0 000.0 000.0 0.0000 000.0 000.0 0000000000 00000 0000.0 000.0 000.0 0.004 000.0 000.0 004.0 0000000000 00000 0000.0 0.000 000 0 400.0 440.0 000000 000 000 z 0 0 0 0 0 0000.0 040.0 000.0 0.000 400 0 000.0 0000000002 00000 0000.0 004.0 000.0 0.000 000 0 040.0 000.0 0000000000 00000 0000.0 0.40 000.0 000.0 000.0 0000000 400.0 040 0 000.0 000.0 000.0 000.0 0000 0000 000.0 040.0- 040 0- 000.0 000.0 000 0 004.0 0000000000 00000 0000.0 000.0 000.0 000.0 040.0 000000000000000 0 0 0 0 0 z 0_0>00 0020002 00000 000000 00000—0xm 00000 000» mocmo spam 32.90000 ULMvcmpm wo:w_..00> 12.90000 lmwpwpm -000000m 0000000 000 0000 00050m 000 000003 -0000000 000000000 1000000 00000000> 00000000 -00000000 . 0000000000 0000000 0000000000 000000 000 00000000 .000000 0000-0000 .000000000 0000000002 00000002 000000000< mFIm 00000 VII-V . . . . . . >00 0000 0 040 0 040 0 40 40 000 0 000.0 0000 000.0 000.0 000.0 00.00 040.0 040H0 000.0 000000000000000 0000.0 00.00 000.0 400 0 000 0 0000X0 0000000 0000.0 000.0 000.0 00.00 000 0 400.0 000000 000 020 000.0 000.0 004.0 00.00 400 0 040.0 000 0 000000000000000 0000.0 40.00 000.0 400.0 400.0 .0000x0 0000000 0000.0 000.0 000.0 00.00 000.0 400.0 000000 000 mwmm w 000.0 000.0 00H00 400 0 040.0 000.0 00000000M0M000w 00 0 000 0 004 0 000.0 .0000X0 0000003 0000.0 000.0 400 0 00.0 . . 0000.0 000.0 000.0 40.00 000.0 Mwm.w ”00.0 0000>00 0000.0 00.0 000.0 . 00.0 “00000000000000 ma 000 o 000 0 .000 xm 000000000 - 0000.0 000.0 000.0 00.00 . 00.00 000.0 400.0 .0 .11-000000000000000 . . 004 0 0000x0 0000000-amz 0mmw.m 000.0 000.0 44.04 040.0 0000.0 000 0 000.0 00.000 000.0 000.0 000.0 000>000 00000 1:: 00.00 000.0 000.0 WWW w 000%00500000000000 0 :w Mwmhmu 0wmmwwz 0000» 000000 0000000xm 000 0.0 xm 00000 - .C .000 .00 0 -0000M0000000 00000000 000.0 20> -0 000.0008 00000.00 . 000 0000 000000 000 - . 000 lewLLOU 0AHwL.wa_.—. COWHMZ -- 0000000000 0000000 160 that between either independent variable and the dependent variable. Furthermore, Total Population contributes practically nothing to the equation. In essence, Left Vote really only needs some benign variable to make the relationship a multiple one so that the critical value required for the Durbin-Watson Statistic can drop enough to make the relationship acceptable. The second equation for Norway in Table 6-l5 differs a great deal from the first. Here the two independent variables, Urban Population and Union Membership, are both highly related to the dependent variable, Revenue, but not so highly to each other. This yields a substantial rise in the multiple correlation coefficient and in the Durbin-Watson Statistic, and a substantial drop in the standard error of estimate. Furthermore, both independent variables are very important for the equation in terms of their contributions to the explanation of the dependent variable: The partial correla- tions and beta weights are roughly equal, and both variables are significant at the 0.0005 level. Union Membership is slightly more important in all of the statistics, but the difference is always slight. Thus, the equation is highly acceptable. The first three acceptable equations for Sweden involve Gross National Product as the dependent variable and Gross Domestic Investment as the primary independent variable, with some other indicator of social mobilization needed to remove the autocorrela— tion, reduce the standard error of estimate, and slightly boost the correlation. All three equations are very similar. First, the 16] multiple correlation is almost perfect, so that the proportion of variance explained in Gross National Product is almost 100 percent. Second, both independent variables are significant at 0.0005, although Investment always has the higher partial correlation and beta weight. Finally, all show high multicollinearity, although it is always less than the simple correlations between the independent variables and the dependent variable, while the standard error of estimate is always substantially reduced. Thus, all three equations can be considered highly acceptable. The remaining six acceptable equations for Sweden involve each of the categories of government expenditures except Health as the ‘ dependent variable, and an indicator of government penetration or l economic wealth as the primary independent variable, with Enfran- chisement needed only, in most cases, to remove the autocorrelation ! from the original relationship. The economic independent variable is always significant at 0.0005 and has a substantially higher partial correlation and beta weight than Enfranchisement, which is significant at 0.0005 only in two equations, Education Expenditures versus Revenue and Welfare Expenditures versus Civil Service. But a none of the equations show high multicollinearity, so that all of them may be considered acceptable. Furthermore, the three equations in which Enfranchisement is significant at less than 0.0l, the two previously mentioned plus the one involving Non-defense Expenditures versus Revenue, will be considered highly acceptable. These three also show the highest multiple correlations, the least amount of. 162 autocorrelation, and the lowest standard errors (except the one involving Non-defense Expenditures). 3. Implications for the Hypotheses These results for Norway and Sweden strengthen some of our earlier assessments of the hypotheses and alter others. The second hypothesis must now be rejected: Our earlier tentative acceptance of it was based on one acceptable bivariate relationship between economic wealth and social mobilization that was not substantially stronger for Norway. Table 6—l5 shows three highly acceptable multiple equations involving indicators of social mobilization and economic wealth for Sweden, but none for Norway. Thus, in retro- spect, the data for the total time period show no acceptable rela- tionships between social mobilization and economic wealth; whereas the data for the post-1905 period show only one acceptable bivariate relationship for Norway and three highly acceptable multivariate relationships for Sweden. Overall, the second hypothesis cannot be accepted: If anything, the relationship between social mobilization and economic wealth is stronger for Sweden than for Norway. The multiple—equation results do lend further support to the third hypothesis: The first equation for Norway in Table 6-l5 adds another acceptable relationship for post-l905 Norway between political mobilization and democratization, leaving Sweden with none. Although that multiple equation is not highly acceptable, it is consistent with the two acceptable bivariate relationships for post- l905 Norway shown earlier in Table 6-3. The equation lends no further 163 support to the fourth hypothesis because Total Population is useful only for removing the autocorrelation from the relationship between Left Vote and Enfranchisement. But the third hypothesis can be accepted for the post-l905 period. The multiple-equation results also further strengthen the sup- port for the sixth hypothesis, but further discredit the seventh: The second equation for Norway in Table 6-l5 adds a fifth acceptable relationship between social mobilization and government penetration to the four already uncovered for both the total and post-l905 periods; but that same equation also adds a second acceptable rela- tionship between political mobilization and government penetration for post-1905 Norway. That equation is highly acceptable, and both of its independent variables are highly significant. Thus, the sixth hypothesis can be accepted for both the total and post-1905 periods, but the seventh hypothesis must likewise be rejected for both periods. The multiple-equation results do alter the earlier assessment of the eighth hypothesis: Five of the last six acceptable multiple equations for Sweden mean that Sweden has as many acceptable rela— tionships between government penetration and government expenditures as does Norway. All of the evidence based on bivariate relationships for both the total and post-l905 periods indicated a stronger rela- tionship between penetration and expenditures for Norway than for Sweden. But all of the acceptable multiple equations are Swedish, in contradiction to the bivariate findings. The fact that the l64 Swedish indicators of penetration need help from Enfranchisement in removing the autocorrelation does not mean that they are any less capable of explaining expenditures than are the Norwegian indi- cators. In each of the equations the indicator of penetration is significant at 0.0005. Thus, the eighth hypothesis can now be accepted for the post-l905 period. - However, the multiple-equation results do not change the earlier assessment of the ninth hypothesis, although they do differ from the bivariate results. Previously, none of the relationships between democratization and government expenditures were acceptable because of autocorrelation. Now, Enfranchisement works in several multiple regression equations predicting expenditures; but all of those equations are Swedish, contrary to the hypothesis. In only two of those equations, Revenue versus Education Expenditures and Civil Service versus Welfare Expenditures, is Enfranchisement significant at 0.0005. But it is significant at less than 0.0l in one other equation, involving Revenue versus Non-defense Expenditures. These three equations were earlier deemed highly acceptable, meaning that there are three acceptable relationships for Sweden and none for Norway. Thus, the ninth hypothesis still cannot be accepted because democratization and expenditures is stronger for Sweden than for Norway. Finally, the multiple—equation results support the earlier rejection of the tenth hypothesis. One more acceptable post-l905 relationship between economic wealth and government expenditures is 165 added to those established in the bivariate results, and it is Swedish. Despite the fact that it is a multiple regression pre- dicting only Welfare Expenditures and requires the inclusion of Enfranchisement to remove the autocorrelation from the original relationship, Gross National Product per Capita is significant at 0.0005 in the equation. And although the equation is not one of the most highly acceptable ones, it is consistent with the earlier bivariate results for the post-l905 period. Thus, the tenth hypothesis can be rejected with more confidence because the results show that, if anything, the relationship between economic wealth and government expenditures is stronger for Sweden than for Norway. II. Acceptable Relationships Figure 6-l summarizes the acceptable relationships established for the post-1905 period on the basis of both bivariate and multi— variate linear patterns of correlation and autocorrelation. As in Figure 5-l, the major dependent-variable concepts of the mobiliza— tion model are listed above their indicators that can be explained by the listed independent variables. The first feature of this figure to be noted is that the two countries have almost exactly the same number of acceptable relationships: eighteen for Norway and seventeen for Sweden. The second interesting feature of this figure is that only four relationships are duplicated: Voter Turnout versus Total Population; Revenue versus Investment; and Total and Non-defense Expenditures versus Gross National Product per Capita. As with the total time 166 Figure 6-l. Acceptable Relationships, Post—l905 Period. N 0 R W A Y S W E D E N Dependent Independent Dependent Independent Variable Variables Variable Variables Voter Turnout GNP per Capita Enfranchisement Representation Representation Revenue Revenue Revenue Civil Service C1v1l Service Political Mobilization Total Population Voter Turnout Voter Turnout Voter Turnout Economic Wealth Agric. Employ. GNP per Capita GNP per Capita GNP per Capita Democratization Total Population, None Left Vote Left Vote Voter Turnout Government Penetration Investment Revenue Agric. Employ. Urban Pop., Union Membership Investment Union Membership Total Population Agric. Employ. Urban Population Investment, Total Pop. Investment, Urban Pop. Investment, Agric. Employ. Investment 167 Figure 6-l (continued). N 0 R W A Y S W E D E N Dependent Independent Dependent Independent Variable Variables Variable Variables .Government Expenditures Total Civil Service Total Enfranchisement, Civil Service Total GNP per Capita Total GNP per Capita Non-defense Revenue Non-defense Enfranchisement, Revenue Non-defense Non-defense Welfare Welfare Death Rate GNP per Capita Non-defense Civil Service Health Education Education Revenue Welfare Civil Service Welfare Objective Security Doctors per Capita None GNP per Capita Revenue Enfranchisement, Revenue GNP per Capita Enfranchisement, Revenue Enfranchisement, GNP per Capita 168 period, this indicates that there are significant differences between the two countries. The most noticeable difference is the presence of acceptable relationships under democratization and objective security for Norway but not for Sweden. Norway also has many more acceptable relationships under government penetration than does Sweden. .On the other hand, Sweden has more entries under political mobilization, economic wealth, and government expendi— tures than does Norway. Another difference is that, in addition to the Death Rate and Doctors per Capita, the variables Representation, Left Vote, and Union Membership appear in the Norway column but not in the Sweden column. Similarly, in addition to the Death Rate, Enfranchisement, and Representation, the variable Civil Service appears as a depen— dent variable for Norway but not for Sweden; whereas Health and Education Expenditures appear as dependent variables for Sweden but not for Norway. Enfranchisement helps explain expenditures for Sweden but not for Norway, which is associated with the fact that nine of Sweden's fifteen acceptable relationships are multivariate, compared to only two of eighteen for Norway. But again, as with the total time period, perhaps the most significant feature of the acceptable relationships for the post-l905 period is the preponderance of economic, monetary measures over social and political measures. Again, the Price Index is the only economic measure not involved in an acceptable relationship. Only eight of the 33 relationships do not involve a monetary measure or 169 Civil Service, which has already been conceptualized as part of the same economic syndrome. In addition, only nine of the 22 bivariate relationships involve a non-economic measure at all, although all of the multivariate relationships involve at least one nonreconomic measure. Furthermore, some of the social measures appearing in the acceptable relationships may be considered conceptually closer to the economic syndrome, particularly Agricultural Employment and Urban Population. This feature strengthens the interpretation that most of the acceptable relationships either involve monetary and related measures that are more reliable than social and political measures, or which indicate an economic milieu of phenomena that is more cohesive than socio-political phenomena. Further interpretation of these results will follow in the next chapter. Another possibility is that social and political phenomena are more often related in a curvilinear fashion. However, a few trans- formations were tried with the LS program on variables whose relationships had residuals that were obviously curvilinear, without acceptable results. An example involves the relationships between Urban Population and Gross National Product, Union Membership, and Revenue for Norway's total time period. Each of these relationships showed residuals that were initially positive, decreased and went negative fairly monotonically, then increased and went positive again fairly monotonically. An exponential transformation was tried for each relationship on the independent variable, Urban Population, and in every case the correlation coefficients and the Durban-Watson 169 Civil Service, which has already been conceptualized as part of the same economic syndrome. In addition, only nine of the 22 bivariate relationships involve a non-economic measure at all, although all of the multivariate relationships involve at least one nonreconomic measure. Furthermore, some of the social measures appearing in the acceptable relationships may be considered conceptually closer to the economic syndrome, particularly Agricultural Employment and Urban Population. This feature strengthens the interpretation that most of the acceptable relationships either involve monetary and related measures that are more reliable than social and political measures, or which indicate an economic milieu of phenomena that is more cohesive than socio-political phenomena. Further interpretation of these results will follow in the next chapter. Another possibility is that social and political phenomena are more often related in a curvilinear fashion. However, a few trans- formations were tried with the LS program on variables whose relationships had residuals that were obviously curvilinear, without acceptable results. An example involves the relationships between Urban Population and Gross National Product, Union Membership, and < Revenue for Norway's total time period. Each of these relationships showed residuals that were initially positive, decreased and went negative fairly monotonically, then increased and went positive again fairly monotonically. An exponential transformation was tried for each relationship on the independent variable, Urban Population, and in every case the correlation coefficients and the Durban-Watson 170 Statistics were both improved only by about five one-hundredths, certainly not enough to make the relationships significant. This implies that the non-acceptable relationships in the mobilization model may be intrinsically non-linear in the regres- sion parameters as well as in the variables themselves (Kmenta l97l: 45l-472). If this is the case, then the best procedure would be to break the total time period down into even smaller portions in order to find the precise points in time at which the parameters change. But the intervals in our data are fairly large, so that analyzing smaller time periods would mean working with an unreliably small total number of observations. Such analyses could be performed on data for every year in the total time period; but such data are beyond the scope of this dissertation. In addition, determining whether the low Durbin-Watson values are due to autocorrelation caused by unmeasured variables or to intrinsic non-linearity would be difficult, because most of our variables increase monotonically over time. Specific transformations designed simply to remove the autocorrelation bodily from the data might also prove useful (Kmenta l97l: 287-292); but they would not be very satisfying theoretically. More will be said about this in the next chapter. III. Summary of the Analysis The hypotheses stated in Chapter Four were tested by comparing acceptable relationships for Norway and Sweden between the sets off indicators of development stated in each hypothesis. This required the computerized use of simple and multiple correlation and regression 171 analysis. A relationship was deemed acceptable if its simple or multiple correlation coefficient was significant at 0.0005, indi- cating a strong relationship, and if its Durbin-Watson Statistic was not sifnificant at 0.05, indicating a lack of significant autocor- relation in the relationship. This analysis was performed for both the total time period covered by the data and the period following the disruption of the union between the two countries in 1905. It uncovered too few acceptable relationships common to both countries to permit direct comparison of the strength of the relationships between the two countries; therefore, comparison was based upon the number of significant relationships for each country. The results of this analysis indicate that only the sixth hypothesis can be accepted for both the total time period and the post-1905 period; that is, the indicators of social mobilization do seem to be more strongly associated with the indicators of government penetration for Norway than for Sweden. Furthermore, only the third and eighth hypotheses can be accepted for the post-l905 period only; that is, the indicators of political mobilization do seem to be more strongly associated with the indicators of democratization for Norway than for Sweden; and there does seem to be no difference between Norway and Sweden in the strength of association between the indicators of government penetration and government expenditures. None of the hypotheses can be considered acceptable only for the total time period. All of the other hypotheses must be rejected for both the total time period and the post-l905 period, at least 172 on the basis of this analysis of this data. The relationships for those hypotheses were either too severely and consistently auto- correlated to permit acceptance of the hypotheses; or they showed no difference between the two countries where they should have favored Norway; or they favored Sweden instead. The next chapter will attempt to interpret these results by examining the methodological problems involved in the analysis and by integrating them with Peters' assessment of the mobilization model for Sweden, Lafferty's comparison of Norway and Sweden, and the descriptive histories of the two countries. Finally, it will attempt to derive some theoretical meaning from the acceptable rela- tionships established in the analysis. CHAPTER SEVEN INTERPRETATION The previous two chapters presented the results of simple and multiple correlation and regression analysis designed to test specific hypotheses concerning the differences and similarities in relationships among time-series indicators of development for Norway and Sweden. The hypotheses were derived from a previous analysis of such differences (Lafferty l97l) and from the developmental histories of the two countries, all organized within the framework of the mobilization model of political development (Peters l970). In the analysis, relationships among the indicators were deemed acceptable on the basis of significant correlation and significant lack of autocorrelation. The analysis uncovered only thirteen acceptable relationships for the total time period, 1875 to l965, and only 35 for the period following the disruption of the union between the two countries, l905 to l965. These few acceptable relationships con- firmed only three of the original fourteen hypotheses. I. Explaining the Failure of the Hypotheses Confronted with these results, the problem of finding meaningful explanations and interpretations now arises. First of all, why were so few of the hypotheses confirmed? One obvious possibility is that 173 174 the previous descriptive accounts and analyses from which our hypotheses were derived bore little resemblance to the actual historical events and processes under consideration. This prOpo- sition is not entirely unreasonable or implausible: Historians frequently describe history without the benefit of "hard" empirical data, such as official national statistics, or at least without thorough quantitative analysis of such data. This is obviously due to their concern with phenomena which often cannot be measured quantitatively. But given our concern with quantifiable phenomena, and assuming that such data are reliable and valid, descriptive accounts of such phenomena could simply have failed to describe accurately the actual history involved. This would mean that quantitative analysis could not be expected to confirm hypotheses based on such accounts. However, in our case rejecting the historical accounts as being inaccurate proves unfounded. In the first place, only certain aspects of the historical accounts can be questioned. Most of our hypotheses were based on the expected timing or coincidence of the movement of our variables over time, derived not only from purely descriptive historical accounts of the events and processes on which those data were based, but also from the recent quantitative analysis by Lafferty (l97l). Specifically, the hypotheses concerning the relationships of social mobilization versus economic wealth and political mobilization were based on Lafferty's findings. But the hypotheses concerning the relationships among those concepts and 175 democratization, government penetration and expenditures were derived by comparing the movement of the mobilization and wealth variables, as reported by Lafferty, with the occurrence of electoral and legis- lative reforms and social legislation, as reported by historical accounts. Information on the various relationships among expendi- tures and their societal impacts was almost completely lacking; thus no difference between the two countries in those relationships was hypothesized by default, so that the lack of confirmation there is not really surprising. But the point is that one cannot seriously question the descriptive accounts of the timing or coincidence of the movement over time of the phenomena under consideration. That is, most historians agree on whgn_the events occurred. One can, of course, question their interpretations of why_those events occurred when they did. For example, as reported in Chapter Four, Lafferty's analysis demonstrated that Norway's more radical political labor movement was ngt_due to any direct social dislocation caused by massive shifts of workers from agriculture to industry, as had been assumed by earlier historians and subsequent labor economists and sociologists. But again, our hypotheses were derived not from historical interpretations but from the actual timing of events. So why were some of the hypotheses not borne out by quantitative analysis of empirical data relating to those events? Some sort of discrepancies must have existed between our data or methods of analysis and the timing of events reported by previous historical descriptions 176 and analyses. Probably the best place to look for such discrepancies is in the single most important source of our hypotheses, namely the work of Lafferty. Furthermore, it might also prove useful to com- pare our results with earlier analyses by Peters (1970; l972) of similar data on Sweden that provided his strongest confirmation of the mobilization model of political development, the conceptual framework within which our hypotheses were organized. A. Comparison with Previous Works l. Peters (l970; 1972) Our data base for Sweden was almost identical to that used by Peters (l970) in his three-way comparison of that country, the United Kingdom, and France. The variables used in both his study and ours were for the most part very similar in operationalization. But the observations or time—points selected were quite different. Whereas Peters used every fifth year beginning with l865, yielding 2lobserva— tions, we used the election years, every third and then fourth year, beginning with l875, yielding 27 observations. In addition, he split his total period into two halves at the l9lO point, leaving only ten and eleven observations in each half, compared with l7 in our post-l905 period. As expected with a larger number of more frequent observa- tions, our relationships generally showed slightly lower levels of correlation and significantly higher levels of autocorrelation. Of the 28 common relationships, eleven were significantly different for the total period, all but two of which showed significantly different correlations as well as autocorrelation. We each had one acceptable 177 relationship, but not the same one. Comparing his post-l9lO period and our post-1905 period, 20 of the 28 common relationships were significantly different, of which only six showed significantly different correlations as well as autocorrelation. Here only two of the different relationships were acceptable to us, whereas l6 were acceptable to him. Interestingly, the relationships that showed significantly dif- ferent correlations as well as autocorrelation were between the indicators of political mobilization and government penetration, and between democratization and penetration.' Furthermore, in the later half-period he accepted several of these relationships while we accepted none. Large differences in autocorrelation are understand- able given our more frequent observations. But why should almost identical data have yielded significantly different correlations for just two specific sets of relationships, just because more frequent observations were used? Why were the differences not on the same order of magnitude throughout the entire data base? There are no satisfactory answers to these questions, except for two possibilities involving differential sampling of time-points. First, the observa— tions selected for the penetration indicators, which were common to both sets of relationships, may have just happened to yield patterns that correlated differently with the indicators of political mobili- zation and democratization. Second, there may have been a lag effect in the relationships between those concepts. That is, the political variables were measured in election years in both data bases. But 178 the penetration indicators were measured for the election years also in our data base, but were measured for usually subsequent years in Peters' data base. Comparison of the results indicates that this lag in his data may have produced somewhat stronger correlations. But why did such a lag effect not emerge in other relationships as well? Except for those two sets of relationships, the correlations were essentially similar. Of course, the extent of autocorrelation in Peters' data base was sufficiently lower to warrant confirming every link in the mobilization model of political development for Sweden, at least by our criteria. Nevertheless, the relationships were usually similar. Besides confirming the link between social mobilization and political mobilization for the post-l9lO period, he also verified the weaker effect of political mobilization on government penetration by partialling out the effects of social mobilization on penetration. He also confirmed the link between democratization and government expenditures only in multiple regression equations with economic wealth, although his indicators of democratization were different. Finally, he also failed to confirm most of the relationships common_ to both data bases on the output side of the mobilization model, namely the effects of government expenditures on personnel services and objective security. Here the lower level of autocorrelation in his data permitted him to confirm only the relationships between health expenditures and the indicators of objective security in the health sector, infant mortality and the death rate, for the post-l9lO time period. 179 In a later work (l972), Peters contributed to the current debate originating in the area of "comparative American state politics“ referred to in our first chapter, over whether political variables or socio-economic variables explain more of the variance in public policy variables. Using one socio-economic variable, Gross National Product per Capita, with two political variables, Civil Servants per Capita and a subjectively-coded Index of Democratization, in separate multiple regressions on sectoral Government Expenditures per Capita, he compared beta weights and found that both the socio-economic and political variables were important in explaining public policy. Specifically, for the total period of l850 to 1965 in France and Britain, the inclusion of both the socio-economic variable and a political variable in the regressions was necessary to remove auto- correlation. And from examination of the residual patterns for these regressions he determined that the influence of the socio-economic variable was declining over time. Sweden presented a special case because the autocorrelation could not be eliminated for the total period of l865 to l965, so it was split in half at the 1910 time- point. Again, the political variable, Civil Service, had stronger effects in the second period than in the first, and was consistently stronger than the socio-economic variable in the second period, although they were roughly equal in the first. These results cannot be compared directly with ours because our indicator of democratization was different and because Civil Service appeared in only one acceptable relationship with expenditures for 180 Sweden. Also, in our results acceptable relationships for the post-l905 period were usually formedby multiple regressions of the two "political" variables on expenditures. That is, Civil Service and Revenue were indicators of government penetration in the original mobilization model. In his article (1972), Peters formed multiple regressions of government penetration (Civil Service) and economic wealth (GNP) on expenditures, whereas our acceptable multiple regres— sions involved government penetration (Revenue) and democratization (usually Enfranchisement) on expenditures. Notice that he disregards the likely possibility of multicollinearity in those regressions, a problem he himself warns against in the introductory portion of the article. But more importantly, our results for Sweden indicated that the socio-economic variable (GNP) was more substantially associated with expenditures than was Peters‘ political variable (Civil Service), and that these results were reversed for Norway, with that political variable being more substantially associated with expenditures than was the socio-economic variable. Nevertheless, these discrepancies are the product of his use in this later article of a different con— ceptual framework for organizing the relationships, namely the rela- tive impacts of political and socio-economic variables on public policy, rather than the mobilization model of political development. 2. Lafferty (1971) Most of the hypotheses of our study were derived by superimpos- ing descriptive and historical writings about Norway and Sweden upon Lafferty‘s (l97l) empirical comparisons and interpretations concerning 181 those two countries, then organizing them conceptually according to the mobilization model of political development (Peters l970). Lafferty's analysis implicitly focussed on the mobilization-model concepts of economic wealth, social mobilization, political mobili- zation, and democratization. However, his methods of analysis did not always involve systematic, quantitative testing of the relation- ships among these concepts, but often emphasized the timing of events in each of those areas. We combined these specific depictions with the descriptive and historical accounts of those events and of government activity to produce specific hypotheses. In retrospect, the problem with this procedure may have involved the brevity of the time-span covered by Lafferty‘s analysis in com— parison with ours. That is, most of the developments he discussed occurred in the period from about l890 to about l930, whereas our analysis covered the entire span from 1875 to l965 and the shorter period from l905 to 1965. Whenever Lafferty did attempt quantitative analysis, his data base contained a dangerously low number of obser- vations, usually decennial or quintennial averages over that brief forty-year period. This, plus differences in variables and operational- izations and in the specific relationships tested, made it impossible to compare correlations directly. But even without these differences, such a vast difference in the time-span covered in the data base might have produced vastly different results. That is, testing hypotheses about the similarities and differences between Norway and Sweden over a long period, based on information concerning the middle 182 portion of that period, required assuming that the early and late portions of the period would not significantly affect the results. Now that the data are at hand, how accurate was thisassumption? We should first point out that it involved two competing dimensions: first, that the relationships would be similar for the two countries in the early and late portions of the total time period; and second, that the relationships would remain stable throughout the entire time period for both countries. The first of these dimensions constituted the initial assumption made prior to the analysis. The second dimen- sion remained tacit until after the analysis. In lieu of correlation analysis of the shorter periods, which would involve too few observa- tions for reliable estimates, the best way to demonstrate this involves examining the residuals of the relationships, that is,the errors in predicting the dependent variables from the independent variables specified in the regressions. These reveal the trends in the data over time and permit detection of shifts in the relationships. Figure 7-l presents a summary of an examination of the residuals in the relationships among four of the six indicators common to our analysis and Lafferty's: Investment, Union Membership, Left Vote, and Representation. The latter is what Lafferty called the Voter Mandate Ratio, the leftist parties' proportion of legislative seats divided by their proportion of the total vote. He also used the proportion of the labor force employed in agriculture, but that variable behaved very similar to Investment in the residual patterns. Both of these variables were indicators of our mobilization-model concept of social 183 Union Membership versus Investment Left Vote versus Investment Norway Sweden Norway Sweden Years Pattern Years Pattern Years Pattern Years Pattern l906- 1905- l906- l905- l9l8 Linear l9ll Random l918 Linear l9l7 Linear l9l8- l9ll— l9l8- l9l7- l953 Random l948 Linear l945 Random l94O Random l953- l948- l945- l940- l965 Linear l964 Linear l965 Linear l964 Linear _,_. Representation versus Union Memb. Representation versus Left Vote Norway Sweden Norway Sweden Years Pattern Years Pattern Years Pattern Years Pattern l906- l905- l906- l905- l933 Random l9l4 Linear l965 Random l9l4 Linear 1914- ' 1914- 1936 Random l928 Linear l933- l936- l928- l965 Linear 1964 Linear l964 Random Representation versus Investment Norway Sweden Years Pattern Years Pattern l906- l905- l930 Random l9l4 Linear l9l4- 1936 Random l930- l936- l965 Linear 1964 Linear Figure 7-l. Patterns of Residuals for Selected Relationships, Post- 1905 Period Only. 184 mobilization. Lafferty also used Gross National Product, which we conceptualized separately from social mobilization, but which he included along with Investment and Agricultural Employment under the general rubric of economic development. Finally, Figure 7-l ignores the period before l905 because of its lack of significant variation in the variables, reflecting the fact that development did not really begin until the early l900's. Figure 7-l shows that for the most part the relationships were similar in Norway and Sweden for the post-War period, and that most of the differences between the two countries did occur in the period covered by Lafferty's analysis, although the period of differences did extend until after the War for some relationships. This suggests that our initial assumption was reasonably well justified, so that our analysis could proceed to cover a longer period of time in order to include more observations while still focussing on the crucial period of difference between the two countries. However, the shifts in the nature of the relationships over time reflect the severe auto- correlation in the data over such a long period of time, which of course could not be detected until the analysis had been performed. Also notice that these shifts occur at widely varying points in time for different relationships, so that it is really inaccurate to talk of distinct periods for all of the relationships in the first place. Such wide variation in temporal shift-points seems to apply to the other variables in our data base as well. The shifting nature of the relationships over time provides a possible explanation for the failure of some of our hypotheses: Those 185 expected relationships simply did not hold up over the entire period from l905 to l965, although others did. Detailed analysis of more frequent data points might show that the variables moved coterminously during some portions of that period, but independently in others. Of course, the residuals indicate that different relationships would be confirmed in different portions of the period. And such analysis must await the collection of more frequent data-points. Finally, the shifting in the nature of the relationships over time requires his- torical and methodological interpretations. B. Methodological and Historical Interpretations The major reason why so few acceptable relationships emerged from our data, and thus why many of our hypotheses could not be accepted, was high autocorrelation, which is almost unavoidable in time-series data. Again, autocorrelation arises when the residual errors in predicting a dependent variable from one or more independent variables are not independent across observations, but rather reflect a systematic disturbance lingering over several observations. In time-series data such a disturbance may be due to a missing variable that should be included in the regression equation; non—linearity in the relationships among the variables included; or non-linearity in the regression parameters over time, indicating different successive equations for that relationship. The extent of autocorrelation increases as the number of observations taken from a given time-span increases, that is, as the observations become more frequent. This was obviously the reason for the high level of autocorrelation in our 185 _ \ data; but the gatur§_of that exaggerated autocorrelation has yet to be determined. Some of the most likely sources of systematic disturbances capable of producing autocorrelation are episodic events that can cause systematic depression or inflation of estimates of a dependent variable by one or more independent variables, similar to the “missing variable“ cause of autocorrelation. Such events must alter the rela- tionship between two or more variables and must have effects strong enough to last over several time—points in order to produce significant autocorrelation. Although the entire examination of the residual pat- terns in our data need not be presented here, some of the results are interesting. One might expect political variables to be most affected by historical events, and the one variable that seemed to behave most differently in comparing Norway and Sweden was Voter Turnout, the proportion of eligible voters actually voting. Sweden's Turnout steadily increased in an oscillating fashion throughout most of the time period, whereas Norway's reached its highest peaks in the mid- l880's and again in the mid l890's, then dropped precipitously at l900, then climbed steadily in an oscillating fashion thereafter. (See Figure'A-Z in the Appendix.) This (early inflation of Norway's Turnout can easily be explained by the occurrence of the constitutional crises concerning first the parliamentary principle and then universal manhood suffrage, whereas Sweden's electoral reforms came later. Norway's drop in Turnout at l900 was due to the poor showing among newly enfranchised voters. This historical difference is significant 187 enough to have affected most of the total-period relationships involving Turnout, causing significant autocorrelation when that early period was included in the analysis. But not all political variables showed such historical differ- ences, and non-political variables did not escape the effects of such differences. Two variables which displayed essentially similar residual patterns in their relationships with other variables in Norway and Sweden were Union Membership and Left Voting. This sup- ports earlier suggestions that the political labor movements progressed similarly in the two countries; but it does not mean, of course, that the two variables formed similar relationships within each country, as was evident by their appearance in different acceptable relationships in the original analysis. Likewise, Total Population, Agricultural Employment, and Infant Mortality all tended to behave the same in both Norway and Sweden, although not necessarily the same as each other; whereas Urban Population, Gross National Product, Government Expenditures, and all “impact" measures except Infant Mortality tended to behave differently across the two countries. The differences for many of these variables are puzzling because their univariate plots, some of which can be seen in the Appendix, appear to be very similar when the two countries are compared. This just illustrates the fact that the interaction among variables is exceed- ingly complex. On the other hand, the way in which many of these variables com- bined to produce autocorrelated relationships is very understandable. 188 All of the variables measured in per capita monetary terms, namely Investment, GNP, Revenue, and Expenditures, were essentially exponential functions over the entire time period. (See Figures A-5 and A-6 in the Appendix.) When correlated with each other, accept- able relationships often resulted. However, when correlated with variables having different shapes over time, severely autocorrelated relationships usually resulted. The same was true of the political mobilization and democratization measures, which were generally logistic and logarithmic curves over the entire time period, approach- ing a ceiling toward the end because they were measured in propor- tional terms. (See Figures A-l and A-2 in the Appendix.) Finally, the three basic measures of social mobilization, namely Total Popula— tion, Agricultural Employment, and Urban Population, tended to be smoothly increasing or decreasing functions over the entire time period. (See Figure A-3 in the Appendix.) However, none of the measures in any of these groups were identical, and the patterns of interaction among them were still complex. In addition to the alteration of the nature of the variables and their interactions by historical events.and processes, other possible explanations exist for the high levels of autocorrelation in our data. The most theoretically appealing possibility concerns "missing variables" not included in our analysis. Because autocorrelation in a relationship between two or more variables represents some sort of systematic disturbance lasting over several time-points, such a dis- turbance might be the result of a variable not included in that 189 relationship. That is, the exogenous variable might interact with the endogenous variables in such a way that the estimates of the dependent variable by the independent variables are systematically distorted, as reflected in the residual pattern. The best method of eliminating such autocorrelation is to identify the variable responsible for it by comparing the residual pattern of the rela- tionship with the univariate distributions of suspect variables, then including each suspected variable or combination of variables in the relationship. Usually the selection of such variables is guided by theory as well as methodological convenience, and care must be exercised to avoid the added problem of multicollinearity, or high correlation among the independent variables in a relation- ship. This is usually a severe problem only when the correlations among the independent variables are higher than the correlations between each independent variable and the dependent variable. That is, only then will the standard error of estimate for the relation- ship actually increase with the addition of new variables. But even more fundamentally, such exogenous variables must be included in one's original data base in order to be even considered as candidates for inclusion in an autocorrelated relationship. If the "missing variable" has been so included and can be identified as the culprit causing the autocorrelation, then the autocorrelation may be eliminated. However, if the "missing variable" has not been included in the data base and thus cannot be identified as the dis- turbance, then the autocorrelation may not be eliminated. In our 190 case numerous multiple relationships were attempted in an effort to reduce the autocorrelation on the basis of residual patterns, and many such relationships succeeded, as reported in the previous chapters. Nevertheless, many relationships remained plagued by significant auto- correlation, possibly caused by exogenous variables not included in the original data base. Such variables may or may not be measurable at this level of analysis. For example, one theoretically important set of variables not included in our analysis that still might be measured is tech— nology, which might explain much of a government's ability to provide public services. As reported earlier, attempts were made to gather aggregate data on the number of patents issued, but they were not available for Norway before about l950. Similar efforts to obtain complete and consistent data on energy consumption bogged down in a miasma of changing definitions of imports and exports of coal, oil, wood, and so forth. Examples of important missing variables which might not be susceptible to measurement at this level of analysis are cultural values and social, economic, and political attitudes that might affect public acceptance of collective rather than individual efforts at social amelioration. The inclusion of such variables as technology and values must await the development of better measurement techniques. Another possible explanation of the high level of autocorrelation in our data is that'the variables are not related to each other in a linear fashion. The method of analysis employed here involved 191 single-period linear correlation and regression; but Peters' (l970) analysis of successive time periods for Britain, France, and Sweden suggested that there is a curvilinear relationship between social mobilization and political mobilization. Although he did not clearly specify whether that curvilinearity resides in the variables or in the regression parameters, it is certainly possible that many of the relationships in our data are curvilinear in the variables themselves. That is, some variables may be expressed as logarithmic, exponential, logistic, or other, more complex mathematical functions of other variables, rather than as straight-line functions. Such relationships may be theoretically expected in such areas as social mobilization versus political mobilization, where increases in the smaller and larger values of social mobilization Variables may not have an effect on political mobilization, but where increases in the middle range of values will have an effect. Plausible theoretical reasons why curvilinear relationships should occur in other areas might also be conjured. Then the variables may be transformed using appropriate mathematical functions in the attempt to specify curvilinear relation- ships. However, such transformations are often not powerful enough to straighten out the residual pattern and remove the autocorrelation completely, as indicated by our efforts in the previous chapter. But the relationships may be curvilinear in the regression parameters rather than in the variables, in that the variables them- selves are still related in a linear fashion, but at different "angles" over time. That is, the regression coefficients, giving the 192 slope of the straight-line prediction from the independent variable to the dependent variable, may be different for different portions of the time period. This was suggested by our earlier discussion of the variation in temporal shift-points for certain relationships during the l930's and l940's. For a logistic-style residual pattern, the coefficient may be small during the early and late portions of the period, but large during the middle portions. Such a pattern could be found by breaking the total time period into smaller portions and testing the relationships in each portion to detect significant changes in the regression parameters. This may yield a better description of the relationship between social mobilization and political mobilization, for example, than a simple logistic-function linear transformation of the variables. But when the variables tend to increase monotonically over time, as most developmental variables do, then it is difficult to tell whether the curvilinearity is in the variables themselves or due to changes in the parameters of the rela— tionship over time (Kmenta l97l: 469—472). Regardless of the cause of the high level of autocorrelation in our data, it could be removed by one of a set of specific transforma— tions of the data known as "de-trending." Such transformations involve adjusting the value of the selected variable for each observa- tion according to the amount of autocorrelation between the residual for that observation and the residual for the previous observation (Kmenta l97l: 287-292). There is nothing unethical about this method, because it merely controls for autocorrelation as though it were a 193 contaminating variable and operates essentially like a regular linear transformation. Nevertheless, in our opinion this method lacks theoretical justification because one does not know the nature of the contaminating variable being controlled. Other methods of removing autocorrelation should therefore be attempted before resorting to de—trending the data. 1 Finally, questions about the quality of measurement at the aggregate level of analysis were raised in the first chapter. One such argument involves simple reliability and validity, that aggre- gate measures often do not really represent the intended concepts, or do so in an inconsistent or unstable way. For example, Michael Drake (1972) has argued that even though Norway is quantitatively an urbanized country (if small towns are included in the definition of "urban"), most Norwegians retain strong ties to the rural subculture through relatives and recreation, often maintaining "second homes“ in the countryside. This suggests that the aggregate measures of urbanization used in our analysis may not have been equivalent for Norway and Sweden. However, Figure A-3 in the Appendix shows that Sweden's proportion of the population living in cities larger than 4 20,000 continued to grow after 1903, whereas Norway's leveled off until after the Second World War. Today, Sweden has roughly 50 per cent of her population living in such cities, compared to Norway's 25 per cent. Thus, Drake was probably referring to official figures which often include small towns under 20,000 in size in the definition of "urban" areas. Besides, rural ties and orientations may be just as strong in Sweden. 194 The fact that most of the acceptable relationships in our analysis were economic in nature and were measured in monetary units suggests that our social and political measures may have harbored more error variance and may thus have been less reliable measures. However, our sectoral expenditures were plagued by error variance due to changes in budgetary classification over time. Also, the larger number of acceptable relationships among the economic variables may also have been due to the aforementioned similarity in shape of their distributions as measured in per capita monetary units. Or, as sug- gested in the previous chapter, economic variables may inherently be more clearly and simply related to each other than are social and political variables, in that social and political phenomena may inherently be more complex than economic phenomena. Another argument is the notion that aggregate measures are too abstract and nebulous to be used as indicators of the concepts under examination, or that the system-level concepts themselves are too abstract and nebulous to describe the social, political, and economic phenomena of interest. Although the first chapter tried to lay these objections to rest, the possibility of their truth still remains in certain respects. The argument just cited, for example, is very similar to the “missing variable" cause of autocorrelation, in that aggregate variables may not interrelate in an acceptable fashion because there are missing links or intervening steps in between the processes they indicate, and those missing variables may not be susceptible to aggregate-level measurement. For example, why should 195 we expect democratization to be related to government penetration without considering the underlying changes in attitude expected of government bureaucrats? Similarly, why should we expect health expenditures to be related to infant mortality without considering whether those funds are channeled into pre-natal and post-natal care? Obviously any definitive explanation of a complex process in any science requires corroborating evidence at several levels of analysis. Nevertheless, if the components of a complex process are in fact related, that relationship will show up even at the highest level of abstraction'ifthe techniques of analysis are appropriate and sufficiently sensitive. Our future task is to make these appropriate techniques of analysis more sensitive. In summary, the failure of some of our hypotheses was due to severe autocorrelation in the relationships, which resulted from curvilinearity in the nature of the relationships over time, missing variables which distorted the relationships, and the tendency of variables with different shapes over time to associate with each other in an autocorrelated manner. II. A Substantive Interpretation Even though the preceding discussion may have provided some explanations for the failure of the original hypotheses, we are still faced with the task of deriving some meaning from the acceptable relationships that did emerge from the data. To do this we must first extricate ourselves from the framework of the original hypotheses and re-examine the results from a different perspective. Probably the 196 most useful strategy is to formulate separate mobilization models of development for Norway and Sweden based on the relationships deemed acceptable in our original analysis, and then compare the two models in terms of the presence and absence of specific links between the general concepts of the mobilization model, and the comparative strength of those links for the two countries. This re-examination of the results will not include the two mobilization-model concepts for the societal impacts of government expenditures. No link was confirmed between expenditures and impacts, and only one acceptable relationship emerged in that entire area: Doctors per Capita versus the Death Rate for post-1905 Norway. Besides, the correlation for that relationship was barely significant, so that it was not worth including here. The re-examination of the results will also focus only on the post-1905 time period, because in both countries development did not really begin until about the turn of the century. This descriptive historical generalization is borne out by the lack of variation in most of the variables in our data base before about 1900, which resulted in a dearth of acceptable relation- ships for the total time period. (See the Appendix.) Also note from Figure 6-1 in the previous chapter that the two aspects of the response of Norwegian political elites to the demands of political mobilization, namely penetration and democratization, were associated with different indicators of political mobilization. Penetration was more substantially associated with Union Membership, whereas democratization was more substantially associated with Left 197 Vote and Voter Turnout. As discussed in Chapter Four, although Lafferty (1971) conceptualized Union Membership as being part of the response of the political labor movement to economic development, and although Peters (1970) conceptualized it as being an indicator of political mobilization, it could conceivably be conceptualized as another indicator of social mobilization, or perhaps as standing mid- way between the two concepts of mobilization. Certainly the size of the labor union movement can be considered another social aspect of the industrialization process, rather than a truly political response to that process. This re-conceptualization of Union Membership as an indicator of social mobilization rather than political mobilization is also supported by the more substantial association of social mobiliza- tion with government penetration for Norway than for Sweden. That is, if the only indicator of political mobilization associated with government penetration is Union Membership in Norway, and if social mobilization is also more substantially associated with government penetration in Norway, then it is reasonable to suspect that Union Membership may be an indicator of social mobilization rather than of political mobilization. Thus, although the relationships of union membership with the other indicators of social mobilization were significantly autocorrelated for both Norway and Sweden, the results support the re-conceptualization of Union Membership as an indicator of social mobilization rather than ofpolitical mobilization. Figure 7-2 presents the links among the concepts of the mobili- zation models of political development confirmed by acceptable NORWAY SWEDEN Figure 7-2. LINKS COMMON, TO BOTH COUNTRIES LINKS PRESENT IN ONLY ONE COUNTRY Figure 7-3. 198 Democratization Political/ ' Mobilization A Government . Government Penetration "T Expenditures Social Mobil 12M Economic Wealth Democratization Political Mobilization Government , Government 1,00””””* Penetration ‘ TWExpenditures Social ,z”””””’* MobilizaM, Economic Wealth Confirmed Models of Development for Norway and Sweden, Based on Acceptable Relationships Among Concept Indicators, Post-1905 Period Only. -Democratization Political Mobilization S N Government ' > Government ‘1000”””’/* Penetration Expenditures Social ,0””’T’T”’T* Mobilizatidn“T§‘*--0_0. 5 Economic Wealth N Democratization Political Mobilization Government Government Penetration Expenditures Social Mobilization . ‘ Economic Wealth Common and Different Conceptual Links in Confirmed Models of Development for Norway and SWeden, with the Country for Which the Relationship Was More Substantial Indicated, Post-l905 Period Only. 199 relationships among the indicators of those concepts, for Norway and Sweden in the post-1905 period. Again, bivariate relationships were deemed acceptable on the basis of significant correlation and lack of significant autocorrelation. For multiple regresSion equations, acceptable relationships here involved only those inde- pendent variables whose beta weights were statistically significant in the majority of acceptable equations for the pair of concepts under consideration. That is, if the variable was usually brought into the equations only to eliminate autocorrelation and did not con- tribute significantly to the proportion of variance explained in the dependent variables, then that conceptual link was not considered here to be confirmed. The specific indicators involved in the con- firmed conceptual links can be obtained from Figure 6-1 and Table 6-15 in the previous chapter. The next step is to examine the similarities and differences between these models of development for Norway and Sweden. Figure 7-3 presents the conceptual links confirmed by acceptable relationships that were common to the models for both countries, and the links that were present in only one country. Each of the links in Figure 7-3 is labelled with the first letter of the name of the country for which the relationships in that link were more substantial. That is, the links labelled "N" had more numerous acceptable relationships for Norway than for Sweden. Links labelled "S" had more acceptable rela- tionships for Sweden than for Norway. Finally, links not labelled showed no significant difference between the two countries in their number of acceptable relationships. 199 relationships among the indicators of those concepts, for Norway and Sweden in the post-1905 period. Again, bivariate relationships were deemed acceptable on the basis of significant correlation and lack of significant autocorrelation. For multiple regresSion equations, acceptable relationships here involved only those inde- pendent variables whose betaweights were statistically significant in the majority of acceptable equations for the pair of concepts under consideration. That is, if the variable was usually brought into the equations only to eliminate autocorrelation and did not con- tribute significantly to the proportion of variance explained in the dependent variables, then that conceptual link was not considered here to be confirmed. The specific indicators involved in the con- firmed conceptual links can be obtained from Figure 6—1 and Table 6-15 in the previous chapter. The next step is to examine the similarities and differences between these models of development for Norway and Sweden. Figure 7-3 presents the conceptual links confirmed by acceptable relationships that were common to the models for both countries, and the links that were present in only one country. Each of the links in Figure 7-3 is labelled with the first letter of the name of the country for which the relationships in that link were more substantial. That is, the links labelled ”N” had more numerous acceptable relationships for Norway than for Sweden. Links labelled "S" had more acceptable rela- tionships for Sweden than for Norway. Finally, links not labelled showed no significant difference between the two countries in their number of acceptable relationships. 200 Figure 7-3 shows that the link confirmed for both countries between social mobilization and government penetration proved to be more substantial for Norway than for Sweden. On the other hand, the links between social mobilization and economic wealth, and between economic wealth and government expenditures, proved to be more sub- stantial for Sweden than for Norway. Similarly, the link between social mobilization and political mobilization was more substantial for Sweden than for Norway. But the link between government penetra- tion and government expenditures was equally substantial in both countries. Most importantly, Figure 7-3 shows that the only Signifi— cant difference between the models of development for Norway and Sweden was the link between political mobilization and democratization, which was present in Norway but not in Sweden. Obviously these similarities and differences between the models of development for Norway and Sweden indicate that, as expected, almost all of the major indicators of social mobilization and govern- ment penetration coincided more strongly with each other in Norway than in Sweden. The same was true of most of the major indicators of political mobilization and democratization. Finally, as expected, most of the major indicators of government penetration and government expenditures coincided equally strongly in both countries with each other. However, contrary to expectation, all of the major indicators of social mobilization except Investment coincided more strongly with only one indicator of political mobilization, Voter Turnout, in Sweden than 201 in Norway. Also, economic wealth unexpectedly coincided more strongly with most of the major indicators of social mobilization and govern- ment expenditures in Sweden than in Norway. Finally, no indicators of political mobilization or democratization coincided at all with any indicators of either government penetration or government expen- ditures. Thus, the major unexpected findings focus on economic wealth, political mobilization, and democratization. However, further re- examination of the acceptable relationships involving economic wealth indicates that the differences between the two countries there are not really very substantial. Figure 6-1 in the previous chapter shows that the indicators of social mobilization for Sweden had to be combined into multiple regressions on economic wealth in order to obtain acceptable relationships, whereas Norway's wealth related acceptably to one indicator of social mobilization by itself, namely Agricultural Employment. Similarly, Sweden had only two more acceptable relationships between economic wealth and government expenditures than did Norway. Furthermore, visual examination of the distributions of economic wealth in the two countries over time reveals that they are remarkably similar. (See Figure A-6 in the Appendix.) Nevertheless, one could interpret the difference in the number of acceptable relationships involving economic wealth as indicating a greater level of "resource-consciousness“ in Sweden than in Norway. That is, the formulation of public policy has perhaps depended more heavily on the availability of economic resources, and the development of those resources has perhaps responded more quickly 202 to industrialization and urbanization, in Sweden than in Norway. The impact of social change on public policy has perhaps been channeled more through the development of economic resources in Sweden, and more through the development of the public bureaucracy in Norway. There were far more acceptable relationships between social mobilization and government penetration in Norway than in Sweden, although the relationship between penetration and expenditures was equally sub— stantial in both countries. Furthermore, Figure A-4 in the Appendix shows that Norway's Civil Service expanded a decade earlier, beginning in 1930, than did Sweden's. Thus, Norway's policy development may have depended primarily on the development of human resources rather than of economic resources, whereas Sweden's may have depended equally on both. This corroborates Peters' (1970) discussion of the intensive financial investment in public policy without bureaucratization in Sweden. But again, these differences do not appear to be great. The major differences between Norway and Sweden and the major unexpected findings focus on political mobilization and democratiza- tion. This might have been anticipated from examination of the dis— tributions of these variables over time, for they appear to be the distributions that differ the most between the two countries. (See Figures A-1 and A-2 in the Appendix.) In both countries indicators of social mobilization coincided with only one indicator of political mobilization, namely Voter Turnout. In Sweden there were three acceptable relationships between social mobilization and Turnout, and in Norway only one. As can be seen from Figure 6-1 in the previous chapter, the indicators involved suggest that in Norway increasing 203 population pressures engendered increases in Voter Turnout, and in Sweden industrialization and urbanization also contributed to increasing Turndut. But apparently in both countries the socialist movement, as indicated by Left Voting, developed independently of social mobilization, or at least it depended on other factors as well. This supports the conclusion of Lafferty (1971) concerning the causes of the radicalness of the socialist movement in Norway. Furthermore, in Norway increases in Left Voting and Voter Turnout coincided with increases in democratization, as indicated by Enfranchisement and Representation, whereas in Sweden they did not. The latter two variables increased fairly gradually and evenly in Norway, but in rather large, isolated spurts in Sweden. (See Figures A-1 and A-2 in the Appendix.) I This runs counter to the general impression one derives from the descriptive historical literature, as expressed at the beginning of Chapter Four, that Sweden seemed to have developed more smoothly than Norway. But more importantly it suggests that the Swedish political elites resisted more strongly the leftist demands for democratic reform until the breakthroughs in 1908 and 1918. The aforementioned delay in the expansion of the Swedish Civil Service until after the Depres- sion also suggests the Swedish elites' stronger resistance to change. All of this supports earlier historical descriptions of the firmly entrenched conservative rulers in Sweden during the late 1800's, but contradicts most descriptions of the progressive Swedish bureaucracy as planning and anticipating social change. Of course, in neither 204 country did political mobilization or democratization have an impact on the development of government social programs. All of this seems to suggest the conceptual separation of the political system from the socio-economic system; the greater cohesive- ness among the concepts of the socio-economic system than among those of the political system; and the greater impact of socio-economic variables than of political variables on indicators of public policy development. This interpretation obviously supports earlier works in the field of "comparative American state politics," discussed in Chapter One, and contradicts others. It also contradicts the cross- national interpretation of Peters (1972) that socio-economic and political variables are both important in determining expenditure patterns, and that the effect of political variables is increasing with time. This latter discrepancy is due partly to Peters' concep- tualization of the Civil Service as a “political" variable, compared with our conceptualization of it as an effect of socio-economic development and a transmitter of the impact of that development on expenditure patterns. Of course, it might also be conceptualized as an effect of government social programs, namely the bureaucratization necessary to administer those programs. Obviously Civil Service is a difficult indicator to conceptualize in this type of research. But even so, our results for Sweden indicated that Peters' "socio—economic" variable, GNP per Capita, had a greater impact on expenditures than did his "political" variable, Civil Service. Furthermore, we rejected any link between democratization and either penetration or expenditures, 205 whereas Peters confirmed such a link for Sweden. This may have been due to the use of different indicators of democratization; but we still must conclude that the effects of truly political variables on expenditure patterns is far weaker than the effect of socio-economic variables in Norway and Sweden. In summary, our results indicate that Norway and Sweden experi- enced very similar patterns of development. Both showed strong effects of social mobilization on public policy development, although Norway seemed to channel that effect more through government penetration and Sweden more through economic resources. Neither country showed any appreciable effect of political mobilization or democratization on either government penetration or government expenditures. In both countries the socialist movement developed independently of social mobilization, whose only impact on political mobilization was increas— ing Voter Turnout, an effect that was stronger in Sweden than in Norway. The only significant differences between the development patterns of the two countries were the closer association between the development of the socialist movement and democratization in Norway than in Sweden and the earlier expansion of the Civil Service in Norway than in Sweden. Both were probably due to stronger resistance of the Swedish political elites to leftist demands for democratic reform and government social programs. III. Suggestions for Future Research Improving the techniques of analysis of aggregate time-series data on the political systems of Norway and Sweden will require the follow- ing steps: 206 1. Further efforts should be made to bring the "missing variables" causing the autocorrelation into the data base. Specifi- cally, some solid measure of technology must be found, whether in the area of energy consumption, research and development funding, mechanization of industry, product consumption, or even electronic communications. Data on mass communications, such as newspaper circulation, might prove useful in their own right. Further informa~ tion should be sought on the nature of political institutions, especially the character of political parties and policy-making bureaucracies, and of the administrative machinery for carrying out government social programs. Perhaps some indication of the effect of socio-cultural cleavages on the developmental process could be obtained through data on the size of specific minority groups and through more detailed figures on the occupational structure than just agricultural employment. Hopefully some indirect indicators of mass and elite attitudes will someday be developed, perhaps through content analysis. 2. Just as some variables need to be added, some can be dropped and others modified. Particularly useless in our analysis were Unemployment, Labor Conflict, and the Price Index. Union Membership should be used as an indicator of social mobilization rather than political mobilization. To reduce the problem of similarity of uni— variate distributions among groups of variables measured in the same way, perhaps all variables, except such basic size variables as Total Population and Gross National Product, should be measured as proportions 207 of something, rather than measuring some variables on a per capita basis. 3. Most importantly, the observations or time-points should be changed. To avoid the problems of missing data during the war years for Norway, the post-War period could be ignored. This is also justified theoretically because the Scandinavian systems were well on the way to becoming “developed'I systems after the Depression, as indicated by the changed nature of the relationships after the War. Data for every year from l875 to l940 should be gathered for both countries. The analysis could then emphasize dividing the total period into portions perhaps even as small as decades in search of' the points in time at which changes occur in the regression parameters for the relationships between the variables. Simple linear trans- formations could again be tried for the larger time portions, and de- trending could be tried as a last resort. All of this exhaustive analysis still might not yield any further substantive information on the process of development or on the differences and similarities between Norway and Sweden, and scientists might then, but only then, conclude that such analysis is not worth- while. But until then hope remains that it will be. BIBLIOGRAPHY BIBLIOGRAPHY Adelman, Irma, and Morris, Cynthia Taft, Society, Politics, and Economic Development (Baltimore: Johns Hopkins University Press, 1967). Alker, Hayward R., Jr., "Research Possibilities Using Aggregate Political and Social Data," in Stein Rokkan (ed.), Comparative Research Across Nations and Cultures (Paris: Mouton, l968), pp. l43-162. Alker, Hayward R., Jr., “A Typology of Ecological Fallacies,“ in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, l969), pp. 69—86. Allardt, Erik, "Aggregate Analysis: The Problem of Its Informative Value," in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, 1969), pp. 4l-5l. Almond, Gabriel A.,"Political Development: Analytical and Normative Perspectives," Comparative Political Studies, I (January l969), pp. 447-469. Almond, Gabriel A., and Coleman, James S. (eds.), The Politics of the Developing Areas (Princeton: Princeton University Press, l960). Almond, Gabriel A., and Powell, G. Bingham, Jr., Comparative Politics: A Developmental Approach (Boston: Little, Brown, and Company, 1966). Almond, Gabriel A., and Verba, Sidney, The Civic Culture (Princeton: Princeton University Press, 1963). Andren, Nils, Government and Politics in the Nordic Countries (Stock- holm: Almquist and Niksell, l964). Apter, David E., The Politics of Modernization (Chicago: The University of Chicago Press, l965). Banfield, Edward C., The Moral Basis of a Backward Society (Glencoe, Ill.: The Free Press, l958). 208 209 Bendix, Reinhard, "The Extension of Citizenship to the Lower Classes," in Reinhard Bendix (ed.), State and Society (Boston: Little, Brown, and Company, l968), pp. 253-257. Blalock, Hubert M., Jr., Causal Inferences in Non-Experimental Research (Chapel Hill, N.C.: University of North Carolina Press, l964}. Board, Joseph B., Jr., The Government and Politics of Sweden (Boston: Houghton Mifflin, 1970). Buchanan, James M., and Tullock, Gordon, The Calculus of Consent (Ann Arbor, Mich.: University of Michigan Press, 1962). Bull, Edvard, Sr., Den Skandinaviske Arbeiderbevegelse l9l4-l920 (Kristiania: Det Norske Arbeiderparti, l922). Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., Elections and the Political Order (New York: John Wiley and Sons, 1966). . Campbell, Donald T., and Stanley, Julian C., Experimental and Quasi- Experimental Designs for Research (Chicago: Rand McNally, l966). Carlson, Bo, Trade Unions in Sweden (Stockholm: Tidens Forlag, l969). Converse, Philip E., "Survey Research and.the Decoding of Patterns in Ecological Data," in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cam- bridge, Mass.: The M.I.T. Press, l969), pp. 459-486. Crozier, Michel, The Bureaucratic Phenomenon (Chicago: The Uni- versity of Chicago Press, l964). Curry, R. L., Jr., and Wade, L. L., A Theory of Political Exchange: Economic Reasoning in Political Analysis (Englewood Cliffs, N.J.: Prentice-Hall, l969). Cutright, Philips, "National Political Development: Measurement and Analysis,"American Sociological Rev1ew, (April l963), pp. 253- 264. Cutright, Philips, "Political Structure, Economic Development, and National Social Security Programs," American Journal of Sociology, LXX (March 1965), pp. 537-550. Cutright, Philips, ”Inequality: A Cross-National Analysis," American Sociological Review, XXXII (October l967), pp. 562-577. 210 Cutright, Philips, "Income Redistribution: A Cross-National Analysis," Social Forces, LXVI (December l967), pp. 180-l90. Deane, Phyllis, "Aggregate Comparisons: The Validity and Reliability of Economic Data," in Stein Rokkan (ed.), Comparative Research Across Nations and Cultures (Paris: Mouton, l968), pp. l7l+185. Deutsch, Karl W., Nationalism and Social Communication (Cambridge, Mass.: The M.I.T. Press, l953). Deutsch, Karl W., "Social Mobilization and Political Development," American Political Science Review, LV (September l96l), pp. 493- Ell. Dogan, Mattei, and Rokkan, Stein (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, l969). Downs, Anthony, An Economic Theory of Democracy (New York: Harper and Row, 1957). Drake, Michael, I'Odd Man Out,” New Society, 22 (28 December, l972). Eckstein, Harry, Division and Cohesion in Democracy (Princeton: Princeton University Press, l966). Eisenstadt, S. N., The Political Systems of Empires (New York: The Free Press, l963). Flanigan, William, and Fogelman, Edwin, l'Patterns of Democratic Development: An Historical Comparative Analysis," in John V. Gillispie and Betty A. Nesvold (eds.), Macro-Quantitative Analysis: Conflict, Development, and Democratization (Beverly Hills, Cal.: Sage Publications, l97l), Forbes, Hugh Donald, and Tufte, Edward R., ”A Note of Caution in Causal Modelling," American Political Science Review, LXII (December l968), pp. l258-l264. Frey, Frederick W, "Cross- National Survey Research in Political Science," in Rober T. Holt and John E. Turner (eds. ), The Methodology of Comparative Research (New York: The Free *Press, l9707jpp. l73— 294. . Galenson, Walter, Labor in Norway (Cambridge: Harvard University Press, l949). Galenson, Walter, "Scandinavia," in Walter Galenson (ed.), Com ara~ tive Labor Movements (New York: Russell & Russell, l952 . le Geertz, Clifford, "The Integrative Revolution: Primordial Sentiments and Civil Politics in the New States," in Clifford Geertz (ed.), Old Societies and New States (New York: The Free Press, l963), pp. l05-l57. I Gillispie, John V., and Nesvold, Betty A. (eds.), Macro-Quantitative Analysis: Conflict, Development, and Democratization (Beverly Hills, Cal.: Sage Publications, l97l), . Gurr, Ted Robert, Why Men Rebel (Princeton: Princeton University Press, l970). Harsanyi, John C., "Explanation and Comparative Dynamics in Social Science," Behavioral Science, V (1960), pp. l364l45. Hays, Wilgiam L., Statistics (New York: Holt, Rinehart, and Winston, l963 . Hennessey, Timothy M., "Considerations of Theory and Concept Forma— tion in Comparative Analysis," paper prepared for the annual convention of the Midwest Political Science Association, Ann Arbor, Michigan, April 24-26, l969. Hirschman, Albert 0., The Strategy of Economic DevelOpment (New Haven: Yale University Press, 1958). Holt, Robert T., and Turner, John E., The Political Basis of Economic Development (New York: Van Nostrand, l966). Holt, Robert T., and Turner, John E. (eds.), The Methodology of Comparative Research (New York: The Free Press, l970). Hymes, Dell, "Linguistic Aspects of Comparative Political Research," in Robert T. Holt and John E. Turner (eds.), The Methodology of Comparative Research (New York: ‘The Free Press, l970), pp. 295-341. - Kautsky, John H., "Introduction," in John H. Kautsky (ed.), Political Change in Underdeveloped Countries (New York: John Wiley, l9625 Kmenta, Jan, Elements.of Econometrics.(Neinork: Macmillan, l97l). Kornhauser, William, The Politics of Mass Society (New York: The Free Press, l959l: Lafferty, William M., Economic Development and the Response of Labor in Scandinavia: A Multilevel Analysis (Oslo: Universitetsfor- laget, l97l). 212 LaPalombara, Joseph (ed.), Bureaucracy and Political Development (Princeton: Princeton University Press, 1967). LaPalombara, Joseph, and Weiner, Myron (eds.), Political Parties and Political Development (Princeton: Princeton University Press, l966). ‘ Lauwreys, J. A. (ed.), Scandinavian Democracy; The Development of Democratic Thought and Institutions in Denmark, Norway, and Sweden (Copenhagen: The Danish Institute, l958). Lerner, Daniel, The Passing of Traditional Society (Glencoe, Ill.: The Free Press, l958). Lijphart, Arend, "Comparative Politics and the Comparative Method,‘I American Political Science Review, LXV (September l97l), pp. 682-693. Lindgren, Raymond E., NorwayeSweden: Union, Disunion, and Scan- dinavian Integration (Princeton: Princeton University Press, 1959). Lipset, Seymour Martin, Political Man.(Garden City, N.J.: Doubleday, l959; Anchor Edition, l963). Lipset, Seymour Martin, and Rokkan, Stein, “Cleavage Structures, Party Systems, and Voter Alignmentsz. An Introduction," in Lipset, and Rokkan (eds.), Party Systems and Voter Align— ments: Cross—National Perspectives (New YorKYT The Free Press, l967), pp. l-64). Loewenberg, Gerhard, "New Directions in Comparative Political Research: A Review Essay," Midwest Journal of Political Science, XV (November l97l), pp. 74l-756. Lorwin, Val R., "Historians and Other Social Scientists: The Com- parative Analysis of Nation-Building in Western Societies," in Stein Rokkan (ed.), Comparative Research Across Nations and Cultures (Paris: Mouton, l968), pp. lOZ-ll7. ' Marshall, T. H., Class, Citizenship, and Social Development (Garden City, N.J.: Doubleday, 1966). McClelland, David, The Achieving Society (New York: Van Nostrand, l96l . McCrone, Donald T., and Cnudde, Charles F., "Toward a Communications Theory of Democratic Political Development: A Causal Model," American Political Science Review, LXI (March l967), pp. 72-79). 213 McRae, Duncan, Jr., and Meldrum, James A., "Factor Analysis of Aggregate Voting Statistics," in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, l969), pp. 487-506. Merritt, Richard, and Rokkan, Stein (eds.), Comparing Nations (New Haven: Yale University Press, l966). Mitchell, Joyce M., and Mitchell, William C., Political Analysis and Public Policy (Chicago: Rand McNally, l969). Morre, Barrington, Jr., Social Origins of Dictatorship and Democracy (Boston: Beacon Press, l966). Neubauer, Deane E., "Some Conditions of Democracy," American Political Science Review, LXI (December l967), lOOZ-l009. Nordlinger, Eric, "Political Development: Time Sequences and Rates of Change," World Politics, XX (April 1968), pp. 494-520. Nordskog, John Eric, Social Reform in Norway: A Study_of Nationalism and Social Democracy (Los Angeles: University of Southern California Press, l935). Ohlin, Goran, "Aggregate Comparisons: Problems and Prospects of Quantitative Analysis Based on National Accounts," in Stein Rokkan (ed.), Comparative Research Across Nations and Cultures (Paris: Mouton, l968), pp. l63-l70. Olson, Mancur, Jr., "Rapid Growth as a Destabilizing Force,“ The Journal of Economic History, 23 (December l963), pp. 529-552. Olson, Mancur, Jr., The Logic of Collective Action (Cambridge, Mass.: Harvard University Press,71965). Organski, A. F. K., Stages of Political Development (New York: Alfred Knopf, l965). Pennock, J. Roland, "Political Development, Political Systems, and Political Goods," World Politics,~XVIII (April l966), pp. 4l5- 434. Peters, B. Guy, "The Development of Social Policy: A Longitudinal Analysis of Social Expenditures and Their Impacts in the United Kingdom, France, and Sweden" (Unpublished Ph.D. Dissertation, Michigan State University, l970). Peters, B. Guy, "Economic and Political Effects on the Development of Social Expenditures in France, Sweden, and the United Kingdom," Midwest Journal of Political Science, XVI (May l972), pp. 225-238. 214 Przeworski, Adam, and Teune, Henry, The Logic of Comparative Social Inguiry (New York: Wiley-Interscience, 1970). Pye, Lucien W., Politics, Personality, and Nation-BuildingA(New Haven: Yale University Press, 1962). Pye, Lucien W., Communications and Political Development (Princeton: Princeton University Press, 1965).- 1 Rye, Lucien W., Political Culture and Political Development (Princeton: Princeton University Press, 1965). . Pye, Lucien W., Aspects of Political Development (Boston: Little, Brown, and Company, 1966). Ranney, Austin (ed.), Political Science and Public Policy (Chicago: Markham, 1968). w Raphaeli, Nimrod (ed.), Readings in Comparative Administration (Boston: Allyn and Bacon, 1967). . Riggs, Fred, Administration in Developing Countries (Boston: Houghton Mifflin, 1964). Riley, Matilda White, "Sources and Types of Sociological Data,“ in R. L. Faris (ed.), Handbook of Modern Sociology (Chicago: Rand McNally, 1964), pp. 1014-1020. Robinson, William 5., "Ecological Correlations and the Behavior of Individuals,“ American Sociological Review, XV (1950), pp. 351— 357. Rokkan, Stein, "Norway: Numerical Democracy and Corporate Pluralism," in Robert A. Dahl (ed.), Political Qppositions in Western Democracies (New Haven: Yale University Press, 1966), pp. 70—115. Rokkan, Stein, "Geography, Religion, and Social Class: Crosscutting Cleavages in Norwegian Politics,“ in Seymour M. Lipset and Stein Rokkan (eds.), Party Systems and Voter Alignments: Cross-National Perspectives (New York: The Free Press, 1967), pp. 367-444. Rokkan, Stein (ed.), Comparative Research Across Nations and.Cultures (Paris: Mouton, 1968). Rokkan, Stein, Citizens, Elections, Parties: Approaches to the Com~ parative Study of the Processes of Development (New York: David McKay, 1970). Rokkan, Stein, and Valen, Henry, "Parties, Elections, and Political Behavior in the Northern Countries,“ in Otto Stammer (ed.), Politische Forschung (Koln: Westdeutscher Verlag, 1960), PP. 103- 136. 215' Rosenthal, Albert H., The Social Programs of Sweden: A Search for Security in a Free Society (Minneapolis: University of Min- nesota Press, 1967). Rostow, Walt W., Stages of Economic Growth (Cambridge, Mass.: Harvard University Press, 1960). Rustow, Dankwart A., The Politics of Compromise: A Study of.Parties and Cabinet Government in Sweden (Princeton: Princeton Uni- versity PreSs, 1955). Salisbury, Robert H., IIThe Analysis of Public Policy: A Search for Theories and Roles," in Austin Ranney (ed.), Political Science and Public Polioy (Chicago: Markham, 1968), pp. 151-178. Scheuch, Erwin K., "Cross-National Comparisons Using Aggregate Data: Some Substantive and Methodological Problems," in Richard Merritt and Stein Rokkan (eds.), Comparing Nations (New Haven: Yale University Press, 1966), pp. 131-168. Scheuch, Erwin K., "Social Context and Individual Behavior,“ in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, 1969), pp. 133-155. Schweinitz, Karl de, "On Measuring Political Performance," Comparative Political Studies, II (January 1970), pp. 503-511. Selltiz, Claire; Johada, Marie; Deutsch,.Morton; and Cook, Stuart W., Research Methods in Social Relations (New York: Holt,.Rinehart, and Winston, 1959). Shively, W. Phillips, "Ecological Inference: The Use of Aggregate Data to Study Individuals," American Political Science Review, LXIII (December 1969), pp. 1183—1196. Stinchcomb, Arthur L., Constructing Social Theories (New York:. Har- court, Brace, and World, 1968). Stjernquist, Nils, "Sweden: Stability or Deadlock?," in Robert A. Dahl (ed.), Political Qppositions in Western Democracies (New Haven: Yale University Press, 1966), pp. 116-146. Storing, James A., Norwegian Democraoy (Boston: Houghton Mifflin, 1963 . Thrupp. Sylvia L., ”Diachronic Methods in Comparative Research," in Robert T. Holt and John E. Turner (eds.), The Methodology of Comparative Research (New York: The Free Press, 1970), pp. 345- 358. . 216‘ Valen, Henry, and Katz, Daniel, Political Parties in Norway: A Community Study (Oslo: Universitetsforlaget, 1964). Valkonnen, Tapani, "Individual and Structural Effects in Ecological Research," in Mattei Dogan and Stein Rokkan (eds.), Quantitative Ecological Analysis in the Social Sciences (Cambridge, Mass.: The M.I.T. Press, 1969), pp. 53-68. . Verney, Douglas V., Parliamentary Reform in Sweden: 1866-1921 (Oxford: Clarendon Press, 1957). Wildavsky, Aaron, "The Political 'Economy.of Efficiency: Cost- Benefit Analysis, Systems Analysis, and Program Budgeting," in Austin Ranney (ed.), Political Science and Public Polioy_ (Chicago: Markham, 1968), pp. 55-82. Wylie, Laurence, Village in the Vaucluse (New York: Harper and Row, 1964 . DATA SOURCE FOR NORWAY Norway, Central Bureau of Statistics, Historical Statistics 1968, Norges Offisielle Statistikk XII 245 (Oslo, 1969). Norway, Central Bureau of Statistics, National Accounts 1865-1960, Norges Offisielle Statistikk XII 163 (Oslo, 1965). Norway, Central Bureau of Statistics, Statistical Yearbook of Nor- moy (Statistisk Arsbok), various years from 1877 to 1969. SUPPLEMENTAL DATA SOURCES FOR SWEDEN Johannson, Osten, The Gross Domestic Product of Sweden and Its Composition, 1861-1955 (Stockholm: Almqvist and Wiksell, 1967). Landsorganisationen i Sverige, Verksamhetsberattelse Landsserkre- tariatets Berattelse for ar 1956 (Stockholm, 1957). Sweden, Central Bureau of Statistics, Statistical Yearbook of Sweden (Statistisk Arsbok), various years. APPENDIX 217 NORWAY .52 Left Vote "\-—.—‘—“\ / .90 .50 .86 Franchise Ln n.1hfi“? I 11111.4J‘IllnliLLJ 1876 1888 1900 1912 1924 1945 1965 SWEDEN .57 ’/‘-\\ Left Vote .45 ,’ ‘\\2-,"‘.52 /\ Ir .88 _39 I .85 Franchise l_.L_L wd'fi'T‘T’j’ . 1.15:1 J . u I L 4 I - L - L4 4 J 1875 1887 1899 1911 1924 1944 1964 Figure A-l. Univariate Distributions of Enfranchisement and Left Vote. 218 NORWAY Representation .90 /\ .84 .82 ’33 {.7E ’Il ‘\ .I“/’\\//\\--v’o85 ,’ ‘70 “ Turnout I \ ,1, .56 .43 Li I I I #1 I L L I. I I I l I I I g. 1876 1888 1900 1912 1924 1945 1965 SWEDEN 1'26 Representation .99 5 1’33"”..‘86‘ 6] /.]\\¢v" . Turnout ,3s,,’>‘\/ /" 19’/~—’ L I If I I n I I I LY I I I I I l I I I I ' 1875 1887 1899 1911 1924 1944 1964 Figure A-2. Univariate Distributions of Representation and Voter Turnout. ‘ 219 NORWAY .072. Labor Conflict '059 Urban ’JLB“, "' o 9.9””--’ Population L..LI.._,1.o907.1...I. ...1o.4,.fl004 1876 1888 1900 1912 1924 1945 1965 ".17 SWEDEN Labor Conflict .O4\V .13 ,z’ Urbar‘L,,.—‘.,/"“ . .06""/“-"’ . .01 A l l Ipopu-llatI-lo? I' I I 1 I I I I I I I I 1 I I I I__._ 1875 1887 1899 1911 1924 1944 1964 Figure A-3. Univariate Distributions of Labor Conflict and Urban Population. 220 NORWAY .024 Civil ' 1 Ser ' v1ce 03 ”[07 c.“ _ .006 ° ,” Union ’05 ,’ Membership W'T'EO? I L.,...n, L., . I 1876 1888 1900 1912 1924 1945 1965 SWEDEN x’" .31 I ,’.28 I I I I I, Union 1.22 .020 Membership ,’ I I I / . Civil .009 [I Sé:I:EEI"’II’--“~———vfi”—’~_“\\_AII '0”) .008 .05 ,I' .008 .006 /\\ I,’ 'OL/ .03 l—lll'mlxllylllllllllll'llll_l 1875 1887 1899 1911 1924 1944 1964 Figure A-4. Univariate Distributions of Civil Service and Union Membership. 221 NORWAY 1984 4561 Revenue 332 / 147 193 I 50] 723 2“ . ’\\I’ Investment 381 423 ,..\,,../"“\/ 353 854 “’v\"’o¢-‘_o 504 L c_ n n l 4. I . I I I l I J I I l I l I I l I l l l l 1876 1888 1900 1912 1924 1945 1965 2743 SWEDEN 4173 I l I I I 12921 R evenue 110 17 ..I 41 ‘-”\v"-"‘-"‘~’*T“--'”285 V'A" Investment I - A - J I l l l I 6? I l I I I I jI I I I l I I I I_l 1875 i887 1899 1911 1924 1944 1964 Figure A—5. Univariate Distributions of Revenue and InVestment. 222 NORWAY 126,11 108711, 1868 / Expenditures AIJLILJVII 1876 1888 1900 1912 1924 1945 1965 SWEDEN 371 Non-Defense Expenditures [1145? I I I 'I / 6783 . ,2] 1875 1887 1899 ' 1911 1924 1944 1964 Figure A-6. Univariate Distributions of Gross National Product and Non-Defense Expenditures. HICHIGQ N STATE UNIV. LIBRARIES | 11111111111111111111111111111| 931 0202041 3 312