Precis of Comparative institutional analysis in large economic systems: usefulness, application, and limits

Transcription

1 1 Precis of Comparative institutional analysis in large economic systems: usefulness, application, and limits Pubished in Istituzioni e Sviluppo Economico - Institutions and Economic Development, vol.2 (1) pp Stefano Solari Department of Economics, University of Padua Via del Santo, Padova, ITALY solari@decon.unipd.it Abstract This paper proposes some reflections on comparative institutional analysis (CIA) and, specifically, on the use of this instrument in analysis of the relationship between institutional configuration of economic systems and their performance. We will firstly outline the debated problem of the desirable modifications of institutional set-ups in different countries. We will argue that the variety of capitalism view represents the most reasonable hypothesis of study when considering the many interactions among technologies, assets and institutions. CIA is supposed to provide tools to achieve some solid statements even in a situation of a plurality of optimal arrangements. We will compare the different theoretical standings of some CIA approaches in economics and socio-economics, showing the main converging principles to frame the study of institutional configurations: complementarity of institutions and viability of arrangements. These principles may also serve to structure the framework of analysis to reduce the indeterminacy of this kind of analysis. Keywords: comparative institutional analysis, institutional change, institutional complementarity, functional analysis, models of capitalism JEL classification: B40 E61 P51 1. The problems raised by the change in western institutional configurations The role of institutional arrangements in the performance of economic systems is a theme which is increasingly attracting the attention of economists. After the Second World War, most industrialised countries adopted institutions that gave priority to financial stability and redistribution 1. The common bundle of institutional arrangements was the Bretton Wood international financial regime characterised by fixed exchange rates and restricted mobility of capital. In this framework the most liquid resource was put under control and each economic system was free to design its internal institutional set-up with a certain autonomy. In fact, many countries put the institutions of the welfare state at the top of their priorities, which were eventually enlarged and made more consistent by centre-left governments in the 70s. 1 This compromise was explicitly theorised by both Keynesians and Social Market Economists (Koslowski, 1998a).

2 2 After the change of international financial regime, the fall of the socialist block and the globalization of markets, the difficulties of keeping autonomous institutions to obtain redistributive effects emerged. European countries were able to partially offset the problems for 20 years thanks to their monetary system 2. The increased openness and connection of economies and the pressures of strong interests connected with financial markets succeeded in eliminating most of the obstacles to capital flows, consequently most of the mixed-economy models of capitalism were put under strain. Moreover, the prevailing technology also changed and small and medium-sized firms gained strong political power due to their increasing economic importance. The latter displayed a total discomfort with the institutions regulating employment and social security, which were designed to fit oligopolistic systems. As a consequence, the institutional configurations of many countries currently appear unfit for the new international situation and strong political influence is exercised by many economic interests to change the institutional relics which gave rise to the Fordist models of capitalism. Finally, the integration of Europe required some rules which have considerably reduced state intervention in the economy, which was one of the pillars of the Italian and French institutional arrangements. The result is a general perceived need for institutional reforms but without a wide consensus on a definite direction, increasing economic uncertainty. Among economists, the prevailing opinion is that the institutional configurations of countries cannot avoid converging towards a liberal finance-led (U.S. kind of) system. However, a consistent group of scholars, expressed in the literature on the different forms of capitalism 3, affirms that convergence is neither necessary nor useful. They argue that there is a rationale for maintaining autonomous institutional arrangements, providing different forms of risk sharing, different distribution of income and investment opportunities resulting in some comparative advantages in international specialisation 4. In both cases economists face the fundamental problem of finding some criteria to evaluate the economic role of institutions, predicting effects of institutional change. The main theoretical issue of this practical problem is how to assess the role of institutional set-ups in the (differently conceived) economic outcomes of a country. A consequent point concerns the measurability of the impact on the economic system of a change in the rules of the game. A field of study is now becoming consolidated under the label of Comparative Institutional Analysis (CIA) which aims at providing some answers to these questions. However, it remains difficult to draw up solid statements on this subject. The production of solid evidence and of 2 The supply shocks of the 70s confused the situation and delayed perception of the issue. 3 See, among the more recent publications, Amable (2003), Hall and Soskice (2001), Whitley (2000).

3 3 some appropriate analytical tools to sort comparative statements is therefore fundamental in order to obtain scientific knowledge to help policy-making avoid the "nirvana fallacy" of comparing existing set-ups with ideal ones 5, which often characterises debates on these issues. It may be objected that studies in the field of economic institutions are probably still not ready to provide clear and indisputable insights into such complex problems. However, the aim of this paper is to illustrate some progress made in this field by CIA. The aim is, first of all, to explore the principles used to organise the evidence concerning institutional configurations and the connections among institutions. Then the issue becomes how to establish a point of reference to make some judgement on the fitness of institutions and institutional change (decided in the polity) without a single, general, although abstract, reference point. 2. Comparative Institutional Analysis: economists chasing after socioeconomics Comparative institutional analysis (CIA) is a field of study which aims at developing new scientific knowledge in the complex topic of the structural differences of economic systems and their performance. According to Stefan Voigt (1999a p. 52) CIA «asks how alternative institutional arrangements affect (economic) outcomes». There are a number of approaches in economics which refer to CIA. First of all, it is a constituent part of New institutional economics (NIE). The studies of Law and economics and of Economic organization 6 began to consider the possible increase in efficiency obtainable by a change in property rights or, in general, by a modification of contractual schemes. As a consequence, NIE shifted its concern from the simple allocative efficiency to the study of how legal and social constraints affect economic performance. Williamson (1985) explicitly introduced the label CIA in the study of alternative governance structures and based it on transaction costs (TC) 7. North shifted this approach to the study of long-run institutional change. In this way CIA was applied to wider economic systems where the institutional setting is complex and some models of stateeconomy interaction are needed to explain institutional change 8. The work of North therefore contributed to a macroeconomic extension of CIA and to a less specific use of the notion of 4 An intermediate view exists, expressed by Freeman (2000), which conceives an optimal model, declined according to a plurality of arrangements due to the costs implied by convergence with the optimal institutions. The latter can be seen as a weak form of the single optimal configuration view. 5 Cf. Demsetz, (1969). 6 See Demsetz (1964); Alchian et al. (1972). 7 Eggertsson (1990), however, has explicitly recognised Transaction Costs Economics (TCE) as an extension of neo-classical economics. See also Klaes (2000) for a methodological evaluation in Lakatosian terms. 8 In North (1981) the polity is the source of institutional change.

4 4 TCs. Frey, Pommerhene and other Swiss scholars introduced the macroeconomic approach to CIA in public finance and constitutional comparisons in order to show the relationship between constitutional settings and public expenditure (and taxation). Voigt (1999a) has recently further extended the CIA approach to Constitutional economics in a way which is closer to North s. On the other hand, Aoki s CIA (1996), which was developed from the study of firms arrangements and subsequently extended to wider economic systems, is not based on TCs. Aoki introduced (and refined) game theoretic explanations of institutions, taken from the multiple equilibria view of economic organisation, to study the interconnections among institutions. However, going backwards in the history of economic thinking, at the roots of CIA lies the literature on comparative systems, which owes much to the German historical school and to Ordoliberalism (Schefold, 1995). This kind of theorisation, together with Social market economy, 9 is interesting for institutional comparison because of the concept of the totality of the economic order and of some methodological intuition given by the need to reconcile the legacy of the German historical school with synchronic theorising 10. Finally, nowadays, it is increasingly fashionable to include institutional factors in growth models to produce cross-country studies which aim at establishing some regularities in the association between the existence of a certain institution and economic growth 11. This is particularly frequent in labour economics where bargaining institutions and employment protection regimes are often contrasted in relation to economic growth 12. This approach can achieve good results when the compared systems are not very different as in the case of Swiss Cantons in Feld and Savioz (1997) 13. In this context, we can assume that the similarity of context is higher than in comparisons between countries, 14 however, it is in the latter case that a CIA is more relevant for decisions concerning institutional change. Looking at the literature in applied economics, which can be defined as CIA, we realise that most of it has been produced by economic sociologists or by economists who consider economy and society as a unitary domain of study 15. In fact, a number of theoretical frameworks and an interesting amount of empirical evidence have been organised according to 9 See Koslowski (1998b). 10 See Labrousse and Weisz (2001). 11 This strategy is able to point out similarities but not to understand the role of differences, or explain the logic of internal coordination of institutional configurations. 12 Calmfors and Driffil (1988) is the milestone in these studies. See Freeman (2000) for a review of the literature. 13 The authors have found good results in the association of direct democracy and economic performance in a study on Swiss cantons. 14 We can assume similar preferences, the same constitutions, a similar territory, the same history, the same development phase, etc. 15 Of the many socio-economic approaches, we devote particular attention to the French Régulation school which will be presented later on.

5 5 such approaches. We can include in this category studies on the different forms of capitalist economies, comparative employment institutions, the comparison of welfare states, the theories on national systems of innovation, the studies on the state s role in the economy, the study of regional political institutions, etc.. 3. The role of institutions in evolving economies The prominence of socio-economics in producing evidence on the role of institutions in economic processes is due to the difficulties of economics in expanding the application of its method to issues broader than the simple allocation of resources. The inclusion (endogenization) of exogenous elements as institutions in the framework of analysis raises some problems which go beyond simple efficiency. In fact, rules represent ethical aspects in themselves which cannot be compensated by increases in wealth or, in any case, are difficult to evaluate. As a consequence, this type of analysis cannot be unravelled without the introduction of further hypothesis. The way we consider institutions is therefore important in order to study the effects of their modification. We can single out two main conceptions of institutions: a) institutions as exogenous elements representing bounds to human choices, which restrict the field of the pre-determined set of outcomes; b) institutions as rules of the game, transcending economic choices but at the same time defining available options. In the former case institutional modification is relatively neutral to outcomes and is defined according to the desirability of results, as in standard Law and economics. In the latter case, institutions display a foundational role in economic interaction because they define property rights on resources, the rules of interaction and also acceptable coordination forms. When studying large economic systems, the latter conception is more relevant. Of course, the modification of economic institutions may be the result of ethical views of society as well as of compliance interests. Eggertsson (1998b, p. 9) affirms, that «institutional policy, therefore, depends not only on the balance of various interests but also on the outcome of the competition between alternative models and ideas about social processes». As a consequence a community may display some high reservation price on the change of some rule because of ethical objectives. However, what in fact determines institutional change is the complex interplay of political and economic interests. There is a further issue regarding the way institutions are conceived to affect choices which is relevant in the study of their change. It regards the relationships arising between institutions. In fact, institutions may be conceived: i) as an additive set of separable units (rules);

6 6 ii) as a system of rules with particular relationships intercurring between them which affect their coordination effect on choices. In the former case institutional change may be operated separately on single rules, directly determining positive or negative effects on economic outcomes. In the latter case, the situation is more complex because also the relationships between rules affect the coordination effect. In this situation the coherence of the whole order becomes important to determine the effect of changes. In general, in this latter case, there is no chance of Paretian improvements, but changes can be made only finding discrete sets of institutional configurations. Moreover, if some rules are rigidly determined because of political-ethical values, this also affects the choice of institutions complementing the general order. Some theoretical research is in progress to formulate justification of the latter hypothesis by the concept of multiple equilibria derived from evolutionary game theory and from the principle of complementarity and supermodularity (Hall and Soskice, 2001; Aoki, 2001; Amable 2003; Pagano, 2002; Topkis, 1998). 4. Institutional comparison: the risk of indeterminacy The problems illustrated above complicate the study of the relative efficiency of rules. Comparing the role of different institutions, application of the same method used for normal economic choices comparative statics produces weaker results (Eggertsson, 1997; 1998a). This is due to stronger ceteris paribus hypotheses needed to take into account part of the elements normally excluded from the standard analysis (institutions). When dealing with large economic systems we face the double problem of coordination and efficiency (institutional arrangements in fact are strictly connected with the concept of coordination). On the one hand, microeconomic efficiency consists in the best allocation of resources within a given opportunity set the latter being also determined by institutional factors and in relation to a defined aim. On the other hand, different forms of coordination result in different opportunity sets for efficient choices 16. If institutional configurations are important elements determining the form of coordination, then we can consider that institutional change does not simply affect opportunity costs, but also the range of opportunities. Therefore, the change in the rules of the game is a rather different issue compared to that of a change in other economic resources. This is an important issue and, in fact, Kirzner (1998) proposes the same coordination as an objective criterion of economic "goodness" abandoning the reference to 16 Institutional change leads first to a change in coordination forms and only at a second stage can we redefine the efficiency problem.

7 7 efficiency 17. Moreover, the co-ordination form is not the only determinant of opportunity sets for economic choices. Other variables such as the distribution of property rights on resources and technology are equally important. Furthermore, some differences in the preferences of actors belonging to different systems may produce, ceteris paribus, different choices which should be differently coordinated. If a change of rules will surely modify the opportunity set of agents, it may also change their strategic behaviour. In fact, preferences concerning choices affecting stock variables and technologies are relative to coordination contexts. This fact implies that in order to evaluate the effects of a change in coordination, we must alternatively assume a general ceteris paribus hypothesis on resources, technology and actors preferences or assume exogenously that actors change their strategies according to institutional change. In the former case, the logic is to assume efficient choices made within the opportunity sets defined by different coordination forms and to compare the latter to establish which is producing best results according to a common principle of judgement. The result is that we would be able to say that a given set of rules provides a best opportunity to increase welfare given a set of agents with some known and invariant preferences, resources and technologies. This approach is best suited to limited contexts where single contractual arrangements may be conceived as instrumental 18. Such instrumentalism may be adopted also at the macroeconomic level for minor legal changes, but results are a good deal more uncertain (Eggertsson, 1997; 1998a). Moreover, it would be disputable and perhaps also contradictory to change some structural factors such as institutions (with large sunk costs) according to short-term preferences. It is more appropriate to think that institutional set-ups evolve in the same time dimension (and presumably co-evolve) with the results of strategic behaviour of the main interests at stake, that is to say, with the dynamics of technology and investment patterns 19. A further aspect is that institutional change implies some cost. As Freeman (2000) argues, if there are costs for modifying institutions, more than one equilibrium would be optimal at the same time. Therefore, comparative statics comparison may be a reasonable framework for applied studies in specific economic contexts but, given the very restrictive hypotheses and the unreachable information required by researchers, it should be used carefully in large systems and for substantial reforms where relationships between institutions acquire relevance. Notwithstanding these difficulties, the applied study of institutional configurations of different countries can provide much information on the available options to improve the 17 Being an Austrian economist he believes in the efficacy of economic processes in achieving the best solution. 18 At the firm level we can treat the form of coordination as an instrumental variable (ceteris paribus) but it is more difficult to make such an assumption for an economic system.

8 8 coordination of economic processes. This has the advantage of avoiding the so-called "nirvana fallacy" of comparing existing institutions with ideal arrangements. Some conscious and wellpondered approximation has to be introduced and some criterion and technique of comparison should be adopted. The solutions vary considerably between the different schools of thought. Constitutional economics, which is strictly influenced by New institutional economics, has theorised CIA as a valuable instrument to measure economic outcomes. Voigt (1999b p. 7) pointed out two conditions in which the effect in rule change can be estimated: 1 - the existence of a commonly agreed measuring rod to compare two situations in relation to their outcomes; 2 - a reliable hypothesis of the form institutional change of type X will lead to an improvement in social welfare of amount Y should be available 20. New institutional economics grounds its hypothesis concerning "welfare gains" considering the change of transaction costs (TCs), which are a metric variable negatively correlated with efficiency and which do not present the disadvantage of mixing qualitative and quantitative variables. However, this kind of comparison implies other seldom explicated hypotheses: i) a "measuring rod of the measuring rod" (TCs) should be assumed, otherwise, if an ideal transaction situation (a market with zero TCs) is assumed as a reference to measure existing TCs, the nirvana fallacy re-emerges 21 ; ii) no complementarity between institutions exists (additivity) 22 ; iii) TCs are imputable to chosen institutions 23 ; iv) preferences are assumed as given (in time and space) and the change of institutions would not modify them, which determines a total disconnection between choices and regularities of behaviour; in any case, v) only a marginal modification can be taken into consideration when estimating a welfare effect. The recourse to empirical associations of TC with empirical features of organisations such as investment specificity, the frequency, the complexity and the degree of interconnectedness of 19 We can therefore affirm that there are choices within the rule set and choices of the rules (Vanberg, 1994), but we must be careful that the two kinds of preferences are not mixed because their interaction does not occur in a timeless space. The sequence order is important. 20 Two more hypotheses are made concerning the fact that there are actors capable of bringing about such an institutional change and their willingness to do so, but they are not of interest here. 21 The first point is important since the point of reference adopted in the evaluation of TC in large systems is not clear. It seems to be important to establish how to measure the change in transaction costs. Many factors are not expressed in monetary terms and their evaluation presupposes some reference point, which may be assumed to be a perfect market interaction, but in this case a selfreferential reasoning emerges. Many factors influencing efficiency of exchanges are anyway not easy to measure, e.g. information asymmetries. 22 The second point concerns the linear additivity of the effects of institutions. If some complementarity among institutions exists, a multiple equilibria situation may emerge which makes welfare effects uncertain (Pagano, 2002; Aoki, 2001). 23 Many forms of uncertainty affect exchanges, which are implicit or latent. The use of TC as a measuring rod to perform CIA works well when governance structures include all the factors influencing TC. If this assumption does not hold, a complex problem of isolation arises playing a major role in the evaluation of governance structures.

9 9 transactions 24 is precisely studied for micro contexts but becomes rather metaphorical in wider systems. The result is that although TCs are seldom observable in large systems, the TC approach remains a pillar of New institutional economics and in certain cases a useful and concise tool in evolutionary studies, in particular the Evolutionary new institutional economics of historians such as North and Eggertsson 25. If we abandon the TC as a measuring rod, the choice is left to an approach based on strong empirical evidence concerning connections between institutional configurations and general economic outcomes (growth, distribution of income, unemployment ). A serious problem of isolation of the significant object of study from the context still remains important because of the necessary choice of the institutions which matter. Institutions include many heterogeneous factors and are closely interlinked so that disentangling the bundle is always a difficult operation of abstraction. In order to overcome these shortcomings linked to the impossibility of good ceteris paribus conditions in applied economics, comparisons should be made in more precise terms and in an encompassing perspective in order to reduce uncontrollable exogenous factors. Voigt (1999a p. 53) suggests at least three types of reasonable comparison to overcome such problems: 1 a comparison of institutional arrangements that are made simultaneously at one particular point in time (by relying on the work of economic historians); 2 a comparison of identical institutional arrangements that have led to different outcomes; 3 a comparison of different institutional arrangements in a single society at different points in time. Such comparisons often take institutions as qualitative variables and economic growth as a metric variable 26. Cross section regressions are currently used as an alternative source of comparison. Their aim is that of finding a correlation between growth and the existence of certain institutions. They may provide some robust facts about the association of institutions and performance. Taken as it is, this type of study makes an implicit hypothesis concerning the additive relationship between institutions, and therefore it presumes the existence of a 24 See Masten et al. (1992). TCs are commonly divided into coordination costs and incentivation costs, the latter further divided into costs due to information asymmetry and to imperfect commitment. 25 Eggertsson has distinguished two categories of NIE which differ in some basic assumptions and lead to different applications of CIA. We can subdivide NIE into a static institutional theory and a dynamic institutional theory. The former is «a relatively homogeneous approach which assumes stable preferences, maximization rationality, and equilibrium; or alternatively relies on functionalism with as if maximization or bounded rationality» (Eggertsson, 1999 p.11). He adds in a footnote that as if maximisation is used to avoid explaining individual decisions and collective choices. The dynamic theory «consists of a portfolio of approaches that explicitly model social learning and knowledge and employ various cognitive approaches and models of rule governed (procedural) rationality" (Eggertsson, 1999 p. 11). 26 The use of GDP as the metric variable is obviously considered to significantly reduce the incomparability of comparison, however, some structural differences in values among countries persist even in industrialised economies. TCs are probably less distorting as a measuring rod, but the difficulty of measurement makes them unpractical.

10 10 single optimal configuration. This kind of technique would have been more robust if institutions directly affected growth without any coordination effect. In fact, this kind of study is generally unable to provide clear evidence. Freeman, for example, found that, in general, the literature on labour economics did not find that institutions have an important effect on growth, even though some evidence of their effect on redistribution and equality exists. Actually, cross section analysis does not perfectly belong to the category of comparative techniques since it shows only similarities and does not provide information on the role of differences. The way to overcome this problem is to use it as a step in the comparative process or to fit it in a more sophisticated process also using qualitative analysis. 27 Qualitative comparative analysis (QCA) 28 is a research technique based on logic which can be useful for CIA to single out evidence of some relationship between institutions and some economic indicator (performance, stability, risk distribution). It explains events by contrasting existing configurations of variables (institutions) and cases when the event has been produced or not. QCA especially helps to single out events produced by a combination of different causal factors and events which may be determined by a plurality of configurations of causes. Both causes and events are expressed in binary terms in a Boolean matrix. Obviously, in the case of institutions, a preliminary taxonomy of the relevant institutions to be assessed has to be assumed and a criterion for discriminating between relevant and irrelevant institutions should be developed. The comparison of cases then adopts the logic of the comparative method per genus et differentiam illustrated by J. S. Mill in A System of Logic (1884) and more sophisticated algorithms for treating large data matrixes. This approach provides fairly robust results (in relation to statistical techniques) but it could be misleading in the case of too narrow a number of variables (institutions considered). In the case of too large a number of variables it can lead to an irreducibility of data to significant theorisation because the geometrical increase of the matrix dimension is 2 k, where k is the number of institutions considered. As a consequence, the taxonomy to classify institutions is a crucial part of CIA based on QCA. Boyer (2001) exploited this technique to demonstrate that the institutional set-up best fitting the "new economy" is not necessarily that of the U.S., the growth of these sectors being higher in Scandinavian countries (with large welfare states). Another deficit of QCA is that it does not directly provide evidence on relationships between institutions. For this purpose cluster analysis is used to highlight some complementary effects between institutions (Amable, 2003). 5. Case-oriented studies and the framework for CIA 27 See Solari (2004a). 28 See Ragin, (1987); (1994).

11 11 A large amount of evidence on the role of institutional configuration on economic processes has been provided by socio-economics and in particular by the followers of Schonfield (1965) 29. This approach is based on a rationale against a best-performance oriented comparison 30. a) No institution can be understood in isolation from the context as a whole. The general institutional configuration defines the role of the single institution. b) Different countries presuppose differences at least in certain kinds of preferences (which are culturally determined) 31 and also in the criteria of evaluation of outcomes, which cannot be easily compared. c) Causes rarely operate in a simple additive fashion as in statistics. Rather, they combine and intersect to produce changes. Moreover, different combinations of causes may produce the same outcome (this fact is mastered by QCA). d) Different economic systems may be in the development process at different time phases. Therefore the same institutions may be playing different roles. In order to avoid such problems of incomparability many researchers have preferred to concentrate on accurate case studies to analyse the many interconnections among institutions which define economic systems. Extensive literature has been developed, now known under the label variety of capitalism. Of the many systems existing, the German, the Scandinavian and the French models have been the most studied since they present the most typical features 32. This approach has some evident advantages since it provides some very good quality information, an articulated structuring of institutions, a good description of the interests at stake and of the actors that carry political demand. Moreover, it has contributed to connecting economic interests, institutional configurations and policies consisting also in processes of reform. Many linkages, interdependencies and synergies have been highlighted which give some depth to economic analysis. On the other hand, this approach presents some shortcomings. Firstly, it is difficult to demonstrate the relative importance of institutions in the working of economic systems if we do not compare different systems. The evaluation of feasible alternative arrangements, which may provide best outcomes, is not possible. Therefore, a viability perspective on institutional configurations can be provided, based mainly on conflict control and synergy of 29 Cf. Hollingsworth et al. (1994); Kitschelt et al. (1999); Soskice (1999). 30 Some of these arguments are taken from comparative statistics debates, see Ragin (1994). 31 In particular, relational assets may play a basic difference. Let s consider the resistance of family structure in Mediterranean countries compared to the higher atomisation of individuals in Northern Europe. 32 See Streeck and Crouch (ed.s)(1996).

12 12 interests, but it is difficult to provide any quantitative economic judgement beyond that 33. Lastly and marginally, results are considered not completely consistent by social scientists used to formalised theorisation and econometric tests, and therefore a problem of scientific communicability exists. Case studies may extend their scientific role showing the complementarities and the contribution of single institutions to the coherence of the whole configuration by the adoption of CIA. The latter may systematically single out similarities and differences of institutional arrangements in relation to their performance. The evidence of recurrent clusters of institutions assumes special importance since it provides arguments against the additive view of institutions and for the variety of capitalism which consists of a given number of viable forms. In fact, the large number of elements may weaken the power of comparison. The crucial problem is to find the right compromise between detail and operationality of the given representation of the economic system. On the one hand, excess simplification of the institutional arrangement may lead to strong results in the comparison, but it affects the accuracy of the evaluation of single special institutions. In this case the risk is that of underestimating the opportunities for change and for achieving new synergies and complementarities. On the other hand, with excess detail there is the risk of losing the help of comparison to fix the performance of certain arrangements. In order to focus on the problems involved in the evaluation of institutional change, the problem of comparison may be subdivided into two parts: 1 - the comparison technique; 2 - the framework in which the economic system is theorised, that is to say, the framework used to order institutions, their relationships and their interactions with other economic variables. As we have already seen, the comparison per genus et differentiam still remains the basic logic used to develop strong evidence. On the other hand, the macroscopic framework representing economic systems to be compared has not found a stable conception. Since institutions are a way of reintroducing a part of the context in economic reasoning, the categories used to model them acquire special importance. There are many taxonomies to help detect relevant institutions. The best suited to our aims are those which provide a general structuring of the whole system. Given the importance of complementarity of institutions for comparison, the classification is as crucial as the correct clustering of single items. A review of the main studies which have considered similar problems provides the idea that there are two basic principles used to connect the general order: micro-based concept of complementarity and the more macroscopic principle of "coherence" or "viability". 33 The viability of the German or French models has often been declared in danger when some institutional change has taken place. At present, they are not in good health but still working.

13 13 6. The different theoretical frameworks of institutional comparison The first frameworks for systematic institutional comparison were probably laid down by the German historical school, however, the concept of economic style as well as that of economic order or Gesamtordnung did not meet with great success in economics mainly because of the marginality role of economic problems and the idiographic method used. From the epistemological point of view, the contribution of structuralism in anthropology to institutional comparison should be mentioned because it affected many economic studies 34. Lévi-Strauss (1958 p. 98) affirmed that structuralism is a method that permits demonstration of how we can discover some differently declined invariant properties behind the heterogeneous appearance of empirical elements. Therefore, this method of generalisation consists of going up to the level of abstraction where «the incomparable becomes comparable», where the objects to be compared are built in such a way that only a single general theory exists, which, by fixing what is common, allows differences to be established at the same time (Théret, 1997 p.197). Institutional arrangements (behaviour regularities) can be considered such an abstract framework. The theory of Régulation represents an economic evolution of structuralism, and it is one of the schools of thought which more intensively adopted CIA. According to Théret, the three main principles of the comparative approach of the Régulation School are: 1) The comparison concerns relations and systems of relations and not system elements; 2) The necessity of comparing systems at a higher level of abstraction which allows the emergence of common structures and differences, avoiding surface comparisons; 3) The need to compare historical developments of structures as well as their synchronic coherence. We add that the Régulation framework has the advantage of completeness, that is to say it concerns the whole socio-economic system and therefore it eliminates most of the dangers of overlooking third variables, any elements that may have a crucial role. This theory is based on the concepts of accumulation regime and regulation mode, both of which lie at a level of abstraction, which allows for comparisons of capitalist economies. In particular, the régulation mode is the whole of the procedures, including single and collective behaviour, which allow: reproduction of fundamental social relations; sustaining and driving of the accumulation; ensuring of the dynamical compatibility of many dispersed decisions. It is normally subdivided into five institutional forms: money and finance, competition and industrial 34 Unfortunately, the most widespread structuralism in economics is that of industry studies which sees economic performance as determined by the industrial structure, which is a poor application of structuralism.

14 14 organization, labour and welfare, state form and intervention, international relations. It includes the generality of institutional forms, which is the backbone of the economy. On the other hand, the elements of the accumulation regime connect institutions to economic growth processes and to stock variables. As a consequence, the Régulation approach is a macroeconomic approach which provides a full taxonomy of institutions and connects them to economic growth through the concept of "coherence" assuring the best economic coordination. 35 The unity of the system to be compared is the national economy. The problem of comparability becomes more complex when dealing with nonmacroeconomic systems, that is to say, when the unity of the economic system is not assured by political boundaries. In this case two helpful theoretical perspectives have been expressed thanks to system thinking (Herbert Simon) and microeconomic subjective game theory (Masaiko Aoki). Simon (1978) exposed a theoretical grounding of CIA connecting institutions to system thinking. His main idea, the functional analysis of institutions, has been taken as a point of reference for CIA by many authors, but never well expanded in detail and, above all, never developed in its original spirit. 36 In Simon s opinion institutions do not necessarily assure, in an evolutionary process, the coherence of the general economic system. However, "functional analysis" can aid the study of their role in the functioning of the entire system. The term functional means that every part has a role in the working of the whole, which is expected to achieve some goals for the actors composing it 37. According to Simon (1978 p. 3), functional analysis is concerned with explaining how «major social patterns operate to maintain the integration or adaptation of the larger system». 38 Consequently, in line with his organizational approach, he introduced a teleological dimension in order to explain system integration. This dimension is strictly connected to social goals. The introduction of a functional perspective does not necessarily lead to "functionalism" and to a teleological explanation (in the sense of end-determined actions). The part is not explained by the whole, it simply has a role in its evolution and in the coherence and viability of its evolution. Individual actors have some (conflicting) preferences concerning its arrangement, but viable institutional compromises are politically reached. In fact, the perspective proposed consists of an open-ended evolution given 35 See Herrmann-Pillath (2001). 36 First of all by Williamson who, it should be remembered, was Simon's student. 37 Wimsatt (1997 p. 103) has pointed out three meanings for the concept of function in biology: ➀ To say that an entity is functional is to say that its presence contributes to the self-regulation of some entity of which it is part ( ); ➁ To say that an entity is functional is to say that under at least some conditions it plays a (presumably causal) role in the operation of some system of which it is part ( ); ➂ To say that an entity is functional is to say that it is being selected for or maintained by natural selection ( ). 38 «Institutions are functional if reasonable men might create and maintain them in order to meet social needs or achieve social goals.» (Simon, 1978 p. 3)

15 15 by the search for a viable co-ordination which finds different forms of institutionalisation 39. What is also interesting in Simon s CIA is that the comparison of discrete forms is the main tool of functional analysis. CIA is therefore seen as a means of exploring viable coordination forms and not of establishing welfare gains. This does not focus on how variables are equated at the margin, or how equilibrium is altered by marginal shifts in conditions. Rather, it focuses on qualitative and structural issues, typically, on the choice between a small number of discrete institutional alternatives (Simon, 1978, p. 6). There are two main implications of Simon s research proposal which lead us back to the points already discussed. The first, which was already illustrated by Hayek (1973), is that an institution is only understandable in relation to the whole economic order and the relationships between institutions are the most important aspect to be analysed. The components of a system should be coherent with the whole, which is not simply given by their aggregation but also by their relationships 40. Secondly, comparing discrete forms implies fixing some categories, some ideal types. Therefore, a framework of analysis is needed which is not a theory but an explicit criterion for ordering experience in discrete forms. We therefore distinguish between a theoretical activity geared to making assertions to be tested and that of developing frameworks to fix categories, a process which involves a procedural rationality on the part of the researcher (Delorme, 2001) who aspires to obtain the intelligibility of the problem studied. The validity of theorising therefore also consists in performing an effective separation between the object of study and its context. Further insights into CIA methodology can be found in the studies of Aoki (1996; 2001). His approach adopts a more microeconomic approach combining «comparative information across various economies with context-specific micro modeling based on recently developed game theory, contract theory and information economics» (Aoki, 1996 p. 2). He suggests accumulating context specific models in order to improve understanding of the functioning of institutions (1996, p. 7). The accumulation of these specific models and the test of their predictive power could improve our understanding of the inter-relatedness of institutions and economics in general 41. On the one hand, the aim of accumulating context specific models makes Aoki closer to single country study approaches, whereas on the other, he formulates a gametheoretic approach to institutional arrangements which is itself an abstract form capable of 39 Therefore we suppose that individuals act purposefully according to their interests and values to find a viable coordination, the latter being the only way to achieve individual aims which can be met only by collective action. 40 Hayek (1973 p. 36) "order" was inspired by the studies of St. Augustine and consequently there is an indirect relationship with Aristotle's principles of final cause. 41 Aoki (1996, p. 3) distinguishes four main forms of links deriving from collective action on economic interactions: markets and money; legal and political rulings by the state; contracts and (private order) organisations; cultural beliefs and social norms.

16 16 setting comparisons between different contexts. His subjective game model is able to grasp the aspects of interlinkages and interdependency of institutions. By introducing a distinction between objective (a consequence function defined by institutions in other domains, laws, technology, etc.) and subjective elements (preferences) in the payoff function he is able to distinguish exogenous and endogenous rules of the game: he reconciles the (Northian) view of institutions as rules of the game and the game theoretic view of institutions as equilibrium of a game. Institutions are therefore seen as some self-sustained system of shared beliefs about how the game is played, leaving open opportunities for endogenous innovation. 42 Moreover, by introducing the supermodularity principle 43 he is able to frame the main co-ordination problems involved in institutions. This game form, defined in various domains, is a good device for analysing the interdependencies of institutions clustering in various economies. In Aoki (2001), an overall institutional arrangement is a synchronous set of institutions across constituent domains in the economy, which are interdependent because of the reciprocal definition of the payoff functions for the agents involved. Complementarity is therefore the result of the coherence of actions across different institutions, given by the fact that a consequence function in one domain is also determined by institutions in other domains, law and technology 44. The interdependence between property rights, public regulation, markets, organisations, contracts, cultural beliefs and social norms, combined with the complementarity effect, leads Aoki s CIA to a multiple equilibria view of economic systems which in a certain sense acts as a micro-foundation of the diversity of capitalism approach and which is much stronger than the view of Freeman (2000). 7. The structuring of comparison The outlined CIA approaches show that there are two main recurring concepts in the logic of institutional comparison, on which the framework of analysis may be anchored to reduce theorising uncertainty: complementarity and viability. Complementarity is a microeconomic principle studying preferential coupling of institutions and therefore concentrates on the basic 42 An institution is therefore defined as «a self-sustaining system of shared beliefs about how the game is played. Its substance is a compressed representation of the salient, invariant features of an equilibrium path, perceived by almost all the agents in the domain as relevant to their own strategic choices. As such it governs the strategic interactions of the agents in a self-enforcing manner and in turn is reproduced by their actual choices in a continually changing environment» (Aoki, 2001 p. 26). 43 Firstly developed by Topkis, then spread by Milgrom and Roberts. See Topkis (1998); Milgrom and Roberts (1992). 44 Moreover, since contracts are always incomplete, non-economic factors strongly influence the functioning of transactions: «the effectiveness of contracts and organizations is supported by a fabric of institutions which define outside options for organizational participants and constrain individual and organizational behaviour» (Aoki, 1996 p. 6).

17 17 mechanisms of coordination between connected spheres of human action. Viability is a macroscopic concept which may be best referred to sustainable socio-economic interaction and therefore it deals with "conflict control" 45. Complementarity permits the study of how discrete structures are built interconnecting basic components, while viability is a principle which sets some constraints on the whole and permits evaluation of whether some functional parts fit the entire arrangement well or not. Actually, these concepts tend to express from different perspectives the coherence of the order (used by the Régulation and by the Hayekian perspective); the former adopts a microeconomic approach while the latter is defined by referring to a macro-social point of view. We obtain different ways in which CIA can be structured according to the order in which presuppositions concerning the kind of relationships founding the definition of the economic system to be compared are assumed. Two different conceptions of CIA may be relevant to our purpose: 1) An organic or functional comparison that takes viability as a precondition for the structuring of the system and as a reference point in the evaluation of its internal dynamics; 2) An organisational comparison of systems functioning which takes the complementarity of institutions as the basic determinant of the compared units, assuming a basic set of interacting institutions as the main object of study. The former kind of CIA (1) assumes the boundaries of the object to be compared as given and system integration (consistent order) is hypothesised as a fundamental constraint in its dynamics. It presumes the existence of a unity of the institutional set-up, that is to say a boundary can be drawn around the economy which is being analysed. This boundary is the polity, because that is the arena where conflicts are reduced by institutionalization of practices. The arrangement of institutional interconnections is on the contrary assumed as something to be discovered since no coherent set of synergies is presumed or taken as a standing point of the comparison. CIA represents a means of consolidating evidence of such institutional arrangements by showing the different institutional sub-components of economic orders. Different systems are evaluated in their ability to produce coordination and avoid conflict, and no direct inference on performance is assumed. A co-evolution of institutions and other economic variables, rather than a direct linear causality, is hypothesised. This approach usually presumes a theory of the state or some insights into its role in obtaining the evolutionary consistency of the economy. A reduction of complexity framework, which may be a taxonomy or other ordering criteria concerning the institutional factors, is adopted to permit the comparison. Institutional forms are detected in the whole economy. 45 Cf. Knight (1992). I have to thank Stefano Palombarini for discussing this issue with me and shedding precious light on the many connected problems.