Conceptual and Empirical Challenges of Evaluating the Effectiveness of Innovation Policies with ‘Behavioural Additionality’ (the Case of IWT R&D Subsidies) Jan Larosse IWT-Flanders Abstract Additionality is a key concept for the evaluation of the ‘effectiveness’ of policy instruments for stimulating R&D and innovation. It is historically linked with the conceptual framework of ‘market failure’ as a rationale for government intervention in R&D or knowledge creation and diffusion in general. It has a theoretical foundation in ‘welfare economics’: when the ‘optimal’ level of R&D investments is not attained, public incentives can ‘make a difference’. Because the social optimal level of R&D cannot be calculated, evaluations try to measure if government incentives (subsidies) are or are not crowding out private investments in R&D. This is done without referring to an optimal level, but taking the existing market conditions as given. These evaluations are rather narrow: additionality equals ‘not substituting’, a negative motivation, and are only interested in the financial effects. But the concept of additionality has more potential for policy evaluation than that, as it is linked with the idea of ‘compensating’ the spillovers of knowledge creation, generated by the special character of knowledge as semi-public good. Knowledge can be used without wearing out and thus contributes to cumulative growth. Dynamic spillovers are in fact the source of productivity gains that are a multiple of the private productivity gains for the private investor. This means that government has an interest to stimulate private R&D because it generates social benefits that go beyond the simple under-investment hypothesis. Additionality has a much broader meaning in this context. Here we have to extend the reference framework of additionality to the conceptual model of the innovation system, because this is grounded in the understanding of innovation as an interactive and cumulative process. The existence of ‘system failures’ provides a rationale for policy to ‘make a difference’ in unleashing the synergy potential of innovation systems that are not performing well because of unbalances, lock-ins, weak connectivity and other systemic dysfunctions that are related to the interaction pattern of actors. Therefore we can use the concept of additionality as a ‘bridging concept’ between the older and the newer models of innovation and government intervention. It has to be extended from the linear model in which ‘input’ and ‘output’ additionality are supposed to be closely correlated to a process oriented non-linear model in which ‘behavioural additionality’ is the key of government policy. Learning takes place in all circumstances, but it is now up to policy evaluation to continuously improve the leverage of policy instruments to increase the learning capabilities on actor level and institutional level and the knowledge distribution capacity of the system. Assessment of the effectiveness of individual instruments (as subsidies) therefore is part of managing innovation systems. Evaluating instruments from the perspective of their (comparative) behavioural additionality is a step towards new systembased evaluation practices, but much work needs still to be done. This paper discusses the challenges. Behavioural Additionality Jan Larosse 1 16/02/16 Conceptual and Empirical Challenges of Evaluating the Effectiveness of Innovation Policies with ‘Behavioural Additionality’ (the Case of IWT R&D Subsidies) Contents 1. Introduction 2. Evaluation and ‘additionality’ in Flanders 2.1. Actual practices of evaluation 2.2. The new perspective of government on innovation 3. About the concept of Behavioural Additionality 3.1. Additionality 3.2. Behaviour 3.3. Kinds of additionality 4. About the measurement of behavioural additionality 5. Policy learning 6. Organisation of the work programme in Flanders Jan Larosse is Scientific Advisor at IWT Flanders, the Innovation Agency of the Flemish government (Belgium). He coordinates the IWT Observatory, the analytical unit of IWT. His work relates to innovation monitoring and innovation studies covering themes such as the knowledge economy, cluster analysis and innovation systems, as well as policy evaluation and additionality. Behavioural Additionality Jan Larosse 2 16/02/16 1. Introduction The 3% target for R&D investments, the increasing budgets for innovation policies in general, and – foremost - the emergence of more integrated and encompassing kinds of innovation policies, call for increased and renewed forms of evaluation of the effectiveness of these governmental efforts. The introduction of the concept of ‘behavioural additionality’ in recent years is an attempt to enlarge the traditional evaluation methods based on ‘input’ and ‘output’ additionality and link them with the policy framework of the national innovation system. The need for of such a concept is demonstrated by the fact that it expresses a ‘catching-up’ of policy and evaluation theory on already widely applied practices of policy makers to explicitly target behavioural changes in the design of policy instruments. Most notorious examples are programmes aiming at fostering co-operation between the actors in the innovation system in order to increase the productivity of the knowledge investments through spillovers. The inability of present monitoring and evaluation methods to adequately assess the effectiveness of the new system oriented policies are an important bottleneck for their generalisation and therefore for performance improvement of innovation systems. System policies capitalise on the synergistic character of the innovation process, while traditional policies (and evaluation methods) only perceive sectoral policy actions towards individual agents on project level. Therefore increased efforts to strengthen the capabilities for evaluation are needed. Because of the recent and still immature nature of the new approach on ‘behavioural additionality’ efforts for conceptual clarification and for standardisation of empirical research methods can contribute to form an international community of practice that may foster the capacity for international policy learning. The work of the OECD to establish a framework for research and for exchange of good practices on this subject is of great importance for national innovation policies. The aim of this document is to contribute to the clarification of the research agenda, based on the experience of IWT Flanders in setting-up an evaluation method for its R&D-subsidies. The following chapter starts from the experience of evaluation of bottom-up innovation stimulation in Flanders. Additionality is embodied in several specific selection requirements. This practice expresses a growing understanding of the complexity of innovation processes but evaluation of the effectiveness of these schemes by their stated objectives cannot cover the new systemic perspective of the role of government. Behavioural additionality appears as a ‘bridging concept’ in the evolution towards third generation innovation policies that takes account of the interaction effects of public and private strategies. The next chapters discuss the opportunities and limits of this new approach on the level of the concept and its measurement. The main message of introducing behavioural additionality is the focus on learning, including policy learning, to enhance innovation performance. The last chapter briefly describes the complementary components of the comprehensive evaluation strategy that is needed to implement this evaluation perspective and subsequent new role for public support agencies in the innovation system. 2. Evaluation and ‘additionality’ in Flanders 2.1. Actual practices of evaluation IWT is the agency of the Flemish government for the implementation of the Flemish innovation policy but has also important tasks in policy preparation. It was established in 1992 to give shape to the new competences in science and technology that were transferred from the federal to the regional governments in Belgium and controlled a budget of 235 Behavioural Additionality Jan Larosse 3 16/02/16 million euro in 2003 (40% of the overall Flemish S&T budget). IWT plays an important role in the new Flemish innovation system. IWT fulfils different tasks; evaluation therefore has many dimensions and levels. Although the question of efficiency (use of given resources) is important to IWT management, is subordinated to the question of effectiveness (attainment of given goals) in the evaluation of stimulation activities such as R&D subsidies. The first question implies the improvement of control and capacity management; the second is oriented towards evaluating desirable outcomes, which is the final policy relevance of the intervention. Do IWT actions make a difference in the improvement of innovation performance? Evaluation culture in Flanders is slowly developing, largely as ‘learning by doing’. Evaluation of effectiveness is not supported with specialised, scientifically based and professionally managed evaluation departments. To explain this learning process in evaluation, and the integration of the additionality dimension, we distinguish the different levels of evaluation practice. Most common in IWT is the project evaluation. IWT is the one-stop-shop for subsidies to industrial R&D in Flanders. The general feature of the subsidy scheme is its bottom-up character: it is a permanently open and non-thematic scheme. IWT has a well-developed set of procedures for project evaluation, based on internal and external referees, to evaluate the ex-ante effectiveness of the project proposals (ex-post evaluation is starting up). Initially evaluation criteria were heavily focussed on the scientific qualities and technological risks of the project, but gradually the economic dimension (‘utilisation’) became equally important, reflecting the shift from an R&D towards an innovation policy focus. This economic evaluation doesn’t only concern the financial feasibility of the project or the commercial prospects for the innovating firm but also the economic return ‘for Flanders’. The most important economic criterion is that the project must be able to generate an economic value added that is at least ten times the value of the subsidy. During the evolution of this IWT evaluation practice also the ‘societal’ qualities of the project – mainly concerning environmental sustainable development - became part of the evaluation and selection procedure, although not on the same level as the two main axes. But because of the formal policy objective to increase the number of projects that support sustainable development, the latter criteria were formally separated from the standard project evaluation in a separate arrangement for ‘sustainable technological development’. This arrangement applies across the existing subsidy schemes and implies an additional evaluation of candidate projects on criteria of ‘eco-efficiency’. The evaluation gives access to extra support in the form of a priority ranking in the subsidy scheme concerned and of a financial bonus of 10% on the project budget. This evolution illustrates that project evaluation in Flanders is closely linked to general policy criteria in a bottom-up innovation policy design. IWT also has an important stake in programme evaluations, because the ‘self-evaluation’ by the management (up to now mostly ex-post evaluation) is an important input for the policy (re)design. Programmes (or support schemes) don’t have a thematic focus but are designed on the basis of characteristics of specific target groups. The ‘SME Innovation Programme’ is also a bottom-up scheme but supports in addition to research projects also feasibility studies and allows submitting a broader range of innovation costs to lower the threshold for innovation in SMEs. Another important programme is the ‘Strategic Basic Research’ Programme that is also non-thematic but organised by a yearly call to form consortia between universities and companies for longer term projects that are strategically important for Flanders from an economic and/or social point of view (also non-technological research is included). Programme evaluation is closely linked to the achievement of these nontechnological objectives. Innovation policy has developed a complete set of support schemes along the ‘innovation chain’ that are put available for bottom-up initiative. But it is felt that programme design has Behavioural Additionality Jan Larosse 4 16/02/16 to fit the different leverage needs and leverage potentials of various types of innovation activities and actors. Evaluation criteria try to discriminate access on that general level. What is still lacking in Flanders is an overall policy evaluation that takes into account the policy mix as a whole (the complementarity of the different programmes) and assesses the performance of the innovation system in an international perspective. This is a component of the organisation of the policy cycle that is not yet formalised, partly because the institutionalisation of the innovation system is not fully completed yet (a formal Ministry of Sciences and Innovation will be established in 2004). But the main reason is that the need for this was not urgently felt because of the strong immersion of policy objectives on the operational levels by means of differentiated ‘bottom-up’ programmes that give the initiative to the business and research actors. Formal top-down objectives of government for the innovation system were inexistent. But the official adoption of the ‘3% target’ will be a leverage to build more ‘strategic intelligence’ for policy making, starting from monitoring and evaluation. The coordination of efforts to promote R&D with the federal government, that still is competent for fiscal instruments to stimulate R&D, is another incentive to assess more systematically the composition of the policy mix and of the innovation system from a strategic viewpoint. In particular strategic research has to orient towards the kind of knowledge economy that has to be constructed for maintaining the welfare position. Guidance for strategic evaluation on choices to be made is not explicitly formulated yet, but without this implicit preferences and priorities shape the agenda. Therefore the question of evaluation has to be linked with explicit analysis of the innovation system and the role of government in improving performance. Evaluation practices in Flanders at this moment are rather piecemeal and pragmatic, e.g. evaluation of the public research institutes or the evaluation of instruments within their own organisational boundaries. They have a fragmented, organisationally narrow perspective and lack a firm theoretical basis. The bottom-up character of policies does not necessarily cause this, because also project evaluation can be put in a broader framework. The stated policy objectives (promoting innovation for SMEs; promoting cooperation between actors and technology transfer from research institutes in particular; promoting environmental technology) are topical and not positioned on a global evaluation scale of the performance of the innovation system and the strategy of government. Therefore the formal selection criteria that try to embody these policy objectives in stimulation programmes and projects can only partly cover the issues of innovation performance on system level. A broader evaluation approach is needed. 2.2. The new perspective of government on innovation Because of the limitations of actual evaluation practices it is of great importance to have a broader view on the behaviour of innovation actors in the perspective of the innovation system, for the evaluation of the success of policy instruments, be it on project, programme or overall policy levels. Firstly, just because there is more than what is captured by the stated policy objectives. More in particular, the effects of policies have to be considered as the result of an interaction between public strategies on the one hand and private strategies on the other; companies responding from their own point of view to the incentives of public instruments. Different instruments also combine mutually through specific interactions (policy mix). Therefore an ‘ex post’ evaluation of the global impact of government stimulation is needed that goes beyond a narrow (one organisation) incentive (one instrument) to effect (one objective) perspective. Policy instruments have to be evaluated as one among many variables that explain the performance of the targeted actors, and the same goes for the policy mix at the level of the system performance as a whole. Does policy ‘make a difference’? How to evaluate results of these interactions? Behavioural Additionality Jan Larosse 5 16/02/16 Secondly, the constrained nature of evaluation at project and actor level of a set of private intentions or results and related public objectives can obscure the broader range of social returns. These are not fully implied in the actual policy set-up. Innovation seldom is an individual activity but the result of complex interactive processes. Policy development is just beginning to discover and explore the huge ‘reserves’ of productivity gains that are linked to the better operation of the linkages in the innovation system. Innovation policy has to manage the innovation system from the point of view of the social return; Flemish innovation policy in particular has to devise a strategy to enhance the performance of the national innovation system to secure welfare creation for its citizens. This role of government as provider of new kinds of (soft) infrastructure for innovation actors has to be matched with appropriate evaluation models. The ambitions of evaluation policy are therefore closely linked with the appreciation of the role of government in present innovation systems. This discussion has become less ‘ideological’, in the sense of dogmatic models that a priori limit this role by virtue of the merits of mythical markets or take it for granted as the emanation of the general interest. The present day policy debates have converged towards the agenda of ‘sustainable growth’, the knowledge-based economy being the leverage for more immaterial types of consumption and for ‘ecological modernisation’ of production. The Flemish government has launched in 2000 a broad social debate (Colourful Flanders) and a legislation-crossing mission statement that resulted in 21 programme points (Pact of Vilvoorde) for the kind of society Flanders wants to build in the 21st Century. Governments are strategic actors in a largely consensual model of goal setting, starting with a given set of assets (history matters!) towards an open future (the reduction of uncertainty). The implementation of this agenda on EU level (the Lisbon goals) by national governments in a globalising economy stresses the abilities of societies for structural adaptation and the strategic role of innovation policies in this respect. In knowledge driven societies that are dependent on their institutional processing capacity of knowledge and information, the role of governments changes from a mere provider of framework conditions to an active and intelligent participant in knowledge creation and distribution; from redistribution of resources to co-creation of resources being one important actor among others in complex innovation processes (not denying the central role of business firms in innovation). How to evaluate the performance of such a government and such innovation policies? There is a need for ‘transition’ from traditional innovation policies (first and second generation) towards new kind of innovation policies (third generation) to keep government policies at pace with the growing demands of the innovation process. At the end a much higher degree of complexity has to be managed. Evaluating methodologies are an important part of the capacity to learn and evolve. But it is more difficult to ‘attribute’ effects to specific actions in a cumulative innovation process of which governments themselves are active participants. At the moment the proper tools for evaluating third generation policies (indicators, methods, organisational set-up) are not yet available. Therefore we need bridging concepts as ‘behavioural additionality’ that have the capacity to ‘internalise’ progressively the new dimensions of interaction between government and other innovation actors in an evaluation framework. On the one hand this ‘behavioural additionality’ is a concept that we know and use in traditional evaluation; on the other hand it exposes new properties that learn us more about the opportunities or limitations of actions by governments to stimulate innovation performance and transform innovation institutions and structures. Behavioural Additionality Jan Larosse 6 16/02/16 3. About the concept of Behavioural Additionality For any practitioner of innovation stimulation behavioural additionality has a great intuitive appeal. The feeling is that allocation of subsidies is more than just a money transfer. There are different levels of interactions that impact behaviour of the firms. The organisational interaction between the agency or administration and the receiver encourages the formalisation of the R&D process in firms. Innovation management is influenced by the necessities of external evaluation and this procedural approach may also advance IPR development. The implicit and explicit conditionalities of the granting agency may impact the course of development of the project as well as the further technology trajectory. Also many examples can be given of the importance of individual interactions of government advisors ‘doing their job’, with sometimes strategic impacts thanks to their expertise and vigilance. But often all these impacts are very ‘accidental’, because their potential is not systematically exploited. They are not visible in evaluation practices, because they have no real legitimacy and theoretical status. Theoretical arguments, mostly economic theories about the conditions for welfare enhancement, provide the broad background for policy design. It is difficult to define a theoretical statute for ‘behavioural additionality’ because of its emergence at the borderlines of an established theoretical framework that is rather closed. The original – and still prevailing – concept of ‘additionality’ is linked to ‘welfare economics’ and the rationale of public intervention in R&D (or knowledge production in general) because of ‘market failures’ (K. Arrow, 1962). Additionality than is the mirror of ‘substitution’ of private initiative, that means a distortion of free market equilibrium and of competition. Additionality equals ‘not substitution’, a negatively defined concept. Additionality is achieved if the coefficient of the subsidy variable in the econometric regression on private R&D investment is positive; ‘the more the better’. But we need to define the role of government in a positive way to better manage innovation systems. This role is more than just compensating under-investment in R&D without crowding out private financing, and has to be analytically decomposed into the relevant policy dimensions. We will see that this is done by linking behavioural additionality to the concept of ‘system failures’. But here theory has followed practice. The search for the ‘enlargement’ of the additionality concept has started with the rapprochement to the ‘subsidiarity’ concept, the division of labour between different governmental levels of intervention. In a study for the European Parliament (Additionality as a principle of European R&D Funding, Merit, 1995) Paul David et all introduced a “larger” definition (as opposed to the “narrow”, financial definition) for evaluation of additionality of a support instrument extended to ‘complementarity’ with support by other governmental levels. In Buisseret et all (What difference does it make? Additionality in the public support of R&D in large firms, IJTM, 1995) the adjective ‘behavioural’ is used for the first time. It describes new dimensions of evaluation of the impact of government support, beyond the “direct effect”, notably in the context of impact on firm strategy. The reference position is “what would have been taken place anyway”, which is a departure from the optimality theorem of welfare economics. But a thorough conceptual discussion has been lacking since, because the behavioural aspect was mainly used to provide an ‘extra’, in surplus of the well-accepted input and output additionality. This position cannot be maintained if the role of behavioural additionality in evaluation of effectiveness becomes more ambitious. Therefore an exploration of the Behavioural Additionality Jan Larosse 7 16/02/16 ‘limits’ of behavioural additionality as a bridging concept between neo-classical and system theoretic approaches becomes necessary. 3.1. ‘Additionality’ The intuitive use of ‘additionality’, in a broader sense than intended in comparative static welfare economics, is grounded in the common guideline ‘making a difference’: the role for government in improving social welfare. The assumption is that there are ‘hidden’ sources of productivity that are not tapped by private actors because the necessary institutional conditions are not gathered, and that these sources or opportunities are growing bigger in the knowledge economy. Knowledge is proclaimed a special type of economic good that has the extraordinary characteristic not to ‘wear out’ in consumption. On the contrary, in usage it grows. But there is a ‘myopic’ theoretical understanding of the functioning of the knowledge economy, focussed on single actors, that restricts policy learning. The way the ‘collective productivity’ of knowledge sharing can be organised in a market economy that is based on individual incentives is not well understood. One way to analyse this ‘hidden’ resource is with the micro-economic concept of ‘spillovers’ of individual economic activities. In welfare economics these spillovers are just disturbing the market, causing under-investment in knowledge production. But what if the productivity gains of these spillovers become a multiple of the productivity gains of the ‘owner’ of the knowledge (as is the case)? Than subsidies are not just legitimate to encourage the individual entrepreneur to take more risks, compensating the danger of imitation by free-riders or compensating bad performing capital markets to take risks. Than new institutional conditions are needed to encourage ‘co-opetition’ to fully reap the social profits of these spillovers. Additionality of subsidies becomes more complex because of this shift of focus. Another way of localising ‘hidden’ sources of knowledge productivity is to start the analysis the other way around, from the system level, to identify synergies in the system that are not fully used. Increasing the ‘connectivity’ of the system is another way of saying that cooperation in innovation has to be stimulated. Other accents are in solving ‘bottlenecks’ or ‘lock-ins’ and balancing ‘mismatches’ between different layers in the innovation system. Both the micro-economic and the systemic paradigm can be complementary in better understanding the conditions for unleashing these hidden resources of welfare production and making government contribute to arrange the institutional preconditions. Additionality therefore can be redefined in different ways for exploiting resources that are not exploited yet (additionalities are everywhere in a knowledge economy!), where a ‘positive’ action of government can make a difference. But there the problems for evaluation also start. In the conventional understanding of additionality there was at least a theoretical reference for the desired state of the system: the ‘optimal allocation’ of resources (Pareto-optimality, as a rule achieved by the free market and the price mechanism). In practice it is unknown how much government subsidies are sufficient to achieve such a hypothetical allocation in case of market failures. What is possible is to try to figure out a possible crowding-out or an additionality effect with ex-post econometric analysis. These econometric models are not comparing actual with Pareto efficient investment levels, but try to determine if subsidies have partly or completely substituted private investment efforts in the existing economic circumstances, what is not the same. ‘Doing better than what would have been without intervention’ only refers to a previous situation. A lot of problems arise also in modelling the determinants of private R&D investments, but still the econometric exercise is believed to be one of the only objective tests of input additionality. If we turn to a broader understanding of additionality in relation to ‘system failures’, the range of additionality potential widens but also looses normative ground in economic welfare Behavioural Additionality Jan Larosse 8 16/02/16 theory. For one thing, a system of innovation doesn’t only deal with knowledge markets but also with other institutional mechanisms to coordinate the production and distribution of knowledge. The science (sub-)system has its own functional logic that is referring to ‘truth’ as an operator and not ‘value’. And even in the processing of knowledge as an economic good, the market is one form of coordination alongside others: hierarchy (intra-firm coordination managed by ‘command’) and networks (inter-firm coordination build on ‘reciprocity’ in stead of ‘exchange’). The ‘optimal’ system cannot even theoretically be determined because of this institutional complexity and its non-linear dynamics. But the evolutionary theory can instruct us about the adaptation behaviour regarding internal and external pressures. In the knowledge society adaptation means learning. And important systemic deficiencies are ‘capacity failures’ of actors, which restrain their ability to adapt and renew. Another problem is ‘path-dependency’ of cumulative learning, because rigidities due to large past investment in existing knowledge infrastructure can cause ‘network failures’ or ‘lock-ins’ that prohibit innovation. Innovation incentives can be effective (‘additional’) in changing behaviour or inducing learning. But the norm of behaviour or desirability of certain outcomes is more a matter of strategic choice than the result of some technical optimalisation. ‘Making the difference’ becomes a kind of quality statement for governments that consider innovation incentives mainly as instruments for managing the innovation system as a ‘learning economy’. Competence building is a central objective. R&D investment is as much knowledge ‘creation capacity’ as ‘absorption capacity’, ability to learn about the already existing or elsewhere newly developed knowledge (which is a multitude of what one creates at his own). This perspective is still weakly developed in evaluation of the effectiveness of R&D subsidies. Behavioural additionality then comes into the spotlights. The concept of additionality has shifted from the narrow context of market failure as a condition for government to achieve hypothetical Pareto optimality (with the restriction not crowding out private initiative) to a context of system failure as a condition for government to improve the system ‘as it is’. This includes the reparation of ‘government failures’ (bureaucratic dysfunctions). Additionality is therefore continuous improvement of the effectiveness of the policy mix (and not the quest for the optimal state); the excursion of institutional learning by governments. But in last resort it must make a – measurable!difference on innovation performance. 3.2. ‘Behaviour’ Behavioural additionality is positioned between ‘input additionality’ and ‘output additionality’ to uncover the ‘black box’ of decision making within firms (the management of the innovation process) that brings about ‘allocations’ (R&D investments). What is of interest is not only to know that there is something ‘beyond’ the immediate impacts on input (R&D) and output (innovation), but to better understand how policy incentives work, such as to learn to improve them. The final effect of innovation policy is aimed at innovation output. Input additionality therefore is only interesting in so far there is a link with output additionality (only unproblematic in the linear innovation model!) and there is a lack of information to evaluate innovation output directly. The time lag between subsidy and innovation outcome is therefore a strong argument for input additionality evaluation. In this respect (improved) R&D investment or R&D cooperation are used as ‘proxies’ for output additionality. But this is not evident if output is not linearly linked to the preceding stages: output is not fully determined by the previous decisions. The number of causal factors increases coming near to the market; it is increasingly difficult to attribute the proper impact to the R&D-investment (let alone the R&D-subsidy). Many output solutions are possible because of the complexity of the process and because the innovation actors learn during the process. Innovation output can therefore be something totally different - by surprise – because of this unpredictable character. Therefore the behavioural additionalities, increasing the competences of the actors and the Behavioural Additionality Jan Larosse 9 16/02/16 conduciveness of the system, are the more important additionalities, because they create a generic capacity that has a lot of spillovers for further innovation. Behavioural impacts are enduring effects because they are internalised and reproduced in the behaviour of agents (not one-time effects) and have structural or institutional impacts on the system. The nature of the innovation behaviour is broader than market behaviour, because innovation systems are broader than markets. It concerns all kinds of social behaviour that affect the working of the innovation system. The behaviour is not limited to natural persons but describes rules of conduct or routines that also are embodied in institutions, e.g. firms (firm behaviour). Different institutions embody different forms of ‘routines’ that are sensitive to other kinds of incentives. Institutional theory describes how markets react to price signals, hierarchies to rules and commands, networks to shared convictions. Innovation policies will have to understand and use the strengths and weaknesses of these different institutional coordinating mechanisms to improve the ‘behaviour’ of the innovation system as a whole. Therefore ‘behaviour additionality’ can claim to be the more general concept that encompasses the other kinds of additionality that all have to do with specific kinds of behaviour, e.g. input additionality covering ‘investment behaviour’. Investment behaviour is also much broader than just the resource allocation in static markets. In dynamic perspective strategy development handling the uncertainties about future markets is very important: ‘portfolio composition behaviour’ is therefore an additional dimension in R&D investment that deserves all attention and is only revealed as a subject for improvement of policy making in so far it is made visible by new conceptual frameworks. Portfolio management theory and strategic management theory have to be ‘translated’ in the multi-institutional context of innovation systems. 3.3. Kinds of additionality Depending on the policy perspective different types of additionality can be distinguished in the literature. The question whether behavioural change is a dimension of these types of additionality or the reverse, that behavioural additionality is structured into these dimensions, is a matter of perspective. - From the point of view of the phases in the innovation process (the ‘arrow of time’), we can make the sequence of input, throughput (or internal process), and output. Sometimes behavioural additionality is narrowed to (internal) process additionality. - From the point of view of the organisational level of the impact, we can see different effects on project level, firm level and system level. - From the point of view of the types of productivity gains (or types of economies) we can distinguish additionalities of scale, scope and speed. - From the point of view of the interaction level of agencies with the innovation actors we can have operational and strategic additionalities, direct and indirect additionalities (spillover additionalities) It might be confusing to substantiate each specific behavioural additionality effect as a separate type (e.g. ‘cooperation additionality’, ‘portfolio composition additionality’), but in this stage of research the advantage of making them visible as subjects for policy action is more important. The more important ‘policy additionalities’, which can be achieved if government can have a socially beneficially impact on private strategies, are also a potential strategic area for making a difference, if the necessary conditions are present (horizontal innovation policy). Behavioural Additionality Jan Larosse 10 16/02/16 Behavioural Additionality in many dimensions Organisational Project level Company / Strategy Process level Input R&D-budget Scale Time Risc System Knowledge spillovers Portfolio composition Throughput - Internal - External Project management Innovation management IPR behaviour Synergy in proj. family Cooperation Alliances Networking / clustering Strategic autonomy Relations with VC/financiers Quality label Output Procesinnovation Productinnovation Training Localisation Strengthening core or differentiation Improved environmental impact Specialised knowledge Human capital / know-how 4. About the measurement of behavioural additionality The broadening of the additionality concept makes it more complex and multiplies the evaluation and measurement problems. The most known problems that have been identified are the problem of ‘counterfactual’ evaluation and the problem of ‘attribution’ of cause and effect. The first problem is inevitable because time is irreversible and conditions are never completely comparable between a situation with and without a subsidy. Control groups of firms without subsidies to measure the impact by comparison are very difficult to compose; the reasons for applying (or for denying) subsidies are linked with other characteristics that already differentiate these firms. The second problem is more general the problem of separating one impact – a subsidy - in a multi-causal phenomenon. These problems are not to be solved with some high-flying measurement technique, but the strategy can be to use complexity itself as a solution method, using the complementarity between different forms of partial solution. Measurement is objectivation without mystification! The evaluation of the effectiveness of innovation policies poses two kinds of specific evaluation challenges: the assessment of the behavioural additionality of instruments on the interactions in the system, and the assessment of the behavioural additionality on social return. This is because we consider the mutual reference of knowledge production (creation) Behavioural Additionality Jan Larosse 11 16/02/16 and knowledge consumption (diffusion and use) as being the main engine of cumulative growth of the knowledge economy The first item – interaction additionality - concerns the opening of the ‘black box’ of the interaction between agency and firms, in particular the effects of instruments on the decision processes and private strategies about R&D (one-time and enduring effects). The different types of behavioural effects of the instrument need to be identified and assessed for their learning leverage. This will preferably be in quantitative form in order to be related to innovation performance. But this evaluation finds also an end in itself, supporting policy design to improve the additionality enhancing qualities of the questioned instrument and the policy mix. How this can be done is by institutional learning in the form of real life ‘experiments’ and studies on the basis of hypothesis testing (thought experiments in policy debate or econometric tests). The use of the control group and of appropriate ways of ‘revealing’ the ‘true’ behaviour are important items for this assessment. The second item – spillover additionality - concerns the estimation of the contribution of policy (through various kinds of behavioural additionality) to the social return (that is the global outcome of the innovation process). Additionality (documented as behavioural changes induced by policy) appears as an extra social return. This means we first have to find better ways to know social return, the combined direct and indirect returns of investments (e.g. in cluster performance evaluation). The problems of disentangling total factor productivity illustrate the challenge of measuring the different components of growth. The challenge is to identify the effect of the behavioural changes on social returns (TFP). Starting with inventorying these changes and with tracing the spillover channels. A related problem in the evaluation of behavioural additionality in terms of social return is that because of the dual nature of knowledge as consumption and production good one has to beware for ‘double counting’. E.g. the ‘soft’ additionalities of learning are often the other side of the ‘hard’ additionalities of financial support if they simply come along without any specific additional intervention. If they would have occurred anyway they are the soft side of the same coin, while additionality targets additional coins! Another related problem is that negative additionalities of support (administrative burden and government failures) have to be subtracted. Again the comparative assessment with a control group or counterfactual scenario seems of great importance to understand magnitudes of additionality. In this assessment we have to be able to compare different instrument on their ‘relative’ and ‘absolute’ additionality. If additionality is relatively bigger in terms of behavioural effect, it is the absolute additionality in terms of innovation output that is the reference! A big change in behaviour in a small actor may result in less absolute additionality compared to a situation were the social return of smaller changes (e.g. in a cluster context) may have a broader basis. 5. Policy learning The potential impacts on the innovation system of new evaluation methods that focus on behavioural additionality are important, sometimes more important than the behavioural impacts of the existing policy instruments that they can reveal. The purpose of evaluation is indeed to change the policy design. In the case of behavioural additionality it is more and more policy practice that defies evaluation methodology. In different ways stimulation policies are internalising behavioural additionality strongly in the policy instruments. E.g. IWT has introduced the instrument of ‘feasibility study’ as a de facto obligatory precondition to apply for an ‘innovation project’; this way encouraging the development of innovation management skills and complementary knowledge building, in the firm and in cooperation with knowledge providers. Another strong example in Finland is the set-up of an integrated support infrastructure (INTRO) to stimulate Behavioural Additionality Jan Larosse 12 16/02/16 the pre-seed market by Sitra, the Finish public venture-capital fund. In cooperation with Tekes and a network of specialised actors it provides the complementary inputs (technological opportunity, finance, business experience, managerial skills) to launch new start-ups. Financing of business-plans or intermediation to find experienced people with commercial background: the behavioural additionality to the new entrepreneurs - but also to all complementary actors - is enormous. Also in less obvious cases there is a potential for improving the behavioural additionality in innovation processes resulting from government intervention, on the condition that policy makers consider each activity in a system context. In the short term this can result in incremental improvements in the existing programmes (what can be a leverage in awareness creation, selection procedures, contract negotiations to follow-up of mile-stones and commercialisation?), but in the longer run the consideration for the interaction effects of government and firm strategies are the basis of improving the system coherence and overviewing the dynamic properties of the system as a whole. Policy learning therefore has to take account of behavioural additionality. In Belgium the policy mix of R&D stimulation by fiscal and subsidy instruments is constrained by an institutional divide between federal and regional governments. A better understanding of the behavioural additionality properties of each instrument can help to find a better ‘division of labour’ and a complementary fine-tuning of the instruments, according differences in their behavioural additionality towards different types of firms. The introduction of new evaluation methodologies that consider behavioural additionality as a touching stone of effectiveness in a system perspective is a factor of knowledge intensification of the management of the innovation system itself and of a transition towards a third generation innovation policy. 6. Organisation of the work programme in Flanders IWT-Flanders is an operational organisation with limited resources for methodological work. The strategy is to cooperate with national and international partners to advance in the elaboration of evaluation methodology and of empirical evaluations of behavioural additionality. In the past the subject has been on the agenda in explicit initiatives such as international conferences. As early as 1996 a Conference was organised in the framework of the Six Countries Programme on the subject ‘R&D Subsidies at Stake: In Search of a Rationale for Public Funding of R&D’, where the discussion on ‘behavioural additionality’ as a rationale was introduced. In 2003 a follow-up conference on ‘Innovation Policy and Sustainable Development: Can Innovation Incentives make a Difference?’ advanced the new approach of behavioural additionality to incentives for innovation that enhances sustainable development. IWT practice also has been the breeding ground for the development of new policy (the arrangement for sustainable technology development) that goes in this direction. This is the reason that IWT can play a role as forerunner in different international initiatives (OECD, TAFTIE). Since 2002 the evaluation of additionality has been put on the agenda of the IWT Observatory by the IWT management. The working programme that is now being constructed tries to exploit the complementarities of different methodologies and partners in view of a comprehensive methodology for policy learning. - Exploratory and case study research. In 2003 an exploratory study was conducted to test a new survey instrument for collecting information of different types of additionality and their effects on performance. This was combined with a case study type of investigation of the differences in additionality potential of different classes of firms. Four categories of firms were targeted according the dimensions of scale and research-intensity (see B. Clarysse et al: Measuring Additionality of R&D Subsidies by Surveys: Towards an Evaluation Methodology for IWT-Studies). Behavioural Additionality Jan Larosse 13 16/02/16 A next step will be to extend this pilot to a large-scale survey and to construct a database of additionality characteristics for statistical analysis. Special attention will be devoted to the construction of a control group database. Structured interviews with firm management are a complementary instrument to deepen the understanding of internal and interaction processes that are important to assess additionality potential. - Econometric work In 2004 the ‘Support Point R&D Statistics’ at the Leuven University to examine the input additionality of R&D-SUBSIDIES will conduct a ‘classic’ econometric research project. The database of R&D and subsidy data will be enriched by all relevant data from the IWT database (e.g. on cooperations) to measure behavioural differences. The specification of the econometric model and the specification of the survey questionnaire will be coordinated to mutually benefit (improving the explanatory value with new behavioural variables). In this work international exchange of experience with other research groups or agencies is very valuable. To extend the empirical base for common work a similar data collection is very useful. - International comparative work Besides exchange of best practice in methodologies and the construction of a common empirical reference base, international comparative work is very important to assess the impact of different institutional set-ups in the organisation of innovation stimulation. Understanding the qualitative particularities that can explain certain strengths and weaknesses, in connection with quantitative results, is a very important way of policy learning. In this process an international community of practice is formed that can be an asset in actual national evaluation exercises. A natural extension of this cooperation could be the set-up of an international peer-review system of evaluators that is another complementary element in the evaluation set-up. - Internalisation in existing evaluation procedures The organisation of an ex-post evaluation is a cumbersome and costly exercise. It is important to make sure that the results are useful and used. The mandate for this kind of exercise coming from the policy makers is vital. The preparation and implementation of the evaluation has to involve the management and the advisors of IWT. They are also an important source of information. But the internalisation of this evaluation will have two directions. First, enhance the instruments and the organisation of the interaction process with the ‘clients’ in a more system-aware sense. Second, include new indicators for the evaluation of behavioural additionality as part of the existing evaluation procedures. This material will feed-in in the evaluation methods above described and increase the learning capacity of the IWT. Behavioural Additionality Jan Larosse 14 16/02/16