Policy Learning and European Integration
本文基于2021年欧盟应对多重危机的实践，提出危机中的学习机制（推断性学习与偶然性学习）决定一体化路径，适合研究欧盟治理与政策变迁的学者判断是否阅读原文。
This was delivered as a lecture to the School of Transnational Governance, European University Institute, Florence, 25 November 2022. There are many ways to describe the European Union (EU) in 2021, but crisis and learning are particularly appropriate. Crisis management and crisis response have marked the whole of 2021 (and even more so 2022), to the point that some have correctly described the growth of crisis-related organizational features, tasks, missions and policies as indicators of the maturation of the EU as polity (Boin and Rhinard, 2022). ‘Crisis’ (in whatever morphology) is here to stay and has impacted on how the EU thinks and works (Mèrand, 2021). To accept, politically and in terms of policy instrumentation, the presence of different and often overlapping crises, and to generate policy responses is now a typical (as opposed to exceptional) ‘way of doing things’ in the EU. In the long term, this is also a ‘way of doing integration’ – a point that was not lost in classic theories of European integration (Ferrara and Kriesi, 2022). A crisis provides a formidable threat to integration. But it also creates a decision-making window of opportunity wider than normal-times windows. How the EU approaches this opportunity is crucial: it can be a disaster or a positive discontinuity, a leap into further integration. After all, ‘Europe will be forged in crisis’ famously wrote Jean Monnet in his Mémoires (1976). Can we then look at integration, to carry on with Monnet, as the cumulative effects of the responses to crises? If there is a connection between crisis and integration, it is reasonable to think that learning is a possible result of the ‘way of managing and responding to crises’. Back to 2021, the year in review seems to corroborate the learning-from-crisis argument. Crisis featured prominently: the turbulent start of the year in transatlantic relations with the attack on Capitol Hill, the total shake-up of orthodox economic paradigms about spending in deficit, the lethal incidents affecting migrants (43 people drowned in January only), the constant attempts to undermine the rule of law, compounded by the discussed leadership of Slovenia when this country assumed the EU Presidency in the second half, the troubled wake of Brexit, and of course the obstinate presence of the Covid-19 virus. At the same time, 2021 was the year of the pan-EU Covid-19 certificate and of the EU institutions coming together in the finalization of ambitious policies for ecological and digital recovery. Resilience is a key word in the EU vocabulary, making it clear that what matters is to withstand and respond effectively to shocks generated by a crisis. The year 2022 started with yet another round of facts pointing to crisis-learning mechanisms – the most important being of course the Ukrainian invasion. During the Euro crisis, it took the EU years to emerge with a new governance architecture to respond to speculative attacks on the sovereign debts of its Member States. With Covid-19, a coherent EU response took months to overcome the haphazard and uncoordinated initiatives of the Member States. In 2022, the EU learned in just a few weeks how to address the invasion and assist Ukraine, diplomatically and military. 1 In the second half of 2022, the EU was hit by the energy crisis – a crisis that can become existential if the ‘return to dirty energy’ significantly delays the plans for a green EU. 2 Yet again, the energy crisis triggered the attempt to forge closer co-operation amongst the Member States. Thus, crisis–learning–integration is, at least prima facie, an attractive triptych to frame the year in review. But in this JCMS Annual Lecture I wish to go further, and put 2021 in a broader, long-term perspective. We can look back at the last 15 years (at least) as a succession of multiple crises hitting different domains and sectors of the European Union (EU), often simultaneously. Think of refugees, the euro area, Brexit and Covid-19. This constant presence of the crisis has crystallized in the notion that EU policy making, and integration thereof, have become crisified (to borrow the terminology of Rhinard, 2019). Crises have not only become the new (empirical) normal of EU policy making, but one key explanatory variable of integration. We then should not be surprised to hear that, as mentioned, learning (or failure to learn, or, in yet another permutation, to learn dysfunctionally and therefore ‘learning to fail’; Dunlop, 2020) is an engine of integration in classic integration theories (Radaelli and Dunlop, 2013) and, beyond ‘the classics’, the failing forward theory (Jones et al., 2016). Between one crisis and the next, slow but historically big movements take place. This happens via spillovers (for neo-functionalists like Haas, 1958) or the decisions of the Court of Justice of the EU (for historical institutionalists like Pierson (1996), who can then detail the transition between t1 and t2 in terms of layering, conversion, and drifting (see Hacker et al., 2015). Modes of governance theories like the experimentalist approach (Zeitlin, 2016) argue that learning is a cause of policy change – and in turn, policy change impacts on the trajectory of integration. This lecture handles together the three elements of learning, crisis, and integration. I theorize the causal relationship with a focus on mechanisms of learning during the crisis as determinant of integration. The claim is that integration depends on how the EU learns in a crisis. The theoretical arguments in support of the claim are anchored on the micro-foundations of policy learning (often overlooked) and theories of crisis management. 3 Crucial to the claim is the argument about two very different mechanisms of learning. To approach our task of connecting learning, crisis and integration we start from mechanisms of policy learning. Learning (without the qualifier of policy) has a formidable pedigree as a theoretical lens in international relations, political science, and other disciplines like cognitive psychology and evolutionary economics. Even if we narrow down to the more specific domain of policy learning, we find an array of conceptual and empirical angles, pointing however to a core set of propositions (Dunlop et al., 2018). We are dealing with a reliatively robust tradition. Apart from being a respectable tradition, another advantage of reasoning in terms of policy learning is that we are not relying on an ad-hoc theory created and leveraged for the EU only. Recent work is discovering and acknowledging the potential of policy learning and its explanatory power when it comes to policy change in a crisis and between crises (Ladi and Tsarouhas, 2020; Serban, 2021). Further, policy learning is not necessarily limited to elites-level learning. We can indeed look at broader mechanisms of learning (Bomberg, 2007) in post-functionalist EU politics. Finally, following the recent literature, crises are not just exogenous shocks. They must be endogenous to the causal argument that learning determines the integration path. The remainder is organized as follows: the two mechanisms of inferential and contingent learning are the core of the Annual Lecture. After having explained and illustrated them in detail with reference to crises, I briefly move to learning between one crisis and the other. Next, I put some flesh on the bones of the claim, by considering the role of time, the relationship between type of crisis and type of learning, the wider political context in which elite-based learning takes place, and the effects of heuristics and misreading a crisis. The conclusions connect the thrust of the Lecture to theories of integration, showing how they contribute to the explanation of integration. Let us then examine policy learning – from now on, when I will only talk about learning in public policy processes, hence I will not need the qualifier ‘policy’. This section presents two different mechanisms of learning. One is inferential learning. This is customary in the political economy/policy analysis literature. Inferential learning has indeed a well-respected tradition in public policy analysis – one can even talk of a family tree that goes back one century (Dunlop et al., 2018, Chapter 1). Inferential learning has significant, fruitful conceptual overlap with political economy and neo-institutionalist theories of change (Hall, 1993). Here we are concerned with the link between policy learning and policy change. The public policy literature of the 1990s has explored this link in depth (Bennett and Howlett, 1992; May, 1992). I argue that inferential learning is a suitable conceptualization of learning for slow-burning crises. The other mechanism (contingent learning) was first identified and strongly corroborated outside political science – yet recently embraced within political science by Kamkhaji and Radaelli (2017). This mechanism applies to fast-burning crises. Micro-foundations must be endogenous to our explanation of learning. They cannot be assumed away without going into excessively abstract arguments. At the micro-level of inferential learning, we find an individual, or organization, that learns for two reasons. One is the repeated failure of existing dominant policy beliefs. This allows individuals and organizations to see in very concrete ways that a given policy (or a wider paradigm that informs a certain type of policies) is manifestly wrong. Experience with repeated failure shows that the cause-and-effect chain presupposed by the policy does not work. The lesson of experience, however, is not sufficient. The second reason is the presence of a viable alternative, visible in the public discourse as a set of relatively coherent set of policy beliefs. In other words, there must be another way of doing things that emerges as a feasible alternative, at the level of elites and in terms of public acceptance. Otherwise, individuals and organizations do not have a policy-making compass when they dispose of their old beliefs. The third factor is the role of agents of learning (an being that move learning to the and In the a classic of inferential learning is the of and about and The that in the and to was the policy of and the that the of in the economy cannot and in the long The of this was possible by the repeated failure of policies and the presence of economic policy by that the crisis of policies was not it was a slow-burning crisis that in the of the There was to see it and on the by The presence of certain of the and of from slow-burning crises. to an an important of inferential learning was the as with the power to the for change. The is, in terms of a to learning to the of (as by The change from learning was the to a In their and have the of how learning this policy change the repeated of the old the of a new and the presence of agents of learning. 1 shows the causal of the micro-foundations in inferential learning (for this reason will find the but can also be an is the of individuals (and in in the look at the policies that have to a crisis, or have in their they to change. change is therefore by from to contingent learning, the in EU in the on the crisis by and the attacks on the sovereign debts in the euro and this we are 2021, we will need to go back in a to this mechanism with a concrete the EU that the micro-foundations of this mechanism have by and many empirical in cognitive psychology and (for see and The decision-making and policy change that EU economic governance during the crisis of the euro area, and particularly in the years and not from inferential learning. There was for from repeated paradigms with and there many different different agents (in to the key and policy of the learning in from the individual, but this we have an in a context of The crisis does not clear is for to be is also not clear what the of the crisis the in a first EU not to cause-and-effect They not have the cognitive and and to reason about the of the of the policy for the of this was not there was a first response in terms of In for the EU this was a fast-burning crisis. to but clear set of to and the of and on public and (as the of at that and in terms of their policy on how to respond to the crisis. of their in a clear to – if there was on to look for and amongst the EU institutions on who should on the the the to be to as of the crisis. The European and the by between of the Member States. The European as the with the In terms of the micro-foundations described in cognitive psychology (see the in Kamkhaji and the EU was in one of the or many in But this time, of a of in some on contingent learning, even old – and we EU and the EU as was certain to to given the of the speculative was therefore a But what should be not be by policy beliefs. the EU in different responding to the by by 2020) in a classic EU not change their as in the empirical of the cognitive of the 2015). The learning mechanism to change was was going back to our a was with the of responses and a governance architecture to the euro area, individuals and institutions to on what possible that was in this chain of key was the to correctly the by the and, how is and learned in the of is is this of the that we can talk about learning as that goes on in the of A to a of and inferential mechanisms of learning back into the to policy change into a coherent new at the of the we are policy The have into who can the learned and the on the of which they this of the EU in the during and crisis has a long even in 2022. is we find the elements of a contingent learning If this things as The of the crisis was what we that the responses a level of the EU. to for was not clear at and 2021). some was and policy learning in the of policies and and connecting with what we just about the euro crisis, learning from Covid-19 was not limited to the domain of to of policies with the and (in the of the plans of the Member States. the EU learns in a crisis what is learned in the crisis. We this in the To a lens to the trajectory of integration, we cannot look at learning by that is, at a crisis without considering what the mechanism and of learning to the EU when the crisis and the connection I just between the and the euro They approach crises as of they the effects of the Covid-19 crisis on is that the by the on the Euro as a result of the need to the European economy the is another of contingent learning. this of contingent learning not a of paradigm and like the one that the of the sovereign crisis. on the learned of the crisis, the EU to generate change in the and put has and learning in the domain of governance of the euro the learning In the contingent learning to be contingent in the of than mechanisms of learning to historically and other. This the learning mechanisms are cumulative and even EU integration can be as a of learning This is the conceptual from which the claim in I that is, EU integration is the result of the succession of different mechanisms of learning. that we have the of contingent and inferential learning, it is to to the in terms of EU integration, between crises? A is that, as mentioned, if crises are the new normal and the and trajectory of integration, they cannot be exogenous to theoretical argument. We cannot just them as to the theoretical is about the of learning. I that the EU can learn in inferential or contingent – or first in contingent when is and in inferential But are of learning. we have to for the that the mechanism is and the of learning is or the Finally, the third is about there can be learning, but what about its and learning is possible in a post-functionalist With we can start our will see in a it a like a with identified by the of learning and politics. To with time, not second of EU is a crisis. The crisis can be can emerge from outside the EU or the EU. can but politically can In if there is a crisis, we must what type of crisis it is and its – that the crisis. They can a fast-burning crisis as in which the learning will be at There also be about what type of crisis is hitting the EU – the of a crisis can an to the to another between one crisis and the to policy change takes during a crisis, it will not In this of (and more beyond the crisis. They stay there during the crisis is et on in their to the crisis of the euro In 2021, they their approach to crises and that the of integration (Jones et al., 2021). approach has the of how one crisis to the a is during a crisis, it the crisis, but it was not an the policy change that to the crisis at t1 will be the cause of a crisis at In the I the causal link is one of the possible ways to connect crises in EU have on the long-term effects of the slow of decisions in the of the policy like Pierson an that is to inferential learning. But in this historical described by the EU institutions do not necessarily and learn they learn that they in a that has but significantly for by decisions by the Court of Justice of the EU. they can indeed first learn and then what they have learned The learned how to the the of decisions of the Court of Justice a visible and of integration. two forward and historical to the more we are on, we that the of a crisis determines the type of learning that is fast-burning crises contingent learning, inferential learning is not possible if not at the of the as a result of contingent learning and Inferential learning, is more to in slow-burning crises. There is also a third that learning in one or the other type of crisis or If this is the there will be integration, or a of crisis. is a of The EU is not to and wider integration. Let us now that there was learning at If is the lesson learned will be and for the crisis. will not away – it is as a Learning can be in a new policy like a policy as the and an (for the or, a new set of dominant and policy core on the digital and ecological The reference to learned in the crisis’ a of possible empirical to and that t2 is a new crisis. If the of correctly the we can see them learning. This is the of from the of the euro and for Covid-19 responses One that learning at t2 with Covid-19 possible a new of learning, with the and Resilience and in find learning in this succession of The new is is of the learning There is a second however, and to see this we factor in politics. learned by elites are politically The lesson in terms of cause-and-effect about policy at But it be feasible or in the political of This be the t1 lesson was an that shows its the new of But this can also be of political described by post-functionalist of integration and 2019). can narrow down significantly the political for the learned to be by and the learned about in the EU. of the rule of and about the of in in the Member have the year in on to or not to some In 2021, not the trajectory for and we can reason that post-functionalist can the EU into an existential crisis. If that happens and things the of can only contingent learning at they are or not with contingent learning at t2 will the integration or into At the of the the political can the lesson learned at t1 like about a in and the of can for further integration, if are by and in the following and elites learn from their and and this their as as their integration and this can take a of its and Let us now a third that t1 was an of learning in a crisis. The lesson is and politically the t2 crisis is not The of EU can be by can be The EU the crisis by the from the The two by to an but the between the at t1 in in a by and the at t2 in created by the and the is with many was in to are not the policy in the we have also an to to at at a when it is not yet – and be for some – for people to to work. The for is The task for now is – public support at the so that can their 2021). In the Member correctly first and then at the EU level et al., – elements of via with policy and to the on policy see and 2022). not to the EU this in terms of the conceptual I have they must be This Annual Lecture has a of European integration in the of crisis, learning, and integration. The claim I is a that of the empirical of the tree illustrated and as to learning for the EU as a whole can integration, and if what of et al., 2022). With this this has the of being on two mechanisms of learning and how they connect crises. does not the role of EU institutions and the Member it is on policy learning than on The arguments the claim are not The can be or integration – it depends on what mechanism of learning in a crisis and how learning (or does not between crises. of post-functionalist and failing forward can be in this The two mechanisms of learning are not They to other. A inferential learning type applies only to slow-burning crises, when there is to But the of EU crises has if the year in we look at the and created by the in have to contingent learning in to inferential learning. crises of different are not necessarily of learning (Ladi and Tsarouhas, 2020) – being in the of the euro crisis and to for Covid-19. as as and learned in the are institutions also The EU is not to the and policy of and 2019). Even when there is the will to the new crisis. that are in with an are more than But what about if we are in the as with the and at the of the euro The word is on one is to learn at the level of another is to and support for the The in and spending of the EU like a (Ladi and Tsarouhas, this change and the governance architecture they are not and that there are here to stay beyond the of the for a of the new of the EU 2022). This in with the political trajectory of the on the of also with the political on EU and how to a mechanism of of this 2022). The of policy learning is the conceptual the of learning and the of the of this was delivered to the on 2022 and to the for the of and 2022. I wish to first of Kamkhaji with I the years on inferential and contingent learning. to the to the and who to for the at of the Finally, I wish to the formidable by the School of Transnational and the wider by European University within the