Once bitten, twice shy: The overgeneralization trap and epistemic learning after policy failure

Why do smart policy makers who try to learn from failure end up overgeneralizing these lessons when facing new crises? This article focuses on the learning that can come in wake of perceived failure, and consequences lesson has for diagnosing tackling subsequent crises. We argue, contrast much existing literature, result area is highly complex there are recognized experts have epistemic authority issue. At same time, tend become overgeneralized which a crisis extrapolate learned past with little consideration whether circumstances similar enough be applicable. Our argument assessed empirically using process-tracing analysis how EU responded Spanish banking 2012. Butz, Adam M., Michael P. Fix, Joshua L. Mitchell. 2015. “Policy Learning Diffusion Stand-Your-Ground Laws.” Politics & Policy 43(3): 347–77. https://doi.org/10.1111/polp.12116. Considine, Mark. “Applying Design Theory Public Policy.” 40(4): 704–24. http://onlinelibrary.wiley.com/doi/10.1111/j.1747-1346.2012.00372.x. Raile, Eric D., Amber N. Charles T. Salmon, Lori Ann Post. 2014. “Defining Will.” 42(1): 103–30. http://onlinelibrary.wiley.com/doi/10.1111/polp.12063. ¿Por qué los responsables políticos inteligentes que tratan de aprender del fracaso las políticas terminan generalizando en exceso estas lecciones cuando se enfrentan nuevas crisis? Este artículo centra el aprendizaje puede surgir raíz un político percibido y consecuencias tiene para diagnosticar abordar posteriores. Argumentamos, contraste con gran parte la literatura existente, resultar área es muy compleja hay expertos reconocidos tienen autoridad epistémica tema. Al mismo tiempo, tienden generalizarse formular una nueva extrapolan aprendidas pasado sin considerar si circunstancias son lo suficientemente similares como ser aplicables. Nuestro argumento evalúa empíricamente utilizando análisis seguimiento procesos cómo UE respondió bancaria española 为何那些试图从政策失败中吸取经验的政策制定者在面临新危机时最终会过度泛化这些经验?本文聚焦于在感知政策失败后可能出现的政策学习，以及经验对诊断和应对后续危机产生的结果。与许多现有文献相比，我们论证认为，当政策领域具有高度复杂性并且存在对该问题具有认知权威的公认专家时，政策失败可能会导致政策学习。与此同时，经验往往过于泛化，即面临新危机的决策者从政策失败中推断出经验，但很少考虑这些经验是否与新危机足够相似而具有适用性。通过对欧盟如何响应 2012 年西班牙银行业危机一事进行过程追踪分析，我们对论点进行了实证评估。 Policies often fail achieve their intended goals. Most cases not unequivocal, but instead subject intense political academic debates about actually “failed” (Blyth, 2013; Bovens 't Hart, 1996; McConnell, 2017). As result, seldom triggered by disputed (Bovens 1996). However, some exceptional cases, decisions made during followed such dramatic visible negative broad consensus emerges was failure. In well-meaning so they avoid making mistakes (Dunlop, Unfortunately, failures cast long shadow over understand future argue this face crisis, significant risk falling into an “overgeneralization trap” involves imposing abstracted causal case onto fit crisis. flawed problem diagnoses at hand, increase adoption ill-suited solutions.1 For example, response 2008 financial ensuing Euro-crisis its wake, governments across adopted austerity policies national level strict EU-level rules budgets. By late 2010s, widespread—though universal—acknowledgement had been contributed lower growth higher unemployment than would otherwise occurred. When countries faced economic due COVID-19 lockdowns March 2020, loomed large understood throughout chose adopt massive fiscal stimulus (“go big, go fast”), even more surprisingly, agreed unprecedented package paid through combination income joint debt (Smeets Beach, 2022). The neglect importance context drawing well established literatures learning, transfer, analogical reasoning public foreign (see e.g., Bardach, 2004; Brändström et al., DiMaggio Powell, 1983; Hood, 2010; Jervis, 1976; Stone, 2012). what less explored intent fall prey overgeneralization trap neglecting core contribution develop process explanation occur plausibility probe study. proceeds three steps. First, we present theory explains why after under conditions. Using recent advances cognitive psychology related as micro-foundations, trap. After contextualized drawn experts. disseminated broader community, stripped then draw without assessing supporting factors enabled it work case, fallen also discuss scope conditions expected tracing study methods used detail empirical observables expect find if chosen case. elaborate further method sources evidence Appendix A. Second, analyze worked theorized. Process studies select positive where cause could allow given operate both (Beach Pedersen, 2019). selected Irish collapse 2010 (recognized failure) were (i.e., technical politically sensitive). finds while initial expert causes focused specific moved policy-making circles transformed abstract “vicious circle” diagnosis particularities erupted spring 2012, quickly diagnosed based manifest structural differences sector. Based diagnosis, decided set motion creation instruments (banking nion) fixed collapse, hand. concludes assessment potential alternative/additional explanations Third, conclusions our literature learning. Given require considerable evidence, space reasons only analyzed one provide found other done trap, extrapolation evaluation suggestions engage systematic contextual before apply failures. section theoretical framework assess theory. article, define broadly “the updating beliefs lived or witnessed experiences, social interaction” (Dunlop Radaelli, 2013, p. 599). Epistemic type issues health “experts” 606). issue and/or experts, shaped interests between powerful actors (Birkland, 2006; Boin 2009; Dekker Hansén, Dunlop Resodihardjo, 2020; Weible want sincere efforts (Deverell, 2010), condition issues. Note does ideational paradigms mental models described scholars change Blyth, Hall, 1993; Jacobs, 2009). building processual model, relatively theorization (but see Birkland, Dunlop, 2017),2 processes whereby works 2018, S54),3 engaged search ideas fields reasoning, psychological research How triggers produce trap? Figure 1 describes multi-phase attempts explain linkage applied context. hypothesized trigger inter-crisis emergence widespread resulted failure; is, high salient costs (Howlett, 544). Recognition theorized create strong incentives expert-driven caused 5; 2015, 221, 2017).4 While resistant (Moyson 2017), strategic goal-oriented shared perception creates desire among crises trying Heikkila Gerlak, 2013). lead authorities within community starting Gerlak (2013) model collective distinguish translation phase individuals interpret information (here failure), dissemination individual-level knowledge becomes group members. following, use simpler term “lesson learning” phase, identify occurred including detailed information) did technical, should degree control production failed (Haas, 1992). disagreements clear learned. types most likely part different (Hall, 1993). Therefore, important exists shares overall paradigm. Where gets interesting travel members (other makers5). research, hypothesize dissemination. Gentner others (2009), Sagi (2012) particulars stored memory individuals, function tends away (Gentner 2009, pp. 1344–1345; effect, general expressed terms “if choice X, consequence Y.” Due salience will interest understanding restrict socially exist, movement individual primarily accept promulgating lesson. defined task-relevant expertise rest (Birnbaum Stegner, 1979; Bonaccio Dalal, 2006, 130). moves outside suggest critical bridge played think tanks policy-oriented academics specialize translating complex, “policy-relevant” knowledge, requires abstraction make them “relevant” situations (Bardach, 2004, 133). final occurs. discrete sets events threaten values, form action (or inaction) short period time (Boin 83–84; 2018; Houghton, 1998, 283–284; Rosenthal 1989).6 problems triggering difficult diagnose because “not instruction sheets” (Widmaier 2007, 748). Before tackled, therefore first figure out causing 2002). boundedly rational 2017; 489), rely short-cuts heuristics) help solutions (Brändström 193; Hacker, 2001; 489; 283–285; Schneider Ingram, 1988; Vertzberger, 1990). focus heuristic makers. analogies sense correspondence current situation past, focusing similarities Markman, 1997; 1344; Maravilla, 194).7 Instead vast number comparable converge 206; 1976, 220–221, 275; Khong, 1992, 36; Moynihan, 2006). Failures immediate resonate particularly strongly (Moynihan, ill-informed might superficial resemblances comparing merely being (e.g., crisis), linkages (X → Y) solution source target (Holyoak Thagard, similarity memory. strengthened innate tendency human mind better detecting 1993, 525–527; 1020). “case of” prior ignored. erroneous inferences happening fix made. good thing overgeneralized. dynamics opposite Jervis (1976) talks his classic perceptions misperceptions. talk overgeneralization, he argues “learn” events, strip those facets depend case-specific (Jervis, 228). He claims retrieval salient—but superficial—features 229). findings, claim fundamentally conditioned relationships produced original A element operationalizing activities associated left 2019; Bennett Checkel, 2015). Table depicts observables. More justifications probative value step-by-step each phases Irish/Spanish ill-fated decision government guarantee major domestic banks September chain led November forced ask International Monetary Fund (IMF) rescue package. governmental taken aftermath Lehman Brothers collapse. U.S. federal guarantees offer bailout, investment bank collapsed, destabilization global markets. Almost immediately, widely (Skeel, 2018). attempted providing several ailing two-year period. expire 2010, markets lost confidence provided state. spilled valuation state bonds, ten-year bonds reaching threshold 7% September. Because asset portfolio many guaranteed included holdings weakened position banks, ever-worsening loss investors. access bond markets, resulting asking EU-IMF 27, 2010. themselves, parliamentary investigation went wrong (Houses Oireachtas, 2016). (EU IMF) Mody Sandri, 2011). system compared government, distress guarantees. situations, vicious cycle spiral control, inability sufficient leading systemic sovereign But coupled identified Gray Ireland, turning actual “lesson.” evidence-based published two IMF economists 2011 (Mody working paper, demonstrates circle (2009) started manifesting itself Ireland course difficulties (Anglo Irish) vulnerabilities ultimately mutually reinforcing Simply put, take-home (X) circle—aka doom loop—that (Y). paper spend discussing avoided beyond sentence mentioning shown delays costly, prompt (although deliberate) needed” drastic interventions strengthen sector (Gray, 31). clearly delineates including: (1) post-Lehman environment, (2) continuing Greece destabilized (29–30), (3) ongoing nature size (21). Is (context) community? composed institutions (European Central Bank [ECB] European Commission) IMF, interacted very frequently (Interviews 2, 3, 4, 5, 9). There close connections Commission officials policy-relevant like Brussels-based Bruegel tank 9), along frequent interactions institutional fora OECD symposiums workshops. Within months publication Sandri (2011) X Y One key indicator authors 2010,” “solvency Anglo-Irish paying back ECB loans,” replaced “state” “banks” “financial systems.” Véron (2012, 2), wrote January that, “One interdependence features systems macroeconomic outcomes.” Pisani-Ferry 6) discusses “only extreme case,” claiming “all Western sovereigns euro “are heavily exposed having banks. Banks own securities.” words, particular potentially affect eurozone must break hurting To feedback loop need sharing borders system. Looking ahead, monetary union needs supported integration unified supervision, single resolution common backstop, deposit insurance fund. (Lagarde, 2012) section, demonstrate initially disagreement May/early June, Events early Spain sort announcement newly elected Rajoy 2 worse revision deficit 4.4% GDP 5.8% GDP. During period, sharp spike news coverage international media outlets read 2). still ambiguity of. Indeed, analysts saw point deficit-cutting measures downturn King, 2012; Smick, assessments suggested weak challenge eurozone, thereby Buti Padoan, instance, reason Moody's downgrade credit rating February increasing poor euro-wide prospects (Moody's, 2012a). began problem. mid-March, analyst commenting rise yields quoted saying, “That definitely concerns me. rally, helps banks' balance sheets. start rising hurts more. It (Milne, worsened, increasingly analogous CEO Bankia resigned, nationalize troubled May 7 injected another €19 billion 25. On 12 injection €28 billion. June 9, requested assistance recapitalize €100 loan debt, exposure. developing rates above 6% May, veer dangerously symbolically 7%-level 3). Rating agencies downgrad