How are accountability and organisational learning related? A study of German bilateral development cooperation

We analyse qualitative data collected from employees at Germany’s two main international development organisations, Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) and Kreditanstalt für Wiederaufbau (KfW) Development Bank, to study how upward accountability and organisational learning interact in the world’s second largest foreign aid system. Goffman’s ‘staging’ heuristic is applied to unpack social practices in these two organisations. We find that employees navigate two separate domains, a frontstage and a backstage. They consider the federal bureaucracy an audience expecting a coherent storyline despite the messy realities of foreign aid. In response, they engage in impression management on a frontstage while shielding their backstages from scrutiny to maximise autonomy. As a result, organisational learning at GIZ and KfW in Goffman’s terms focuses on collective efficacy at satisfying accountability expectations through staged performances. We relate these insights to the hierarchical structure of Germany’s foreign aid system, the role of organisational interests and prevailing professional norms.


Introduction
There has been a convergence among international development researchers in recent years that 'accountability' and 'organisational learning' both seem critical for development agencies to achieve impact.Yet, there is disagreement on whether and how accountability and organisational learning condition each other.Some scholars see them as complementary and mutually reinforcing (Picciotto, 2018) and propose that rigorous evaluations benefit both (Clements, 2020).Others have pointed to trade-offs (Boyle and Tyrrell, 2021) and express scepticism, worrying that accountability requirements stifle organisational learning.In their view, the accountability and learning 'seesaw' needs to be rebalanced towards learning by reducing central oversight and decentralising decision-making autonomy (Honig, 2015;Kogen, 2018).
This ongoing debate on the relationship between accountability and organisational learning has been informed chiefly by an Anglo-American lens.For instance, Picciotto (2018: 369) quotes the former Director General of the World Bank's Independent Evaluation Group claiming that 'as an evaluator for over 20 years, I have never struggled to reconcile accountability and learning.For me, they are two sides of the same coin'.Similarly, in December 2018, the United States Agency for International Development (USAID) agency tweeted, ' [W]e see accountability and #learning as linked and mutually supportive' (USAID, 2018).Analogously, on its USAID Learning Lab (2023) website, the agency states, 'You don't have to choose between accountability and learning -they go hand in hand'.
The contrasting view similarly reflects primarily US and UK perspectives.Kogen (2018: 99) argues that 'accountability and learning have been inappropriately conflated in the documentation of two of the world's major donors -USAID and the UK's former Department for International Development (DFID) -stemming from the assumption that accountability, by definition, improves projects'.Honig (2019) compares eight USAID and DFID-funded projects to explain statistical variation in evaluation results and argues that accountability based on quantitative reporting can limit project-level adaptation through learning.Springer (2021: 11) warns that 'M&E [monitoring and evaluation] systems take on managerial powers' that crowd out organisational learning.Hoey (2015: 9; emphasis added) criticises such managerial 'overreach' in the US context, lamenting that performance indicators required by USAID 'encourage a focus on accountability . . .rather than learning'.
A third, reconciliatory position holds that accountability and organisational learning need to be conceptualised differently.Lumino and Gambardella (2020) criticise that existing evaluation research frames them too narrowly.Instead, they propose to think of accountability 'as a relational and multidirectional concept' where each accountability relationship 'generates a specific form of learning' (Lumino and Gambardella, 2020: 149).Gujit argues that if development organisations were less focused on assigning short-term responsibility and had more of a long-term perspective, there would no longer be a 'tug-of-war between learning and accountability' (Gujit, 2020(Gujit, [2010]]: 277).However, no matter how these different positions view the relationship between accountability and organisational learning, they agree that the two influence each other directly.
In this article, we analyse in-depth interview data collected from 15 employees of Germany's two largest state-sponsored implementing organisations to study how upward accountability and organisational learning interact in the world's second largest donor system (BMZ, 2022;OECD, 2021).While taking concerns about limited organisational learning in German development cooperation seriously (Dörrbecker, 2023), our research seeks to neither assess German development cooperation's evaluation capacity nor judge its quality.Instead, we apply existing definitions of accountability as comprising answerability as well as enforceability (Fox, 2007;Goetz and Jenkins, 2005;Schedler, 1999) to a case of interorganisational hierarchies.'Upward accountability' in this context denotes answerability of the two implementing organisations to their main principal, the Federal Ministry for Economic Cooperation and Development (BMZ) to enable the latter to enforce rules and, potentially, also foster learning.
To study learning within international development organisations (Yanguas, 2021), we focus on the detection and correction of errors or problems to better achieve organisational goals (Argyris and Schön, 1978;Visser, 2016) through transforming individual knowledge into organisational knowledge.We understand such organisational learning as a 'process through which organizations change or modify their mental models, rules, processes or knowledge, maintaining or improving their performance' (Chiva et al., 2014: 689).Importantly, we regard 'organizational learning as both a situated and a social process' in which politics and power relations directly mediate interpretative processes (Vince et al., 2002: 2).In our case, the overarching goal of both organisations is to deliver bilateral foreign aid objectives ranging from standardised project implementation to promoting social change.Organisational learning in this context comprises situated and social processes of transforming individual experience into organisational knowledge on development practice.Related scholarship has so far not focused on the nexus of organisational learning and accountability, with Christian's (2020) work on internal criticism being a notable exception.
Unlike other bilateral foreign aid donors, Germany relies heavily on bilateral technical assistance, with an average of 35 per cent of its bilateral aid budget between 2013 and 2021 1 being spent on this modality.Germany also sets itself apart from other donors through a unique governance structure.The Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) is the BMZ's primary implementer for technical cooperation, and Kreditanstalt für Wiederaufbau (KfW) Development Bank is the principal administrator of German bilateral financial aid.For each organisation, its annual allocation from BMZ has more than doubled over the last decade.For GIZ, it amounted to €3.4b in 2020; for KfW, €4.1b. 2 Notably, BMZ's allocations to GIZ and KfW are not publicly tendered.The main treaty ('Generalvertrag') and articles of association ('Gesellschaftsvertrag') governing BMZ and GIZ relations are not publicly accessible, and BMZ rejected a recent formal parliamentary inquiry about their contents (Deutscher Bundestag, 2019).BMZ also restricted public access to GIZ's independent audit reports after one of Germany's opposition parties referenced GIZ's 2017 audit in its criticism of federal development policy.
From an institutional perspective, GIZ and KfW thus form a non-competitive oligopoly of state-sponsored implementing organisations operating under limited public transparency requirements.Both organisations frequently work with subcontractors who are directly accountable to GIZ and KfW.Different from these subcontractors, however, GIZ and KfW receive by far the highest allocations and do not compete with other organisations at their level.They also receive binding policy guidance from BMZ, a ministry independent from the German foreign office that oversees the majority of Germany's foreign aid.This setup of one ministry with two major implementing organisations functioning as 'accountability consolidators' differentiates it from the United Kingdom and the United States where development policy is set in the foreign office or by a subordinate agency and where implementing organisations that compete against each other are directly accountable to the political principal (Dietrich, 2021: 99;cf. Calleja, 2017;Esser, 2018). 3The positionality of GIZ and KfW vis-àvis BMZ is thus more central to understanding interactions between accountability and learning in the German system than accountability relationships between these two 'consolidators' and their subcontractors.
Considering this peculiar context, we wonder how upward accountability and organisational learning are related specifically at GIZ and KfW.Are they mutually reinforcing, does accountability crowd out learning, is there a need for reconceptualisation, or do none of these apply?In developing our response, we seek to enrich ongoing debates in public administration and development studies through an organisational sociology lens.We do not intend to produce a representative assessment of the quality of German bilateral development cooperation.Instead, our exploratory approach helps us to unpack social practices of upward accountability and organisational learning as we put empirical findings from an under-researched setting into conversation with evaluation scholarship.

New public management scholarship: Context-dependent interactions
In public administration scholarship, accountability and organisational learning are viewed as interrelated, ranging from mutually supportive to inverse.In this field, New Public Management (NPM) has for decades constituted the orthodoxy that scholars have been wrestling with (Brinkerhoff and Brinkerhoff, 2015;Shahan et al., 2021).NPM is understood to promote 'the adoption by public organizations of the management and organizational forms used by private companies' (Laegreid, 2014: 324) and is credited with producing 'a shift from a bureaucratic ethos of office to a managerial regime' (Laegreid, 2014: 324).In such a managerial regime, 'accountability and learning are both part of the same effort to monitor and evaluate performance' (Greiling and Halachmi, 2013: 387) and are fostered through incentives.
Working with these conceptualisations raises the question of how the NPM literature balances accountability and organisational learning objectives.Greiling and Halachmi (2013: 382), for instance, have argued that learning opportunities embedded in accountability processes are 'more important than having a great number of accountability arrangements that generate the same kind of information again and again to establish blame that cannot be contested'.At the same time, they stress that learning does not automatically follow from accountability (Greiling and Halachmi, 2013: 388;cf. Gujit, 2020cf. Gujit, [2010]]).The main reason is that 'the kind of data needed to establish accountability are different from what is needed to entice organizational learning or because [of] the defensive posture of agencies' (Greiling and Halachmi, 2013: 388).
Other NPM scholars have framed the relationship between accountability and organisational learning as competitive in that 'some accountability arrangements can actually hinder organizational learning' (Ebrahim, 2005: 57).They posit 'that there is an inverse rather than positive relationship between established accountability mechanisms and organizational learning' (Schillemans et al., 2013: 410).If so, then organisations face a dilemma of having to choose between the two.Yet this proposition of a rival relationship has been challenged by scholars who hold that organisational learning hinges on how accountability is conceptualised.Jarvis (2015), for example, suggests that their relationship is multi-layered and differentiates between four 'purposes for accountability', of which organisational learning is only one.
He also demands recognition 'that error-free administration is simply not possible' (Jarvis, 2015: 459).
In addition to highlighting different purposes, NPM scholars differentiate between different audiences for accountability.Broadbent and Laughlin (2003: 3) have introduced the 'distinction between "managerial" and "political/public" forms of accountability'.For them, political accountability denotes duties and incentives for office holders to explain decisions and outcomes 'downwards' to the public.In contrast, in the context of managerial forms of accountability there is an assumption that the person or being who delegates responsibility (often referred to as a 'principal') to another (often called an 'agent') can and has power to exert pressure over the performance of the latter.(Broadbent and Laughlin, 2003: 3) Considering the institutional position of Germany's two largest implementing organisations, we focus the present inquiry on managerial ('upward') forms of accountability.Nonetheless, we are mindful that political and managerial accountabilities likely condition each other.Laegreid (2014: 9) warns that a 'pursuit of managerial accountability can exact a price in the shrinkage of a sense of political accountability'.Furthermore, focusing on strengthening top-down managerial accountability may not be appropriate for organisational settings that operate laterally rather than vertically (McNulty and Ferlie, 2004).A resulting analytical question for evaluation researchers is: What form(s) of accountability are observable in which specific organisational contexts, and how does their interaction affect organisational learning (if at all)?

Development studies: Accountability crowds out learning
The distinction between political and managerial accountability equally features in international development scholarship, albeit implicitly.In development cooperation, 'the lines connecting global institutions to the people affected by their projects and decisions are often so long and thin as to be considered negligible' (Hale, 2008: 76).Political accountability in donor countries presumes that taxpayers oversee aid budgets.In partner countries, it supposes that citizens get to assess the use of development funds by their governments.In practice, however, there is a 'broken feedback loop between recipients and decision-makers' (Martens, 2005: 643;McGee, 2013;cf. Springer, 2021).For development bureaucracies to bridge this gap, a managerial accountability architecture is required to establish oversight mechanisms within and between the organisations constituting a foreign aid system.Crucially, these mechanisms need to be built on data collection and analysis; they cannot rely solely on organisational reflexivity (Van Wijk et al., 2020).The intermediary position held by development bureaucracies then explains why research on accountability in the context of bilateral development cooperation should focus on them.
Although some international development scholars are confident that '[a]ccountability and learning go together' (Picciotto, 2018;cf. Pollard and Lindkvist, 2020), others have observed critically that accountability has been prioritised to the detriment of organisational learning.In line with former USAID administrator Natsios' (2010) diagnosis of a 'counter-bureaucracy' obsessed with measurement, Kogen (2018: 104) has expressed concern about 'a misplaced dependency on accountability' and argues, 'Accountability does not easily contribute to a larger conversation about aid, . . . it only provides legitimacy for funders and practitioners to keep pursuing current development projects' (Kogen, 2018: 105).
As a result, development researchers advocate disentangling the 'intricate web of accountabilities' (De Silva Godage, 2021: 7;cf. Esser and Ha, 2015;Wellens and Jegers, 2017) and, where possible, reducing requirements for managerial accountability to enable organisational learning.In this context, Honig and Gulrajani (2018: 72-73) have stressed the importance of autonomy as one key factor for improving aid agencies' performance.At the same time, they warn that 'autonomous agents are more capable of acting badly as a result of reduced constraints' (Honig and Gulrajani, 2018: 73).Recognition of the presumed benefits and risks of autonomy thus indicates that organisational learning is determined not only by incentives but also by social interactions within and between organisations.This is where both NPM scholarship and development studies intersect with sociological perspectives.

Organisational sociology: Accountability and learning as social phenomena
Instead of focusing on incentives, the sociological study of organisations foregrounds 'the techniques, procedures, tools, strategies, and methods the actors use when constructing and organizing a meaningful world' (Ho, 2012: 30).In this vein, accountability and learning are not necessarily related, let alone reinforcing.Whether accountability spurs or constrains organisational learning rather depends on social dynamics in specific organisational settings.Hence, sociological approaches to organisational learning focus on the social emergence of what Elder-Vass (2010: 122-130) terms 'norm circles', which in turn determine 'organizational roles' (Elder-Vass, 2010: 164-167;cf. Seidl and Mormann, 2014).Although these roles do not override organisational members' multiplex relations with other social units, such as family, religious groups, or political parties, they co-determine their behaviour (Elder-Vass, 2010: 159).As a result, organisational norms, both explicit and tacit, attain causal power.We therefore need to pay close attention to how socialisation and sanctioning works within organisations -through incentives but also social interaction.
The late Canadian sociologist Erving Goffman's frontstage/backstage heuristic is useful to study accountability and learning both in and among organisations (Goffman, 1959;Samra-Fredericks and Bargiela-Chiappini, 2008).Although Goffman's primary object of inquiry was intraorganisational interaction (Ringel, 2019: 710), global affairs scholars have since applied his theory to interorganisational behaviour.Andreas (2008) employs 'frontstage/ backstage' to analyse divergent actions by local and international actors during the siege of the Bosnia-Herzegovinian capital of Sarajevo (1992Sarajevo ( -1996)).Explaining the silence on failed projects in development organisations' reports to donors, Nguyen (Ho, 2012: 60) points to the 'frontstage' where organisations work to avoid 'signs of inconsistency [that] can weaken a group's reputation'.
Organisational researchers have adopted 'Goffmanesque staging' (Solomon et al., 2013: 196) to cover an array of organisational impression management and resulting limitations to accountability even though it does not travel effortlessly (cf.Fine and Hallett, 2014;Ho, 2012: 31).Desportes et al. (2019: 37) concede that '[t]he frontstage/backstage perspective has obvious analytical shortcomings in the study of aid'.Unlike the performances of a play or, as in Goffman's empirical settings, restaurants, casinos, hospitals and asylums (e.g.Smirl, 2006;Zaman, 2003), where interactions are one-off, stages are physically separated and the focus is on speech, foreign aid relations are continuous, stage boundaries are metaphorical and overlapping, and organisational presentations comprise performances as well as written documentation (cf.Hoyle and Wallace, 2008).
These limitations notwithstanding, we propose that applying the frontstage/backstage heuristic to the German setting is analytically insightful.Unlike Desportes et al. (2019), we investigate distinct realms denoting intra-organisational processes as the backstage, and interorganisational processes as the frontstage.We regard this separation of frontstage and backstage as made possible by the institutional context of our case, which is distinct from most setups in the field of international relations (cf.Kranke, 2020;Schimmelfennig, 2002;Van Der Veen, 2011).Similar to Solomon et al. (2013), we argue that both stages are important to create and perpetuate a perception of corporate accountability vis-à-vis public audiences.
Seen through this prism, organisational learning comprises collective efficacy at satisfying audiences' accountability expectations while engaging in effective backstage strategising.What some scholars have labelled as 'hypocrisy' (Brunsson, 2002) decouples talk and resulting decisions from actions and their impact (or lack thereof).It increases the degrees of freedom for the organisation in its quest for legitimacy and stronger positions in the market.In this scenario, an organisation does not turn to accountability to bolster its legitimacy and competitiveness, but to hypocrisy.Indeed, '[a]n organization that could not behave hypocritically would find it more difficult to work in a world of conflicts than one that could' (Brunsson, 2002: xv).Hoyle and Wallace (2008), too, point to organisational hypocrisy as a source of agency.Under these conditions, learning in organisations centres on effective decoupling of what is done from what is reported, which requires skills at framing and presenting past actions in ways that satisfy external accountability expectations without jeopardising future contracts.

Methodology
We treat Germany's non-competitive oligopoly of two state-sponsored implementing organisations as a single case.Since our inquiry is not evaluative but rather informs an ongoing debate among evaluation scholars, we study interactions within Germany's bilateral aid system instead of focusing on distinct characteristics of one specific organisation.We recognise that its particular institutional structure sets Germany apart from other national foreign aid donors, which invariably limits the generalisability of our findings.At the same time, focusing on such an outlier case can generate new insights to discussions that have been informed primarily by Anglo-American empirics.Single-case studies also have the distinct advantage of allowing for deep dives into the interplay of causal dynamics over time (e.g.Cumming, 2018;Essex, 2013;Hanrieder and Kamradt-Scott, 2017;Mitchell, 2014;cf. Pennings et al., 2006: 39-41).
We analyse qualitative data from 15 in-depth interviews.Interviewees were selected purposively with three objectives in mind: similar representation of both GIZ and KfW, inclusion of headquarters and in-country perspectives, and gender equality.Individuals were encouraged to speak in their own capacity as opposed to representing any particular job function or their organisation as a whole.All informants received confidentiality guarantees.Fourteen interviews were conducted in German and one in English, each by both authors who then transcribed each interview separately and translated all German data into English.Both authors compared and, where necessary, adjudicated their individual transcripts of each interview.All data were then anonymised and depersonalised.
Interviews followed the same format comprising an introductory explanation, five questions prompting respondents to provide examples of knowledge management, organisational learning and development effectiveness from their work experience, and a list of unstructured optional follow-up topics.The data on accountability and learning in interorganisational relationships between the implementing organisations and BMZ thus emerged inductively.Individual responses across both organisations revealed a high degree of uniformity and -to our surprise -lacked organisation-specific characteristics.We interpret this interorganisational conformity as confirmation of our systemic approach that combines interviews as one data set instead of slicing the data by organisation.
Consolidated interview transcripts were analysed in two steps.First, we screened all transcripts with a view to clustering the data thematically.We generated 15 analytical categories, 4 14 of which we considered relevant to the topic of accountability and organisational learning for development effectiveness.In a second step, we wrote summaries of each thematic section.These summaries form the basis of this article's data section along with illustrative quotes.All quotations are thus taken from cross-validated data and comply with confidentiality assurances by not being attributed to any particular respondent or organisation.We are mindful that our data represent subjective self-assessments and cannot be verified positivistically.
Our methodology has three limitations.First, our interview data are not representative of the two organisations since we selected interviewees purposively.We mitigate this limitation through a two-pronged approach: cross-validation by peer researchers and expert peer reviews prior to manuscript finalisation (Creswell and Miller, 2000).The four experts who provided comments on our analysis all possess firsthand experiences with the German foreign aid system.The second limitation lies in the number of interviews being relatively low.We justify our cut-off point with qualitative data saturation (Francis et al., 2009;Guest et al., 2006).Recent research on data saturation confirms 'that the stopping point for an inductive study is typically determined by the judgement and experience of researchers' (Guest et al., 2020: 14).Twelve of the aforementioned 15 analytical categories had emerged after our ninth interview.Only three additional categories resulted from interviews 10-12.The remaining three interviews did not generate any new analytical categories.
The third limitation concerns subject sensitivity pertaining to political and managerial accountability in the German context.Respondents in this sector can be reluctant to share critical insights.We are therefore grateful for the informants' openness in offering critical assessments of their respective organisations and the sector overall.We interpret their willingness to disclose information which, although not formally classified, might be considered sensitive as an indication that the interviewees themselves wish to contribute to meaningful reforms.We are positive that strict confidentiality and all interviewees being elites preclude negative repercussions.On this basis, we trust that our analysis can contribute to an improvement of both accountability and organisational learning at the systemic level without placing responsibility or blame on any individual or organisation.

Thematic analysis of interview data
In the empirical sections that follow, we do not associate any of the data with specific interviewees to prevent distinct voices from emerging (thus possibly violating anonymity assurances).Instead, we use broad indications ('all', 'most', 'several', 'few', 'none') to describe how prevalent certain observations are in the data.We also refrain from discussing the data or connecting them to relevant theories to allow for an empirical overview to emerge before we present our discussion in the subsequent section.

The institutional structure of German development cooperation
Interviewees classify the German system as 'quasi-monopolistic' or 'duopolistic'.Although the federal government formally maintains a tendering system, direct competition for contracts is reportedly a rare exception.Indirect competition between GIZ and KfW occurs insofar as financial and technical assistance allocations in the federal budget fluctuate, reflecting political negotiations.Nevertheless, several respondents point to internal pressure to 'be competitive', which results from GIZ's and KfW's self-images as market-oriented organisations.Analogously, management in both organisations is said to frequently stress the need to 'act economically'.Interviewees note that such orientation towards economic efficiency promotes 'thinking in silos' as it leads to 'competition at all levels: individually, between organisational units, but also between institutions'.In this context, knowledge constitutes a precious resource that must not be shared: As long as we compete to control the narrative, knowledge remains a resource that I don't give away -even to the BMZ, because they get suspicious that we only want to land the next contract.This causes real differences; it could work differently if there were more trust.But the BMZ is afraid of [our organisation] becoming too big and tries to counteract that with competition.This paradoxical perception of competition where almost none exists is, in part, due to the system's institutional structure.Interviewees describe their own organisations as 'de facto . . .subordinate authorities' of the BMZ or 'simply' the ministry's 'extended workbench'.They go as far as referring to themselves as 'quasi-civil servants' given their high job security.A few respondents refer to a 'golden cage culture'.For GIZ specifically, there is a 'triple linkage' to BMZ which is GIZ's 'main client, shareholder, and one of the BMZ undersecretaries sits on the supervisory board'.Although the governance setup of KfW Development Bank is less interwoven with BMZ, the ministry is the development bank's main client while also acting as its principal.
These structural characteristics affect how staff at both organisations perceive the quality and limitations of their professional relationship with the ministry.Across both organisations, several interviewees speak of their moral commitment to 'world improvement' (Weltverbesserung) and regard additional competition as neither viable nor necessary.The majority of interviewees also reflect self-confidently on their own technical knowledge, including the implementation of aid projects, and their ability to innovate.One interviewee quips, 'Some of us like to promote the myth of "we do not make mistakes, and we do not apologise"'.At the same time, staff seek to avoid conflict by filtering information before passing it to senior levels, a practice referred to by one respondent as 'the internal washing machine'.
In contrast, most respondents doubt that BMZ officials understand the 'realities' of development cooperation.The ministry is perceived by many to set 'high expectations [despite] little technical knowledge'.Interorganisational relations with 'the client BMZ' are characterised by 'tussle', 'low trust', 'political pressure' and a reluctance to consider external criticism.In interactions with the BMZ, too, 'the internal machine kicks in: How is the meeting managed and who participates? . . .These formalised communication channels render it difficult to talk openly with each other -they are institutional safeguards'.This perception is expressed in behaviour towards the BMZ whereby GIZ and KfW staff, according to one interviewee, act 'guardedly towards their clients, which raises expectations that we will not make any mistakes'.
Such aversion to admitting problems extends to the realm of political accountability.A few interviewees voice concern that taxpayers do not receive sufficient information about the reality of German foreign aid.They posit that implementing organisations are under limited public pressure to demonstrate results, as they report mainly to BMZ.Although two respondents argue that implementing organisations should be robust enough to handle criticism from the 'parliamentary arena', both organisations are characterised by many to act defensively and share as little information as possible.

Perspectives on accountability in Germany's development cooperation system
All interviewees view development cooperation as 'complex' and argue that it cannot be captured adequately in existing reporting formats.The German foreign aid system is portrayed as predicated on a 'control model' fixated on goals set ex-ante.Some criticise this contractual approach whereby 'you have to make a promise of impact in advance' instead of proceeding flexibly and adaptively.Many interviewees identify disbursement pressure as the overriding organisational priority.One of them summarises this pre-eminence of disbursement in contract management as a 'bureaucratic rule of three' comprising 'commitment -review -payment'.'The hard constraints are the contracts', and what allegedly matters most is that budgets are expended on time.The mechanism underlying this disbursement pressure is that 'we have a general contract with BMZ, and we receive compensation: the bigger the project, the higher the payment'.
Against this background, most interviewees mention 'accountability' only in the context of the relationship between implementing organisations and BMZ (i.e.managerial upward accountability).They report that demands for this form of accountability have increased in recent years due to low levels of trust and resulting scepticism.Several surmise that BMZ might be dissatisfied with the information it receives for accountability purposes.These interviewees assume a dynamic whereby some units in BMZ are 'driven more towards accountability, by the fear that [we in the implementing organisation] will end up doing what [we] want.They [BMZ] see us as not controllable and say, "the effects are not happening"'.In these interviewees' perceptions, accountability demands of BMZ have been increasing.
According to a few interviewees, the ministry sets objectives that are motivated politically: 'The BMZ usually overloads the logframes [logical frameworks, a tool to plan and manage aid projects] with political bullshit'.Implementing organisations purportedly find themselves in perpetual negotiations with BMZ to reconcile political expectations with the 'reality on the ground'.Coincidentally, several respondents view existing indicators as problematic because they regard them as focusing attention on 'cost-benefit monitoring' and 'value-for-money'.These priorities allegedly spur a dynamic where 'a lot of resources are used so that the minister has numbers', but the latter do not represent what projects do (or fail to) 'achieve on the ground'.
In addition, several interviewees describe BMZ's managerial accountability requirements as 'too formalised . .., too structured, too set in stone', dominated by hierarchical needs within the ministry.As a result, reports amount to a 'ritual', and 'nothing can be learned' from them.BMZ is alleged to 'always want the reports' even though its staff 'don't know what they want with them'.Several interviewees also share the suspicion that BMZ officials 'don't read any reports at all'.In these interviewees' individual perceptions, BMZ sets priorities but does not have the ability to follow through.Based on such scepticism, the majority of interviewees advocate reducing managerial reporting requirements and favour less stringent oversight by BMZ.
In contrast, only a few interviewees concede that while the ministry should embrace that foreign aid is inherently unpredictable, implementing organisations should in turn improve the rigour and quality of their reporting.One interviewee explains, Our indicators and logframes are often wrong and incomprehensible.I think that in about 50% of the projects, we change at least one of the outcome indicators halfway through the project cycle.But even then, it is difficult to map outcomes.We can really only triangulate.This observation is corroborated by another interviewee positing, 'If the BMZ accepts -in some cases at least -poor logframes we have drawn up, then our staff say, "Why should I even make an effort?"'Accordingly, implementing organisations' internal quality assurance mechanisms are perceived as unsuited to enhance upward accountability towards BMZ.Instead, units in charge of quality control are said to focus on protecting organisational interests rather than scrutinising documents' internal validity.
On balance, most interviewees thus frame the accountability relationship between BMZ and implementing organisations as dysfunctional.They describe a vicious circle connecting increasing accountability demands with decreasing development impact: 'if there is no impact, there will be even more control and even more bureaucracy'.Another respondent warns that 'the wish for more control, more rules and more promises of impact inevitably clashes with reality'.As a result, several interviewees worry that these expectations could have the inverse effect: 'with even more indicators, we will be ending up with even less impact'.

Views on individual and organisational learning
Turning to views on learning, interviewees differentiate between different types and purposes.These include 'learning at the individual or organisational level', 'learning as knowledge acquisition or as context navigation' and 'learning for private or professional benefit'.At the individual level, learning is described as an arduous journey across different sources of information and training formats: 'Everyone . . .wastes a lot of time and effort at the start'.Most interviewees argue that knowledge acquisition has to focus on tacit knowledge as the most critical form of knowledge in development organisations.Therefore, informal formats such as conversations with experienced colleagues, individual research and overtime efforts play a pivotal role for learning.One respondent explains, I see organisational learning as a distance that information must travel.Ideally, it starts with evaluation and ends with implementation.However, this feedback loop between evaluation and use is the exception and rather a 'nice-to-have' for the organisation, that is, not indispensable.
A few interviewees share positive experiences of 'working across organisational boundaries' in communities of practice, for instance on methods, and of 'spontaneous situational learning'.Yet such examples remain the exception.Overall, respondents rarely mention successful organisational learning.One of them surmises that learning has 'probably taken place somewhere' but struggles to identify an example.Another interviewee reports feeling 'confident that a learning process exists, but we cannot prove it'.According to several interviewees, this elusiveness of organisational learning conditions how GIZ and KfW present themselves to BMZ and other external audiences: 'To the outside world, we always portray ourselves as better than we are'.Importantly, elusiveness in this context does not suggest an absence of organisational learning in the German bilateral aid system as such but reflects scant evidence of learning in the relationship of GIZ and KfW with BMZ.
The data contain multiple explanations for the elusiveness of organisational learning.These explanations comprise internal as well as systemic factors.Internally, supposedly innovative 'crazy ideas' by staff reportedly fall victim to the organisations' quality checks and are 'safeguarded to death'.The 'internal washing machine' is mentioned again; it reduces complexity and 'hinders learning for complex problem solving'.
Interviewees also link elusive organisational learning to interorganisational factors outside of their control.One interviewee stresses that there are 'no binding debriefings' in BMZfunded projects because 'as long as the next job or money do not depend on it, it's of no use anyway'.Other respondents mention time lags and friction losses between ex-post evaluations and the commissioning of subsequent projects as obstacles.Some even blame BMZ for a lack of organisational learning in GIZ and KfW, maintaining that pressure from the ministry to expend allocated budgets prevents learning in implementing organisations.One interviewee remarks that, as a result, 'the learning taking place in the implementing system is not compatible with the commissioning system'.The implied rationale suggests that learning on how to spend money crowds out learning on improving development impact.
For some interviewees, internal and external constraints to organisational learning are rooted in an inherently inverse relationship between accountability and learning.They explain this tension with the persistence of a bureaucratic administrative logic that prioritises standardisation and documentation for upward accountability.Accordingly, several interviewees draw a link to evaluations of German development cooperation, which allegedly remain 'focused on the needs of accountability' and are 'not geared towards learning'.
While only one interviewee explicitly calls for 'learning and accountability to be mutually supportive', several respondents offer suggestions on how to reconcile the two.One demands openness so to 'accept that not everything works 100 percent' as a basis for 'greater willingness to discuss failures jointly' between implementing organisations and the ministry.Another interviewee asks for 'less formalised and more regular exchanges at the working level between BMZ and implementing organisations beyond departmental boundaries'.Notably, a few interviewees offer positive anecdotes of 'open-ended exchanges among self-organised groups of motivated staff' and 'peer-backing for courageous pilot approaches'.Such formats are seen to promote mutual understanding and, over time, rebuild trust since 'co-learning does not take place when prejudices collide'.

Practices of Goffmanesque staging
Despite such constructive thinking, most interviewees describe the alleged tension between accountability and organisational learning in the German foreign aid system as unbridgeable.Several refer explicitly to a 'frontstage' where they perform accountability to satisfy bureaucratic reporting requirements, in contrast to GIZ's and KfW's 'backstage'.They allude to a decoupling of the bureaucratic realm 'on paper' from the supposed 'real work of development cooperation'.This practice allows for what one respondent terms 'defensive accountability' towards BMZ to maintain room for manoeuvre for project managers who need to 'play both games'.
On the frontstage, project managers try to 'keep their projects as simple as possible' and seek to 'fulfil their logframes, but always in parallel with the actual project life'.According to one respondent, No project is carried out as it is designed and written up. . . .We write stuff down linearly.Everyone should know that no project is done this way. . . .It happens in secret, so to speak.Hidden achievements cannot be evaluated; they cannot be presented.There is what happens on the stage, and then there is what happens backstage.
Even though the duality of front-and backstage produces 'high transaction costs [that] cannot be in the interest of the BRH [German Federal Auditor's Office] and accountability', several interviewees consider the ability to uphold the schism between the two stages a key indicator of managerial talent.One of them explains, From day one of the project, the manager has to figure out what can actually happen -also in light of bureaucratic requirements.She has to translate what in-country stakeholders need and how the project can fit into a format that she can report on.How good or bad is she at putting it on stagethat's an art.How does this kind of juggling work?How do you get the best possible impact despite compulsory documentation and reporting requirements?
Several interviewees share examples of how managers leverage the front-and backstage duality to their advantage.In line with critical assessments of BMZ's technical competencies, most of these respondents argue that BMZ officials regularly fail to detect quality deficiencies during project planning and implementation.Nonetheless, GIZ and KfW project managers 'in the field' try to maintain 'a good relationship with BMZ officers' to engage in 'realistic expectation management' through 'established channels, mostly informal'.In some cases, these relationships can be relied upon to adjust plans through direct, informal communication, for instance when crises -such as war, drought, recession -occur suddenly and require flexibility.In other cases, such relationships can reportedly be abused, for example when 'retroactive plausibilisation' of unmet targets is deployed to obscure lacking performance by altering previously agreed-upon objectives.

Discussion
The interview data suggest upward accountability by implementing organisations to BMZ as predominantly performative.GIZ and KfW formally comply with BMZ's information requests while guarding knowledge to protect their business.Applying the frontstage/backstage heuristic allows us to identify this behaviour as a case of 'Goffmanesque staging' (Solomon et al., 2013).Even though almost all interviewees report frustration about the presumed need to keep the frontstage performance going, it fulfils a self-protective function and is in fact regarded as indispensable.Such dualistic and seemingly hypocritical behaviour is an example of organisational decoupling of reporting from actions to improve legitimacy and competitiveness (Brunsson, 2002), although GIZ and KfW are not even competing in open markets.We propose that the observed 'hypocrisy' can therefore only be functional in terms of bolstering GIZ's and KfW's legitimacy in the eyes of BMZ.This finding differentiates the German case from Anglo-American contexts where accountability and learning reportedly compete directly.Whereas in the latter, learning is crowded out by competition-induced accountability pressures, it is the elusiveness of learning that characterises German bilateral development cooperation.
The decoupling of performative upward accountability from organisational learning does not need to be problematic: in theory, accountability could be effective while organisational learning might be vibrant, even if they do not condition each other.In our case, however, both accountability and organisational learning appear limited.Learning in implementing organisations occurs chiefly through individual experiences and social exchanges.Many interviewees point to such learning dynamics, but not in relation to accountability processes.Rather, learning seems amorphous; it allegedly happens but cannot be pinpointed.Where lessons are learned, they do not appear to factor into accountability processes, which are formalistic rather than formative.
Our finding resonates with research on intra-organisational reporting on foreign aid impact (Eyben, 2010).Project-level staff in international development organisations are known to reconcile official strategies and reporting prerogatives from headquarters with their social entanglements 'on the ground'.Our analysis suggests that in the German case, this formalistic approach to accountability cuts across organisational boundaries.It unfolds in an institutional setup characterised by scant public attention.In this context, GIZ and KfW are central components of a boiler system that transforms public resources into development projects.This system's purpose is to channel and expend these funds with as little interference as possible.All of its parts -GIZ, KfW and BMZ -share an interest in upholding this setup where volume trumps quality, an interpretation validated by recent research on BMZ funding for technical and vocational education implemented by GIZ (Fontdevila et al., 2022).This dynamic again is different from interactions in Anglo-American organisational settings where government principals have a broader range of implementing organisations to choose from and work with.
The German case helps us appreciate existing scholarly literature on the relationship between accountability and organisational learning in a new light.Although most interviewees apply the seesaw image of accountability and learning to this case as well (cf.Greiling and Halachmi, 2013;Gujit, 2020Gujit, [2010]]; Kogen, 2018;Schillemans et al., 2013;Springer, 2021), our data remain silent on how tightened or relaxed upward accountability would have an inverse effect on organisational learning.Some interviewees wonder whether less stringent reporting requirements might unleash creativity.Others posit that the dearth of organisational learning drives BMZ's escalating demands for managerial accountability.Each of these hypotheses rely on this seesaw logic.Yet the strongest theme emerging from our data points to a decoupled relationship.While the German case does not necessarily cast general doubt on accountability and learning being related, we believe that extant research has yet to consider the possibility that organisational interactions may lead to their intentional decoupling in practice.
We see the reasons for decoupled accountability and organisational learning as manifold, and our data and method caution us against formulating strong causal propositions.We speculate that one aspect could be the lack of competition in the German foreign aid system, which renders frontstage performances functional for organisational legitimacy, but not for competitiveness.Although GIZ and KfW cannot be out-competed, they both face an overriding necessity to maintain an image of professional competence.Another explanation could focus on decoupling reflexivity from organisational self-interest to ensure survival.Accountability and organisational learning both require transparency and intra-organisational interactions, which can limit managerial degrees of freedom.At the same time, if neither accountability nor learning are corporate priorities, then senior leadership interest in any kind of relationship -whether inverse or synergetic -will understandably be limited.
A final proposition centres on decoupling being indicative of an organisational repertoire or habitus (Bourdieu, 1990).Most interviewees explicitly acknowledge the existence of frontstage performances where different configurations of accountability and learning can be enacted.From a sociological perspective, BMZ would not serve as a passive audience but constitute an integral part of the performance.The audience would then be the German parliament and the wider public as ultimate arbiters of legitimacy.In this scenario, BMZ, GIZ and KfW coproduce accountability performances to ensure the German foreign aid system's legitimacy as a whole.This is not to suggest collusion.Instead, we assume a shared social norm that foreign aid flows must not be jeopardised by public criticism.
The function of accountability -to provide 'legitimacy for funders and practitioners to keep pursuing current development projects' (Kogen, 2018: 105) -would then be similar to what has been observed in the Anglo-American context, Yet, the (inter-)organisational trajectory of coproducing such legitimacy looks different in the German setting.In Anglo-American aid systems, accountability performances would likely be scrutinised publicly.In Germany, however, such performances have so far unfolded in a more inward-looking political setting.

Conclusion
In Germany's foreign aid system, the world's second largest in terms of bilateral flows, upward accountability and organisational learning are neither rivals nor reinforcing; rather, they are decoupled.What does this imply for the study of accountability and learning in other development bureaucracies, and for evaluation research more broadly?
We believe that one of the key messages from existing scholarship -that is, overbearing accountability requirements stifling development impact by hindering organisational learning -appears incomplete.Respondents in these institutional settings portray upward accountability as problematic and argue for greater autonomy (Honig and Gulrajani, 2018;Valters et al., 2016).In our case, interviewees do not push back against greater accountability but rather stress the importance of performative enactments of accountability to safeguard autonomy.The relationship they draw therefore does not run between accountability and learning, but between presentation on the frontstage and autonomy on the backstage (Hoyle and Wallace, 2008).This explanation could be the result of structural forces; they may just as well be symptoms of self-interest or path-dependent socialisation into a professional field.Any combination of these three potentially complementary factors could help elucidate other cases as well, particularly those in institutional settings where interorganisational trust is limited and public scrutiny is comparably low.Recent research suggests that the Spanish foreign aid system could also be classified in similar terms (Yanguas, 2023).In the German case, respondents characterise interactions with BMZ as marked by low trust, scepticism and excessive formality.They suggest that less scripted exchanges could help promote mutual understanding and rebuild trust.Trust thus emerges as a connective tissue between accountability and organisational learning meriting additional research.
Scholarship on the lives of foreign aid workers (cf.Denskus, 2014;Fechter and Hindman, 2010;Mosse, 2005) demonstrates that critical self-reflection is common among these practitioners.The same has been documented for senior-level staff in German federal bureaucracies (Veit et al., 2018).A challenge for organisational leaders that emanates from this co-existence of internally reflexive but collectively protective professional communities is two-pronged.It comprises creating organisational incentives that reward unbiased data analysis, including of experiences where targets were missed, while supporting interorganisational spaces that render interactions between these communities less threatening and more frequent.Reflecting specifically on aid evaluation systems in Sweden and Norway, Reinertsen et al. (2022: 372) remind us that trust can induce organisational learning in 'closed-off, internal spaces' but that these, in turn, hinder public scrutiny and accountability.Based on our data, we posit that it could be worth exploring to which extent inter-(rather than intra-) organisational spaces for learning can serve as steppingstones for increasing public tolerance for error -especially in Germany where public scrutiny of foreign aid is less intense than in Scandinavia.
Finally, we envision further research on the effect of competition on the relationship between accountability and organisational learning.Springer (2021) has analysed how competitive settings can restrict learning.Our case illustrates that in one non-competitive national market at least, organisational learning does not flourish either.Comparative research on different configurations of competitiveness and organisational legitimacy and their implications for accountability and learning is needed to better understand variation in organisational impact.