Abstract
The recent surge in popularity of ‘accountability’ in public administration and international development seems in part divorced from centuries of conceptual and empirical work done in related disciplines of finance and accounting, and in political science. This article brings together the core meaning of accountability as used in hundreds of previous works, and seeks to bring order to the litany of subtypes in this literature. An organizing scheme with three dimensions (source of control, strength of control, and direction of relationship) captures all the existing varying types of accountability. The resulting typology also clarifies that varying subtypes have not only different actors and characteristics, but also seek to uphold varying values and are facing different challenges. These have important implications both for research and the (im-)possibility of translating findings from one subtype field to another; as well as practical implications for the policy world.
Accountability has several different forms depending on the actors (e.g. citizens–politicians; politicians–bureaucrats; or judges–citizens). These types of accountability seek to protect different values, and are accompanied by varying challenges. Yet, everything is not accountability: it is but one of many possible ways to constrain the (mis-)use of power. This article clarifies the core idea of accountability. It then depicts the full range of subtypes with their different characteristics and problems. This can function as a guide for policy makers and practitioners when seeking to address weaknesses in accountability of varying actors based on acknowledging their differences.
The concept of accountability has a long tradition in both political science and financial accounting, but only more recent prominence in public administration and international development. In accounting, the concept’s long tradition is strictly limited to financial prudence and accounting in accordance with regulations (e.g. Barton, 2006; Normanton, 1966). Yet, the underlying principle of delegating some authority, evaluating performance, and applying sanctions is essentially the same as in its long tradition in political science. Thus, Locke’s theory of the superiority of representational democracy built on the notion that accountability is only possible when the governed are separated from the governors (Grant and Keohane, 2005; Locke, 1980 [1690]). It was also fundamental to the fathers of the American constitution (Dubnick and Romzek, 1993; Finer, 1941; Friedrich, 1940). The central idea is still with us: when decision-making power is transferred from a principal (e.g. the citizens) to an agent (e.g. government), there must be a mechanism in place for holding the agent accountable for their decisions and tools for sanction.
In the last 10 to 15 years, the concept of accountability has become fashionable. To illustrate this growth: when Schmitter and Karl (1991) reminded us that accountability was central to most definitions of democracy, their claim was met with indifference (Schmitter, 2004). A decade and a half later, a quick search in any of the academic search engines generates a dizzying number of entries. Unfortunately, this proliferation has resulted in a myriad of meanings and dimensions ascribed to the concept of ‘accountability’. Donors, their consultants, and their partner governments have picked up on this trend in their focus on ‘good governance’ and have added peculiarities to the landscape.
One may reply ‘so what?’ since ‘essentially contested concepts’ (Collier et al., 2006; Gallie, 1955) have produced fruitful theoretical debates and spurred empirical research. 2 This article is not an argument against such informed engagement. Rather, it is an effort to escape the dangers of creating ambiguity, vagueness, and collective semantic confusion. There is an important difference between a debate on the advantages and disadvantages of a few clearly delineated alternative definitions, and uncontrolled, myopic, and unrelated conceptual diversification.
A recent example of this is the recent addition of the ‘Utrecht School’ of accountability. It deserves special attention because it has become very influential in the European literature on public administration and law. While it admittedly launches a new formulation of the definition of accountability (Bovens, 2007: 450, 2010: 951), the change of words offers nothing new to the classic definition touched upon above (and discussed in greater detail below). It is reinventing the wheel and giving it a new name. Proceeding to the operational definition, the classic understanding of the components of accountability is again recast in new terms where accountability is introduced as a ‘social relation’ between ‘an actor and a forum’. The main achievement is to obfuscate the distinctiveness of accountability from other types of constraints on actors’ power to act autonomously. When the term ‘sanction’ finally is misunderstood to denote only punishment (deviating from the proper meaning of the word in English), the paraphrasing becomes misleading.
The above is harsh but necessary. In a ‘dialogue of the deaf’, constructive theoretical and conceptual advancement is no longer being achieved. Reviewing a substantial portion of the literature I have counted well over 100 different ‘subtypes’ and usages. A series of clarifying contributions on the organization of various usages of a set of central concepts in the social sciences (e.g. Collier and Levitsky, 2009; Gerring and Barresi, 2009; Kurtz, 2009) have serviced the profession with some order and guidance. The present article is an effort to contribute in the same vein with regard to ‘accountability’.
The first order of business for this article is to establish what type of concept accountability should be treated as. I argue that rather than a radial or family resemblance concept (as some of the current usages indicate), the concept of accountability is better conceived of ordered in a classical sense to a more general concept of constraints on power.
Second, it is argued here that the internal structure of accountability should also be treated as the order of a classic concept and that it thus has several classic subtypes for which the problems and prospects of accountability differ. For example, the question of what is required for a society to imbibe a general culture of expectations of accountability from the state and its political institutions, is very different from the specific issues relating to vertical political accountability between citizens and representatives; or what factors and incentives make for effective vertical bureaucratic accountability. It cannot be assumed that findings in the area of one subtype of accountability are relevant for another.
This article thus first discusses the structure of the concept of accountability and then takes a closer look at the subtypes of political accountability. Summarizing a vast literature on accountability in finance and accounting, political science, public administration, and international development studies, the article presents a synthesis of subtypes of accountability that is meant to cover all theoretical possibilities. Finally, the article illustrates the implications in terms of the inappropriateness of using the findings from one subtype of accountability to another.
Accountability: classic, radial or family resemblance concept?
What kind of analytical concept is the construct of accountability best seen as? Classic concept formation is hierarchical and aims for mutually exclusive definitions, hence defining attributes are co-jointly necessary and sufficient. All defining characteristics must be present in an object in order for it to be classified as an instance (Collier and Mahon, 1993; Coppedge, 2005; Goertz, 2009; Munck, 2001; Sartori, 1984; Schedler, 2001).
It follows that the more, and the more precise, characteristics one attaches to a concept, the fewer the objects it will apply to. 3 This is what some philosophers have held to be a ‘law of inverse variation’ (e.g. Cohen and Nagel, 1934). Sartori (1970: 1040) refers to this as the ‘ladder of abstraction’ or better ‘ladder of generality’ (Collier and Mahon, 1993: 846). The further up the ladder, the fewer defining characteristics (intension) and the larger number of objects it refers to (extension). When a phenomenon in a new context does not squarely fit the definition of accountability, one moves up the ladder of generality to a superior category, thus avoiding conceptual stretching. For example, one can argue that accountability belongs to a general category of ‘methods of limiting power’ (others being, for example, devolution of power, violence, economic pressure, public shame, and anarchy). Classical concept formation thus follows a neat and simple set of rules not only for categorization but also for comparison. In deciding on what basis a set of cases should be compared, the defining characteristics are used for classification.
Subtypes of classical concepts, however, can be numerous but are complete instances of the general category. 4 Classical subtypes are therefore nominal (not more or less, just different) versions of the concept in question. Religions such as Islam and Christianity are not more or less religions but nominally different instances. If we think of various kinds of accountability as subtypes in the classical sense, this applies to instances of various types of accountability (political, financial, legal, bureaucratic). None of them represents more or less accountability but different types of accountability. This opens up the possibility for two particularly important comparisons that nonetheless need to be recognized as yielding different types of conclusions. First, units such as countries, regions and other localities can be compared with regard to which subtypes of accountability are present (and to what degree). Particular configurations or strengths may be found associated with better outcomes in terms of the provision of public goods, for example. Second, one can compare units to inquire into the relationships, possibly causal ones, between various subtypes of accountability. Does strong electoral accountability enhance bureaucratic accountability, for example? But for any of these comparisons to yield accumulative knowledge the profession needs clearly demarked and commonly shared definitions of accountability and its subtypes.
Second, if one rather wishes to measure ‘levels of accountability’ between systems, it is appropriate and fully legitimate to shift the unit of analysis (cf. Marsteintredet, 2007: 3–5). Pick a set defined by something interesting such as local political systems, relationships between courts and district assemblies, or citizen–bureaucrat relationships in the land sector. Accountability can then be conceptualized as qualities that these units of analysis have to a greater or lesser degree. Thus, the universe of cases has the potential of displaying certain qualities – in this case accountability relationships.
The third option is, as the Utrecht school suggests we do, to recast the existing concept. Reading the now vast literature on accountability with various typologies covering virtually everything from policy networks in the international community (Grant and Keohane, 2005) to community governance of crime in some particular localities in Britain (Benyon and Edwards, 1999), this seems unfortunately to be a trend.
Family resemblance
But there are at least two more fundamental alternatives. Scholars (e.g. Marsteintredet, 2007: 5; Putnam, 1981: 19) have drawn upon Wittgenstein’s (1968) investigation of the nature of language and the idea of family resemblance. This entails a principle of ‘belongingness’ to a category that is different from classical categorization in that there is not a single attribute that all objects share – like members of a human family (Goertz, 2006). In Collier and Mahon’s (1993: 847) words: A category defined in a particular way, may fit a number of cases reasonably well, but on close examination it can become clear that for most cases the fit is not perfect.
It should be said that the main disadvantage – and this applies also to the World Bank’s index – is a possibility of descending into the abyss of meaninglessness. If, for example, accountability has five attributes (a principal, an agent, an area of discretionary responsibility for the agent, an obligation of the agent to inform about and justify decisions to the principal, and a right of the principal to sanction the agent), two problems arise:
Family resemblance implies that having, say, four criteria in any combination can ‘compensate’ sufficiently for the missing one. One would have to take a careful look at the empirical implications in order to justify such permissiveness. If we allow for one missing attribute, why not also one and a half, or two, and then three? It remains unclear whether allowing for one missing attribute but not more is justified and how close an object must be on each attribute to be classified as an instance of the concept.
In effect one risks what Pitkin (1967) sees as a reality of many concepts in political science: various phenomena are collated under a single term (Adcock, 2005: 26–27). Too much rigidity may prevent interesting analyses but excessive flexibility may thus make for essentially meaningless concepts and hence comparisons.
Let me illustrate: let us say that we are studying accountability as a key factor in explaining variations in the extent of public goods provided. If all sorts of phenomena that constrain the exercise of power are labeled accountability relationships one may end up with the conclusion that accountability plays little or no role at all simply because several mechanisms that constrain the exercise of power (e.g. violence or anarchy) may not produce many public goods at all.
For comparative analytical purposes it is crucial to remember when using family resemblance categories that they are always nominal. It thus imposes stringent restrictions on the types of comparative analyses one can carry out. Along with the complications discussed above it also provides the argument for why accountability is not a family resemblance concept.
Radial categories
The third main type is radial concepts. Collier and Mahon (1993: 848–9) borrowing from Lakoff (Lakoff, 1987: 47, 74–76, 83–84) exemplify with ‘Mother’ understood as a ‘prototype’ of which there are very few objects. Yet, subtypes such as ‘birthmother’, ‘stepmother’, or ‘nurturing mother’ have more empirical referents than the superior concept. Collier and Mahon argue (correctly) that in classical categories, sub-types add attributes to the defining characteristics of the superior category, typically leading to fewer empirical referents. But the authors also argue (incorrectly) that with radial categories all attributes are present in the superior concept while the subtypes have only some, thus fewer characteristics. Therefore the ‘law of inverse variation’ is reversed and there are more ‘genetic mothers’ in the world than ‘mothers’ in the full sense (Collier and Mahon, 1993: 851).
Their mistake is not realizing that in radial categories no relationship between extension and intension exists per se. The ‘sub-type’ include all defining attributes but with either a negative or positive load. Take ‘stepmother’ defined as a female married to the father. Yet, the definition of stepmother also includes the characteristic that she is not a ‘birthmother’ and not a ‘genetic mother’. Empirically, it is easy to picture a universe with more nurturing mothers than prototype mothers. But one can also imagine a situation with more prototype mothers than, say, stepmothers. Radial ‘superior’ and ‘subtype’ categories belong to the same level of abstraction, or generality as it were, because they include the same number of defining characteristics (see Marsteintredet, 2007).
Unlike classic and family resemblance concepts, radial categories not only have affirmative characteristics but also negations of other radial categories as their defining characteristics, hence they must be treated as such in comparative work. Accountability does not fit readily into the scheme of radial concepts. As argued in detail below, the core concept is not a ‘prototype’, and the ‘subtypes’ do not include negative loads on varying characteristics as part of their definition. Rather, it seems more reasonable to argue that there is a core concept of accountability with a restricted set of defining characteristics that are also shared by all subtypes. The subtypes, in turn, have additional characteristics that differentiate them from each other as different instances of accountability, however different in other ways that are important to consider when analyzing causes and consequences.
The fundamental notion of accountability
The fact that there is now an increasing number of varying definitions used in a literature ranging from ethnographic interpretations of idiosyncratic local meanings to highly technical financial auditing techniques, should perhaps caution us to not even try to find common denominators. 6 It would also be impossible to discuss all the attempts in the literature to define and operationalize accountability. Yet, a discussion of some of the varying meanings can help us narrow down and identify core attributes of the root concept. This will also clarify why accountability should be treated as a classic concept.
Bentham’s principle that ‘The more strictly we are watched, the better we behave’ (quoted in Hood et al., 1999) perhaps best captures the idea behind the necessity of accountability. Nietzsche thus indicted that we give an account only when it is requested, and only when that request is backed up by power (from Butler, 2005: 11). This is the reason that accountability has for long been a key issue in constitutional scholarship (Smith and Hague, 1971: 38). At a very fundamental level then, accountability is closely associated with authority though not necessarily political authority. A puppet acting as an extension of someone else’s will is not a legitimate object of accountability. That is why accountability is different from ‘responsiveness’. Only actors with some discretion to make authoritative decisions can be the objects of accountability relationships (e.g. Christiano, 1996: 219; Hyden, 1992; Thomas, 1998). In Burke’s succinct statement on representative accountability: ‘Your representative owes you not his industry only, but his judgment; and he betrays, instead of serving you, if he sacrifices it to your opinion’ (Burke, 1774, cited in Brevold and Ross, 1960: 148). Accountability hence is associated with the act of discretionary governing, typically understood as the authoritative allocation of resources and exercising control and coordination (e.g. Dahl, 1971; Kooiman, 1993; Marsh and Rhodes). This in turn points toward the need for an identifiable locus of authority, as famously argued by Mill: Responsibility is null and void when nobody knows who is responsible. … There must be one person who receives the whole praise of what is well done, the whole blame of what is ill. (Mill, 1964 [1861]: 332)
Having authoritative responsibility is thus an important aspect of accountability. This, of course, does not negate the possibility that having overlapping layers of accountability relationships creates a stronger bulwark against the undesirable use of authority as argued by e.g. Day and Klein (1987), Finn (1993), Fisse and Braithwaite (1993), Romzek and Dubnick (1987), and Stone (1995). But it explains why accountability has always been central to democratic theory to the extent that it is sometimes posited that accountability necessitates democracy. What Locke referred to as all men are, or ought to be considered equal as political beings 7 (Locke, 1970 [1689–90]: 322), and what Dahl calls the ‘idea of intrinsic equality’ (1989: 85), is the fundamental condition of rule by the people. 8 It is the equal access to being part of the decision-making process, rather than the liking of the substantial decision by all, which satisfies the right to self-government. Yet, the etymological understanding of democracy leaves out the other side of the coin (Sartori, 1987: 30). Rule of the people is exercised over the very same people, and to be workable any modern form of national democracy must be representative, 9 thus representation became grafted upon democracy (Hindess, 2000). This understanding of democracy induces a particular kind of vertical political accountability (Schedler, 1999), and with modern notions of democracy in larger more complex political systems, accountability has taken on paramount significance (e.g. Moncrieffe, 1998; Schmitter and Karl, 1991).
Yet, there are non-democratic, as well as democratic types and mechanisms of accountability (Grant and Keohane, 2005) and accountability as an idea pre-dates modern democracy and is still wider in scope than just ‘democratic’ accountability. For now we thus have to stay on the more general level. At that general level, there seems to be agreement on four characteristics of all types of accountability. Unless I am mistaken, more or less all the literature referred to in this article agrees that the following elements should be included as defining characteristics of any form of accountability:
An agent or institution who is to give an account (A for agent); An area, responsibilities, or domain subject to accountability (D for domain); An agent or institution to whom A is to give account (P for principal); The right of P to require A to inform and explain/justify decisions with regard to D; and The right of P to sanction A if A fails to inform and/or explain/justify decisions with regard to D.
These defining characteristics may be expressed in various ways
10
but seem to capture the core of the concept. It should be noted at the outset that none of these conditions specifies that these relationships have to be formally codified or that the agents and institutions involved are formal institutions or hold an official office. Even if the individuals involved are indeed office holders such as bureaucrats in a state body, their accountability relationship may be in part or wholly informal.
Conditions 1 and 2 together mean that an identifiable person or office must have some discretionary power over a certain domain, and that domain is subject to accountability. This neither implies that the agent or institution is elected or otherwise democratic, nor does it mean that the agent or institution is necessarily accountable for all of the domains over which it has discretionary decision-making power. Every common-sense use of the term assumes that a certain measure of authority, or responsibility, is vested in an individual, group, or institution. Conditions 3 and 4 together mean that there is another agent or institution with a de facto right to require the agent or institution A to explain and justify decisions and actions with regard to the specified domain. To say that A is ‘accountable’ if and when there is no one s/he has to answer to, just does not make sense in any intuitive meaning of the word. Again, it is not assumed that P is elected or acting necessarily in the public interest (see Philip, 2008: 3). Another important thing to note at this stage is the wider applicability of this conceptualization to many others, in particular those that build on democratic principal–agent theory. Accountability is then usually thought of as P (the people) holding A (elected official) accountable for actions with respect to P itself. However, in many accountability relationships P is holding A accountable with respect to x, y or z denoted above as D. For example, native English speakers would agree that accountability is a proper word also for ministers that can hold top-level bureaucrats accountable for their decisions and actions with respect to lower-level bureaucrats or other state agencies. In other words, it should not be assumed that those affected by the decisions of A are necessarily identical with P who has the right to hold the decision-maker(s) accountable.
Finally, condition 5 stipulates the crucial condition of the right for P to sanction A if the request to inform and/or explain and justify actions is not honored to P’s satisfaction. More restrictive definitions that exclude condition 5 are rare, but Smulovitz and Peruzzotti (2000), in their conceptualization of ‘societal accountability’ for example, exclude any right to sanction agents and institutions. This effectively removes any obligation of decision-makers and actors to inform, justify and explain their actions beyond what they themselves are comfortable or feel obliged to disclose. This reduces the notion of accountability to mean little if anything in terms of holding to account since the decision-makers then are holding themselves to account only as much as they like.
Regarding condition 5, a few things should be noted. The right to sanction is limited to the right to punish a failure with regard to A providing information and justification in the most general conceptualization. Some definitions of accountability (e.g. Schedler, 1999) take this further to require that the right to sanction A’s decisions and actions is part of the definition. But an important distinction should be made between the right to sanction A for failure to provide the information requested and justifications for decisions and actions taken, and the right to sanction agents or institutions (A) for the content or effects of such decisions and actions. At its core, accountability only necessitates the right to sanction A for failure to provide information and justify decisions. The right of P to also sanction the content of decisions and actions by A is a possibility that if present adds additional leverage for P but is not strictly necessary for the concept of ‘accountability’. There are many examples of everyday usages of the word and of real-world institutions of accountability that can illustrate this. Take the ombudsman office found in many contemporary democracies. It acts as a principal, monitoring elected officials as well as bureaucracies, and has the right to request that those agents provide information about their decisions and actions as well as calling on them to justify their actions. Ombudsman offices typically do not have the right to sanction the agents for their actions as such but can use the courts to sanction offices if the information requested and/or justifications are not provided. Similarly, in many situations elected officials can demand that bureaucrats provide full information about their decisions and actions and to justify them, but as long as the holder of a bureaucratic office has not acted in direct contravention of the formal rules and instructions, the elected official (the P in this case) is not allowed to sanction the bureaucrat with downgrading, loss of office and so on.
The right to sanction failures by A also has another implication. There must exist a set of criteria for measuring accountable behavior (Knouse, 1979; Schedler, 1999). If there are no standards or measurable expectations of A with respect to A’s duties in the domain (in terms of information, justification, and perhaps decisions and performance), there can be no accountability. If one has no clear picture of what is acceptable and what is unacceptable behavior, it cannot be evaluated and sanctioned. This means further that there must be some form of evidence of accountable behavior. Together these implications are crucial in drawing a line between instances of empirical phenomena that belong to accountability and those that do not. In private client–patron networks, for example, there are often only blurred or poorly defined expectations of behavior on the part of the patron, whereas the client can be confronted with quite direct demands. The client’s right to demand information and justification from the patron is sometimes very weak or non-existent in these highly asymmetrical power relationships, and then it ceases to be an accountability relationship. If the client can request such information, there are often no or few means available to verify it, and then the level of accountability is low. No one really knows what the patron does for the various individual clients except the patron him/herself. Thus, even if the client or a set of clients can regularly extort some form of compensation, protection, or personal favors from the patron, that does not make such a relationship an instance of accountability. Again, this resonates with common-sense and everyday usage of the term accountability. To say that a landlord is ‘accountable’ to servants or serfs does not resonate with established usage to a native English speaker. If and when the patron–client network is at least partly transposed into a more public form where the asymmetry of power is more level, for example when the patron becomes an elected official, the clients get principal status as voters and can demand information as well as justification on explicitly stated standards such as promises made during election campaigns. There will also be more verifiable measures of behavior such as facilitating accountability.
I am acutely aware that the above still leaves substantial room for interpretation in terms of what ‘sufficient degree’ means. However, I am not sure this can be resolved at this level of generality. What is sufficiently explicit and measurable in terms of duties of information and justification, and in terms of verifiable indicators, must probably be established in context-specific analyses of specific subtypes of accountability. The principal argument, however, is therefore important to keep in mind.
The five conditions also translate into a stylized time-line of accountability as depicted in Figure 1. P must first transfer decision-making power over a particular D to A. Then A acts in this capacity and P can thereafter require A to provide information and justification for these actions; and if A fails to do so, P can apply sanctions.
Time-line of accountability
Sub-types of accountability
Within the superior, or root, concept extant formulations of different (sub)types of accountability can be organized along three dimensions. A basic notion is the distinction between the source of the accountability relationship (the accountability holder, or principal) being internal or external to the one being held accountable (e.g. Gormley and Balla, 2004; Radin and Romzek, 1996; Romzek and Dubnick, 1987). For example, in a bureaucracy where superiors are holding subordinates accountable for their tasks the source is internal. In the relationship between voters and representative, the source is external to the legislature or executive being held to account.
But it is not only the source of accountability that is important; the degree of control that the principal exercises over the power holder is also of consequence. Studies of voting have, for example, spent considerable energy on trying to find out exactly how much control voters in fact exercise over representatives (e.g. Cox, 1997; Powell, 2000) and the effects of varying influence. In the extreme, there are fundamental differences between financial auditing exercising extremely detailed control based on specific rules and regulations, and patron–client accountability where control is typically diffuse and highly contextual.
The third critical dimension in distinguishing types of accountability regards the spatial direction of accountability relationships (e.g. Schedler, 1999). Shareholders exercise an upward form of vertical accountability when they hold business executives accountable for the company’s performance. When politicians hold bureaucracies and their leaders accountable for the implementation of decisions they have taken, the accountability relationship runs downward in a vertical fashion. Finally, when the legislature engages in executive oversight or the constitutional court reviews acts adopted by the legislature, this is a form of accountability that runs horizontally ‘among equals’ (O’Donnell, 1998).
Types of accountability
This table is an elaboration and synthesis of typologies found in Gormley and Balla (2004); Grant and Keohane (2005); Mulgan (2003); Radin and Romzek (1996); Schedler (1999); Scott (2000); and Smulovitz and Peruzzotti (2000).
In all organizations we can find informal forms of accountability relationships that run both upwards as in client–patron accountability and downwards as in patron–client accountability. Both are characterized by a low degree of control with adjacent information scarcity and limited means of monitoring and evaluation. A client can request information and hold the patron accountable for delivery of the kind of benefits promised as part of the bargain but is usually limited to that. If the client is a more powerful broker as representative of an important group of clients, for example, or has a special favored relationship with the patron, that client can often hold the patron accountable for certain issues outside of the direct exchange relationship. The exact dynamics typically revolve around the importance of the loyalty of the client to the patron since the main leverage for the client tends to come from threatening (implicitly or explicitly) to exit. The reverse is also true. In the patron–client form of accountability, the patron can usually hold the client accountable for a wide range of things given that what the patron can offer the client is important enough be it material rewards, personal safety, career opportunities, or improved status and fame. In poorer societies, material rewards and low-skill jobs are naturally more powerful tools for a client than in more economically developed countries. Add to that the informal norm that someone who helps to sustain your living is a ‘father’, to whom you always have to be loyal, which is not uncommon in poor nations, and the power of the patron to hold the client accountable is usually stronger than the client’s. Nevertheless, the patron’s right and ability to require very specific information regarding all facets of the client’s activities is fairly low compared to the types of accountability discussed above. The individualized nature of patron–client forms of accountability and the absence of formal rules along with the exit option for clients adds to the fact that findings from studying clientelism are unlikely to be generalizable to other forms.
A sometimes formal (as in academic and professional organizations) or informal (as among peers and cohorts of professionals) form of horizontal accountability is professional/peer accountability. The degree of control is low since peers can only require information about very narrow segments of the activities of their colleagues and usually not particularly specific information, but the P (the peers) are internal, and accountability relationships are horizontal, focusing like audits on safeguarding the organizational or occupational reputation. Among the factors that distinguish this subtype of accountability is the respect for colleagues and voluntary self-submission to being held accountable that comes with professional peer reviewing. Academics evaluating each other’s work and results, and lawyers’ associations certifying and monitoring peers, are examples. Thus, the dynamics of this subtype of accountability are also not easily transferred to other spheres, and findings from studies of peer professional accountability are unlikely to be useful for generalizations to other subtypes.
Accountability relationships with external Ps are fundamentally different. Representational accountability in democratic political systems has the citizens as P and their elected representatives are A in an upward vertical relationship, and the degree of control is relatively high. Several forms of action are available to citizens for the purposes of requiring information and holding elected leaders accountable: Voting is one that can have dramatic consequences for representatives; phone-calls, meetings, demonstrations, letters, emails, writing to newspapers, and mobilization of community and action groups are others. The nature of elected office gives citizens a powerful position and they can require specific and detailed information often including private issues. Yet, in modern states with large citizenries, large-scale collective action is required to exercise effective accountability. Thus collective action problems frequently affect the P’s ability to use formal rights to request information and to hold elected representatives accountable that distinguishes this form of accountability from many of the other subtypes. In addition, political institutions differently mediate accountability relationships. First-past-the-post electoral rules in single-member districts, for example, facilitate a closer and more direct accountability relationship between citizens and their representative to the national legislature but effectively also undermine the viability of using elections to demand new policy by voting for new or smaller parties. This exemplifies that generalizations regarding both the causes and the effects of representational accountability are unlikely to travel well to other types.
Fiscal and legal accountability, respectively, can be both vertical-downward and horizontal. In both cases the degree of control is typically high and very detailed. When fiscal accountability runs vertical-downward an external P such as a legislature holds various ministries, departments, and agencies fiscally accountable. Another case is when a state bureaucracy holds an implementing agency that can be a non-profit business fiscally accountable. But when the legislature holds the executive accountable for fiscal prudence and budgetary constraints, the relationship is horizontal rather than vertical. We find a similar situation with regard to legal accountability. Most instances of legal accountability run vertical-downward. Various judicial institutions being the external P hold citizens, businesses, politicians and others accountable for lawful behavior. But when judicial institutions hold other state institutions accountable such as in the case of a constitutional court, or in some countries the Supreme Court hearing cases of the lawfulness of executive decisions and acts by the legislature, the accountability relationship is horizontal.
An external P does not always go hand in hand with a high degree of control, however. Societal accountability is characterized by actions taken by civil society and media aimed at forcing political, bureaucratic, business and legal decision-makers to give information on, and justifications for, their actions. The strength of control is typically relatively weak in these cases but also varies with contextual factors such as legislation (e.g. freedom of information act or not). It is distinguished from representational accountability in that the Ps are more or less ‘self-appointed’ in their role and first have to convince the As and the surrounding society of this arrangement. Yet, it is therefore tenuous and circumscribed by the fact that the ‘right’ of the P to hold the A accountable is in the end based on voluntary action, and the degree of control therefore becomes low and rather diffuse. Political accountability is the vertical-downward variant of external accountability with relatively weak control. Politicians’ degree of control over the administration is by its very nature relatively weak. Bureaucracies are large and handle thousands of issues every day and politicians are few, and have several competing priorities as well as severe time and cognitive constraints. In consequence, there are only a very limited number of bureaucratic decisions and processes that politicians can attend to in any detail, giving the bureaucracy significant discretionary powers. There is also evidence that bureaucrats seek to block, slow down, divert, and capture politicians. In turn, politicians may bring in unelected advisers to support them against the permanent officials. Reputational accountability’s most significant expression runs horizontally among peers or peer institutions that are external to the agent. It is another form of diffuse and even indirect accountability whereby the agent’s reputation among other equals can be damaged if they are deemed to act contrary to established norms and procedures. It is highly dependent in informal norms among participants, both agents and principals, and thus even within this category the causes and effects of successful accountability are likely to vary significantly.
Before moving on to accountability mechanisms, a final note is important. Responsiveness is often proclaimed to be one of the key aspects of accountability (e.g. Barton, 2006; Kelly, 2003; Schedler, 1999; Walker, 2002). There is a crucial difference here between the two types of vertical accountability. In vertical-upward forms of accountability this is exactly the point. Shareholders, clients, citizens, and societal organizations are the Ps who after delegating decision-making power and discretion to agents such as the executive, patrons, and politicians, monitor their behavior and hold them to account for their failure or success in providing information and justifications for their decisions (and sometimes the outcomes of those decisions as well). If dissatisfied, the Ps can sanction the As by way of ‘throwing the rascals out’ or simply expose the A’s actions and failures. The intended after-effect is to make agents responsive to the wishes and interests of the principals.
Vertical-downward is in principal very different. In these cases, the P (who transfers power to the agent) is not directly involved in the accountability relationship. Instead it is the A who in turn holds an implementing agent (I) accountable for its actions toward P, or even a principal behind P. For example, a president (P) appoints and transfers decision-making power to a head of an agency (A) that holds the bureaucracy of that agency (I) accountable for carrying out directives toward citizens (principals behind P) who elected the president (P). In this case, you may argue that the ultimate principal are the citizens, but even if we limit ourselves to consider the president as the P, the triangular relationship has consequences. The I is held accountable by A but is not supposed to be fully responsive to A, but rather consider P (or even the principal behind P, the citizens) and still act within the law and bureaucratic regulations. This means that lower-level bureaucrats are in fact not supposed to be responsive to A or P if and when A’s or P’s directives go against either bureaucratic rules or the law.
This is particularly evident in the case of legal accountability. High-ranking judges are typically appointed by the executive and/or legislature with the mission to uphold the law. In a democratic system it is reasonable to argue that citizens are the ultimate P who elects the politicians that in turn appoint judges. In exercising their mandate, however, judges are not supposed to be responsive to either politicians or citizens. Indeed, they are expected to do the exact opposite: to exercise their decision-making power independently of the interests and wishes of various Ps in the interest of the rule of law. Judicial institutions, generally, are held accountable for being non-responsive. Similarly, vertical accountability in the form of periodic voting may, for example, be a blunt instrument of policy-specific accountability since voters typically can only choose between a few alternatives that represent complex mixes of a number of policies from defense to child care. But vertical accountability exercised in the form of frequent interactions between the legislators elected in single-member constituencies and constituents can be very effective in achieving policy-specific responsiveness. In short, responsiveness may be the desired outcome of some types of accountability but far from all, and should not be understood as integral to accountability itself.
Making sense of accountability for empirical analysis
Where does all this leave us? A key argument of this article is that accountability as an analytical concept can be saved despite the current state of conceptual confusion. Identifying five key characteristics to denote the conceptual core of accountability, this article argues that there are three additional characteristics that translate into a typology of 12 types of accountability. The differing characteristics of these 12 types highlight the uncertainty (often ignored by policy-makers in particular) of making generalizations between types with regard to causes and effects of accountability. In addition, it should be noted that most accountability theory assumes linear cause–effect relationships, as well as rational and informed decision-making with the aim of producing collective goods. These are neither viable empirical assumptions nor theoretically consistent with the notion that the object of accountability can have as primary focus to maximize the non-objective functions such as to maintain relationships and loyalties (see Philips and Berman, 2007). Hence, accountability should not be equated with always being good.
There is also a second reason why effective accountability cannot be assumed to have only desirable outcomes. Responsive agents can take actions and decisions that are easily ‘sold’ or popular rather than doing what is right or necessary (Maskin and Tirole, 2004: 1035). In the extreme, agents in strong accountability relationships may, for example, undermine the functioning of a society’s economy and the rights of minorities and other groups.
Taken together, the above reflections have important implications for empirical research on accountability. First, we cannot scale or even rank order the various types of accountability. The differences between them are nominal and it makes no sense to use qualitative or quantitative techniques designed for scale or ordinal variables in analyzing outcomes.
Second, each type of accountability has its designated functions and is compatible with certain situations only; no one is a panacea for all kinds of problems of restraining power. Legal accountability is designed to insulate agents from the influence and responses of principals, for example, while bureaucratic accountability is suited to detailed and very direct response on the part of subordinates. Representative accountability allows for broad mandates allowing sufficient (sometimes excessive) discretion to agents to deal with unforeseen and complex matters in accordance with their judgment. An empirical analysis that mistakes different types of accountability for more or less accountability is unlikely to be very helpful.
Within each of the types of accountability, one can find many variations in levels. It is possible to construct measures of levels (ordinal or scale, qualitative or quantitative) for each of these types. Yet, thirdly, it remains crucial not to assume that a measurement instrument, criterion, or approach used for one type of accountability is necessarily very useful for gauging levels of accountability in another type. We need to develop separate measurement schemes and instruments for each type.
Finally, and perhaps most importantly, one has to be very cautious in terms of making generalizations from the study of one type of accountability to another. This is further reinforced by the fact that many contemporary students of politics theorize complex and varying models of causation where equifinality, multifinality, multiple conjectural causal models, path dependencies, increasing returns, and diffusion figure prominently (see George and Bennett, 2005; Gerring, 2007; Hall, 2003; Mahoney and Goertz, 2006; Ragin, 1987). A path dependency argument, for example, negates the independent effect of individual factors as well as the causal ontology of multiple conjectural causation as far as the factors aiming to create a particular outcome are modeled as current and historically independent. Even if studies are carried out on the same subtype of accountability but are based on different causal ontologies, the results will not be a sound basis for comparative generalizations. We just have to be more careful with making generalizations about different forms of accountability.
Empirical reality has a way of making things complicated. In most societies and political systems various accountability relationships have been established at various points in time with the effect that even single institutions have multiple layers of various types of accountability. This makes it hard to discern and disentangle when and where a particular type of accountability relationship is engaged, and then trying to establish the level of accountability produced. That it is not easy does not justify not doing it, however, since the alternative is fraught with risks of systematic measurement errors (conflating the presence of one type of accountability with zero accountability, for example), undermining the validity of the analysis.
In a more practical sense, it stands to reason to believe that problems and prospects of accountability differ. What is required for a society to imbibe a general culture of expectations of accountability from the state and its political institutions is one thing; the specific issues relating to vertical political accountability between citizens and representatives is another; how horizontal political accountability in terms of the legislators providing effective oversight of the executive a third; what factors and incentives makes for effective vertical bureaucratic accountability is a fourth question, and so on. It cannot be assumed that what works in one area will also work in another. There might even be causal relationships between them such as the prevalent logic of vertical political accountability (exercised through voting and other means), that can impact on how much horizontal political accountability is provided in a particular political system. If, for example, vertical political accountability is centered on local and perhaps even personal issues, there are few incentives for legislators to invest time and resources in the provision of executive oversight. Conversely, an electorate putting a high premium on an honorable and well-monitored executive will impact positively on horizontal accountability. A situation in which both vertical and horizontal political accountability are present is in turn likely to impact on the nature and level of bureaucratic accountability. In short, it is vital to be aware of the full range of possibilities so as to be clear when it comes to research design and analysis.