University of Birmingham The politics of public services : a service characteristics approach

Summary. — Politics is widely regarded as aﬀecting and being aﬀected by the performance of public services, yet little research diﬀer- entiates between services in exploring these eﬀects. The article addresses this gap by proposing a framework for understanding and comparing the politics of diﬀerent services. It identiﬁes how the nature of the good, type of market failure, tasks involved in delivery, and demand for a service—hitherto regarded largely as economic and managerial concerns—aﬀect political commitment, organizational control, and user power. Policy responses can be targeted to address service characteristics where they present opportunities or constraints to better services.


INTRODUCTION
Recent analysis of the politics of public services has been pursued in two main directions. On the one hand, political interests, incentives, and institutions may affect the performance of public services (Bakker, Kooy, Shofiani, & Martijn, 2008;Keefer & Khemani, 2003;Mcloughlin, 2014a, chap. 5;Pritchett & Woolcock, 2004). On the other, the processes and outcomes of service delivery may themselves affect politics, and even the legitimacy of the state itself (Mcloughlin, 2014b;Mcloughlin & Batley, 2012). On either side of the debate, very little research differentiates systematically between services in seeking to explain political causes and effects. Political economy analysis is usually generic, assessing the effects of political context as if it conditions all services to behave similarly. Likewise, in policy circles, "service delivery" is often referred to in the aggregate as though it addressed common issues of politics and performance, regardless of sector. By contrast, the sectorial silos of health, education, etc. are typically so insulated that they pursue their specialisms without regard to common challenges that recur across all services.
This article presents an analytical framework for understanding why services and tasks within them differ in the types of political dynamics they tend to attract. There are predictable reasons why, even within the same political environment, a service such as hospital health care is likely to raise different issues of politics and accountability compared, for example, to urban water supply. At least part of the explanation 1 lies in the characteristics of these services. At their core, these characteristics concern the nature of the good being delivered, the type of market failure being addressed, the tasks involved in delivery, and how a service is demanded and consumed.
Service characteristics have previously been identified as having mainly economic or managerial implications, but they also affect the politics of service provision. Specifically, service characteristics may combine to have powerful effects on the incentives for politicians to commit to services, on relations of control and monitoring between political actors and provider organizations, and on the level of citizen pressure for services and how this is voiced.
Although service characteristics set broad parameters that condition relations of power, they do not determine political processes and outcomes. Their effects are not immutable: institutional context and human agency may modify them. Analyzing the structural problems and opportunities in the nature of services can be the basis for change. Policy responses and organizational reforms can be targeted to address service characteristics where they present a barrier to better services. For example, if the reason why certain services are neglected can be pinned down either to politicians' difficulty in claiming credit for their delivery or to users' difficulty in understanding their benefits, then a more focused response can be devised. The analysis may be more or less formal: either by a process of participatory analysis and awareness-raising among stakeholders that allows them to develop their own responses to problems and opportunities, or by a more deliberate process of analysis and reform. Since the approach indicates similarities as well as differences between services, it can also be used to identify possibilities for transfers of practice between services facing similar constraints.
In the next section, the paper sets out the theoretical foundation of the idea that services and tasks within them have characteristics that influence not only their management but also the forms of politics and governance which develop around them. Section 3 presents the evidence of the effects of characteristics on the incentives of politicians, on organizational accountability and control, and on the organization of demands by users or citizens. Section 4 explores the implications of the service characteristics approach for future analysis and for the design of policy and organizational interventions to improve services. The conclusion summarizes the case for applying a sector characteristics framework to inform future research and policy design.

IDENTIFYING SERVICE CHARACTERISTICS
Scholars have analyzed the characteristics of goods and services, and tasks within them, from a range of disciplinary perspectives. Perhaps the most enduring framework-which identifies why under some conditions markets cannot operate or fail to work efficiently and public intervention is necessary-has its origin in welfare economics (Stiglitz, 2000). A scarcer source of more general frameworks can be found in the field of organization theory and public management. James Q. Wilson's formative analysis in 1989 categorized government activities on the basis of their observability (or visibility) and measurability of their outputs and outcomes. Pollitt (2006) added budget weight, to identify why government activities attracted political attention or interference. Pritchett and Woolcock (2004) and Pritchett (2013) differentiate types of task according to the intensity of transactions between agents, the level of agents' discretion, their stakes and their need to act innovatively. The World Bank has played an important part in creating frameworks that link economic and task-related factors with service performance, for example in its World Development Reports for 1994, 1997, and particularly 2004.
The service characteristics framework presented here draws together and builds on these earlier scholarly and policy-oriented works, piecing together common service or task descriptors and recurring themes about their effects, to form a single comprehensive framework. The characteristics listed in Table 1 2 and described below were distilled not only from the frameworks described above but also from findings of research principally on the health, education, water and sanitation sectors in developing countries (Mcloughlin & Batley, 2012).

(a) The nature of goods and services
In economics a fundamental distinction is made in the nature of goods and services, based on whether they are considered rival (meaning if a good is used by one person, it cannot then be used by another), and/or excludable (meaning whether or not it is possible to exclude any individual from its benefits) (Stiglitz, 2000). At a general level, public goods are non-excludable and non-rival, whereas private goods are rivalrous and excludable (Stiglitz, 2000). 3 To illustrate, most public health functions are considered public goods-for example in a public health information campaign, where the benefits are collective, it would be impossible to measure how much any individual has consumed, the service is not used up by its consumption and it is not possible to exclude (or make individuals pay for) its benefits (Batley & Larbi, 2004). Cancer treatment, on the other hand, is considered a private good because the benefits are largely consumed individually, rivalry is inherent between consumers competing for a limited pool of resources, and individuals can be excluded (Khaleghian & Das Gupta, 2005). Rivalry and excludability are fixed characteristics, inherent in the nature of the good, and they determine how particular goods can be financed and delivered. There is no market incentive to produce public goods, since users cannot be excluded, levels of consumption cannot be determined, and there is no way of charging for their use. Individuals are often reluctant to contribute to the support of public goods, creating the basis for what is known as the "free rider problem" (Stiglitz, 2000).
(b) Market failure characteristics All public services have market failure characteristics, which essentially means an unregulated market will either under-provide them or, as in the case of public goods, fail to provide them at all (Besley & Ghatak, 2007). These characteristics therefore generate the rationale for, as well as the likely form of, state intervention (Batley 1996;World Bank, 2003). In any given sector, the state may intervene-either by taking over its delivery or by controlling it through indirect roles 4to reduce information asymmetry, to ameliorate the negative effects of monopoly tendencies, or to produce positive externalities and prevent negative ones. Some goods, and functions within them, are particularly associated with some types of market failure more than others. Information asymmetry, for example, is chronic in highly professionalized services, such as education and health-care (compared with, say, waste collection), where citizens cannot easily make choices based on an evaluation of the quality and efficiency of the services offered (Batley & Larbi, 2004). Conversely, information asymmetry in favor of clients can also generate challenges to the effective functioning of health insurance markets. Urban piped water supply is a classic example of a natural monopoly, resulting from high investment costs and enormous economies of scale, making it very difficult for alternative suppliers to compete (Nickson & Franceys, 2003). Another type of market failure occurs where individuals do not necessarily understand or appreciate what is in their own, or the wider public interest (Walsh, 1995). Where this failure is perceptible, governments may compel individuals to consume certain goods, often called merit goods, to generate what would otherwise be missed opportunities for positive externalities. Immunization falls into this category, because it not only has private benefits for the vaccinated individual, but also contributes to the protection of the wider public (Khaleghian & Das Gupta, 2005).

(c) Task-related characteristics
The nature of the good and the rationale for state intervention might be regarded as higher order characteristics but, within a service, there are also task-related characteristics that affect relationships of control and accountability between actors in the processes of production and delivery. Services can be distinguished by the difficulty of monitoring their outputs, which depends not only on the capacity of government to do the monitoring, but also on the nature of the task being undertaken (World Bank, 2003). Here a distinction is often made between types of process or output that are easily observed, or visible and therefore easily measured (e.g., waste collection or road maintenance), and those that are more difficult to observe (e.g., sewerage disposal). Imprecise objectives such as "good education" are much less easily measured than good power supply (Besley & Ghatak, 2007). When service outcomes are neither easily observed nor measurable, they present the problem that providers cannot easily claim that the benefits to users are attributable to their efforts. As we will show in Section 3, problems of attributability arise also for other reasons than visibility and measurability so it is presented here as a distinct characteristic. Certain types of task are highly discretionary and transaction-intensive-such as curative care and classroom teaching-making them particularly difficult to standardize and control (Pritchett & Woolcock, 2004). As well as being discretionary and transaction-intensive, teaching is an example of a task that might be described as highly variable, in that it has to be customized to serve different user needs-in this case, the different aptitudes, motivations, and learning styles of pupils (Bruns, Filmer, & Patrinos, 2011). Variability is therefore naturally greater where there is a high level of heterogeneity of user need, the most illustrative example of which is individually oriented clinical services (World Bank, 2003). Lastly, services that have a high technical content and where knowledge is scarce are associated with provider autonomy. This signifies that agents may become dominant over principals (citizens and policy-makers) through the strength of their organization and expertise (professional groups, organized labor, and contractors). This may be compounded where the service is locally monopolistic-for example, urban water supply and hospital health-care (Batley & Larbi, 2004).

(d) Demand characteristics
Services also have demand characteristics, which are to do with the effect of the nature of provider-user transaction on users' demands and providers' responses. Frequency and also predictability of use are key differentials in relations between producers and consumers. Individual needs for health-care are generally episodic, unpredictable, and highly variable, so, while quality of care becomes urgent at times of illness, it is otherwise, in the normal course of life, a less continual concern than education (Nelson, 2004). A distinction should be made between on-going chronic ill-health and the disruptive shock of critical illness. Unpredictability of treatment means consumers may be unable to adequately plan and insure, and ultimately have less choice at the point of need. In some services, consumers naturally have less choice than in others, and therefore less influence over what and how services are provided. Piped water is the classic example of restricted choice between alternative providers, whereas health services usually present some opportunities for individual choice but inadequate information about the quality of alternatives (Batley & Larbi, 2004). On the other hand, water supply presents collective opportunities for the expression of consumer demand. Water consumers are defined territorially in that pipes or boreholes typically serve a geographically distinct area and people experience the same service regularly. This creates greater possibilities for people to collectively organize around them at local level (Nickson & Franceys, 2003). Services that are consumed individually or collectively in a restricted geographic area are targetable, in the sense that providers can choose to focus their service on particular beneficiaries (Keefer & Khemani, 2003).

POLITICAL EFFECTS OF SERVICE CHARACTERIS-TICS: SOME TESTABLE PROPOSITIONS
Service characteristics have conventionally been regarded as technical matters with largely managerial implications, affecting the likelihood and form of state intervention. Yet they also help generate the political profile of services-shaping the types of institutions, incentives and power relationships that surround them. Characteristic features of the type of good being produced, the market and its imperfections, the tasks essential for delivery, and the nature and degree of demand for or consumption of the service, each have predictable effects on some of the most persistent political challenges widely observed in delivery. These are principally ones of skewed or weak political commitment to provision, unequal power relations between policymaking and provider organizations, and restricted or immobilized citizen capacity for oversight. Below, we set out specific propositions about the effects of service characteristics on these political challenges.
The analysis is based on two sources. These are, first, a rigorous review of published research addressing how service characteristics affect politics in a range of developing country settings (Mcloughlin & Batley, 2012) and, second, a series of focus groups with sector specialists. The focus groups invited between 10 and 20 UK-based practitioners, policy-makers and researchers from health, education, water and sanitation sectors to discuss and contest the significance of service characteristics, based on their own analysis and/or experiences in the field (see  for an overall synthesis). 5 This dual data collection enabled the systematic capture of both explicit and tacit knowledge about why and how service characteristics matter for the politics of delivery. The resulting propositions do not cover each of the characteristics equally, but rather reflect the key recurring themes that emerged.
Though to a degree already tested, the primary goals in presenting them here are to suggest that they be further tested across a range of settings and to demonstrate their joint political effects. A summary of findings is provided in Table 2.
(a) Political commitment to provision Even in the same political and institutional environment, certain services are consistently under-or over-provided relative to others-for example, water supply over sanitation, and infrastructural over social services. Part of the explanation for this lies in the characteristics of different services. Political incentives increase with the possibility of excluding some users and targeting the service to favor particular individuals or groups, where user benefits can be clearly attributed to political intervention, and with increased visibility and public profile. To the extent that politicians can know and are willing to act on user preferences, these characteristics influence calculations about the degree to which providing a particular service offers scope for winning support, or for servicing clientelistic relationships.

(i) Excludability
Politicians' incentive to provide or improve a service is greatest where it has high excludability and offers private over public benefits (e.g., household water connections versus mains sewerage). This is because excludability opens up the possibility of targeting goods and services on favored political con-THE POLITICS OF PUBLIC SERVICES: A SERVICE CHARACTERISTICS APPROACH 277 stituencies, or on individuals willing to pay a rent, often with the effect of exacerbating inequality of opportunity. The result is a prevailing political dynamic of "it's our turn to eat" (Wrong, 2009). Local governments, by virtue of being closer to voters, may have a stronger political incentive to respond to these preferences (Besley & Ghatak, 2007). Low excludability has the reverse effect of weakening political commitment to provision. In particular, non-excludable and therefore non-targetable public goods are often neglected because they offer limited scope for servicing favored constituencies (Keefer & Khemani, 2003). Likewise voters' preferences are typically for highly targetable private transfers such as jobs, cash, subsidies, and in-kind transfers, as opposed to the general condition of broad public goods (Akin, Hutchinson, & Strumpf, 2005;Khemani, 2010).
(ii) Information asymmetry Information asymmetry, where users are less able to judge the quality of a service, makes it difficult for citizens to assign credit or blame for performance to political actors. This breakdown in long-route accountability can in turn weaken political incentives to improve provision, or skew incentives toward short-term and easily attributable results. As Keefer and Khemani's (2003) widely cited study shows, information asymmetry in the health and education sectors are particularly acute cases, because certain aspects of their quality and efficiency are especially hard to judge (e.g., to observe infection in hospitals, or judge the quality of teacher performance). For similar reasons, politicians find it difficult to build credible, long-term political platforms for some aspects of these services, because citizens will not have confidence that they can deliver on promises (Keefer & Khemani, 2003). The quality of water supply is much harder to assess than its quantity, underpinning the political incentive to install and extend systems rather than ensure their functionality (Mason, Harris, & Batley, 2013).

(iii) Visibility
High visibility-or the more observable, in a physical sense, the processes and outputs of service delivery are to citizens and government-can enhance political incentives for provision. The potential political returns from responding to visible problems and producing visible outputs are greater than those from tackling lower profile challenges or improving systems and processes that are obscured from public view. This has the effect of skewing incentives toward producing visible hardware, or easily imagined and quantified expansions in access (i.e., school construction or the creation of more spaces) over backroom, quality-improving functions. Physical buildings and infrastructure are inherently "noisier signals" of political effort than recurrent expenditures for less visible outputs like human capacity building (Keefer & Khemani, 2003) that are harder for citizens to observe (Harris, Batley, Mcloughlin, & Wales, 2013a). Invisibility of water pipes versus more monumental over-ground infrastructure can lead to long-term, deliberate underinvestment (Bakker et al., 2008). In turn, within the water sector, infrastructure often gets prioritized over investments in maintenance or recurrent costs (Mason et al., 2013). Particularly in democracies, low visibility public goods such as health, sanitation or education are likely to get less voter and political attention than famine relief or defense during war (Mani & Mukand, 2007).
In the same way, it may be politically valuable to prioritize highly visible social problems over invisible ones. This is why democracies are sometimes characterized as generally better at intervening in the "spectacular suffering" of famine, than they are at preventing the "unspectacular suffering" of chronic malnutrition (Khaleghian & Das Gupta, 2005). The prominence of a service in national debates can be manipulated by political elites, through media channels and information campaigns (Eldon, Waddington, & Hadi, 2008;Harris, Batley, & Wales, 2014). Nevertheless, visible physical assets are a more enduring and everyday symbol of political effort, and they are also easier to mobilize around, for example through ceremonial openings.

(iv) Attributability
As already noted, politicians may find it more convincing to claim their role in building a hospital or generating employment for clinical staff than in reducing malnutrition. This is not only due to visibility, but also because malnutrition is affected by exogenous factors, not least individual choices (i.e., diet or lifestyle) that are beyond government control (Mani & Mukand, 2007;World Bank, 2003). Sanitation programs that depend on collective action by communities may be more effective than government subsidies, but they offer less opportunity for political attribution, and therefore attract less political support (WSP, 2011). Uncertainty about the level of consumer uptake of certain water, sanitation and preventive health interventions (e.g., bed-nets, vaccinations and point-of-use water treatment) makes it difficult to calculate their cost and benefits, reducing their political appeal (Whittington, Jeuland, Barker, & Yuen, 2012). Development agencies sometimes have to guard their funds against diversion High excludability: possibility of excluding some users and targeting services to favor particular individuals or groups. Low information asymmetry: benefits can be clearly attributed to political intervention.
High visibility: outputs are physically visible or problem has high public profile. High attributability: outcomes are clearly attributable to political intervention.
Low discretion: tasks are easy to specify in advance.
High measurability: delivery is standardized and outputs are measurable. Low provider autonomy: limited specialist knowledge and organization by providers. High territoriality: clearly defined boundaries of consumption.
Low rivalry: (perceived) competition and variability of treatment. Low monopoly: choice or exit option for users. Low information asymmetry: users well informed about rights and quality. High frequency or predictability: creating unity of demand and common experience. High territoriality: users concentrated in defined areas and service provides focal point.

278
WORLD DEVELOPMENT to more tangible outputs, or find themselves substituting for governments' underinvestment in them (Whittington et al., 2012). Likewise, in education, the persistent problem of achieving quality of education compared with achievements in improving access is not only because of the visibility of quality as an output, but the uncertainty about how to achieve it and therefore to get credit (Harris et al., 2013a). Attribution problems also occur when the benefits of investment reveal themselves beyond the expected timeframe of political office. Long delay before the realization of some types of benefit means that current users lack information on which to judge political or professional performance. In education, improvements in quality also take longer to become evident (in test scores) than expansions of access (Harris et al., 2013a). Another example is waste-water sanitation where, even if the potential long-term benefits (e.g., reductions in diarrhea and infant mortality) are well understood by political actors, investments are often not made because citizens do not see explicit links to improved health, and in practice these benefits take longer than an electoral cycle to reveal themselves. The cumulative political effect is that politicians may estimate the costs of production of urban sanitation to be higher than its political rewards or returns (Winters, Karim, & Martawardaya, 2014).

(b) Organizational control
The fields of economics and management have explored why several features of services complicate problems of reporting, controlling and monitoring provider performance. These are more than managerial concerns. They also affect the capacity of political actors and users to assert influence over front-line staff and delivery organizations, and the balance of power between these actors. Problems of organizational control may be compounded where front-line bureaucrats can exercise high discretion, where procedures and outputs are hard to measure and specify in advance, where the providers of the service have accrued organizational autonomy through specialist knowledge, and where consumption is geographically highly dispersed or fragmented. These characteristics vary by service and tasks within them.

(i) Discretion of front-line bureaucrats
High bureaucratic discretion shifts the balance of power away from politicians, policy-makers, managers and possibly users, in favor of providers, contractors and field staff. Measurement and control are difficult where discretion is exercised by front-line staff, often in the process of interacting with users of services (World Bank, 2003). Teachers, for example, must use their own judgment across tasks that are highly discretionary and transaction-intensive, recognizing the variable needs of pupils, and engaging in repeated and frequent interactions with them to produce the required results. It is therefore difficult to specify in detail sufficient to be monitored, the way in which teachers are expected to operate (Pritchett & Woolcock, 2004). Curative care also poses a challenging combination of being both discretionary and transactionintensive. Discretion is unavoidable in difficult-to-specify services such as clinical practice and classroom teaching (Gauri, 2013).

(ii) Measurability
Policy-makers and managers will be more able to monitor and control providers where procedures and outputs of delivery are easily measured (e.g., infrastructure construction). Some tasks within services, for example vaccination, are rela-tively straightforward to measure and monitor because they can be standardized (Bruns et al., 2011). In other services, policy-makers may suffer information asymmetry about the performance of providers much as client-users do. For example, in waste collection or networked water supply, information is available to measure both process and outcome; in medical diagnosis, procedures of technical experts cannot be determined by policy-makers but there are clear, measurable outcomes; in health education, neither processes nor their directly attributable outcomes are easily measured (Abma & Noordegraaf, 2003;Besley & Ghatak, 2007;Pollitt, 2006). To add to this complexity, even within a specific service, employees may perform multiple tasks with multiple measures of performance (e.g., test scores and creativity in education). Incentives then have to be carefully managed so that effort is not skewed toward producing more measurable processes, outputs or outcomes (Besley & Ghatak, 2007).

(iii) Provider autonomy
Provider knowledge may also give them effective capacity to organize and control in their own interest. This is more likely in the case of functions where there are few actors with the necessary professional capacities-for example, doctors, water engineers or qualified teachers. Reforms in highly skilled sectors can be undermined by the unwillingness of governments to challenge professional associations, particularly where they have power not only as employees but also as managers of the service (Batley 2004). The negative effects of professional dominance on health care regulation and outcomes are widely noted (Balabanova, Oliveira-Cruz, & Hanson, 2008;Harris, Wales, Jones, Rana, & Chitrakar, 2013b;Mills, Bennett, & Russell, 2001;Palmer, 2006). The significance of provider interests (professions, unionized labor, managers, business interests) varies depending on whether those groups are powerful, not only locally but also across the entire sector, and even within government and political parties. There is unlikely to be much incentive for politicians to agitate for service improvements where front-line providers, for example teachers, are also political organizers, and where politicians rely on them to deliver votes (Devarajan, Khemani, & Walton, 2011). In most countries, teachers are the largest single group of state employees, affording them and their associations significant political power (Harris et al., 2013a), and capacity to resist monitoring and control (Duflo, Hanna, & Ryan, 2012).

(iv) Territoriality
Clearer and well defined territorial boundaries around the consumption of a service improve the capacity of the state to monitor performance and hold providers to account. Networked water may be easier to regulate than the informal market of non-networked providers due to its geographical "containment" (Bakker et al., 2008). On the other hand, the fragmented, informal marketplace of urban, non-networked water creates difficulties in obtaining comprehensive data on the scale and diversity of provision (Franceys & Gerlach, 2011). The diversity and geographically dispersed nature of small-scale, informal health providers, operating across differently defined territories, makes monitoring through visits and inspections a formidable challenge (Palmer, 2006).

(c) User power
Direct accountability of service delivery organizations to users implies that citizens can both organize demand for services and monitor their performance (World Bank, 2003).

THE POLITICS OF PUBLIC SERVICES: A SERVICE CHARACTERISTICS APPROACH 279
Analysis of the effects of service characteristics suggests that the collective power and bargaining position of citizens are greater where the service produces publicly consumed goods that are not rivalrous in the sense of being divisive, there is choice rather than monopoly in supply, users are able to assess the quality of the service, the service is used predictably and frequently, and the territory in which the service is consumed is small enough to allow easy encounters between users.
(i) Rivalry Goods that are highly rivalrous can suffer particularly acute collective action challenges. In curative health, for example, where consumption is largely individual and private, problems of collaboration between users arise from the heterogeneity of needs (distinct health issues) and high levels of variability and discretion in treatment. Prospects for collective action are dampened not only by perceived competition, but by different user experiences . Compounding this, competition for attention not only between individuals but also groups of users may open the way for more powerful actors (medical professionals, pharmaceutical companies, and global funders) to set service priorities . At the other end of the spectrum, public goods that are non-rivalrous attract collective action problems because private gains are either unappreciated, or depend on collective effort that is not guaranteed. In sanitation, for example, households can be motivated to adopt the private goods of toilets for reasons of dignity, convenience or prestige (Allan, 2013), but are likely to be less readily mobilized around the non-rivalrous public good of a defecation-free environment, partly because collective gains may be jeopardized by non-adopters (Mason, Batley, & Harris, 2014).

(ii) Monopoly tendency
Monopoly acts as a structural constraint on the exercise of user power relative to providers, with the effect of undermining citizen capacity to make claims or exercise oversight. While an absence of choice, and therefore an inability to exit, may on the one hand make the exercise of voice more pressing (Golooba-Mutebi, 2005) it may simultaneously undermine it.
Where there is no alternative provision, the decisions of monopoly providers regarding access to or exclusion from services can become a powerful tool to quell citizen agitation over pricing or quality. This has been particularly prevalent in the monopoly supply of piped water, where documented experiences have shown that even where users attempt to exert power, or assert their right to participate in decision-making, these efforts can ultimately be stumped by the providers' capacity to turn off supply (O'Reilly and Dhanju, 2012). In other instances, even non-networked water vendors may periodically establish local oligopolies, and make arrangements with the official network provider to maintain the water scarcity that supports their power over consumers (Ahmed & Sohail, 2003;Mason et al., 2013;Swyngedouw 1995).

(iii) Information asymmetry
User power to demand service improvements is reduced when they have limited or imperfect information about the quality of the service they receive. Information is not a panacea for user accountability (Lieberman, Posner, & Tsai, 2014), but it may be a necessary condition. This is partly why, within the same local political economy, social accountability tools have been observed to be less well attended in sectors where users do not feel well informed to judge the quality of delivery (e.g., clinical health), compared with others where the outputs are immediate and easily assessed by intended recipients (Wild & Harris, 2012). The acute incidence of information asymmetry between providers and users of curative health services reduces scrutiny especially by those less educated about the risks and potential of treatments (Harris et al., 2013b). Similarly, inequality of power and information between parents and teachers can mean that the expected long-term benefits of increased choice or participatory reforms are never realized in practice (Gershberg, González, & Meade, 2012).
(iv) Frequency and predictability of use Regularity in the use of a service may be important in creating opportunities for interaction among users, improving their ability collectively to demand service improvement. Demand for drinking water is frequent and predictable, creating scope to coalesce around service issues, especially where there is a shared provider. Nevertheless, unity of demand may be undermined by the variety of legal and illegal sources of urban water supply, differentiating experience and creating room for dispute (Mason et al., 2014). Similarly, the use of education services is highly predictable and frequent, but these conditions for collective action may be weakened by rivalry for access and by the fact that the pupils must rely on parents and teachers to make decisions on their behalf (Harris et al. 2013a). Frequency and predictability clearly apply to sanitation generally and the disposal of human excreta most specifically. However, in this case, unless social norms of shame can be overcome, discussion and collective action are likely to be inhibited or left wholly unspoken (Mason et al., 2014). Most clearly, infrequency and unpredictability apply to the need for and experience of curative health services. Infrequent individual consumption of health care, low predictability and high urgency lead to low capacity to gain information, exercise choice, bargain and collectively organize (Chaudhury, Hammer, Kremer, Muralidharan, & Rogers, 2006;Harris et al., 2014).

(v) Territoriality
The capacity of citizens and clients to organize, aggregate demands, and advocate for better services improves when the population of users is concentrated within a clearly defined territorial space. Users who share a day to day experience of, or problems with, a service that operates in a clear locality may be more likely to be able to organize by virtue of their shared consumption (Batley, 2004). Education is strongly territorial in the sense that schools often become a focal point for the local community. This enables users to frequently and informally exchange information about the service (Harris et al., 2013a). Likewise a visible boundary of consumption in the case of piped water can facilitate collective mobilization to address problems (Kacker & Joshi, 2012). Group deliberation or problem-solving are logistically aided by a discrete territorial boundary of consumption. This can in principle empower users where it helps them to aggregate heterogeneous preferences and interests into a coherent set of demands (Gauri, 2004). Participatory mapping of sanitation problems can itself be a visual reminder of territory, encouraging community members to articulate the inadequacy of sanitation in poor urban areas (Pervaiz, Rahman, & Hasan, 2008). The presence of a single point of contact between users and providers is positively associated with the likelihood of collective action. A public distribution system for subsidized food in Delhi provided a focal point for users to solicit support from local officials (Centre for the Future State, 2010). On the other hand, confusion about which branch of government is responsible for urban sanitation in Indonesia dissipates demand (Winters et al., 2014). 280 WORLD DEVELOPMENT The capacity of users to organize may be hampered when the size of the space in which a service is consumed is large. Services that are set up to rely on community maintenance through the creation of consumer groups may suffer if the sheer size of the territory over which these groups preside exceeds their physical capacity to get out and monitor it. Accordingly, some research has attributed the relative success of rural piped gravity water schemes partly to their relative size, with larger ones being unmanageable and more likely to fail than smaller ones (Kleemeier, 2000).

IMPLICATIONS FOR ANALYSIS AND ACTION
Analyzing service characteristics-either as a free-standing exercise or as part of a wider political analysis-adds value to an overall understanding of the politics of service delivery in two ways. First, its comprehensiveness offers the possibility of a finer grained analysis of the politics of specific services than has previously been applied. Second, policy interventions may be targeted to explicitly address the political effects of service characteristics where they present opportunities or constraints to better services.
(a) The combined effects of service characteristics Service characteristics are experienced not separately as analytical categories but together in lived experience in which they combine to have cumulative effects on the political profile of any service. A policy analyst aiming to find opportunities for intervention to change the political dynamics of service provision needs to understand the cumulative effects of characteristics on relations between stakeholders at different points in the production and delivery of a service. The analysis presented here has generated some specific, testable propositions in this regard: Politicians' incentive to provide or improve a service is greatest where it offers exclusive and targetable private rather than public benefits (e.g., medical treatment rather than sanitation); where it benefits users directly rather than through external effects on the wider population (water supply rather than dis-ease vector control); where citizens have information to understand the benefits and results are visible in the short-term, and can therefore be clearly attributed to politicians' action (e.g., construction of schools rather than improvement of teaching standards).
Policy-makers and managers are more able to monitor and control providers where providers do not have a high degree of professional dominance, and where they exercise relatively little discretion (as in standard procedures for vaccination); where procedures and outputs (e.g., infrastructure construction) are easily specified and measured; where the service delivered is a public good offering few opportunities for disruptive rent-seeking (e.g., public health campaigns); and where managerial information is accessible because the service offers direct benefits (rather than external effects) to definable clients within clear territorial boundaries.
Users' power and capacity to organize collectively is greater where a service is used frequently and predictably within a limited territory (e.g., piped water or primary schooling by comparison with hospital health care), and users are able to assess the quality of the service, allowing the formation of shared opinion and demands; where there is choice rather than monopoly in supply; and where the service offers easily visible and short-term benefits to known beneficiaries (a health center rather than a public health campaign). However, collective organization is easily diverted by users' competition for private benefits (such as access to household water connections) that can be targeted on favored people by the exercise of providers' discretion.
Together, service characteristics affect the political salience of a service. If we understand salience to mean that there is an incentive for political leaders to provide services, that they can be delivered, and that recipients are able to offer political (electoral or clientelistic) returns, it is the ultimate political determinant of provision (Mcloughlin, 2014a, chap. 5). While political salience will also be affected by contingent contextual factors (institutions, political settlements, vested interests, scarcities, and crises), it is at least partly a product of the service itself. Table 3 sketches the political profile of four services: tertiary curative healthcare, public preventive healthcare, community sanitation, networked drinking water, and primary education. This is a general statement, ignoring the fact that services are complex and contain many sub-elements (for example, the construction of water networks and the delivery of water); and our assessment of the effect of characteristics always depends on their interplay with context. The table is illustrative of typical conditions. It should be noted that, depending on the characteristic (e.g., monopoly or visibility), a high rating may signify a negative or a positive political effect-as indicated in Table 2.
In this estimation, drinking water has the highest rating for political salience because it is both easily attributable to political effort (targetable, visible, and measurable) and presents possibilities for organized demand (frequently and predictably used within a limited territory). Primary education is not easily attributable after the initial construction of a school has been accomplished, but its predictable and territorial usage may foster organized demand for sustained improvement. Tertiary (hospital) healthcare may, similarly, be most politically attractive in the construction phase, but offers a much weaker basis for sustained demand for quality treatment. Programs of preventive healthcare and community sanitation have strong public goods qualities but produce personal benefits that are not easily understood or visible except in the long-term; they present the dilemma of requiring a public health crisis or a bold political gamble before conditions are created for any visible political payoff.
Simply recognizing that services have their own profile can help policymakers and practitioners to think differently, escape their silos, and appreciate what is distinct about their sector. Sector specialists in government and development agencies as well as in academia occupy spheres that generate their own internal preoccupations and world-views, limiting recognition that they may understand their own sector better by comparison with others. Awareness that services can be systematically compared may change mind-sets.
The framework not only enables understanding of what is distinct but also of what is shared between service sectors. Cross-sectorial learning opportunities can result from the discovery that services face similar challenges in relation to political commitment, organizational monitoring, and collective action by users. Identifying similarity as well as difference suggests the possibility of exchanging experience, transferring practices and perhaps even collaboration. This may be possible not only for different sectors within which some services operate in similar ways (for example, client-oriented services like health centers and schools), but also for services within the same broad sector where there are complementarities. For example, environmental health, a public good which is likely to suffer a low level of demand, may be promoted at health centers and hospitals which offer more attractive private services ).

(b) Policy and organizational responses
Even without deliberate policy intervention, the political context in which services are delivered may enhance or mitigate the effects of service characteristics. For example, the dominance of provider groups and their resistance to monitoring may be increased in national contexts where professional groups (such as doctors, teachers or engineers) have strong influence in local and national politics; monopoly is strengthened where it is legally defended. Information asymmetry between providers and users may be reinforced where educa-tional standards are low. The possibility of user organization may be reduced where the prevalence of political clientelism legitimizes the targeting of services on favored clienteles. Service characteristics should always be understood as indicating likely tendencies which are modified by their interaction with the wider, and especially the political, context.
A process of participatory analysis among stakeholders using this framework may itself raise awareness and generate pressure for change, even if it is not followed by the deliberate formulation of reforms. Analysis can alert policy-makers, providers, citizens, service users and activists to problems and opportunities, help explain why these occur, and enable incremental adjustments in practice. For example, in most sectors, expert providers have opportunities to assert control over politicians, managers and users-but awareness of this risk also enables its avoidance. Campaigning organizations may come to recognize the incentivizing effects of openly crediting politicians with achievements in the delivery of social services, in order to balance the pressure on politicians to prioritize more easily attributable capital investment projects. Citizens might become more aware of their opportunity to organize around local territorially based services, such as schools, that can also act as a focal point for considering less local services, such as tertiary healthcare.
Ultimately, this approach provides a basis for the systematic analysis of structural problems and the deliberate derivation of possible responses. Recognizing how and why politics differs by service, as well as by political context, implies there is no single approach to addressing political constraints. Rather, the approach highlights why certain policy responses-for example social accountability tools or user monitoring-may achieve better results in some services than in others (Batley & Wales, 2015;Grandvoinnet, Aslam, & Raha, 2015). For example, community scorecards may be most effective for highly visible services with low information asymmetry; information campaigns may be more vital in services with attributability problems; community monitoring may have limited user empowerment effects where there is a structural problem of monopoly. In this way, analyzing service characteristics can identify the possibilities and limits of the suite of possible policy interventions.
Below, possible organizational or policy responses to the effects of service characteristics on political commitment, organizational control and user power are addressed in turn. Each of these effects and responses is exemplified by a sectorial case. The reservation should be repeated that the possibilities are always conditional on the political and institutional context.

(i) Political commitment
In urban piped water supply, the political incentive is often to favor the visible installation of systems, and the subsidization of private household connections and consumption. Likely to be neglected are less visible and targetable public goods functions such as maintaining the infrastructure of water supply, and improving the disposal of waste water. Politicians are likely to be most ready to hear the demands of current users precisely because the latter are networked through the service and have connections with provider organizations. On the other hand, citizens who do not already have access to piped water lack a basis for organizing and are easily neglected.
Policy responses might include using the budgetary process to ring-fence public goods functions, so as to ensure that demands for subsidized tariffs are not privileged over maintenance of the water supply system and waste disposal. In 282 WORLD DEVELOPMENT principle, politicians would be distanced from detailed allocative decisions, for example by contracting out water supply and delivery, though the experience is often that political decision-making penetrates such arrangements. Clientelistic targeting of water connections can be challenged by ensuring the public reporting of policy commitments, expenditure priorities and performance. The well-organized demands of current users might be balanced by encouraging community associations and NGOs to support residents of under-served areas to organize themselves as targetable constituencies for political attention (Mason et al., 2013).
(ii) Organizational control Curative healthcare particularly at the tertiary level is notoriously beset by factors that make providers dominant in their relationship with policymakers as well as users. Supply-side interests, such as pharmaceutical companies and global funds, can have a powerful influence on decisions of practitioners, favoring certain medical conditions and types of treatment. Medical professions in many countries have a strong presence in politics and policy-making as well as in service delivery organizations, and are usually self-regulated. Information asymmetry allows them to exercise a high degree of discretion in the interpretation of policy priorities and in deciding appropriate treatments, with little scrutiny where outcomes are not easily measurable.
The first step to redressing these imbalances is to recognize what is often seen as forbidden territory: challenging undue provider dominance. A basic condition for containing distortions is the clear public statement of budgeted expenditure plans to protect policy priorities, followed up by performance monitoring and public reporting. Restrictions can be put on the staffing by medical professionals of line ministries, regulators and hospital administrations. Pressures for over-prescription based on information asymmetry can be countered by publicity campaigns on appropriate medicine usage. 6 The local monopoly status of hospitals can be checked by independently monitoring and publicizing their comparative performance, based partly on users' assessments of the more visible aspects of their treatment, such as health worker attendance and the quality of personal care. Professional standards can be enhanced by publicizing cases of good practice, legitimizing pride in the profession .

(iii) User power
Primary education seems to offer good conditions-frequency, predictability and territoriality-for users to meet, share experience and organize collective action. On the other hand, the relationship between users and providers is complicated by other factors: provision is rivalrous if there are insufficient school-places or pupils have to compete for teachers' time; the direct users depend on their parents to act for them; there are information asymmetries both between parents and between parents and teachers about what comprises quality education; and the benefits of education are long-term and not easily visible to current users.
The policy response might be to formally recognize and encourage the opportunities presented by primary schools to act as a natural point for community organization on issues including but going beyond education. A national program to improve education might use schools to host independently convened discussions between children, parents and teachers in order to make the features of quality education more visible: informed by statistics and real-life examples of the long-term returns to education. Pro-active schools would increase their standing by acting as future models. The monitoring capacity of parents might be mobilized to monitor the very visible and collectively shared problem of teacher absenteeism (Harris et al., 2013a).

CONCLUSION
We have proposed a systematic framework for comparing public services on the basis of a range of characteristics that are validated in the literature and applicable to all service sectors.
More than purely managerial or technical concerns, the intrinsic characteristics of specific services affect power relations between key actors involved in their provision (politicians, policy-makers, providers, potential and actual users of services). Service characteristics act not just individually but also in combination to (re-)produce persistent political constraints to effective services: low visibility and poor attributability underscore weak political commitment; discretion, measurability and territoriality reduce the power of policymakers over delivery organizations; and users' power collectively to organize so as to demand better services is weakened by rivalry, monopoly and unpredictable use.
These effects are predictable but not immutable. Either political context or deliberate policy interventions may modify them. Analysis of how characteristics play out in a particular service in a specific context provides the basis on which interested actors (whether policy-makers or anyone) can decide how to intervene to improve services. Recognizing how and why politics differs by service and task illustrates why there is no general solution to the political constraints to better services, even within one national context. In the policymakers' toolkit, some tools are better suited to services with some characteristics rather than others.
Services are not distinct in all respects but may have similarities as well as differences. Our view is that this provides serious possibilities for cross-service learning, where services which are in some respect similar perform more or less successfully in the same country context. Comparison and the identification of common features could provide a bridge for transfers of experience across service silos.
As a framework for research, this approach is ready-made for application and testing by others. The propositions set out in Section 3 are based on the available evidence, and could be regarded as hypotheses for further research. There is no good reason why the characteristics should not be relevant for other goods and services aside from the ones tested here, and for more and less developed countries. Future research could test how far characteristics can be seen as intrinsic factors, and how far they vary in their political effects by context. Additional characteristics may also be identified. Indeed, evidence of the political effects of some of the characteristics we have identified-targetabil ity, externalities, transaction intensity and variability of treatment-is still quite sparse, and warrants further empirical research.