Who loves input controls? What happened to “ outputs not inputs ” in UK Public Financial Management, and why?

Why do frequently criticized input controls survive in the management of public spending while apparently more enlightened output/outcome controls come and go? The question matters, because output/outcome controls are often assumed in public financial management and related literature to lead to superior policy performance as compared with input-focused approaches. We tackle the question by applying qualitative push – pull analysis to compare one key type of input controls (administration cost [AC] controls) with one much-discussed form of output/outcome controls (performance targets linked to spending allocations) in one major country case, the United Kingdom, over two decades. Drawing on documents and in-depth interviews with 120 key political and bureaucratic players, we conclude that bureaucratic inertia at most only par-tially explains the survival of input AC controls in this case. The push/pull factors associated with the politics of blame and credit made the political players fair-weather output controllers but all-weather input controllers.


| INTRODUCTION
How far should public spending control focus on outputs or outcomes, as against the traditional approach of concentrating on inputs? That issue has long been debated by scholars and practitioners in public administration and public financial management (PFM) (see, e.g., Hood, 1991;Pollitt, 1990;West, 2011, p. 11;Wildavsky, 1978). Indeed, much of the thrust of "New Public Management" thinking from the 1980s was to put more emphasis on output or outcome controls, on the assumption that such controls could motivate and enable public managers to focus on achieving results that citizens care about. So what is it that limits such a shift to an output/outcome emphasis in spending control in practice? Why do frequently criticized input controls survive in the management of public spending while apparently more enlightened output/outcome controls can be precarious?
Existing PFM literature offers some clues, identifying at least three general conditions as conducive to a serious shift from an input-focused to an output-focused or outcome-focused approach to spending control: bureaucracies that are not stifled by institutional inertia; sufficient honesty and integrity in the public spending system to make relaxation of input controls possible without risking substantial corruption, fraud or embezzlement; and other political and economic conditions assumed to be favourable to such a shift.
How then do we account for cases in which these plausible conditions are apparently met, but switches in emphasis from input to output/outcome control are at best limited, half-hearted or reversed? The United Kingdom (UK) in the 1990s and 2000s constitutes just such a case. Over that period a watered-down form of output/outcome controls was applied to central government spending and then largely abandoned, while input controls on administration costs (ACs) survived throughout the period, even though they were much criticized and had some obvious shortcomings. Given that the conditions mentioned earlier broadly applied to the UK at least for a time in the late 1990s and early 2000s, the UK case allows us to explore and isolate other factors that limited the shift from input to output/outcome controls.
We do so by analyzing the advantages and disadvantages of input and output/outcome controls, from the perspective of key bureaucratic and political ("veto") players 1 involved in the UK's public spending control system from 1993 to 2015. We elicit those perceptions from archive documents and in-depth interviews to get at aspects of the spending control process that may not show up in studies based on published official documents. We frame the perspectives of those players into a "push/pull" analysis of a type derived from migration studies. Such an analysis aims to clarify what made traditional input controls attractive to those players and thereby limited the "push" away from an input focus. It also allows us to identify what made output/outcome controls attractive to those players and thereby contributed to the "pull" toward greater emphasis on such controls.
We find that the input controls, despite their acknowledged shortcomings, were seen as useful by some of the bureaucratic players. Those controls also provided credit claiming opportunities for the political "veto" players, both in times of fiscal expansion and restraint. By contrast, output/outcome controls seem to have been generally popular with the bureaucratic players, while the political players found them useful only in times of fiscal expansion.

| THE LONG DEBATE ON OUTPUT VERSUS INPUT CONTROLS IN PUBLIC SPENDING
As mentioned earlier, the "inputs-versus-outputs" issue in budgeting and public spending control has a long history in public administration and PFM. On the one hand, some studies point to the limitations of "performance" budgeting and management, for instance because of problematic validity and reliability of performance statisticsissues that can reflect honest perplexity and inherent limits of measurement as well as deliberate "gaming" or strategic manipulation of the key numbers (see Bevan & Hood, 2006;Elton, 2004). Such issues lead some PFM scholars and practitioners (such as Ho, 2018;Schick, 2014) to argue for a continuing role for input controls over spending.
Indeed, a generation ago Wildavksy (1978, p. 508) thought the traditional input-focused approach to budgeting was "simpler, easier, more controllable, more flexible than modern alternatives." On the other hand, advocates of "performance budgeting" have argued for decades that a shift toward more emphasis on objectives-outputs or outcomes 2 -in budgeting and spending control can lead to better policy or service performance. The argument is that placing the control emphasis on outputs or outcomes can incentivize public managers to deliver "public value," by giving them freedom to deploy their budgets to focus on whatever leads to their stated objectives (such as crime reduction) rather than concentrating their efforts on staying within a set of different subcategories of inputs within their overall spending limits (Schick, 2014;Sterck, 2007;Sterck & Scheers, 2006;Ter Bogt et al., 2015).
From that latter perspective, the input controls that have traditionally dominated PFM, which work by putting "ring-fences" around different kinds of input expenditure (such as capital versus current, administration versus programme, protected versus unprotected) and only allow those ring-fences to be surmounted by specific approvals by the legislature or budgetary controllers, can be seen as a recipe for wasteful or ineffective spending. Commonly cited examples of such negative effects arising from an input-focused approach include the hiring of expensive consultants to get around staff headcount limits and end-year spending splurges in response to annuality "use-it-or-lose-it" rules (Andrzejewski, 2019;Fichtner & Green, 2014;House of Commons, 2007).
But despite such arguments for more focus on output/outcome-oriented "performance" approaches to improve policy or service delivery, numerous studies have reported only low or sporadic engagement by elected politicians in performance budgeting (or in public service performance information in general), with little or no corresponding diminution in input controls and "ring fences" (see for instance Kristensen et al., 2002;Peters & Savoie, 1996;Ter Bogt et al., 2015). And the Organisation for Economic Co-operation and Development (OECD, 2008) has noted that those countries that claim to have adopted "performance budgeting" have rarely gone beyond what it terms "performance-informed" budgeting. By "performance-informed budgeting," OECD means aspirational output or outcome targets which are published alongside budget numbers and which have some relationship to allocations, but stop short of "direct" performance budgeting, in which the sums allocated link directly to expected or actual results, for example through funding that is conditional on performance or through careful unit costing of "production functions." So how can we account for this rather patchy and limited shift from an input to an output/outcome focus in the planning and control of public spending in the developed world? As mentioned earlier, PFM literature points to at least three inhibiting conditions may account for that result. They are: sheer bureaucratic inertia (the tendency to stick with familiar ways of doing things that are deeply rooted in organizational culture; Krause, 2009); pressure to limit misappropriation, embezzlement, fraud etc. in the face of challenges to the "regularity" of public spending systems; and conditions of political or fiscal crisis, which can sometimes produce external pressure for change but can also serve to limit the political "bandwidth" or "headspace" necessary to move to a new style of control (Curristine & Flynn, 2013, p. 246;Krause, 2009;Lopez-Acevedo & Krause, 2012;Schick, 2014, p. 10;Wildavsky, 1978).

| THE UK CASE 1993-2015: ANALYTIC SIGNIFICANCE
Here we explore a case which is not obviously explicable by those three "inhibiting conditions." The UK from the early 1990s to the mid-2010s saw the survival of one much-criticized form of input controls over spending while one often-lauded form of output or outcome controls came and went. The instance of input control survival is that of "AC" limits as a subcategory of overall departmental expenditure limits, while the case of output or outcome control waxing and waning is that of various efforts to develop performance targets alongside resource allocation in the 1990s and 2000s, most notably in the form of so-called "public service agreements" (PSAs) and related targets that were negotiated with and monitored by UK Treasury spending controllers from 1998 to 2010. Table 1 provides a summary timeline of what happened to those two types of spending control in the UK over two decades. As the table shows, AC input controls (dating in their modern form from the mid-1980s) survived throughout the period for all departments (apart from the Ministry of Defense from 1994 to 2004) despite their recognized shortcomings and several reviews.
That contrasts with what happened to output/outcome controls over the same period. From the late 1980s a growing number of "executive agencies" responsible for specific service delivery operations (such as prisons or insolvency services) were spun out of out of policy ministries and subjected to performance targets. A "Citizen's Charter" in 1991 specified a set of performance standards to be applied across all government providers. John Major, Conser-vative Prime Minister at that time wrote, "We had to end the excessive focus on financial inputs rather than service outputs" (Major, 1999, p. 261).
Neither of those two developments amounted to more than at most "performance oriented" budgeting in practice. But a much closer link between performance standards and budgetary allocations was foreshadowed in a 1995 White Paper (official statement of policy) announcing plans for a new "Resource Accounting and Budgeting" (RAB) system across central government. Schedule E of that White Paper provided for output and performance analyses (OPAs), requiring central government departments to state the specific objectives relating to their budgetary allocations and report progress against those objectives in departmental accounts, which could then be exposed to audit and parliamentary scrutiny. The New Labour government that succeeded the Major government in 1997 scrapped the Citizen's Charter, wound back the autonomy of the executive agencies, and swiftly replaced the Major government's plans for OPAs by a set of "PSAs" (Public Service Agreements) to be agreed with and monitored by Treasury spending controllers.
PSAs, which committed government to meeting high-stakes performance targets, were closely linked to the Labour party's key election pledges, such as the reduction of waiting times for health care treatment (Panchamia & Thomas, 2016). In some cases at least they commanded high-level political attention, and they were described by Labour Chancellor Gordon Brown (1998), in language highly redolent of performance budgeting, as "effectively a contract with the Treasury…money will be released only if departments keep to their plans." PSAs were developed and refined in subsequent iterations of spending reviews in 2000, 2002, and 2004, with increased emphasis on outcomes rather than outputs. But their numbers were greatly reduced in a 2007 spending review that marked the beginning of a fiscal squeeze in New Labour's final years in government, and after the 2010 UK general election, a new Conservative-Liberal Democrat collation scrapped PSAs altogether (as the Conservatives had pledged to do in their election manifesto) and replaced them by departmental "business plans" which had a much lower profile.
Can this pattern-survival of input controls on ACs, together with waxing and waning of output/outcome PSA controlsbe explained by the three factors identified in the PFM literature as likely to inhibit moves to performance budgeting, as mentioned earlier? Table 2 offers a summary assessment of how the UK during this period rated on those factors.
As for "bureaucratic inertia," there is no standard comparative metric for this slippery concept. But the UK is commonly said to have been one of the "vanguard states" of the New Public Management approach under successive governments since the 1980s (e.g., Hood & Dixon, 2015, p. 15;Pollitt, 2013), with its focus on public service performance measurement and "managerial" reforms over much of its public sector. Bureaucratic inertia did not prevent other major policy changes in the country over the period considered here, such as privatization, devolution, and wholesale changes in the welfare benefit and student finance regimes. Indeed, a key in-depth internal review of the UK Treasury's approach to public spending control in 1994 conducted by Treasury officials described "outputs not inputs" (along with "focus" and "clarity of responsibility") as an "obvious and noncontroversial" principle for the way such control should work (HM Treasury, 1994, p. 33, §2.8-9). Furthermore, as a counter-force to any inertia within the bureaucracy that might have existed for adopting output controls, there were key officials within the Treasury in the late 1990s who had experience of crafting such controls in the form both of "conditionalities" for overseas aid and performance targets for executive delivery agencies.
As for "regularity" challenges to public spending control, it seems hard to argue that fears about incompetence, dishonesty or misappropriation were what principally limited the UK's shift to output/outcome controls over the period considered here, still less that such factors account for the survival of the AC input cost control system. It is true that T A B L E 2 UK output controls 1993-2015 and three conditions for a shift from input to output controls mentioned in international public financial management (PFM) literature

Condition for increased emphasis on output controls
Assessment: Was the condition broadly satisfied?
Evidence from published or publicly available sources A policy-making system not frozen by bureaucratic inertia

Mostly yes
The UK was widely seen as a "vanguard state" of New Public Management or economic rationalist approaches to public administration Regularity challenges not severe enought to seriously inhibit a switch to output controls

Mostly yes
If not at the very top, the UK FM system consistently rated high on regularity in international comparisons Sufficiently favourable economic/fiscal conditions and political stability Yes, at least for part of the period The late 1990s and early 2000s up to the Iraq war constituted a period notably free from political and fiscal/ economic crisis nontrivial levels of waste, fraud and abuse were recorded in parts of its public finances (National Audit Office, 2003, 2008 and that "qualifications" of departmental accounts by the public auditor were far from unknown (Hood & Dixon, 2015, p. 161). 3 But nevertheless over the period considered here, the country consistently scored in the upper levels of international ratings of honesty and integrity in government (see for example Transparency International, 2019).
Nor is it clear how retention of AC input cost controls might have acted as a limit or deterrent to misappropriation, given that much of the fraud and abuse noted earlier centered on the payment of welfare benefits.
As for economic and political crisis as a third possible inhibiting factor for a shift away from output to input controls over public spending, there were two major fiscal crises in the UK involving severe spending restraint during this time period, namely recession and currency crisis in the early 1990s, and global financial crisis and recession after 2008 (Hood & Himaz, 2017, chapters 9 and 10). There were some notable political "bumps in the road" as well, for example during John Major's Conservative government, which saw a leadership challenge in the Conservative party, numerous highprofile backbench revolts in Parliament (including one on the 1994 Finance Bill) and the eventual loss of the government's parliamentary majority in its final months. That political turbulence evidently did not prevent the Major Government from developing a measure of output controls for executive agencies and government departments, but the rise and later fall of PSA output/outcome controls over the period are broadly consistent with the third condition. Those controls were indeed first applied to central government departments during a "benign" period of low inflation and steady economic growth making for buoyant tax revenue, together with political stability in the form of a stable majoritarian government that enjoyed high approval ratings, a huge parliamentary majority and mostly weak opposition. Similarly, those controls began to fade and then go into reverse after that key "window of opportunity" passed and fiscal consolidation came back onto the agenda. But it is the survival of AC input controls, despite well-known shortcomings, that is not so readily explained by any of the three conditions depicted in Table 2.

| APPLYING PUSH-PULL ANALYSIS
To account for that observed pattern, we therefore need to go beyond the sort of broad-brush "conducive conditions" analysis summed up in Table 2, and get inside the system for a finer-grained view of how the key players saw the issues. So we now turn to published and unpublished documents and 120 interviews for a qualitative analysis of how the key political and official players viewed the "push" and "pull" factors associated with those input and output controls over the period considered here (more details about sources are given in Appendix).
"Push/pull" analysis is an approach originally developed for studying migration. In that context, "push" factors are what causes any given population to leave its home, while "pull" factors are those that positively draw such a population to a new area (Calhoun, 2002). A similar analysis can be applied to the choice of policies, institutions or reforms, where "push" factors consist of dislike for, or unattractiveness of, old solutions, while "pull" factors denote whatever attracts policy makers to new solutions. Such analysis has been applied in various institutional or policy contexts (see, for instance, Balakrishnan & Sayeed, 2002, pp. 16-18;Gitlin & Margonis, 1995;Zmud, 1984).
The pull exerted by one alternative can constitute the inverse of the push exerted by another, but does not invariably or necessarily do so. For instance, the theoretical advantages that can constitute a "pull" for one hitherto untried solution may be accompanied by "push" in the form of real-world negative experience of another. That is exactly how Jamasb et al. (2005, p. 4) explain moves in a number of developing countries from state ownership and centralized organization of electricity production to regulated private ownership. While there were "rationalist" theoretical arguments for the superiority of the latter approach that constituted "pull" factors, what was needed to trigger decisive change from one system to the other was a decisive "push" in the form of underperformance in state owned enterprise. The question of what shapes the balance between input and output/outcome controls over public spending is an analogous issue. Moreover, as indicated earlier, there is a long-standing set of arguments for the advantages of the latter, but less analysis of the "push" factors associated with input controls over public spending. Table 3 summarizes the results of our analysis of that material, distinguishing between the "push" and "pull" factors applying to the official or bureaucratic players from those applying to ministers and their political entourages- the key political "veto players" in a parliamentary system.
That push-pull analysis suggests both that there were limits to the "pull" exerted by output controls, particularly on the political "veto players," and that the "push" factor away from AC input controls was weaker than that 1994 "outputs not inputs" mantra quoted earlier might suggest. Both for the political players and for some of the officials, those AC input controls seem to have been more important to preserve than the conventional "rationalist" critique of such controls may account for.
4.1 | Bureaucratic "pull" and "push": Enthusiasm for PSAs, semicovert support for AC controls Most of the former officials we interviewed, both from the Treasury and spending departments, expressed emphatic enthusiasm for the idea of control by outputs as a general principle. They were attracted by the rationalist and purposive logic of focusing on the delivery of measured results, which fitted with the economics-based training and outlook of many former Treasury officials. Many of those echoed the view of the 1994 report quoted earlier that described "outputs not inputs" as an "obvious and non-controversial principle" for spending control. For example, one former senior official (55) described the move to PSAs and subsidiary targets as "helpful and positive," because they meant that Treasury spending control ceased to be just a process of "beating people up about money" and could also focus on negotiating and tracking policy and management issues.
Similarly, there were plenty of expressions of enthusiasm for output controls in principle from spending department interviewees. A former Finance Director of one of the biggest spending departments (68) said output controls were "definitely a good idea," and that view was echoed by others (102, 119). A former departmental Permanent Secretary (executive head) (14) said PSAs were "a great idea," because they gave Permanent Secretaries a raison d'etre that went beyond just "keeping the Minister happy, handling the scandals, getting bills through Parliament, etc." And one former Treasury official (77) went so far as to describe the development of output controls as "utterly transformative of everything government did"-hyperbolic language seldom used by senior UK civil servants reflecting on bureaucratic processes.
It is true that some of our official interviewees were less positive about how the PSA-type output controls had worked out in practice as distinct from the theoretical attractions of output controls. Several former Treasury officials commented critically on the absence of a close link between spending allocations and agreed objectives (the hallmark of "direct" performance budgeting in the OECD terminology noted earlier and what the plan for OPAs in the mid-1990s had aimed to provide). One (14) thought that, far from being "effectively a contract with the Treasury," the PSA performance targets were "a PR exercise." Almost all thought PSAs had greatly improved over time after the initial targets, many of which were lengthy, unmeasurable, and focused on outputs rather than outcomes. Such attitudes seem to represent the very opposite of "bureaucratic inertia," since they imply preferences for taking the output-control approach further rather than toning it down.
What about the "push" factors associated with AC input controls for the official players? In a mirror-image of attitudes to the principle of focusing control on outputs, many interviewees both from Treasury and spending departments (e.g., 6,57,68,90) at least initially expressed distinctly negative attitudes to AC input controls.
"Another ring-fence!" scoffed one former departmental Finance Director (57) at the mere mention of the phrase "ACs." On the Treasury side, several interviewees who had worked in spending control teams (e.g., 42, 58, 62, 104) indicated they did not take such controls very seriously, saying they were a joke or hardly mattered and that AC cost allocations were often settled last-and perfunctorily-in the final stages of spending reviews. One (120) said the AC controls were "stupid" and "pointless" and "got in the way of making sensible decisions." Several recalled that Treasury officials themselves had proposed scrapping the controls or applying them on an "earned autonomy" basis, exempting departments the Treasury considered to be well-managed from the control regime altogether but retaining reserve powers to put poorly performing departments "in the clinic." There were at least three major negatives about the operation of the AC controls for the official players. The AC controls partly duplicated other Treasury controls on public service pay and they were seen as a poor way of comparing or controlling "overhead costs" across government or over time because, despite frequent tinkering to include or exclude particular items or organizations, they were so inconsistently applied in practice. For example, some satellite bodies were subject to Treasury AC cost controls but not others; some inspectorates (such as those for schools) counted as programme costs while others (such as those for social services) scored as ACs, and the vast central bureaucracy of the Ministry of Defence (MoD) dropped out of the AC control regime altogether for a whole decade from 1994 to 2004 because the MoD operated its own way of distinguishing between "support" and "operational" functions. AC cost controls might have provided useful information for genuine benchmarking among departments and agencies if they had been cut down to cover just core information technology, human resources, and finance functions, but that radical step was never taken, such that ACs represented an awkward tension between the general costs of running any given public organization and the costs only of "back-office" operations, with a ragged, everchanging and inconsistent boundary between what was considered "frontline" and "back office." Indeed, the AC regime as it operated in practice was seen as often hampering rather than enabling Treasury efforts to limit the tendency of upper-level officials in spending departments to "top-slice" the budgets of those further down the food chain. For instance, in the early 2000s Treasury budget controllers thought there was "bloat" in the upper reaches of the prison service. But because the funding of both prison headquarters staff and those working in the prisons themselves counted as AC costs and the former tended to top slice the overall AC budget, 4 any squeeze on the department's ACs would be likely to hit hardest at the operational rather than HQ level-exactly the opposite of what the Treasury was trying to achieve.
Third, the way AC controls worked in practice was thought to provide perverse incentives for departments to over-outsource functions in order to reclassify the relevant costs as "program" rather than "administration," even when there was no value for money improvement from outsourcing. 5 On the face of it, such shortcomings look like substantial "push" factors for both Treasury and departmental officials, making it surprising that such an apparently flawed cost-control system survived both in times of fiscal expansion and consolidation. But when we dug a little deeper, both in documents and interviews with former officials, we found evidence of some "shy," passive or "contrarian" support for retention of the AC cost control regime which limited the "push" factor for that regime. First, some interviewees (116, 55) pointed out that AC cost controls provided a domain for Permanent Secretaries to preside and negotiate over in their role as managers of their departments.
And according to Treasury documents from the early 2000s, the AC control regime also provided departmental Permanent Secretaries and Finance Directors with a convenient way of bolstering financial discipline within their departments. In common with other input controls, AC controls enabled Finance Directors to respond to spending demands from their mission-driven colleagues by saying the Treasury would be unlikely to accept such demands, without the social awkwardness of having to say "no" directly to their colleagues (a classic "system maintenance" intermediation practice in bureaucracies; see Parry et al., 1997). 6 Such support for AC ring-fences might be considered "shy" or "passive," in that it was mostly not convenient for such views to be expressed too explicitly, and that it seems to have manifested itself in quiet toleration of the system and negotiations over departmentally specific classification issues (over who or what was to be counted as ACs or a special departmental system in the case of the Ministry of Defence) rather than a concerted and determined push from spending departments as a group to demolish the control regime in the name of "outputs not inputs" managerial freedom. So the "push" factor for AC controls seems to have been weaker at official level outside the Treasury than might at first appear.
As for Treasury officials, interviews (116) and documents indicate that at least some of the Treasury's central general expenditure policy spending control group concluded in the early 2000s that even if they could not be radically reformed, AC cost controls were on balance probably worth keeping in some form, if only for the reasonapparently flippant and contrarian-that no-one else had a good word to say about them, in public at least. Perhaps that position indicates some bureaucratic inertia, albeit with a twist. But the negative side of such controls was clearly recognized within the Treasury, the options of outright abolition or a move to an "earned autonomy" approach were seriously canvassed at that time, and we found no indication in documents or interviews that the official Treasury was collectively adamant about keeping those controls.

| Push and pull for the political players: Fair-weather output controllers, all-weather input controllers
What were the "push" and "pull" factors associated with input and output controls for the top-level "veto players," in the sense of ministers and the political civil service ("Special Advisers," who hold office during the tenure of their Ministers and whose job is to provide a partisan political input into policy and administration)?
As for "pull" factors exerted by output controls, interviews and secondary material suggest that a key political attraction of the output approach for Treasury and other ministers was the credit-claiming opportunities it presented. Most of the interviewees (e.g., 14, 48, 82) who had been involved in the development of PSAs in place of the OPA system in 1998 stressed the political importance of showing voters that the Labour government's increased spending on public services, partly funded by a substantial hike in National Insurance Contributions (social security tax) after the 2001 general election, was delivering demonstrable and measurable benefits, particularly in electorally salient domains such as health care, education and crime and justice. One interviewee (82) stressed the link between the PSAs and "conditionalities" in overseas aid payments.
At the same time, interviews with those veto players indicated there were limits to the political usefulness of output controls as credit-claiming opportunities. One former senior Labour minister (49) thought the political value of the PSAs was a wasting asset, with a useful life of only about 2 years. That assessment seems to imply that once the political novelty of the new output-metrics approach had worn off and the inevitable crop of negative media stories about fiddled numbers or missed targets had begun to accumulate (see Hood & Dixon, 2015), their political credit-claiming value declined.
The other and related political downside of the PSAs was the risk they presented to incumbent ministers if the targets were missed (Lopez-Acevedo & Krause, 2012). A graphic example of that risk is the fate of the Education Minister Estelle Morris, who resigned in 2002 after a series of fiascos, including a missed PSA target over improving school literacy and numeracy standards (Curtis, 2003). Partly because of that downside political blame risk, numerous interviewees thought output targets were more politically workable at a time of increasing public spending-when it was likely to be easier to show voters that incumbents were delivering "more for more"than in times of fiscal consolidation when public service performance standards are more likely to fall (reflecting an observation by Lopez-Acevedo & Krause, 2012). It is notable that PSA output targets got no mention in the 2010 Labour Party general election manifesto (other than a brief pledge to cut down on central targets), and that two former Labour ministers (61, 49) we interviewed were doubtful if the party would have continued with PSAs had it won the 2010 election.
As for the "push" factors associated with AC cost input controls, we have already mentioned that New Labour ministers in the late 1990s and early 2000s were dismayed about the steep increase in such costs and some of the quirks in their application. As already noted, functions that scored as ACs when carried out by directly employed staff in departments often came to be classified as programme costs when they were performed by private-sector contractors, loading the scales against in-house bids in outsourcing competitions within departments. This practice provoked complaints from public service labour unions, who as funders and backers of the Labor party had links with some key Ministers. More generally, the application of the AC control system was a time-consuming and contentious responsibility for the Treasury's public spending ministers (Chief Secretaries) in dealing with their senior party colleagues heading spending departments, and the inconsistencies in the classification were hard for them to defend.
Still, against those "push" factors making some ministers dislike the AC cost control system, there were two powerful blame-avoidance and credit-claiming considerations that working in the opposite direction, at least for Treasury ministers. And, in contrast to PSA-style output/outcome targets, those two factors limiting the "push" apparently applied both in times of fiscal expansion and of fiscal contraction.
One is that input controls on ACs provided Treasury ministers with credit-claiming opportunities to demonstrate to middle-ground voters how tough they could be on "back-office bureaucracy." Several sources indicate that AC cost controls had high political and parliamentary salience throughout the period covered by our study. For example, in the mid-1990s, Conservative Chancellor Kenneth Clarke claimed credit in three successive budget speeches for major cuts in ACs and civil service staffing numbers. In the following decade Clarke's Labour successor Gordon Brown (2004)  office" costs to put more resources onto the "frontline" of service delivery.
The other element working to restrain the "push" factor associated with the AC control system for the political players seems to have been a perceived political blame risk associated with any decision to scrap AC input controls.
Abandoning the controls without discontinuing the reported numbers would have run a political risk that any increase in such numbers would provide ammunition to political opponents aiming to show that incumbents were "soft on bureaucracy." On the other hand, scrapping both the controls and the numbers could cause the incumbents to be accused of "burying the evidence" of burgeoning bureaucracy. Nor were such blame risks imaginary: ACs were regularly raised in hostile questions by opposition parties in Parliament, meaning that Treasury ministers and their advisers could not avoid getting into the minutiae of the way the numbers were presented 8 and reclassifications were challenged by opposition parties in parliamentary debates as well. 9 Accordingly, from the perspective of the political "veto players" the survival of the AC controls at that potential "window" for a shift from input to output controls looks less like a case of simple "bureaucratic inertia" than of "inertia politics" (in the term used by Rose & Davies, 1994) to denote political strategies of sticking with inherited arrangements to limit blame), combined with positive credit-claiming opportunities.

| CONCLUSION
Returning to our opening question, "why do frequently criticized input controls survive in the management of public spending while apparently more enlightened output/outcome controls come and go?,' this push-pull analysis of the case of the UK over two decades suggests three general conclusions.
First, the waxing and waning of output/outcome-oriented controls, in the form of performance targets linked to spending allocations, fits with observations in the existing literature that such controls are more likely to be F I G U R E 1 References to "public service agreements" and "administration costs" (or similar) in House of Commons debates and questions 1992-2015 (data not available for 2005/2006) emphasized in times of fiscal expansion than contraction. Indeed, as shown earlier, the UK's move to government-wide "performance-oriented budgeting" (in OECD terminology) is broadly consistent with three of the conditions for such a shift suggested by the international PFM literature, and fitted precisely into a politicoeconomic "window" conducive to such a shift in emphasis. There is a striking contrast between the waxingand-waning pattern of performance budgeting in the UK and the more sustained shift to a "performance approach" over recent decades by the public auditor (in the form of "value for money" auditing). The analysis here would suggest that those contrasting institutional paths could be explained by the different "veto players" in those two institutional arenas, exposed to different blame/credit pressures-government ministers and SpADs served by civil servants in one case, a cross-party group of backbenchers interacting with an autonomous auditor in the other.
Second, what is not so easily explicable from the perspective of that literature is why centrally-operated input controls on ACs were retained in the UK's public expenditure management system, despite widely recognized defects and limitations of such controls from a "rationalist" or managerial perspective and declared ambitions, especially in the 1990s, to move from an input to an output focus in spending controls. We find that it was more the "inertia politics" of credit and blame rather than bureaucratic inertia simpliciter that was the decisive factor in producing that result.
Third and more specifically, the qualitative push/pull analysis applied here revealed evidence of reluctance among both officials and political veto players to let go of AC input controls at any point during the two decades considered here, although (as with the output-oriented controls) there was plenty of "finessing" of the reported cost numbers to fit a desired political narrative, of bearing down on bureaucratic "waste." The balance of perceived credit-claiming opportunities as against blame risks associated with administrative cost input controls seems to have been more stable for those political players over the economic cycle than it was for output/outcome-oriented controls.
Finally, what are the implications of this case study for the application of "performance budgeting" in other countries? Further comparative research is clearly needed. But there is no obvious reason why the pattern depicted here, of a political logic of "fair-weather" output control (reflecting credit-claiming opportunities in times of increasing spending) and "all-weather" input control (reflecting blame and credit imperatives to demonstrate toughness on "bureaucracy"), might somehow be unique to this case. Indeed, it can be argued, as we suggested earlier, that the politico-economic conditions of the UK in the late 1990s and early to mid-2000s were notably free from the three factors commonly mentioned in the comparative PFM literature as liable to hold back a decisive move to full-blooded performance budgeting and relatively less emphasis on input controls.
And if such a move was so limited in the UK, given its frequent portrayal (or caricature) as a "vanguard state" of the New Public Management movement, it raises the question of when, if ever, large, institutionally complex democracies of a comparable type can be expected to go beyond the "fair weather/all weather" approach to output and input controls observed here.