Thinking and Working Politically: Learning from Practice. Overview to Special Issue

Over the last 15 years, a set of ideas now referred to as &#8220;thinking and working politically&#8221; (TWP) has coalesced into a &#8220;second orthodoxy&#8221; about how to take context into account when implementing development interventions. This approach stresses the importance of obtaining a better understanding of the local context (&#8220;thinking politically&#8221;) in order to support local actors to bring about sustainable developmental change (&#8220;working politically&#8221;). However, the evidence base to justify this new approach remains thin, despite a growing number of programmes which purport to be implementing it. Officials in development agencies struggle with putting it into practice and it is unclear how TWP differs&#8212;or not&#8212;from similar approaches, such as Problem Driven Iterative Adaptation (PDIA) and Doing Development Differently (DDD). This Special Issue sheds light on what TWP means in practice by examining a set of initiatives undertaken by both development partners and government departments in Nigeria, the Occupied Palestinian Territories, China and India. This overview article outlines, in brief, each of the Special Issue's four papers and then draws out five lessons&#8212;for funders and for practitioners&#8212;from across all the papers. Our five lessons are: (1) the fundamental importance of undertaking political economy analysis (PEA) to adapt programmes to their contexts; (2) the importance of having a realistic level of ambition for interventions; (3) the need to support local ownership&#8212;not just &#8220;agreement ownership&#8221; (between a donor agency and government) or local &#8220;management ownership&#8221; of the programme, but critically &#8220;driver ownership&#8221; by generating trust with the key local actors driving change; (4) the need for a more effective set of tools for measuring results in complex programmes that attempt to achieve improvements in long&#8208;run governance; and, (5) that although the political economy of donors is often seen as a barrier to applying TWP, the articles show how much can be done with a TWP approach if the analysis takes into account the political economy of donors as well as that of the local context. We conclude with a set of operational recommendations for donors and implementors, as well as suggestions of avenues for further research.


| O3
MCCULLOCH and PIROn countries. It shows the extent to which these initiatives were successful, and attempts to disentangle the reasons for the success or failure of various initiatives-linking these to the broader narrative of TWP. In so doing, we hope to shed light on what TWP means in practice-its strengths and weaknesses as an operational concept; provide some lessons for practitioners and policy-makers; and identify recommendations, including where further research is needed.
The next section outlines, in brief, each of the four case studies. We then draw out in more detail five lessons-for funders and for practitioners-from across all the articles. Finally, we conclude with some recommendations, both for how funders and practitioners might apply the lessons identified, as well as for future research.

| The Facility for Oil Sector Transparency and Reform in Nigeria
The article by Lopez Lucia, Buckley, Marquette, and McCulloch on the Facility for Oil Sector Transparency and Reform (FOSTER) programme in Nigeria is one of a growing range of studies that attempt to draw lessons from both successful and unsuccessful interventions. The FOSTER programme consisted of a wide range of related but discrete interventions over a five-year period, all designed to improve the quality of natural resource (and specifically oil sector) governance in Nigeria. These included: work to strengthen the Nigerian Extractive Industries Transparency Initiative, technical assistance to government departments, public information tools to increase awareness of the problems in the sector, and work with parliament to support the development of legislation that would facilitate improved management of the sector, among many other interventions.
The large number of interventions provides an excellent opportunity to examine why some activities were successful and others failed within the same overall programming context. The article draws on extensive interviews, programme reviews and newspaper articles to understand the process through which interventions were implemented, the substance of what was done and the way in which the staff of the programme ensured that work responded to the shifting political context.

| How political economy analysis has influenced DFID's programming in Nigeria
The article by Williams, Owen, Duncan, Kingsmill, and Paterson in this issue also focuses on Nigeria, but with a much wider ambit. The authors provide a detailed historical account of the way in which political economy thinking has informed DFID's governance programming since Nigeria's return to democracy in 1999. It is a revealing story, since it shows how, over almost two decades, the successes and failures of programmes have fed back into a shifting understanding of what it means to think and work politically. DFID Nigeria was an early pioneer of political economy analysis (PEA)piloting one of the first Drivers of Change studies in the early 2000s that pointed to the institutional basis of Nigeria's problems and the need for a social and fiscal compact. At the same time, it led DFID to abandon optimistic assumptions about the strength of reform commitment following the return to democracy, and challenged the belief that building capacity would be sufficient to strengthen publicsector performance.
The result was that, from 2003, DFID's portfolio shifted to recognize the complexity of the political economy of the country by helping Nigeria better use its own resources. In particular, it innovated with issues-based programmes which engaged with all the relevant stakeholders on an issue in an attempt to build a stronger social contract in which politicians would be held responsible for O4 | MCCULLOCH and PIROn the provision of public services rather than the distribution of patronage benefits. Williams et al. describe how this approach was then put into practice in a series of governance programmes at the federal and state levels, and how each in turn influenced thinking about how to think and work politically.

| A dual-track approach to Public Financial Management reforms in the Occupied Palestinian Territories
Pijuan describes the challenges, and surprising success, of introducing public financial management (PFM) reforms in the Occupied Palestinian Territories (OPTs). The article reflects on the two distinct waves of PFM reforms in the Palestinian Authority and how they were shaped by the macro political context, notably the Second Intifada and then the Hamas government from 2006 onwards. It then describes one programme, the Palestinian Governance Facility, which started in 2012 during a period of political turmoil. Three different Ministers of Finance during the course of the first year made it very difficult for the programme to align with a government strategy for reform. In response to this challenging context, the advisory team took a dual-track approach. It invested time and effort in gaining in-depth understanding of the political dynamics in the OPTs, discussing how change processes were likely to occur and constantly scanning the context to identify opportunities for reform. It then designed an approach which attempted to tackle the long-term problems with the planning and budgeting system. At the same time, the Deputy Minister of Finance identified escalating external health referral costs as a major political problem for government. The team therefore picked a single therapeutic area, cardiac catheterization, as an area to pilot reform. It conducted a PEA, breaking down the referrals problem into its root causes, and drew on the expertise of a UK cardiologist to get all the OPT's cardiologists to agree a new set of referrals guidelines. This led to a modest, but useful, improvement in the efficiency and cost of such referrals.

| Thinking and working politically by the climate change departments in China and India
Most of the literature on TWP has focused on donor-funded programmes (Laws & Marquette, 2018). The article by Harrison and Kostka shows that similar strategies are adopted by agencies within developing country governments to influence wider government agendas. The authors use research on climate change mitigation-and, in particular, energy efficiency-in China and India to explore how government agencies seek to overcome challenges of limited capacity and competing priorities by bundling climate change together with more immediate priorities. Despite significant differences between the two countries in terms of their level of development, state capacity and emissions levels, they show that both used similar tactics, which they refer to as framing and bundling, to reconcile climate change mitigation with competing policy priorities, and to leverage existing institutional structures to achieve their policy objectives.
Framing is a device to reduce the potential controversy around national policy positions by packaging priorities in ways that seek to reconcile competing priorities. For example, since 2012, air pollution in China resulted in a shift of its framing of climate change issues towards a focus on sustainability and environmental stewardship (Gippner & Torney, 2017). By contrast, India's framing of climate change issues focused on the "challenge of sustaining its rapid economic growth while dealing with the global threat of climate change" (Government of India, 2008). Harrison and Kostka show that these framings then provided the basis for bundling strategies that sought to harness existing policy | O5 MCCULLOCH and PIROn mechanisms (policy bundling) and align the interests of stakeholders with energy efficiency objectives (interest bundling).

| LESSONS LEARNED FROM THE CASE STUDIES
The case studies provide a rich set of lessons for development practice. We pick out five that we see as having more general applicability: 1. These articles reconfirm the fundamental importance of "thinking politically" through PEAs, but go beyond this to highlight how this can be applied in different ways to help programmes adapt to their contexts. 2. The case studies show how "working politically," by moving from analysis to operational implications, requires deciding on a realistic level of ambition and doing the things that can be done in a particular context and point in time. 3. The articles illustrate how supporting domestic policy change requires careful attention to working with the right counterparts or coalitions to generate sufficient trust with those who will be driving change. 4. We reflect on whether these case studies are able to demonstrate that TWP is more "successful" than conventional approaches. While the articles all point to "islands of success" that can be achieved through TWP, we still lack an effective set of tools for measuring results for complex programmes that attempt to achieve improvements in long-run governance. 5. The political economy of donors is often seen as a barrier to rolling out TWP. However, the case studies show how much can be done with this approach if the analysis takes account of the political economy of donors as well as that of the local context.
We elaborate on each of these further below.

| Thinking vs assuming
'Thinking' politically requires understanding the local context and the competing factors that drive or block change-from human agency (the interests, incentives or ideas of individuals or organizations) to institutional or structural factors (North, 1990). This explicit focus on understanding the political economy contexts of development interventions is probably the main difference between TWP and PDIA/DDD approaches (Teskey, 2017), with the latter more attractive to organizations that are less comfortable with an explicit reference to power and politics. Such analysis is not limited to external aid programmes. For example, as Harrison and Kostka show, domestic agencies responsible for setting or implementing policy need to consider their pathways for influence and room for manoeuvre. While they may not use formal or structured analysis, an understanding of the wider "rules of the game" and how these rules are followed in practice by different actors guide officials' tactical choices. Similarly, Williams et al.'s overview of the DFID Nigeria country programme illustrates how political economy analytical tools have evolved to provide advice of greater operational relevance for development programmes. One of the first comprehensive "Drivers of Change" studies was commissioned by DFID in the run-up to the 2003 elections to help DFID decide how to support Nigerian reform efforts (Heymans & Pycroft, 2003). It argued that, in the absence of changes in the institutional rules and structural conditions, capacity building and formal system reforms would only make modest contributions. This analysis influenced O6 | MCCULLOCH and PIROn DFID's Country Assistance Plan 2004-2008 and individual programmes were then designed to operate within the political economy constraints identified. The analysis also generated insights on the space for development partners to promote change at the federal and state levels, not only to design programmes, but also to decide in which sectors not to engage, for example if policy capture was intractable. PEA started to be undertaken and used regularly by state-level programme teams and their stakeholders, such as the State Accountability and Voice Initiative (SAVI) and State Partnership for Accountability Responsiveness and Capability (SPARC), rather than one-off detailed studies undertaken by consultants disconnected from programmes (Derbyshire, Fraser, & Mwamba, 2014;SAVI 2015). It led to greater differentiation between state programmes and thoughtful adaptation to the reform space.
Lopez Lucia et al. also demonstrate the benefits of using PEAs to inform interventions, the need to combine both informal and more structured analysis and the need to choose the right level to focus PEAs. FOSTER commissioned quarterly PEAs to understand policy areas in the oil and gas sector that might be amenable to reform, as well as broader changes in the socio-political context. They were based on stakeholder analysis, power-mapping, key informant interviews and quantitative analysis, and were validated through stakeholder workshops to consider interventions for programming. The FOSTER core team also undertook informal assessments as part of their daily work, focusing on some of the key individuals in the sector. Reflecting on some of the less successful activities, the FOSTER team identified the need for intermediate analysis-not simply at the level of a single intervention nor at the overall socio-political context-but at the "cluster" level (that is, around a particular area for reform of the oil sector). This would allow an analysis not just of change makers ("agency") on which informal assessments focused, but also of their incentives and influence given the broader context ("structure") which was occasionally weak for FOSTER.
The articles all offer a cautionary note: PEAs are not a panacea, but part of an attempt to make reforms supported by external actors more likely to succeed. As Williams et al. conclude, PEA must tread a careful line between the hard-to-eradicate expectation that it is a short-cut to "picking winners" and communicating its real value: shaping programming in a way which continually enables winners to be picked.

| Doing the things that can be done
The articles in this Special Issue all illustrate how analysis of the context and opportunities have influenced the levels of ambition of programmes. In other words, they have led to more realism and a focus on doing the things that can be done. Many of the case studies describe situations in which there are deep-seated political reasons why the system being addressed is dysfunctional. In the OPTs, the wider regional conflict, limited Palestinian self-governance, territorial fragmentation and associated political manoeuvrings severely constrained possible PFM improvements. In Nigeria, the entire political settlement, based on rents generated by oil and gas, and the patronage system sustained by it, limited the willingness to reform public administration, both in general and in the oil and gas sector.
This focus on the "doable" in TWP programmes stems from two realizations. First, some aid programmes suffer from a degree of hubris, promising to stimulate growth and end poverty in one short, small intervention. The TWP approach encourages a more honest reflection about a programme or institution's ability to influence the wider environment. Occasionally, it may be possible to do things which are transformational-in the sense of durably changing the rules and behaviours associated with a particular policy objective. In most cases the scope for change is more limited; programmes that have identified and focused on what can be changed have been more successful than those with grander but unrealistic ambitions (Levy, 2014).

MCCULLOCH and PIROn
In all our country examples, programmes did not try to influence the broader political context, but they could still achieve some significant practical improvements when the policy environment was conducive. Working politically entailed developing a deep understanding of that context in order to be able to seek to achieve progress. The individual articles show how the various programmes focused on targeted reform areas, such as enhancing the transparency of fuel subsidies (FOSTER in Nigeria), improving the accuracy of budget data (OPTs), streamlining referral procedures (OPTs) or setting up energy service companies (India). The level of ambition depended on the room for manoeuvre, which changed over time. For example, Lopez Lucia et al. show that, while the Jonathan Administration in Nigeria (2011Nigeria ( -2015 was characterized by policy paralysis and uncertainty -and in some cases wilful resistance to fundamental oil and gas sector reform, the 2015 Buhari election was seen as an auspicious moment for radical petroleum reform and FOSTER therefore adjusted its approach to these different contexts (see also Lopez Lucia, Buckley, Marquette, & McCulloch, 2017).
Second, the focus on achieving practical (though less ambitious) results can provide a window for longer-term reforms. Reforms often occur obliquely, starting from an initiative that was not designed to tackle the more fundamental problems but, for precisely that reason, was more feasible. Occasionally, small "pockets of effectiveness" can be created which can be platforms for wider reform initiatives. For example, Pijuan notes that, after 2012, there was no longer a strong political drive to improve PFM in the West Bank. The Palestinian Governance Facility used a step-by-step "strengthened approach to programme budgeting" to generate a wider programme of PFM reform. It started by raising awareness of planning and budgeting issues with Ministry of Finance staff, which resulted in the constitution of a team of "reform entrepreneurs" within the Ministry, who introduced first new systems and processes, and then a complete budget preparation cycle by 2015. Thus, Pijuan shows how a significant improvement was achieved even though senior management and political support were still tenuous.
3.3 | Identifying the right partners and coalitions to support "driver ownership" Booth and Unsworth (2014) argue that successful TWP programmes are both "locally owned and politically smart." Donor programmes managed by foreign private-sector organizations face more difficulties than national organizations (such as a government department or a local non-governmental organization) because they are, by definition, not rooted in the country context. By contrast, local actors are more likely (although by no means guaranteed) to demonstrate the required political expertise, as Harrison and Kostka put it: "… leaders who are well embedded in their localities or policy spheres are more likely to get things done by making context-specific policy adaptations." But what does local ownership look like in practice in TWP programming? At one level, the term is used to mean that there is a formal agreement to a programme, for example with the Minister of Finance of the Palestinian Authority or with the Nigerian government at the federal or state levels. However, the case studies suggest that such an agreement is neither necessary nor sufficient for success.
In some of the TWP programmes described, local ownership refers to "management ownership." In Nigeria, this was done by drawing on national teams with the relevant technical knowledge (public administration for SPARC, the oil and gas sector for FOSTER) as well as a deep knowledge of the political context and excellent networking capacities. By 2015, FOSTER was implemented by an entirely Nigerian in-country team based in Abuja (though overall management responsibility still lay with an international contractor). The core FOSTER team remained relatively stable and could therefore O8 | MCCULLOCH and PIROn develop deep relationships and knowledge of the key actors in the sector. Lopez Lucia et al. note that "the FOSTER team had the freedom to think, operate and communicate in a politically informed way on a day-to-day basis." However, Lopez Lucia, Buckley, Marquette, & McCulloch (2017) say that "Finding the right people was not an easy task" given the required skills and competencies, and they also raise the important question of how "programmes better retain and reward talented, in-demand country staff over the life of the programme?" However, the experience from our case studies suggests that, while "agreement ownership" and "management ownership" matter, the most important type of ownership is "driver ownership"-that is, driven by a group of local actors who are committed to a reform agenda and would pursue it regardless of external support. Ownership as "driver" is most obvious in the China and India case studies, where central agencies navigate their domestic contexts in order to implement their own policy priorities. However, the case studies show that it is key to externally supported programmes too. For externally funded teams to support locally driven reform, it is essential that they generate trust with the recipients of the assistance. Trust matters for two reasons. First, it is needed for stakeholder engagement, which is central to success. Pijuan discusses the importance of earning the trust of key officials in the OPT; and trust was a critical element of the effectiveness of the FOSTER programme in Nigeria. It was also essential for the formation of informal coalitions in China and India which were not just about aligning interests but also about knowing who to approach, how to approach them and being able to secure their trust.
In one sense, this is hardly surprising-all effective projects require trust between the client and the provider. However, what is striking is not just the importance of trust, but the fact that the ability to build trust is seen by several authors as more important than technical skill. Certainly, technical skill was required for all of the programmes-but in the situations described in these articles, it was not the most important factor or even, in some cases, essential. This is because, when deep technical skill is required, it can be brought in, as Pijuan's example of engaging a UK cardiologist in the OPTs shows. Conversely, without a close and trusting relationship with the partner, no amount of technical skill was able to solve the challenges that the programmes faced.
Second, trust provides a licence to tackle the more fundamental issues which are often difficult because of political sensitivities. Building trust by successfully tackling more prosaic issues can provide openings later to make progress on more critical issues. Sometimes these can go hand in hand-for example Pijuan describes how the programme in the OPTs attempted to tackle the long-term problems with the planning and budgeting system, whilst simultaneously delivering something practical in the short-term [on service delivery]. Tackling costly external medical referrals provided stronger legitimacy to continue to address long-term PFM issues.
The articles also show that supporting "driver ownership" is not about generating trust with a single individual, but rather enabling coalitions with individuals and organizations that can support change; finding ways to work with them over time and, perhaps more controversially, identifying how to overcome potential resistance, as in the Harrison and Kostka article. It provides a detailed examination of the informal coalitions that have been central to the implementation of national climate change mitigation measures in both China and India. Ministries and agencies dealing with energy efficiency have had to build coalitions of support with stronger agencies, local government officials and the private sector through "policy bundling," combining the less popular issues on which they are working with other policies that may have greater salience to key political players. In India, this has been achieved by framing actions on climate change mitigation in terms of the "co-benefits" they bring, such as promoting energy security through renewable energy or financial savings made from achieving greater energy efficiency. Coalition formation also requires "interest bundling," bringing different interest groups around a particular policy objective in "win-win" situations. In China,

| O9
MCCULLOCH and PIROn "creative manoeuvres" by local leaders include strengthening formal and informal incentives to bring the interests of enterprises in line with those of the state, to achieve centrally set targets.
How can external aid agencies support the creation of effective coalitions? Finding the right counterparts is central. The case studies describe the advantages and challenges of working with counterpart government institutions. Unsurprisingly, where donor programmes have a strong and supportive counterpart, rapid progress is possible, as was the case between 2002 and 2006 in the OPTs according to Pijuan, under a technocratic and reform-minded Minister of Finance who was able to introduce comprehensive PFM reforms centred on budget execution to control salary payments, a political and donor priority. However, the case studies mostly document the far more common occurrence of considerable instability in leadership and influence of key counterparts. It is hard to make progress when, as Pijuan recounts for the OPTs, you have three Ministers of Finance in one year. Programmes faced with such uncertainty need to build alliances with a range of different actors, navigating the political space to find pathways through which progress might still be made despite the unpromising formal context. Pijuan describes how the Palestinian Governance Facility took time to understand the political economy context and build a web of relationships to enable it to operate. It concluded that '[t]he ideal coalition involves flexible, aligned donors and relevant "institutional entrepreneurs" from within local organizations, with direct access to politicians. This coalition should be involved in analysis and reform implementation. If high-level government buy-in falters, a coalition of technocrats may help broker and construct problems from which to gain political access." Lopez Lucia et al. present the radically different solution adopted by FOSTER: not having a government counterpart at all. Rather, the programme was designed to work with any counterpart agency that was interested in undertaking reforms. Almost uniquely among aid programmes, the choice about whether to work with an agency (or not) was vested in the local team. They were then able to choose counterparts who were serious about making progress and, critically, withdraw from working relationships with agencies when changes in personnel or policy made it clear that no further progress would be possible. Most donor projects are not designed with this level of flexibility (and many governments would not permit it), but Lopez Lucia et al. show that this ability to switch between potential counterparts in government and to work outside the government entirely was a key component of the success of the first phase of the programme.
Regardless of the agreement between the donor and the government, or the local management of the programme, the case studies make clear that success depends on reforms being driven by a coalition of local actors with a strong sense of ownership of the changes that they wish to bring about.

| Redefining what success looks like
Probably the most important operational question in terms of the wider adoption of TWP over more conventional approaches is whether it does indeed deliver better results, not just more sophisticated processes of analysis, adaptation and relationship management. A recent review of the evidence found that TWP studies 'rarely focus on outcomes, instead focusing on the reform and/or programming process instead. Few studies discuss criteria for "success" or how they are measured' (Laws & Marquette, 2018, p. 7).
The case studies in this issue grapple with this challenge. While they do not offer formal counterfactuals, they are able to compare different approaches to policy goals or regions within the same country over a similar period. In the OPTs case study, Pijuan contrasts traditional "gap-filling" budget support aid to the OPT administration with efforts to change how a system operates: 'if donors had merely continued to cover the external referrals cost, this would have provided more revenues for interests that abused the system, increasing their stake in its continuance and ability to block reform'.
He can also demonstrate the financial gains from the reduction in external medical referral expenditure in the OPTs. Williams et al. point to efforts to demonstrate results by comparing Nigerian states supported by TWP programmes to those that were not. Although the overall performance of reforms supported by SPARC was mixed, an econometric study pointed to better PFM outcomes in SPARCsupported states compared to those that did not have SPARC support (SPARC, 2015).
The more fundamental question that TWP raises in terms of results is how they are defined in the first place: not just in terms of measurable increases (such as the number of schools built or nurses trained), but also whether 'institutional relationships have begun to shift in ways that make such breakthroughs more feasible in the future' (Booth & Unsworth, 2014, p. 7). Reviewing the available public evidence on DFID Nigeria's governance programmes, the Williams et al. article concludes that programmes which adopted a TWP approach did lead to a wide range of results, but that most were limited in scale and should be considered to be "islands of success" rather than examples of "transformational change." The distinction between short-term and long-term results is essential, with results in short-term service delivery often presented as an entry point to tackle more challenging institutional issues and vested interests over a longer period (as in the OPTs example). Results may well be "second-best" outcomes that are shaped by the need to navigate multiple constraints and combine different policy objectives, rather than providing a clear and direct response to a particular policy issue. As Harrison and Kostka note, well devised strategies focusing on changing incentives may not yet deliver longerterm climate mitigation objectives, such as short-term targets in China that do not produce sustainable emissions reductions or the production of energy audits rather than the introduction of efficiency measures in India. Should such results be counted as successes-because they provide an entry point for further reform-or failures, because they do not deliver the objectives originally specified?
If success is defined in terms of institutional changes that make the achievement of long-term policy objectives more likely (Booth & Unsworth, 2014) and, at the same time, there are no predetermined institutional models to be adopted, 1 then judging whether a programme or reform process is institutionally on track becomes very difficult. Indeed, it requires another level of political analysis to see whether and how incentives are changing and how success is defined in a specific context. The case studies illustrate the innovative tools used to assess progress in such situations (such as a more flexible use of logframes or outcome harvesting in the OPTs and Nigeria). Several also point out the critical role of failures to refine the most appropriate approach. For example, Lopez Lucia et al. explicitly examines both failures and successes (a separate study, Lopez Lucia, Buckley, Marquette, and McCulloch (2017), provides more details). This portfolio approach tried 'a range of approaches that seemed promising with the knowledge that not all would necessarily be successful'. Failure is not regarded so much as a risk as an inevitability. The challenge is not to avoid failures, which would be impossible, but to ensure that the programme undertakes a broad enough set of initiatives to maximize the likelihood that some initiatives will be successful. This approach goes along with a critical and adaptive approach that attempts to identify when initiatives are not working as early as possible and adjust as one goes along, to "fail fast"-in the now common phrase- (DiPiro & Chisholm-Burns, 2013). However, as Woolcock (2009) points out, reform trajectories are not necessarily linear and can even follow a J-curve in which things get worse before they get better, making early judgements about what interventions are most likely to achieve the long-run institutional objectives even more fraught.

MCCULLOCH and PIROn
The key lesson is that we are still some way from an effective set of tools for measuring results for complex programmes that attempt to achieve improvements in long-run governance, but that the practical approaches adopted in these case studies provide some examples of methods that may be helpful.

| The political economy of donors
Finally, the case studies confirm an uncomfortable lesson that funders tend to ignore, though noted in the literature (Brinkerhoff, Frazer, & McGregor-Mirghani, 2018), that the political economy of donor agencies themselves influences the chances of a TWP programme succeeding. The Nigerian and OPTs programmes depended on DFID being willing to operate differently throughout. This entails a higher appetite for risk, the willingness to learn and adapt programmes over time, and delegating decision-making powers to local teams over which reforms and stakeholders to support. At the same time, the articles show how these programmes were not immune from other pressures which made the adoption of a TWP approach more difficult within DFID (Valters & Whittey, 2017). These include the need to demonstrate quick results to satisfy the immediate concerns of ministers and the UK public, or shifts in the allocation of UK resources for reasons unrelated to the opportunities to support change (such as the expansion to Northern Nigeria in 2009 and changes in state-level engagement in 2016 in ways that did not reflect PEA recommendations on the prospects for reform -as analysed by Williams et al. in this issue).
Pijuan highlights the mismatch that can result from the donors wanting to endorse TWP while remaining embedded in non-TWP norms and incentives, such as the tension between the criteria used by funders to award contracts, and the importance of trust, as described above. Typically, local knowledge and the ability to build an effective and trusting relationship are not as important in donor "scoring criteria" for contracts as technical prowess.
Similarly, Lopez Lucia et al. suggest that an important part of building trust is the extent to which the implementing organization is seen as "arms-length" from the donor (Booth, 2013). Local partners want to know that the implementing team is "on their side" and not merely representatives of a foreign funder. The FOSTER programme went to the extent of adopting a "discreet" approach, with a policy of not branding its activities to enable it to form a wider range of partnerships. 2 This raises the issue of to whom the implementer is accountable if it is both arms-length from the funder and does not necessarily have a fixed counterpart. This distance also puts implementing organizations in a difficult position since they must try to fulfil the requirements of their funders, while at the same time being seen to be somewhat independent of them. To make this work, significant trust is required by the funder in the implementing organization, enabling it to experiment and explore different approaches (Brinkerhoff et al., 2018).
Learning from failure through a mixed portfolio also poses challenges for donors. A portfolio approach may be rational for a private investor seeking to minimize risk and maximize gain-but the transparency required of aid programmes means that any failures will receive scrutiny and, very probably, criticism. This makes aid officials averse to risk, particularly if a failure poses broader reputational risks for the donor. In both Nigeria and the OPTs, development interventions were subordinate to the wider diplomatic relationship between the UK government and the respective governments. This has implications both for the types of interventions undertaken and the manner in which they are implemented. In particular, funders keen to preserve good diplomatic relations are rarely willing to use aid to "serve as an asset for those seeking to disrupt entrenched elites" (Yanguas, 2018, p. 211), O12 | MCCULLOCH and PIROn even where such disruption might promote developmental objectives. Even if programmes do not seek to disrupt the status quo, the risk of failure can lead to a desire by the funder to control the detailed delivery of programmes through milestones and payment by results. Such behaviour runs counter to the flexibility and risk-taking that has been critical to the successes observed in the case studies.
However, the overall lesson on this issue from these case studies is not that the political economy of donors does not influence programmes, but rather the reverse, that it is remarkable how much can be done with a TWP approach if the analysis takes account of the political economy of donors as well as that of the local context.

| RECOMMENDATIONS
The lessons that we have drawn from these case studies indicate two types of recommendation: those for practice (by donors and implementers) and those for further research.
Our recommendations for practice are threefold: First, the case studies reconfirm the importance of doing political economy analysis. While methodological debates continue regarding the most appropriate types of analysis for different purposes, all of the case studies support the idea that careful, explicit, objective analysis of the political, economic and social context in which interventions or reforms are being undertaken is time well spent, as much because it can help stop doing things that have no chance of success as because it points to approaches likely to succeed. Funders should ensure that PEA is embedded in programmes as an ongoing process rather than simply as an upfront assessment or part of an evaluation.
Second, while there is widespread agreement that "ownership" is key to success, there is often much confusion about what ownership means. We distinguish three different types of ownership: (a) agreement ownership-such as a compact between a donor and recipient government, (b) management ownership-where management autonomy is provided to a team of local development entrepreneurs, and (c) driver ownership-where local actors who are already strongly motivated to drive the reforms take leadership of the initiatives. While traditional aid programmes have focused on agreement ownership, and PDIA has stressed management ownership, we suggest that the successes of a TWP approach arise primarily when local drivers lead the way. Donors should focus more on how they can best support the efforts of the coalitions that are driving reform, and not just attempt to instil ownership into those with the formal responsibility for implementing it.
Third, we recommend that donors take seriously the need to revise their models for contracting teams to manage programmes to put much greater focus on the skills needed to build trusting relationships (if necessary, at the cost of strong technical skills). The case studies provide convincing evidence that the ability to build trust with local partners is central to facilitating positive change, yet many funders still provide strong incentives for implementing organizations to form teams based on years of technical experience.
In putting forward recommendations for further research, we are conscious that the recent review by Laws and Marquette (2018) of the evidence on the integration of politics in development practice has already produced a comprehensive set of recommendations on this issue, with which we concur. We therefore highlight only three recommendations that derive from the case studies in this Special Issue.
First, further research is needed into how success is defined and measured for interventions that are designed to achieve long-term governance or institutional reforms. Recent years have seen two diverging narratives in development practice-one focused on the need for locally driven, politically feasible institutional reforms, the other stressing evaluation of impact and payment by results. Yet, as | O13

MCCULLOCH and PIROn
McCulloch, Barnett, Duncan, Kingsmill, and Kydd (2017) point out, the methods typically used for the latter can create an obstacle for the achievement of the former. The solution is not the abandonment of quantitative rigour where feasible, but rather the development and application of rigorous qualitative methods that are appropriate for the complex, non-linear, "small-n" problems that are often associated with institutional reform.
Second, the implicit claim of advocates of TWP is that it is more effective at achieving long-term reforms than more traditional approaches. However, the evidence base for this is thin (Piron et al., 2016;Yanguas, 2018;Laws & Marquette, 2018). There is an urgent need for research that could help to identify the kinds of problems and the sorts of context where a TWP approach is more or less effective. This might require comparisons between TWP and non-TWP approaches in the same sectors and countries.
Third, we need to find a way of learning more from failures. The incentives facing donors and implementers make them very reluctant to admit failure (Saito-Jensen & Pasgaard, 2014). However, systems and processes can be put in place which make it easier to learn from failure. For example, studies can be published after a period of time has passed, new programmes can be required to explain how they have taken into the account lessons from previous programmes prior to approval, and funders and implementers can create "safe spaces" in which staff can speak openly and honestly about what went well and what did not without fear of sanction.
We hope that the lessons and recommendations derived from the case studies in this Special Issue help to strengthen the evidence base and improve the ability to think and work politically in development practice.