The rise of multi-bi aid and the proliferation of trust funds

Since the end of the Cold War, development assistance has been transformed in various ways. The aid allocation and the aid effectiveness literature intensively discuss the qualitative reorientation from geopolitical towards actual poverty reduction objectives and the donors’ related geographical and sector choices. The introduction of the Millennium Development Goals (MDGs), the World Bank’s poverty reduction strategies (PRS), and the new principles for aid developed in the context of the Paris Declaration have all triggered important dynamics that also found a corresponding reflection in the literature. In parallel, there has been a much more silent revolution of funding mechanisms, widely discussed within aid agencies, but so far without any significant analysis in the academic literature. While donor countries traditionally face a binary choice between two channels for official development assistance (ODA), namely, the bilateral and the multilateral channel, they now increasingly opt for a combination of the two, generally called ‘multibi’ aid. In this context, they channel funds to an international development organization (IDO), a multilateral agency that implements development activities,2 but without providing the IDO with the authority to spend these funds at its own discretion. Owing to this earmarking to specific areas in which the funds may be used, multibi aid differs substantially from traditional core funding to multilaterals. A further difference consists in the voluntary nature of multibi contributions that provide much more flexibility to the donor government. Multibi financial contributions to IDOs are typically independent of longterm agreements at the international level. In contrast, they are released on a shortterm basis, and they usually flow into separate trust funds with individually designed governance and management structures that may also allow for membership of private organizations, for example, foundations or nongovernmental organizations (NGOs). These trust funds can be established for individual donors, but also for a group of donors that jointly finance a common priority area. They are administered and often directly or indirectly crosssubsidized by the IDOs, but not (or only partly) subject to the IDOs’ own governance and monitoring mechanisms. One of the numerous examples for such parallel governance is the Nordic Trust Fund (NTF). According to an informal communication by the trust fund manager, the US$20 million multidonor fund was entirely driven by Scandinavian countries to mainstream human rights activities into World Bank projects. Neither the fund itself, nor its freestanding projects formally required approval by the board of executive directors, although – because human rights is at times a controversial topic – the Nordic executive director reached out to the other board members and sought their ‘no objection’ to establishing the NTF. Formally, the NTF work programme

is simply based on an approach paper and an administrative agreement, jointly agreed among the donors and the World Bank as an administrative agent who chooses the individual projects (for more information on the NTF, see also World Bank, 2014a).
Starting from nothing in the early 1990s, this new type of aid has grown massively over the past 20 years. With a volume of US$19 billion in 2012, multi-bi aid today amounts to almost 60 per cent of the volume of multilateral aid and to roughly 20 per cent of bilateral aid (see section 4, Figures 30.2 and 30.3). In some organizations, the dependence on multi-bi aid has become extreme. For instance, the share of multi-bi finance has grown to 80 per cent in 2010 for the United Nations Development Programme (UNDP) (UNDP-EO, 2012: xii), and to 85 per cent for the World Food Programme in 2012 (Tortora and Steensen, 2014: 8). At the World Bank the overall number of active trust funds was more than 900 in June 2013 (World Bank, 2014b).
Hence, with the growth of multi-bi aid over the last years, the multilateral system has been witnessing a proliferation of new (semi-)autonomous institutions, blurring the distinction between bilateral and multilateral aid with potentially important consequences for aid effectiveness, donor fragmentation and international governance, and posing a severe challenge to the management capacity of existing IDOs. Concerns have grown that multi-bi aid limits the ability of IDOs to create independent knowledge as it opens the door for undue donor influence, or that it increases transaction costs through excessive auditing and additional reporting to donors (The 1818Society, 2012. In the following, we review the different donor reports and policy papers as well as the very few academic studies on multi-bi aid in order to provide an overview of the reasons for the development of this new trend and the opportunities and challenges it presents (section 2). As the limitations of the existing literature are primarily due to a lack of comparable data, we then introduce a new dataset developed in consultation with statisticians from the Development Assistance Committee (DAC) at the Organisation for Economic Co-operation and Development (OECD). Based on these data, for the first time, overall trends in multi-bi aid can be described in detail for a prolonged time horizon and at a global level, rather than only for individual IDOs. We start by a general description of the data and a discussion of underlying definitions relevant for any comparative analysis (section 3). We then present some initial descriptive statistics and simple statistical tests to provide the reader with a first idea about the empirical evidence for some of the problems typically discussed in the context of multi-bi aid (section 4). In particular, our analysis seeks to shed light on three questions: (1) is multi-bi aid additional to existing funds or does it substitute for traditional aid channels? (2) Do multi-bi funds flow into substantially different areas than multilateral aid, and how strongly are they earmarked by donor governments? (3) Is multi-bi aid as volatile as some observers fear? We find that multi-bi aid can support harmonization among bilateral donors, mobilizes additional donor resources, and complements existing multilateral activities. However, earmarked funds burden the multilateral system with additional transaction costs and risk a 'bilateralization' of multilateral aid. Section 5 provides a further critical discussion on these issues, and section 6 concludes.

LITERATURE REVIEW: THE RISE OF MULTI-BI AID AND THE PROBLEMS IMPLIED
The two defining characteristics of multi-bi aid, namely that funding is voluntary, and that funding is earmarked, can be attributed to distinct historical developments. The trend to provide voluntary funding to international organizations in order to complement their core budgets is much older than multi-bi aid. In fact, voluntary funding emerged as early as the 1960s when Western donors advocated the expansion of multilateral activities into development, but faced the opposition of the USSR regarding a general UN budget increase for that purpose (Graham, 2015: 178f.). This disagreement led to the creation of the several large UN development funds and programmes, for example the UNDP (Auerbach and Yonekawa, 1979: 510), which were largely financed by voluntary contributions. Generally, de-colonization broadened the membership of the United Nations (UN) General Assembly, leading to the relative decline of the Western vote share, and voluntary payments represented a way to re-establish the larger contributors' influence through the backdoor. Both the new funds and programmes, and the existing UN specialized agencies (such as the World Health Organization or the International Labour Organization) soon depended on periodic replenishment and annual voluntary contributions and hence on the donor countries' individual funding decisions (Graham, 2015: 181f.;Panneels and Beringhs, 2005: 2). The rise of multi-bi aid started in the early 1990s when earmarking emerged on top of voluntary funding. Its rise is related to changes in the geopolitical power structure, to new challenges requiring international cooperation, and a lack of capacity as well as legal constraints of the existing multilateral institutions in addressing these challenges.
The existing literature presents two distinct narratives that both start with the end of the Cold War. The first primarily looks at bilateral aid agencies and their attempt to engage in a major reform of development assistance for a new era in which geopolitical considerations would not be so important any more. A large consensus emerged that effective development assistance would require greater coordination and harmonization of aid between donors and the recipient countries' ownership of all development interventions. This recognition eventually led to structural commitments by bilateral donors expressed in the 2005 Paris Declaration (reconfirmed and monitored at follow-up conferences in 2008 in Accra and in 2011 in Busan). Joint trust funds hosted by IDOs appeared to be the ideal instrument to support a common cause by a group of donors (Barakat, 2009: 108;Barakat et al., 2012: 2f.;Woods, 2005: 394). A trust fund provides a forum to coordinate, harmonize decisions, and avoid the fragmentation of individual donor interventions -while financing specific issues or countries. Moreover, governance of the funds could include all relevant stakeholders, not only sovereign donors and multilateral organizations, but also private donors (NGOs or foundations), recipient representatives, and civil society actors, so that decision-making processes would ensure greater ownership (Barakat et al., 2012: 15;UNDP-EO, 2012: 59f.). Trust funds can also provide the opportunity to include the so-called 'new' donors that do not otherwise participate much in the exchange between donors (Eichenauer, 2015). For the legal aspects of trust fund governance see Bantekas (2009).
When the international attention started shifting towards global public goods, for example, in the areas of security or climate change, trust funds also appeared as an ideal B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account means to pool resources (Heimans, 2004;IEG, 2011). Trust funds allowed groups of countries to jointly become active in areas in which they would not venture alone, and in which the IDOs administrating the funds would not be willing or, given their legal restrictions, not be allowed to engage directly (IEG, 2011: 5). For example, the World Bank may only engage with member state governments so that territories such as West Bank and Gaza, or previously Kosovo, cannot obtain core funding. In addition, in difficult contexts such as fragile states and post-conflict countries, the use of trust funds allows donors (including IDOs) to pool the risk of their intervention (Guder, 2009: 37). Finally, trust funds enable countries to act fast, for example, in situations of emergency, without having to wait for lengthy board decisions within the IDOs' governance structures (IEG, 2011: 6;Reinsberg et al., 2014;World Bank, 2013a: 9). The second narrative takes a less functionalistic perspective, but focuses on the political economy within donor countries. From this perspective, the end of the Cold War removed the key strategic rationale for aid, which in turn weakened the aid proponents' base in national donor constituencies. The general public became more critical of foreign aid (IEG, 2011: 1), and NGO criticism of multilateral institutions, notably with regard to intransparent governance and neoliberal reform agendas, increased, too (for example, Weaver, 2008). A frequently deplored aid fatigue started to emerge, which donors may have tried to counter by making their aid more visible, and attempting to link it more clearly to perceived national interests, for example, in the areas of trade, investment, or migration. As opposed to traditional multilateral aid, multi-bi aid allowed individual donor countries to earmark their funds to expenditures considered relevant in these areas, and to further present their individual efforts to their national constituencies -something that would be impossible with traditional multilateral aid. At the same time, this ' bilateralization' (Mahn, 2012: 3;Thalwitz, 2013: 1) of contributions to multilaterals could be easily obfuscated to appease those parts of the public who believe that multilateral aid is generally more altruistic and hence more effective than bilateral aid in reaching its development objectives -a belief that is widely shared in the academic literature. 3 For the general public, the distinction between multilateral aid and multi-bi aid is indeed extremely difficult. In fact, even for experts, this is tricky, as even in the official international statistics on aid flows compiled by the OECD-DAC multi-bi aid can appear as multilateral aid when it is channelled through what we will call 'pass-through multilaterals' (see section 3).
A further advantage of multi-bi aid with respect to the risk of public scrutiny should be that some of the critical stakeholders such as NGOs or business representatives can be included in the governance structures of the new trust funds (Forman and Segaar, 2006: 207f.;IEG, 2011: xiii). Generally, joint endeavours between different actors (be they bilateral donors, IDOs, NGOs, or private-sector representatives) also imply joint responsibilities. In the context of fragile states and high corruption the above-mentioned advantage of 'pooling the risk' may then effectively boil down to 'pooling the risk of exposure' (Guder, 2009: 37), that is, to reducing the risk of public criticism in case of programme failure.
Finally, the creation of trust funds allows donor governments to appear engaged and active on new topics of international concern. Hence many large funds were created in the wake of international summits recognizing new development challenges (Panneels and Beringhs, 2005: 3;World Bank, 2011: 9). This notably holds for the Global Environment Facility, established in the aftermath of the Earth Summit in Rio de Janeiro, the Global Fund to Fight AIDS, Tuberculosis and Malaria, following agreement of the G8 on a big push to fight these three diseases, and the Global Agriculture and Food Security Program initiated by the G8 in L'Aquila to act against the global food crisis (see, Smyth, 2011;Triponel, 2009). The creation of a new fund signals a successful outcome of the meeting, even if actual financial commitments are not always ensured. Also, if donor governments really intend to make strong commitments in a particular area, trust funds are equally useful because the financial transfer to the fund will immediately be accounted for and the immediate engagement can be demonstrated to the national public, even if in reality, appropriate projects have yet to be designed. 4 The explanations for the rise of trust funds provided above mainly relate to incentives for bilateral donors. One may still wonder about why the IDOs accept that managing such funds comes with so many strings attached. For some organizations, these resources are relevant for their very existence, but more generally, there is a large and ongoing debate about whether these new funds are at all additional (for the debate within the World Bank, see, for example, Huq, 2010;IEG, 2011: viii, xvi, 26; for an assessment, see also section 4). At the same time, they are known to create substantial transaction costs that are often not fully covered by administrative fees, and they generate deficits in supervision and management oversight that may eventually harm the reputation of the IDO as a whole (IEG, 2011: xf., 58ff.). The alumni organization of the World Bank has even identified the proliferation of trust funds as one of the five major threats to the development of the organization (The 1818 Society, 2012: 2).
However, trust fund agreements are usually negotiated by individual operational departments that -independently of any assessment for the IDO as a whole -tend to welcome multi-bi aid as a means to grow and to become more independent from upper management (Michaelowa and Michaelowa, 2011: 262f.;Reinsberg, 2014;Tortora and Steensen, 2014: 25f.). In fact, they tend to actively compete for these resources with other units, with other IDOs, or also with the private sector (Michaelowa and Michaelowa, 2011;UNDP-EO, 2012: 47;Woods, 2005: 400).
While these considerations only explain the existence of trust funds, the existing literature also tries to assess the effects of this new type of development assistance on aid effectiveness. Some of the problems for aid effectiveness are directly linked to the problems for the individual organizations already mentioned above. Based on statistics for individual IDOs, on interviews with numerous stakeholders and on individual-country case studies, the literature shows a strong consensus on the important administrative and management problems that are yet to be settled, and highlights the risks this implies not only for the IDOs, but also for effectiveness in the use of this new type of aid (see, for example, IEG, 2011: 49-64).
Regarding the effect of competition for funds between aid agencies, opinions are divided. The UNDP-EO (2012: 47) reports cases in which this competition was perceived to raise efficiency. Most studies, however, suggest that 'shopping for funds' by individual IDOs (and different units within IDOs) has rather led to raising inefficiencies, notably due to duplication and internal conflicts (for example, Browne and Weiss, 2012: 8ff.;IEG, 2011: 43f.;Woods, 2005: 400). Michaelowa and Michaelowa (2011: 263f.) also suggest that there may be some harmful competition with the private sector, when crosssubsidized IDO units drive private banks or private consultancy firms out of the market for the management of such trust funds and related services.
Regarding the objectives of the Paris Declaration, the existing assessments are mixed. There seems to be evidence that despite great expectations, recipient government ownership or even participation in decision making processes has not been generally improved (Barakat et al., 2012: 33f.), except maybe for those trust funds that are specifically conceived to complement resources targeted to individual country programmes (IEG, 2011: 42). Similarly, the effect of trust funds on harmonization and coordination among donors appears to be ambiguous (Barakat et al., 2012: 34f.;IEG, 2011: 43). While considerable consultations between donors takes place within such trust funds, these trust funds usually do not replace the existing bilateral and multilateral activities, so that the new trust funds only constitute additional donors on top of the existing ones (Barakat, 2009: 112). Overall, it does not seem that donor fragmentation diminished in the wake of the Paris Declaration (Kharas and Linn, 2008;. Indeed, the sheer number of new trust funds does in itself represent a proliferation of donors leading to an even more fragmented international development system. The consequences for developing country governments who face the ever larger number of donors are intensively discussed in the more general aid fragmentation literature (see, for example, Acharya et al., 2006;Dreher and Michaelowa, 2010;Kilby, 2011;Knack and Rahman, 2007). In addition, the problems arising in this context relate to the problems considered in the even more general literature on the proliferation of international organizations. In fact, studies that generally discuss the proliferation of international organizations or the proliferation of donor agencies and related problems for international governance (for example, Johnson, 2014) and for the aid architecture (Kharas, 2009;Reisen, 2010) often consider IDO trust funds as separate international organizations.
As a further problem of multi-bi aid, many studies mention the influence that individual governments obtain over IDOs. As mentioned above, influencing IDOs is often considered a central argument why sovereign donors provide multi-bi aid. In fact, in a recent DAC survey, this is the reason most frequently given by respondents from the different bilateral donor agencies (Tortora and Steensen, 2014: 16). The problems induced by this behaviour can be considered from a purely normative and a more practical perspective. Normative arguments focus on the bypassing of official decision making bodies that further undermines the already much disputed democratic governance structure of multilateral organizations Sridhar and Woods, 2013: 17). From a more practical perspective, this implies that bilateral donor influence may distort the priorities of the multilateral agencies determined by their governing boards (for example, Eichenauer and Knack, 2015). This is what Sridhar and Woods (2013: 4) define as 'Trojan multilateralism'.
Another frequently mentioned problem is the voluntary nature of multi-bi funding that is generally expected to introduce a high volatility of aid flows. As noted by Tortora and Steensen (2014: 25), multi-bi aid is 'intrinsically unpredictable'. This may represent more than a simple management problem (for example, UN 2012: 37ff.; UNDP-EO 2012: 52). However, to some extent, the risk of volatility comes as the unavoidable flipside of the flexibility of multi-bi aid that is very much valued by the donor community. While it provides the freedom to react, at a national level, to economic and budgetary constraints as well as to the influence of certain lobby groups (a flexibility that may be valuable for national governments, but less so for IDOs and recipient countries), it also provides the possibility to quickly react to emergency situations and to newly recognized international challenges (which is much more broadly considered to be an advantage that other types of aid cannot provide so easily 5 ).
There is yet another feature of multi-bi aid that may be considered as an advantage and as a disadvantage at the same time. As mentioned above, the possibility to support programmes in countries that may otherwise have difficulties to obtain any funding (such as fragile states or countries emerging from civil wars) is usually seen as a useful complementarity to traditional aid (IEG, 2011: 8, 29f.). However, there may be well-justified board decisions prohibiting direct IDO interventions in this area and similarly important reasons why bilateral donors do not individually venture into the field. If joint interventions do not lead to an actual reduction of the risks involved with these interventions, but only to a reduction of the risk of exposure to a critical public if 'things go wrong' (owing to pooled responsibility), this may lead to the financing of activities that are too risky from a global development perspective.
To a large extent, the judgement about multi-bi aid generally depends on one's assessment of the alternatives, namely, bilateral and multilateral aid. If one believes that both are effective as they are, there is hardly a need for anything in between. If one believes that multilateral aid is not sufficiently effective (for example, because of its slow adjustment to new requirements), then one will support multi-bi aid if it comes as a substitute for traditional multilateral aid. If one believes that bilateral aid and hence bilateral donor influence on IDOs is problematic because it undermines agreed international governance structures and distorts aid allocation towards donor interests rather than recipient needs, then multi-bi aid will be welcomed if it substitutes for traditional bilateral aid, but not if it substitutes for existing multilateral funding. Some of the arguments to be considered in this discussion can be drawn from the general literature on aid allocation and aid effectiveness (see, for example, Milner and Tingley, 2013, on the decision between bilateral and multilateral aid) and also from more general considerations on countries forming organized subgroups below the international level in order to solve problems that they, as a group, consider important ('minilateralism', see for example, Kahler, 1992;Kellow, 2012;Ruggie, 1992).
In our context, the challenge so far has been that no internationally comparable data was available to even describe the ongoing trends. Data was available only for individual IDOs, whereby each used somewhat different definitions. Even if these definitional differences were ignored, the datasets could not be aggregated because they partially overlap as some IDOs pass substantial amounts of funding through to other IDOs, and some IDOs do not provide the relevant information. When trying to assess multi-bi aid internationally, authors had to rely on very crude approximations (see, for example, Adugna, 2009). The lack of data can explain why there has been hardly any independent and rigorous academic research on the topic despite the obvious political relevance of the phenomenon at hand (Bakarat et al., 2012: x, xi).
However, recently, the OECD-DAC started coding the different channels through which bilateral donors fund international development programmes. On this basis, and in collaboration with DAC statisticians, Eichenauer and Reinsberg (2015) established a new database that allows us to appropriately describe the international trends. This database will be described in the following section along with the necessary definitions and conceptual ideas underlying our coding decisions. Given the new data, we will then be able to provide B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account some evidence on a number of relevant questions such as: (1) is multi-bi aid additional to existing funds or does it substitute for traditional (pure) multilateral or bilateral aid?
(2) Do these funds really flow into substantially different areas than multilateral aid, and how strongly are they really earmarked by donor governments? (3) Are they as volatile as some observers expect? In addition, the data will even give us some evidence to assess the plausibility of the different narratives regarding the emergence of multi-bi aid.

CONCEPTUAL CLARIFICATIONS AND DEFINITIONS
The OECD-DAC records two types of databases. On the one hand, its development statistics include data on the flows of multilateral aid and bilateral aid by all DAC members. 6 On the other hand, the OECD-DAC maintains the Creditor Reporting System (CRS), which records individual aid activities -let us call them 'projects' for simplicity -along with detailed project descriptions and some information on the relevant institution through which aid flows are channelled for implementation (OECD, 2014a). The channel codes are intended to indicate whenever aid is not directly given to a recipient country, but rather channelled through another organization without being multilateral aid. Together with information on earmarking, the channel codes provide the basis to distinguish multi-bi aid from other types of aid. Donors have started systematically reporting channel codes since 2004 with information becoming more and more frequent in subsequent years. Nevertheless, the database still contains many missing values, very general entries, and simple coding errors by the contributing donors. Using the additional information from the individual project descriptions, Eichenauer and Reinsberg (2015) added and updated the channel codes wherever they were missing, incorrect or imprecise. This allowed us to eventually obtain a comprehensive time series with detailed channel codes for all DAC donors from 1990 to 2012. 7 We further added information on the degree of earmarking. For this purpose, we evaluated project descriptions with respect to three potential dimensions of earmarking: (1) a thematic dimension, (2) a geographic dimension, and (3) an institutional dimension. Within each dimension, we distinguish between no earmarking, soft earmarking, and tight earmarking depending on the degree of flexibility left to the IDOs that eventually manage and implement the projects. On the thematic dimension, for instance, tight earmarking implies that the exact project is specified, whereas soft earmarking only specifies the sector or general theme to which aid must be allocated. For geographic earmarking, we distinguish between the definition of the country (tight earmarking) and the definition of the broader region (soft earmarking). For institutional earmarking, soft earmarking refers to contributions to specific (sub)divisions of the implementing IDOs, while tight earmarking involves the delegation of staff from the bilateral donor to the IDOs to carry out a concrete task.
Finally, we coded the available information on whether multi-bi aid reflects the initiative of a group of bilateral donors resulting in multi-donor trust funds (MDTFs) or individual undertakings leading to single donor trust funds (SDTFs). Interestingly, most of the above cited literature on multi-bi aid seems to focus on MDTFs and the narrative on the rise of multi-bi aid as a consequence of donors' willingness to enhance cooperation and harmonization obviously cannot explain any potential increase in SDTFs. Our B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account data will show whether multi-donor undertakings are really as dominant in multi-bi aid as this literature may lead us to expect. Our coding regarding MDTFs versus SDTFs was based on two types of information. First, knowledge about the nature of the trust funds identified through the channel code allowed us to identify the most important MDTFs. Second, any mentioning of collaboration with other donors in the project description led to a coding as an MDTF. This procedure may still leave some MDTFs unidentified and hence bias our shares towards an overestimation of SDTFs. Yet, a comparison with data from the World Bank, where detailed information for the type of trust funds is available, increases our confidence in the appropriateness of our coding. The full Eichenauer and Reinsberg (2015) dataset on multi-bi aid and a detailed description of all the above-mentioned additions to the CRS is available at www.aiddata. org.
Interpreting the data requires some further conceptual clarifications. First, it should be noted that when simply adding up the different types of aid (bilateral, multilateral and multi-bi) there is some risk of double-counting. Traditionally, the problem was limited to the European Union (EU), which is counted as a bilateral donor while, simultaneously, bilateral EU member contributions to the EU are counted as multilateral aid. With the ascent of multi-bi aid, the problem multiplied because numerous new multilateral trust funds appeared that also channelled their funds to further IDOs for actual implementation. The Global Environment Facility (GEF) for instance, created in 1991 as a programme within the World Bank, but gaining more autonomy in 1994 in the wake of the Rio Earth Summit, generally contributes its funds to the World Bank, the UNDP or the UN Environment Programme (UNEP) for actual implementation (GEF, 2013). Once the DAC recognized the GEF as an independent agency, the GEF's contributions to the actually implementing IDOs were accounted for as bilateral aid. Simultaneously, the bilateral donors' contributions to the GEF that provided the financial basis for the GEF's funding of Word Bank, UNDP or UNEP programmes were reported as multilateral aid. Hence simply adding up all financial flows leads to double counting ever since the GEF was recognized as an autonomous organization whose own aid allocation also appears in DAC statistics. In contrast, had the GEF been continued as a simple programme within the World Bank, related contributions would have been counted only once, namely as bilateral, earmarked aid channelled through the World Bank. This is no rare example: in summer 2014, there were more than 90 of these 'passthrough multilaterals', that is, multilateral funds funding the activities of existing IDOs. They are often called 'global funds' by the OECD or 'financial intermediary funds' by the World Bank. As explained above in the example of the GEF, whenever the DAC members consider these organizations as sufficiently autonomous from their parent organizations (for example, the World Bank or the UN), the contributions of bilateral donors to these organizations are coded as multilateral aid, while the pass-through organizations' own contributions are considered as bilateral aid. These new multilaterals -along with the GEF one might want to mention the Global Fund to Fight Aids, Tuberculosis and Malaria (GFATM) or the Global Partnership for Education (GPE) -generally earmark their financial contributions when they provide them to the IDOs implementing the projects. Hence eventually, aid channelled through these organizations mostly turns into multi-bi aid.
We solve the double-counting problem by first identifying all aid that, at some point, becomes an earmarked contribution to an IDO. This is what we report as multi-bi aid. When it is initially channelled through pass-through multilaterals, we subtract the corresponding amounts from the sovereign donors' multilateral contributions. Hence, we count the amount only once, namely when it flows from the pass-through multilateral to the IDO, and not when it flows from the sovereign donor to the pass-through multilateral. Second, we divide the remaining aid flows between (pure) bilateral and (pure) multilateral aid. When we talk about multilateral or bilateral aid, we hence refer to commitments net of all earmarked contributions to IDOs. This is illustrated in Figure 30.1. 8 Unlike an alternative definition of multi-bi aid that only looks at whether a bilateral donor earmarks its contributions or not, our classification has the additional advantage to avoid discontinuities in the time series when a trust fund is considered to have 'graduated' from a sub-entity of a larger IDO to an autonomous pass-through multilateral (such as in the case of the GEF mentioned above). In this case, the simple re-categorization of the trust fund within the CRS would otherwise lead aid initially considered to be multi-bi (earmarked for a specific use within the larger IDO) to be suddenly regarded as multilateral (an unearmarked contribution to the independent trust fund). Yet, the purpose of the contribution would be just as narrowly defined as before. When including the second step of the chain in Figure 30.1, line (3) in the consideration of earmarking, and defining aid as multi-bi whenever earmarking takes place anywhere in the delegation process, this problem does not arise.

Notes:
As opposed to 'pass-through multilaterals', we define 'international development organizations' (IDOs) as those multilateral organizations directly involved in project implementation. Pass-through multilaterals are only involved in financial administration. They are usually rather narrow in scope and their funding is virtually always earmarked when they pass it on to the IDOs. Hence, funding channelled through passthrough multilaterals is considered as multi-bi aid.
All non-earmarked contributions to multilateral donors -be they core-funding or voluntary contributionsare considered as multilateral aid. Bilateral aid is development assistance provided under the direct responsibility of donor country agencies.

EMPIRICAL EVIDENCE ON TRUST FUNDS AND MULTI-BI AID
We can now present some initial descriptive statistics and basic statistical analysis based on our new dataset. We leave it to future studies to make use of the data for a more refined analysis (for example, Michaelowa et al., 2014;Eichenauer and Reinsberg, 2015). As a first step, we present the general development of the three different types of aid over time in absolute volumes (Figure 30.2) and in relative terms (Figure 30.3). Note that all representations of aid volumes based on the CRS activity database are somewhat distorted as the entries in this database are generally incomplete throughout the 1990s and only sum up to about 70-80 per cent of total aid during this period. Yet, we can assume that this holds for all types of aid in about equal proportion, and if data in Figure  30.2 were adjusted accordingly, the notable rise of multi-bi aid would be equally obvious. The figures show that as a share of total ODA, multi-bi aid still seems rather modest, but when compared to multilateral aid, its relevance becomes more than obvious. Both in absolute terms and as a share of other aid, the most remarkable increase in multi-bi aid happened in 2005, the year of the Paris Declaration, where it more than 0 20 40 60 80 100 120 140 160 1 9 9 0 1 9 9 1 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 2 0 0 1 2 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2   Figure 30.4 provides an overview of the main IDOs that implement multi-bi aid. 61 per cent of all multi-bi aid is channelled through the various UN organizations and, with a share of 20 per cent, the World Bank forms the second most important channel. Other relevant IDOs are the regional development banks (RDBs) and the EU. The data presented refer to the period since 2006, that is, once multi-bi aid volumes have really become significant. It should be noted that some IDOs like the World Bank have to deal with an even higher share of multi-bi aid, but only as financial administrators without responsibility for implementation. Indeed, as funds are handed on, several institutions may have to deal with the management of the same funds at subsequent points in time. In order to avoid double counting, we attribute these earmarked funds only to the final implementers (the IDOs), and not to those organizations that only act as pass-through multilaterals.
Source: Eichenauer and Reinsberg (2014). the multi-bi channel for more than 20 per cent of total ODA, closely followed by Finland, the Netherlands and the United Kingdom, all of which have more than 18 per cent of multi-bi aid. Nineteen out of 24 DAC member countries spend more than 10 per cent through the multi-bi channel. Are the countries with high shares of multi-bi aid more engaged than others in collaborative efforts in line with the Paris Declaration? While different donor motives for multi-bi aid will be examined in detail in Eichenauer and Hug (2014), Eichenauer and Knack (2015), and Reinsberg et al. (2014), Figure 30.6 already suggests that for many donors, it is not the perspective of coordination and harmonization that dominates. As it turns out, most trust funds are SDTFs based on contracts between a single donor and an IDO. While the share of MDTFs might be slightly underestimated due to our coding procedure (see section 3 and the note to Figure 30.6), it is obvious that the vast majority of multi-bi aid is channelled through SDTFs. Since contributions to MDTFs are usually larger than those to SDTFs, the contrast is particularly strong when we look at the number of aid activities. This should be relevant when considering the effect on the transaction costs arising for IDOs and pass-through multilaterals. Hence, even if SDTFs individually may need less administrative care (Reinsberg, 2014: 9), as a whole, the administrative burden may still be higher than for the much smaller number of MDTFs.
Unlike large MDTFs, which are primarily created to address new global problems, SDTFs might diverge less from IDOs' internationally agreed priorities. Single donors    Note: Single-donor trust funds are assumed whenever an activity is not clearly identifiable as a multi-donor partnership.
Source: Eichenauer and Reinsberg (2014 Notes: Single-donor trust funds are assumed whenever an activity is not clearly identifiable as a multi-donor partnership. For multilateral aid, see the note to Figure 30.1. However, here, multilateral aid refers only to joint outflows of EU, UN, and IDA, due to poor availability of disaggregated data for other IDOs. Source: Eichenauer and Reinsberg (2014).

Figure 30.7 Comparing the distribution of funding by SDTFs, MDTFs, and multilateral aid, 2006-12
Regarding sector allocation, the above reasoning is clearly incompatible with the evidence. In fact, MDTFs seem to be closer to the distributional choices of multilateral aid than SDTFs. This is primarily due to the large share of humanitarian assistance in SDTFs, notably from the US. Regarding regional choices, we do observe the expected focus of MDTFs on global programmes, while SDTFs are more in line with multilateral aid. As can be further observed from our data (not shown here), divergences also vary considerably between different IDOs. As the UN is the main organization to carry out emergency assistance, this special focus of multi-bi aid distorts the UN's aid allocation profile more than that of other IDOs. At the same time, the regional allocation of multi-bi funds channelled through the UN corresponds almost exactly to the UN's allocation of core multilateral aid. The World Bank, however, hosts most of the funds directed towards new global activities, so that its overall regional orientation is driven away from its otherwise stronger focus on region-or country-specific programmes in sub-Saharan Africa and South and Central Asia. At the EU, multi-bi aid shifts the focus towards neighbouring countries in Europe and the Middle East/North Africa.
Rather than looking only at the broader regions, we can also look at the specific choice of recipient countries (Table 30.1). When comparing the major recipient countries of multilateral, multi-bi and bilateral aid, we first observe that numerous countries, namely, Bangladesh, Democratic Republic of Congo, Ethiopia, Nigeria and Pakistan, appear among the top-10 for all three types of development assistance. As opposed to what might be expected, nothing indicates that multi-bi aid stands somewhere in between bilateral and multilateral aid when it comes to the choice of most favoured recipients. Except for Afghanistan (which is also among the top-10 for multilateral aid) the remaining countries that figure on the list for multi-bi aid are not part of the preferred recipients for either bilateral or multilateral aid. These are Sudan, Iraq, Zimbabwe and China. (To  simplify the reading of Table 30.1, all such countries that appear among the top-10 in only one category of aid are marked in bold.) Overall, the list confirms the strong orientation of multi-bi aid towards humanitarian and security issues. China is a special case whose importance seems to be driven primarily by trust funds combating environmental problems including global climate change. Some caution is, however, necessary when interpreting the differences between bilateral, multilateral and multi-bi aid (see also Reinsberg, 2014). First, given the above cited arguments about the dependency of IDOs on voluntary contributions, it may well be that the allocation of multilateral funding that we observe is already adjusted to the requirements of the major bilateral donors. In this case, the remaining differences with respect to bilateral and multi-bi aid do not reflect the full influence of multi-bi aid. At the same time, the influence could also be overestimated, namely when IDOs adjust the allocation of their core funding to the supplementary aid inflows of multi-bi aid (fungibility). For instance, as more humanitarian aid comes in through trust funds, the UN might decide to shift some of its other resources, initially foreseen for humanitarian aid, to other fields. In principle, it could thereby re-establish the previous balance for its overall resources.
While the exact effect of earmarking on the allocation of funds is hence difficult to determine, earmarking is nevertheless used not only widely, but also intensively (see Figure 30.8). It appears that at least the bilateral donors (or their national constituencies) believe that earmarking is important and set up binding constraints to the IDOs' choices. Notes: For the underlying definition of multi-bi aid, see Figure 30.1. For geographic earmarking, soft earmarking includes (sub-) regional earmarking while hard earmarking refers to specific countries. For thematic earmarking, soft earmarking refers to specific sectors or fields of intervention, hard earmarking to specific projects. For institutional earmarking, soft earmarking is a contribution to a (sub)division in the organization, while hard earmarking implies the delegation of staff for a concrete task.
Source: Eichenauer and Reinsberg (2014).  Figure 30.8 presents the intensity of earmarking in three different non-exclusive dimensions. Note that it refers to multi-bi aid only, so that some sort of earmarking is always present. Institutional earmarking is rare, and geographic earmarking is most widely spread. In fact, with just over 1 per cent, the share of multi-bi aid that is only earmarked thematically or institutionally, but not geographically (light grey part of the column for geographic earmarking), is quite insignificant. For 85 per cent of multi-bi aid, geographic earmarking goes beyond broad regional specifications and indicates concrete country choices. Most donors further combine geographic and thematic earmarking.
While earmarking sets clear limits to the IDOs' discretion in aid allocation, it is the voluntary nature of multi-bi aid that raises concerns about the potential volatility of these funds. As mentioned above, just as bilateral aid, these funds can, in principle, be increased or decreased at very short notice at the donor government's discretion. At the same time, even when financial contributions are not agreed upon for an intermediate to long period at the international level, there are constraints related to the relations with partner countries or to the expectations of national constituencies, or even national administrative processes that may hinder ad hoc adjustments.
On the basis of our data, we can calculate the coefficients of variation (standard deviation divided by mean) for bilateral, multilateral, and multi-bi aid across time for each country. Owing to the tremendous increase in multi-bi aid, the coefficients for this type of aid are very high for the overall period covered by our data . However, when we split the period in two equal parts and look at the second half (that is, from 2002 to 2012), 9 the coefficients of variation appear to be quite close to the coefficients of variation for multilateral aid. This is true despite a persistently strong general increase in these funds, and despite the fact that the period includes phases of strong economic downturn, notably in the aftermath of the global financial crisis in 2008. Despite the fact that bilateral aid is totally at the discretion of national governments (at least as much as multi-bi aid), the coefficient of variation for bilateral aid is generally the lowest. The median value of the coefficient of variation for all 21 DAC members that reported to the CRS during the whole period of analysis is 0.22 for bilateral aid, 0.39 for multilateral aid and 0.41 for multi-bi aid. This shows that, even if bilateral aid is more flexible than multilateral aid in principle, in practice, this does not necessarily lead to higher volatility. This argument also holds for multi-bi aid.
An alternative way to assess the volatility of multi-bi aid is to directly test the effect of economic variables on aid volumes. This procedure is more reliable, notably because the general upward trend over time can be controlled for, which also allows us to include the full time period from 1990 to 2012 into our analysis. When running corresponding regressions using either the donor countries' growth rate, the budget balance or the unemployment rate as independent variables (either for the current year or lagged by one year), we never obtain any significant coefficient indicating that economic downturns systematically reduce multi-bi aid. Table 30.2 shows donor country fixed effects regressions in which the three lagged economic variables and a control for a linear time trend (measured by the variable 'year') are included simultaneously.
The results hold here as well. The only marginally significant coefficient in the first regression points in the opposite direction, implying that, if anything, multi-bi aid is reduced when the budgetary situation improves. This may be an artefact of reverse causality. In any case, there is no indication that an economic downturn induces multi-bi aid B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account to shrink. Further controlling for the level of bilateral and multilateral aid (Regression 2) does not change anything to this result. The insignificance of the economic variables is just the same as for multilateral aid (Regression 3). Only for bilateral aid, can significant effects of growth and unemployment be observed (Regression 4). As opposed to the simple coefficient of variation, they do suggest that bilateral aid reacts to economic downturn.
In sum, while multi-bi aid may bring about a greater risk of volatility, at least up to now, there is no evidence for that in the overall flows. A more refined analysis could examine a similar question for variations between different areas of multi-bi aid along with donors' shifting thematic or regional priorities. From a recipient or IDO management perspective, sudden shifts across different areas of multi-bi aid may be as disruptive as general shifts in and out of multi-bi aid. However, this issue has not been identified as a significant problem in the literature, which suggests that we do not need to go into the details here.
Finally, let us consider the additionality of multi-bi aid. If multi-bi aid substitutes for the traditional forms of either bilateral or multilateral aid, then, to find general approval, multi-bi aid will have to prove to be at least equally effective. If it comes on top of the existing development assistance, and we assume some positive returns for aid as a whole, then the mere fact that the new instrument mobilizes new resources will make it  acceptable. The question about additionality to traditional aid can also be asked because the global public goods addressed by most of the large MDTFs, such as climate change or global security, are not themes that fall under the classical classification of development assistance (Woods, 2005). This lies in the very nature of a global public good with its characteristics of being non-rival and non-exclusive at a global level. Take the example of climate change mitigation. The worldwide effect of a certain amount of emission reductions is exactly identical wherever it happens. When reducing emissions through interventions in the US or in Switzerland, nobody would even think of calling this development aid. However, when comparable projects are implemented in Senegal or India, they tend to be considered under that category. Nevertheless, India and Senegal would have benefited as much, had the intervention taken place in the US or in Switzerland (Michaelowa and Michaelowa, 2012: 40f.). In addition, industrialized countries are responsible for the bulk of the emissions accumulated so far, so many would argue that whatever they invest in climate finance (that is, support to developing countries' climate change policies) should be additional to traditional development aid. This also corresponds to the 2009 Copenhagen Accord, which includes a collective pledge by industrialized member countries of the United Nations Framework Convention on Climate Change (UNFCCC) to provide 'new and additional resources'. Ever since, reaching actual financial additionality has remained a matter of concern (see, for example, Stadelmann et al., 2012: 128f.).
In any case, as long as multi-bi aid is additional, much of the criticism and fears related to this new instrument will be quenched. Those that remain concerned about the risk of volatility, for instance, will also be appeased if the basis of traditional multilateral aid remains intact.
The coefficients of bilateral and multilateral aid in Table 30.2, Regression 2 can be used as a first indicator for the relationship between the different types of aid. The coefficients are positive and significant, which is consistent with a hypothesis of complementarity (that is, additionality) rather than of substitution. As the regression specification controls for time-invariant donor characteristics (donor country fixed effects), the result does not simply reflect cross-country differences in the preference for development assistance that would codetermine all types of aid. Yet, the interpretation of Regression 2 is problematic because causality could run in both directions. In fact, sceptics of the additionality of multi-bi aid would argue that multi-bi contributions influence spending on bilateral and multilateral aid, rather than the other way around, and we tend to agree with this perspective. In this case, bilateral and multilateral aid should be considered as the dependent variables while multi-bi aid would be the explanatory factor. Whichever way, we should be careful to interpret the coefficients merely as correlations and not as causal effects. In any case, it seems worthwhile to note that the correlation is positive.
Running fixed effects regressions with bilateral and multilateral aid as dependent variables, and multi-bi aid as an explanatory factor, we can further refine the analysis by interacting the multi-bi aid variable with the time trend. 10 This allows us to distinguish between a level effect and a trend effect.
In the regression for bilateral aid, we obtain positive correlations regarding both trend and levels, but none of them is significant. In the regression for multilateral aid, both relevant coefficients are significant, but go in different directions. It appears that countries using more multi-bi aid tended to contribute less multilateral aid in the initial year than B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account expected given their time-invariant country characteristics. However, over time, contributions to multilateral aid increased significantly more in those countries that simultaneously spent relatively high amounts through the multi-bi channel. Given the overall positive correlation between multilateral and multi-bi aid, this trend-effect dominates the initial level-effect. The trend-effect is presented in Figure 30.9, which shows the estimated annual increase of multilateral aid for different levels of multi-bi aid. For example, in the absence of multi-bi aid (left side of the graph), the annual increase of multilateral aid for an average donor would be zero, while the annual increase would be around US$200 million when the donor spends around US$6 billion through the multi-bi channel (right side of the graph). This suggests that multi-bi aid expedites the positive growth trend of multilateral aid.
In sum, at least in this simple analytical framework, there is little indication of multi-bi aid substituting for traditional forms of aid or reducing their annual growth. To the contrary, the traditional aid flows are positively correlated with multi-bi aid, which indicates additionality and points at a complementary use of the different channels. This suggests that the rise of multi-bi aid has not simply led to a re-allocation of aid within a fixed budget. Further differentiation between a level effect and a trend effect reveals that the  overall positive relationship between multi-bi and traditional forms of aid is driven by the trend-effect, that is, by the association of multi-bi aid with a higher growth in traditional aid types, significantly so for multilateral aid. Obviously, these results should only be taken as an initial and rough assessment of these relationships. Based on the new data, other studies can carry out more refined analysis including appropriate control variables beyond the fixed effects, and establishing proper causal relationships. This, however, is beyond the scope of this general introduction to multi-bi aid.

DISCUSSION: FACTS AND CONSEQUENCES
In the following, we will summarize and further comment some important facts that can be derived from the above presentation of our data.
First, multi-bi aid has risen exponentially over the past 20 years. While bilateral and multilateral aid are still considerably larger, the volume of multi-bi aid now exceeds 50 per cent of multilateral aid and has become the dominant form of financial flows to a number of multilateral agencies. Multi-bi aid has therefore led to quite some turmoil within the multilateral aid system, and challenged existing governance structures and management procedures.
Second, there seems to be a strong potential for multi-bi aid, notably to enhance donor cooperation and harmonization. However, the data show that the overwhelming majority of trust funds are financed only by a single donor. As opposed to a widespread belief, our data indicate that these SDTFs are not necessarily closer to the IDOs' own funding priorities than MDTFs. Transaction costs for IDOs are high and an unresolved question is whether the efficiency gains at the level of the bilateral donor agency can outweigh the efficiency losses at the level of the IDOs. Given the large number of SDTFs, the share of multi-bi aid by a specific sovereign donor is not a good indicator of this donor's willingness to engage in joint development efforts. One needs a closer analysis of the individual trust funds financed by this donor to distinguish the underlying objectives. In fact, among the 10 donors using multi-bi aid most extensively according to Figure 30.5, only the United Kingdom and the Netherlands stand out for a strong general engagement in donor coordination (as measured, for example in the leadership roles taken up in DAC working parties 11 ).
Third, while donors tend to make intensive use of earmarking, usually combining geographic and thematic restrictions when providing their multi-bi contributions, the overall geographic and sectoral priorities are not always as different between the different types of aid as one might expect. While we do see a stronger focus of multi-bi aid, notably SDTFs, on humanitarian aid, and of MDTFs on global public goods, it is also remarkable that five out of 10 top-recipients are identical across these bilateral, multilateral and multi-bi channels, and that for the UN as a whole, the regional distribution hardly differs between its multilateral and its multi-bi resources. To what extent this may be a matter of anticipatory obedience given that multilateral aid is often voluntary -especially for the different UN agencies -so that IDOs might be under pressure to follow the directions of major donors anyway, is not clear. In any case, this result suggests that the earmarking restriction may not always be binding and that, at times, they are more of a signalling B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account device to the national public than an effective attempt to influence multilateral agencies and to introduce new development orientations. Fourth, while by definition, the stability of multi-bi financial flows and their additionality to traditional aid cannot be ensured, our (admittedly rough) initial empirical analysis does not show much reason for concern in these areas. Our results suggest that so far, volatility has not been significant, and that additionality has been respected. Rather than reducing other types of aid, multi-bi aid has added to these, and has even slightly increased the growth path of multilateral contributions.
The primary concerns that may remain on the basis of this analysis appear to be the following: 1. The strain on the multilateral system (in terms of additional cost, governance and management problems). 2. The bilateralization of multilateral agencies.
With respect to (1) numerous reforms are already under way. In particular, these reforms try to limit new trust funds to those initiatives exceeding a certain minimum level of resources, to ensure full cost recovery (thereby avoiding cross-subsidization), and to adopt comprehensive strategic planning, risk management, and budgeting processes simultaneously including all core and non-core resources (Tortora and Steensen, 2014: 26ff.). These reforms are foreseen in the UN's 2012 Quadrennial Comprehensive Policy Review, but also, framed in a slightly different way, at the World Bank (World Bank, 2013aBank, , 2013b. Some administrative steps have become fairly concrete. In spring 2014, for instance, the World Bank increased the threshold for the creation of new SDTFs by 100 per cent of the original minimum value of contributions to limit the exponential growth of these funds (World Bank, 2013c).
However, not all of the intended reforms can be introduced by the IDOs' upper management alone. They require the collaboration of internal operational units and of bilateral aid agencies. As long as bilateral agencies enjoy the flexibility to make ad hoc decisions about financial resources channelled through IDOs, and the operational units gladly accept these funds to strengthen their own independence within the IDOs, it may be difficult to truly implement substantial reforms.
With respect to (2), the current reform discussion shows that donors are reluctant to cede control to IDOs, as this would undermine some of their benefits of multi-bi aid. Except for those areas where the bilateralization is directly linked to the management and supervision problems addressed under (1), there seems to be a widespread acceptance of things as they are. It is quite indicative in this sense that in the recent DAC survey already mentioned in section 2, the majority of DAC members openly admitted that a key motivation for using multi-bi aid was to influence the IDOs' development agenda, to align the organizations' agenda with their own bilateral priorities, and to introduce new fields of activity (such as in the area of global public goods) (Tortora and Steensen, 2014: 16). No relevant stakeholder in the multilateral system seems to oppose these attempts by bilateral donors. It seems that the IDOs' upper management do not have any clear incentive to discourage such developments. This holds both for a reorientation towards specific bilateral priorities as for the orientation towards security issues, climate change, and other global challenges. At the World Bank, for instance, the mission creep from poverty reduction towards global public goods seems to enjoy a wide acceptance, and the '2013 Trust Fund Annual Report' proudly presents the achievements in these areas (World Bank, 2013a: 13ff.). Obviously, global public goods are important issues to be handled at the international level, but the emerging division of labour between IDOs and the induced consideration of support in this area as development aid are not in line with the preferences of many developing country governments (see, for example, the debate about the role of the World Bank in climate finance - Reinsberg et al., 2014: 19). While these countries are represented in the formal decision-making boards of the IDOs, they have little influence over multi-bi aid.
In the light of the discussion of 'post 2015', and the presumable introduction of 'Sustainable Development Goals' to replace the current MDGs, these trends will become even more important. It will primarily lie in the hands of international NGOs to act as watchdogs to ensure that at least financial additionality, which, as presented above, does not represent a problem at the moment, will also be maintained in the future, when development assistance and the provision of global public goods will be even further intertwined.

CONCLUSION
This chapter has provided an overview of the evolution of multi-bi aid, that is, earmarked voluntary contributions to international development organizations. Over the past 20 years, this new type of aid has grown to almost US$20 billion in 2012, which corresponds to almost 60 per cent of multilateral aid and 20 per cent of bilateral aid. This development has challenged the traditional aid architecture. While it opened up considerable opportunities, notably with respect to achieving donor cooperation, harmonization and recipient government ownership, it has also generated a number of new problems.
Following an extensive literature review discussing the new phenomenon from various perspectives, we provide a descriptive empirical analysis based on a self-coded dataset that considerably expands the information on multi-bi aid available so far. The results show that the potential advantages of the new instrument have only partially been used. For instance, the largest share of multi-bi aid is channelled through single-donor trust funds, while the opportunity to use these funds for joint endeavours of different donors is the exception rather than the rule. Generally, multi-bi aid turns out to be a very complex phenomenon, whose multiple facets are not easily assessed. In certain respects, the same characteristics that render it an attractive instrument in the eyes of some render it problematic in the eyes of others. For instance, what some would praise as greater flexibility, others would dismiss as an encouragement of ad hoc decision-making, which, in addition, circumvents the internationally agreed decision-making structures within multilateral agencies.
Among the many problems discussed in the literature, it is the strain on the multilateral system (in terms of additional cost, governance and management problems), and the bilateralization of multilateral agencies that appear most important to us. The latter can be expected to further increase with the growing international attention to the provision of global public goods in the context of the new development goals for post 2015.
While volatility is also perceived as an important structural risk regarding multi-bi aid, B. Mak Arvin and Byron Lew -9781783474578 Downloaded from Elgar Online at 11/14/2015 12:16:04PM via communal account our data suggest that up to now this type of aid has not been significantly more volatile than multilateral aid. Moreover, we can show that multi-bi aid has so far been a complement, rather than a substitute, for traditional aid. Whether it will also remain additional in the future is of course an open question.

NOTES
1. This chapter is Open Access and distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. The chapter was prepared as part of the interdisciplinary research programme 'The Proliferation of Multilateral Funds'. It is partially based on an initial draft by Eichenauer and Reinsberg (2013). Financial support by the Swiss Network for International Studies is gratefully acknowledged. 2. See Eichenauer and Reinsberg (2014) for a detailed definition of IDOs. 3. There are many aid effectiveness and aid allocation studies that support the view that multilateral donors are indeed more altruistic and more efficient in reaching development objectives (see, for example, Birdsall and Kharas, 2010;Easterly and Pfutze, 2008;Headey, 2008;Knack et al., 2011;Maizels and Nissanke, 1984;Powell and Bobba, 2006). 4. IEG (2011: 6), for instance, quotes a Dutch official mentioning that the government had decided a 15 per cent increase in spending on aid for education, but that 'it was hard to spend that large amount bilaterally', and that trust funds were regarded as a good solution to this problem. 5. In principle, humanitarian funds could of course be filled with sufficient resources in the first place, so that ad hoc contributions in emergency situations would not be necessary. However, this seems to be difficult in practice since sovereign donors do not want to make such long-term commitments that would limit their own flexibility (Bakarat, 2009: 110f.). Indeed in the context of emergency aid, the flexibility to react fast is necessary only under the assumption that donors cannot be convinced to give up some of their own freedom of action in the first place. 6. In 1990, the start of our coding period, there were 21 DAC members, namely: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Italy, Korea, Japan, Luxembourg, the Netherlands, Portugal, Spain, Sweden, Switzerland, the United Kingdom, the United States and the European Union. Currently, the DAC has 29 members. 7. We thank the OECD-DAC statistician Willem Luijkx for regular exchange and for his support in ensuring the consistency between our coding and the extant codes for the more recent years. 8. Note that the OECD-DAC avoids double-counting within any given database because the available OECD-DAC statistics either take the perspective of the bilateral donor, or the perspective of the recipient country. Yet, to study multi-bi aid as a whole, the different databases need to be combined, and this is when the problem of double-counting arises if no adjustments are made along the lines mentioned above. 9. If we shortened the period even further (for example, to the period from 2006 to 2012 as in most of the figures presented in this section), this result should become even clearer. However, we would lose a lot of observations on which to base the statistical interference. 10. The two regression equations are then specified as follows: A it 5 b 1 year t 1 b 2 multi-bi aid it 1 b 3 year t · multi-bi aid it 1 a i 1 e it , where A it refers to traditional aid flows of donor i in year t, a i is a donor-specific fixed effect, and e it an idiosyncratic error. All aid variables are measured in constant 2012 US$ million just as in Table 30.2 above. (Technical note: By rescaling the variable 'year' to 'year -1990' (5 0, 1, 2, . . . 22) in this regression equation, b 2 can be directly interpreted as the level-effect of multi-bi aid. b 3 shows the trend-effect, that is, the increase in the trend of multilateral aid related to an additional million of multi-bi aid. The trend corresponds to b 1 for multi-bi aid50.) 11. See Reinsberg et al. (2014: 46) for the coding of this variable.