A Multilateral Donation that Boomerangs Home: Analysing the Donor State Advantage in UN Procurement

Abstract Bilateral aid procurement is politicised and strongly favours suppliers from donor countries. Does multilateral development assistance eliminate the procurement bias favouring donor countries because international bureaucrats make procurement decisions? Existing evidence from the World Bank, which delegates procurement responsibilities to aid recipient countries, cannot answer our theoretical question. Using official data from 20 UN organisations during the 2013–2018 period and applying regression and mediation analysis, we find that the procurement of international organisations still favours donor countries when international bureaucrats make procurement decisions. We identify donor state representation within the UN staff as a key stepping stone linking donation to procurement bias. In contrast, member states whose nationals are heads of a UN bureaucracy do not enjoy procurement advantage, suggesting that UN procurement bias operates through an informal bottom-up channel. Our paper contributes to the debates on the independence of international organisations in the context of multilateral development assistance and procurement.


Introduction
Local procurement of goods and services for development projects can be a way to 'spend the development dollars twice' (Bontjer, Holt, & Angle, 2010;Zhang & Gutman, 2015). However, development dollars tend to boomerang to donor states through procurement. For example, 95 per cent of bilateral donations from the United States (US) to Haiti after the 2010 earthquake went through US firms (Walz & Ramachandran, 2013). Strong public support for awarding foreign aid contracts to domestic firms underlies the biased procurement decisions made by the US government (Christiansen, Heinrich, & Peterson, 2019). This donor advantage in procurement (for example, Michaelowa, 1997;Morrissey, 1993;Younas, 2008) is an example of how the politics of donor states can complicate bilateral aid (Neumayer, 2003;Winters & Streitfeld, 2018).
If bilateralism fails to promote local procurement, is multilateralism more effective? Milner (2006) believes that multilateral aid encourages local procurement because, unlike bilateral aid, multilateral aid cannot be tied to the purchase of goods or services from a particular donor at the procurement stage. Milner's argument is consistent with the well-known theoretical justification of multilateralism, namely, that the influence of donor states that fund international organisations (IOs) can be insulated by the international bureaucrats who are expected to manage the IOs independently and professionally.
Does the evidence suggest that multilateralism encourages local procurement? Procurement by multilateral institutions is still an understudied research topic. McLean (2017) provides evidence that local procurement is common in World Bank development projects. Heinzel (2021) further shows that local procurement is more likely when the Bank staff supervise development projects in their home countries. The available evidence appears to support the argument that multilateralism encourages local procurement. However, the World Bank is not a representative multilateral institution to study procurement bias. After the World Bank commits a certain development project to a recipient state, the recipient state's government plays a 'key role in administering the procurement process' (McLean, 2017). The World Bank can still supervise the procurement process, but the domestic politics of the recipient state directly influence any decisions regarding procurement. Currently, there is little evidence on procurement in a truly multilateral setting.
This paper proposes to fill the gap in the research on multilateral procurement using data from 20 United Nations (UN) organisations over the 2013-2018 period. In our context, the power of procurement allocation lies in the hands of UN bureaucrats. A brief overview of our data, as shown in Figure 1, demonstrates that multilateralism preserves the strong procurement bias in favour of donor states, such that the states that make larger voluntary donations to UN organisations tend to receive more UN procurement orders than those that make smaller donations.
Why does the UN bureaucracy fail to insulate UN procurement processes from the influence of multilateral donors? Inspired by the literature on representative bureaucracy (Bauer & Ege, 2016) and Heinzel's (2021) work on the World Bank procurement, we argue that donor state nationals represent a key intermediate force through which donor states influence routine IO bureaucratic decision-making processes such as procurement allocations. Even if IO bureaucrats are recruited through merit-based procedures, their decisions may still be affected by their allegiance to their home countries (Weiss, 1982) or by more subtle personal preferences (Clark & Dolan, 2021) or country-specific expertise (Heinzel, 2021). For example, Clark and Dolan (2021) suggest that the US's preferences have permeated routine decision-making at the World Bank not through direct intervention but because the World Bank staff voluntarily 'design programs that are compatible with U.S. preferences', according to their interviews with development elites. Not coincidentally, North America is the most represented geographic region of origin among the World Bank staff.
By analysing official data from UN systems, we suggest that a country tends to receive more procurement funds from a UN organisation if its nationals account for a higher share of the organisation's professional staff. We further demonstrate that individual states' representation in UN organisations' professional staff is positively associated with their voluntary contributions to each organisation. To provide evidence that donor state nationals are a key instrument mediating donor state influence in multilateral organisations, we study the traditional mediating analysis of the organisational staff composition on the relationship between voluntary contributions and UN procurement orders. We find that countries that make more voluntary contributions to a UN organisation, thereby with more staff representation, receive more procurement orders. Mediation analysis suggests that donations influence UN bureaucratic decisions primarily through the channel of bureaucratic representation. We also provide evidence that our findings are not driven by the bilateralisation of multilateral aid (Bayram & Graham, 2017).
Many scholars believe that big powers influence international bureaucrats through controlling the leadership appointments (Nay, 2010). We examine whether states may affect UN bureaucracy by appointing nationals to leadership positions but find no strong evidence for the top-down argument. The head of international bureaucracy has no direct impact on the recruitment of bureaucrats or the allocation of procurement contracts. Even if donor states successfully place more nationals in leadership positions, it would not be possible for donor states to affect routine decisions through the top-down channel. Taken together, our findings suggest that multilateral institutions cannot insulate their procurement decisions from the influences of donor states because of the bottom-up influence of bureaucrats from these states. Procurement is only one example of how IO bureaucracy decisions may affect international development. There is a broad consensus that IO independence in general is a good thing. Scholars debate the independence of IOs because they disagree about whether IO independence is achievable or how it could be achieved. A large literature suggests that international organisations ought to be independent and efficient in development projects since it is staffed by career bureaucrats with policy expertise (Martens, Mummert, Murrell, & Seabright, 2002;Milner, 2006;Milner & Tingley, 2013). Independent IOs can achieve better development results than bilateral donors in terms of efficiency and welfare enhancement. Realists do not challenge the values of IO independence. They challenge the feasibility of IO independence in a world dominated by big powers. Multilateral aid execution by IOs and bilateral aid execution by states Note: A small constant was added to all values in our data.
should have similar patterns since IO decisions have to reflect the interests of powerful states (Copelovitch, 2010;Kersting & Kilby, 2016;Kilby, 2013;Mearsheimer, 1994;Stone, 2011). Specifically, Stone (2011) stated that the bureaucratic staff can constitute a channel of influence for member states. Mearsheimer (1994) argued that the diversity of the staff is significant enough to represent the balance of power in the international system. Finally, the delegation literature reconciles the above two approaches, emphasising the possibility that big powers may be willing to delegate important decisions to possibly independent IOs (Abbott & Snidal, 1998;Koremenos, Lipson, & Snidal, 2001;Pollack, 1997). This paper instead emphasises potential problems associated with IO independence. Our research is close in spirit to a small literature that discusses the pathologies associated with IOs and professional IO staff (Autesserre, 2014;Barnett & Finnemore, 1999;Karim & Beardsley, 2016). The contribution of our paper is unique as we highlight national staff composition as a plausible mechanism through which IO independence could lead to undesirable outcomes. Clark and Dolan (2021) touches on the topic briefly, but they do not test the mechanisms systematically. We carefully test the influence of both lower-level staff and the head of IO bureaucracy, and uncover a surprising pattern that lower-level staff plays a more important role in our context. Our findings do not necessarily support realists' approach to IOs, which are more compatible with a top-down view of IOs (for example, Copelovitch & Rickard, 2021;Honig, 2018). We show that lower-level staff can still plan their roles even in organisations with strong topdown controls (Honig, 2018;Woods, 2007), possibly thanks to bureaucratic rules of decisions. We thus present a nuanced view of bureaucracy independence.
The findings in this paper have direct policy implications for IO staff recruitment. The donor-centric focus of IOs is a long-standing complaint (Easterly, 2009). One clear example of this problem involves staffing. Recent positive signs suggest that IO staffing is shifting closer to the broad representation required by Chapter XV of the UN Charter. Research by Parizek and Stephen (2021a) indicates that such a shift is particularly strong in large IOs with high political and societal visibility. Although low-income countries appear to be over-represented in the staff of some UN organisations, this occurs only on a relative scale (Badache, 2020). Despite such positive improvements, the majority of UN staff originate from rich donor states (Thorvaldsdottir, 2016). Nationals from emerging powers remain under-represented in international secretariats compared with those from established powers, and the UN lags behind the International Monetary Fund and World Trade Organization with respect to the representation of emerging powers (Parizek & Stephen, 2021a). Using evidence regarding multilateral procurement, this paper emphasises the importance of recruiting nationals from developing countries and empowering them with bureaucratic power as a key mechanism through which IO bureaucracies can better serve the interests of developing countries.
The remainder of the paper is structured as follows: Section 2 reviews the existing literature on this topic and develops our theoretical hypotheses; Section 3 describes our data and methods; Section 4 reports our main results; and Section 5 presents our conclusions.

Donor state advantage in multilateral procurement
According to the UN Procurement Practitioner's Handbook (United Nations, 2020), UN bureaucrats are given considerable flexibility in operational procurement planning and requirement definition ( Figure 2). The procurement officer is responsible for developing an individual procurement plan for each case. This officer also supports the requisitioner in market studies on supply sources, pricing, availability, and the extent of competition. However, the opacity of public procurement makes it difficult to identify possible violations of procurement rules.
In general, donor states have both incentives and abilities to influence IO bureaucratic decision-making processes (for example, Kilby, 2013; Thorvaldsdottir, 2016), including procurement decisions. The selection of bureaucratic staff appears to be a crucial pathway of donor state influence (Stone, 2011). IO leadership positions are perceived to be particularly crucial in this regard (Nay, 2010), as such posts carry decision-making powers and often include a strong diplomatic function. M. Copelovitch and Rickard (2021) argue that the managing directors' identities, and specifically their ideologies, systematically influence the conditions attached to international financial rescue, to the extent that the institutional structures of many IOs are designed to minimise the leaders' discretionary influence. Powerful member states, therefore, seek dominant positions in the IOs' secretariats to increase their influence over the IOs' decisions, policies, priorities, and projects (Par ızek, 2017). In certain cases, the heads of organisations may become servants of member states rather than servants of the international community (Troy, 2021). The distribution of senior positions by nationality may provide information about the influence of states in an international system (Novosad & Werker, 2019).
Traditionally, member states try to influence IO bureaucracy by controlling the head of IO organisations. Major donor states effectively always control the nomination of the bureaucracy head. The US and the EU ensure that their nationals always control the major financial IOs, namely the World Bank and the IMF. There is a norm that the IMF chief should always be a European, whereas the World Bank chief should always be a US citizen. Emerging powers increasingly demand IO leadership positions. China particularly craves the bureaucratic heads of the UN organisations. With the support of developing countries, Qu Dongyu, the former deputy minister of China's Ministry of Agriculture and Rural Affairs, was elected as the director-general of the Food and Agriculture Organization (FAO) in June 2019. China is currently the only country leading four specialised agencies, among the 15 specialised agencies of the United Nations. In comparison, France, the United Kingdom, and the United States combined direct the same number of these types of agencies. China is ready to use all weapons in its arsenal to fight for UN leadership positions. Some US officials revealed that China threatened to cut-off key exports from Argentina, Brazil, and Uruguay if those countries did not back Qu Dongyu (Fung & Lam, 2020). It is alleged that China defeats Western-backed candidates and proposals in the United Nations by providing loans and other assistance to the developing countries in Africa, the Pacific, and other regions. For example, Cameroon suddenly withdrew the candidate for the leader of FAO in 2019 after China forgave about $78 million in debts owed by the Cameroonian government. However, ordinary bureaucrats, rather than the heads of IOs, are expected to play a more influential role in routine decision-making about procurement. According to our data, UN organisations make thousands of decisions related to procurement each year. It is impossible for agency leaders to track such a large number of decision-making processes; therefore, rulebased delegation to bureaucracy is necessary. Delegation tends to increase the effective power of bureaucrats, despite the existence of rules. Research by Lang and Presbitero (2018) shows that subjective staff judgement plays a crucial role in assessing the risk of debt distress, a process that is supposed to be based on a mechanical and model-based rule. Bureaucrats tend to have more opportunities to influence opaque procurement decision-making processes. Their decisions may explicitly or implicitly reflect the preferences of their home states (Clark & Dolan, 2021;Thorvaldsdottir, 2016). The influence does not have to be illegitimate. Heinzel (2021) shows that the increased local procurement under the supervision of the World Bank staff working in their home countries is due to country-specific expertise rather than capture by local interests. For donor states that try to influence the routinised decisions of IO bureaucracy through informal channels (Stone, 2013), placing more nationals on IO staff appears to be an effective strategy.
Donor states, particularly those that provide voluntary donations to IOs, have strong leverage over IO staff recruitment. Both mandatory and voluntary contributions are crucial to UN operations. Mandatory contributions help fund the UN's regular budget, which covers administrative costs, a few programs, and peacekeeping operations. In contrast, many UN organisations, such as the UN Children's Fund (UNICEF), the Office of the UN High Commissioner for Refugees (UNHCR), and the World Food Program (WFP), mainly rely on voluntary contributions. The UN Junior Professional Officers (JPO) Program is also established under voluntary financial support from some member states (United Nations Human Rights, 2021). Voluntary contributions are strongly associated with staff recruitment, such that the more an IO depends on voluntary donations in a given year, the more it reduces the ratio of permanent staff among its total workforce in subsequent years (Ege & Bauer, 2017).
In summary, this paper presents a two-stage argument ( Figure 3). First, countries that make more voluntary donations to an IO tend to place more nationals on their staff. Second, an increase in donor state nationals on the staff of an IO helps the donor state to receive additional procurement contracts from the IO. We are not arguing that donor countries donate to IOs to win contract money; for example, it would be absurd to argue that the US donated to postearthquake reconstruction efforts in Haiti to win procurement contracts. Rather, donation and procurement serve different political purposes.
We summarise our main argument into the following hypotheses: Hypothesis 1: When individual countries make voluntary contributions to an IO, the shares of professional staff from these countries tend to increase. Increased national representation among IO staff, in turn, enhances procurement decisions that benefit the donor countries.
In comparison, we propose the alternative hypothesis regarding the influence of the head of an IO bureaucracy on both staff recruitment and procurement.  A multilateral donation that boomerangs home 175 Hypothesis 2: The country of origin of the head of a UN organisation tends to receive higher shares of bureaucratic representation and more procurement than other countries.

Empirical design
We conducted empirical research according to the following steps. First, we studied whether a country's representation in the staff of a UN bureaucracy affects the number of procurements allocated to that country. Inspired by studies (Ege & Bauer, 2017;Thorvaldsdottir, 2016) noting that a country with more voluntary contributions to a UN organisation tends to place more nationals on the staff of that organisation, we designed the following approaches to search for possible mechanisms. First, using the fixed-effect model at the organisation-country-year level, we separately studied how voluntary contribution affects staff representation and how staff representation affects procurement allocation. Second, to further study the underlying mechanism, we used staff representation as a mediating variable and measured its influence on the relationship between voluntary contributions and IO procurement. Third, we studied how the heads of UN organisations affect bureaucratic decisions within the organisations by exploring whether these organisation heads recruit more staff from their countries of origin than from other countries. We also studied whether the head's country of origin influences the allocation of procurement contracts. We used the following official data from UN organisations at the organisation-country-year level: voluntary contributions during 2010-2018, 1 the national breakdown of the professional staff during 1997-2019 (UN Systems Chief Executive Board for Coordination, 2021a, 2021b), and procurement information during 2013-2019 (United Nations Global Marketplace, 2021). We also collected data on all UN organisation heads during 1997-2020. 2 In total there are 131 officials from 60 countries. The main findings are based on an analytical sample of 20 organisations for which complete sets of the above data are available during the 2013-2018 period.
Our baseline model used staff (the number of international professional staff from a country in an organisation divided by the total number of international professional staff in that organisation in each year) or procurement (the yearly procurement of an organisation to a country divided by the worldwide yearly procurement of the organisation) as the dependent variables, and voluntary contribution (the yearly voluntary contribution of a country to an organisation divided by the yearly voluntary contribution of all countries to the organisation), staff, or home country (whether the head of the IO originates from a given country or not) as the main explanatory variables.
All of our regressions included control variables and fixed effects at the country, UN organisation, and year levels. Robust standard errors are clustered at the country and organisation levels.
To explore possible mediation channels through staff recruitment, we use staff as a mediating variable, voluntary contribution as an explanatory variable, and procurement as a dependent variable. To define staff, we focussed on staff in the category of International Professional, who are often bureaucrats with decision-making powers.
We do not claim that our mediation analysis is causal. After reviewing a number of developments in the field, Keele (2015) concludes that causal mediation analysis requires 'strong untestable assumptions even when conducted in the context of a randomized control trial', Rubin (2013) rejects the enterprise of causal mediation entirely, saying it hardly works within the Neyman-Rubin counterfactual framework. One strong assumption proposed by Imai, Keele, and Yamamoto (2010) is called sequential ignorability, which implies that researchers should control all the pretreatment variables that affect both mediator and outcome variables. To shed light on the possible mediation channel in our context, we perform a traditional mediation analysis, yet also follow the spirits of Imai et al. (2010) as closely as possible by controlling variables that affect both procurement and staff recruitment. We find an exhaustive list of possible control variables from the literature, in total 44 variables from 7 papers. Details about the major control variables of each paper can be found in our supplementary materials. We decided to include only eight of these control variables in our mediation analysis. As our regression already controls for country fixed effects, all those control variables that do not vary over time are discarded. We discuss the control variables that are used in our mediation analysis below, with justifications of why they are included.
First, the education level and population in a country may impact the number of qualified candidates for IO and qualified suppliers for procurement (Badache, 2020;Thorvaldsdottir, 2016). Therefore, we included control variables, Average Years of Education and Population (ln). Also, the extent to which an IO is active within a country may influence local people's familiarity with an IO, which may influence the number of potential candidates and suppliers. Therefore, we included Local IO Activity as proposed by Parizek and Stephen (2021a). We also included GDP per Capita (ln), GDP Growth (annual %), and Trade Openness. These measures of economic capacity have a clear impact on both staff recruitment and procurement in our context (McLean, 2017). To ensure the fair allocation of development assistance, IOs take corruption into account during the process of contract bidding (McLean, 2017). This variable may affect recruitment as well, though both directions are possible. We included the Level of Corruption of each country accordingly. Finally, we control the Level of Democracy, since countries with a higher level of democracy may have higher levels of political globalisation and be over-represented in IO and in the procurement process (Heinzel, 2022).
For those control variables, Level of corruption, and Level of Democracy during 1997-2019 were obtained from the V-Dem Dataset (Coppedge et al., 2020). GDP per Capital (ln), GDP Growth (annual %), Trade Openness, 3 and Population (ln) during 1997-2019 were obtained from the World Bank's World Development Indicators (World Bank, 2021). Average Years of Education during 1997-2019 was obtained from Human Development Reports of the United Nations Development Programme (2022). We calculate Local IO Activity following Parizek and Stephen (2021b) using our data. 4 The summary statistics of our sample are shown in Table 1. Appendix A contains a list of the organisations and countries that appear in our sample. Table 2 reports the impacts of bureaucratic representation on procurements. Column 1 shows how voluntary contribution affects bureaucratic representation. Column 2 shows how representation affects procurements. Columns 3 and 4 report how the country of origin of an organisation's head affects bureaucratic representation and procurements, respectively. Column 1 of Table 2 reports the regression coefficients for the impacts of voluntary contribution on bureaucratic representation for all countries in our analytical sample. The control variables in the regression included the level of IO activity, GDP per capita (log), level of corruption, level of democracy, population (log), GDP growth, trade openness, and average years of education. In Row 1 of Column 1, the estimated coefficient shows that a 10 per cent increase in the share of voluntary contributions from a country to a UN organisation is associated with a 0.65 per cent increase in the proportion of that country's nationals on the professional staff of the organisation. The estimated regression coefficient is significant at the 1 per cent level. Row 4 of Column 1 shows that when voluntary contribution and other control variables are held constant, a 10 per cent increase in the level of a country's local IO activity is associated with a 0.01 per cent increase in the proportion of that country's nationals on the professional staff of the organisation. Again, the estimated regression coefficient is significant at the 1 per cent level. Other control variables are not statistically significant at conventional levels except local IO activity and GDP per capita.

Baseline results
Column 2 of Table 2 reports the impacts of bureaucratic representation on procurement, using the same set of control variables as in Column 1. Row 2 of Column 2 shows that a 10 per cent increase in the proportion of a country's nationals on the staff of a UN organisation is associated with a 14.40 per cent increase in the share of procurement received by the country from the organisation, which is significant at the 1 per cent level. Row 4 of Column 2 shows that the coefficient of local IO activity is significant at the 10 per cent level. When staff representation and other control variables are held constant, a 10 per cent increase in the level of a country's local IO activity is associated with a 0.01 per cent increase in the share of the procurement received by that country from the organisation. Again, other control variables are not statistically significant at conventional levels except local IO activity and GDP per capita. Columns 3 and 4 of Table 2 show the results pertaining to the country of origin of an IO head. Row 3 of Column 3 shows that the head's country of origin does not enjoy more bureaucratic representation than other countries. Row 3 of Column 4 shows that the head's country of origin does not receive a higher share of procurements than other countries. In these two regressions, when the IO head's country of origin is held constant, in both Columns 3 and 4, Row 4 shows that a 10 per cent increase in the level of a country's local IO activities remains associated with a 0.01 per cent increase in the share of that country's nationals on the professional staff of the organisation (significant at the 1% level) or 0.02 per cent increase in the share of procurement received by the country from the organisation (significant at the 5% level). Row 5 of Column 3 shows a 10 per cent increase in the GDP per capita (ln) is associated with a 0.01 per cent increase in the share of that country's nationals on the professional staff of the organisation.
Summarising our results in Table 2, we show that countries that make larger voluntary contributions to UN organisations tend to have greater representation in the organisation staff and subsequently receive more UN procurement contracts than other countries. However, the country of origin of the organisation head does not receive a higher share of staff recruitment or procurements than other countries. Table 3 presents the results mediation analysis obtained using the Baron and Kenny (1986) method. In our mediation analysis, Voluntary Contribution is the independent variable, Procurement is the Dependent variable, and Staff is the mediator. We use the same set of control variables as the baseline regressions.

Mediation analysis
From Table 3, the relationship between Voluntary Contribution and Procurement is completely explained by Staff. First, Row 3 of Column 1 shows that the total effect of voluntary contribution towards procurement is 0.215, which is significant at the 10 per cent level. Second, from Row 3 of Column 2, the voluntary contribution significantly affects the mediator, Staff, at the 1 per cent level. Third, from Row 2 and Row 3 of Column 3, for given levels of Voluntary Contribution, Staff significantly affects Procurement at the 1 per cent level. Meanwhile, the direct effect of voluntary contribution on procurement is not statistically significant. This shows that Staff fully mediate the impact of Voluntary Contribution on Procurement and the possible causal mediation effect is 0.089 (0.065 times 1.365).
Summarising our results in Table 3, Staff as the mediator fully mediate the impact of Voluntary Contribution on Procurement. A summary figure can be found in our Supplementary Materials.
The mediation relationship is summarised in Figure 4, which reports the estimated mediation effects and calculates its standard error using bootstrapping method. We show that staff representation fully mediates the impact of voluntary contribution on procurement, using bootstrapping procedures with 1000 bootstrap replicates and a 95 per cent confidence level (CI). The possible average causal mediation effect (ACME) is 0.089, which is significant at the 5 per cent level. The bootstrapped unstandardised indirect effect is 0.089, and the 95 per cent CI ranges from 0.043 to 0.130 for 1000 bootstrapped samples. The average direct effect of voluntary contribution on procurement is not significant (95% CI ¼ À0.055 to 0.309).
In summary, our evidence from regression and mediation analysis indicates that staff representation is a key pathway through which donor states receive more procurement contracts from UN organisations than other states. In contrast, the countries of origin of the heads of UN organisations are not associated with increased representation of nationals from that country in the organisation or increased procurement contracts issued to that country.

Bilateralisation of multilateral aid?
Voluntary donations to multilateral institutions are often earmarked and resemble bilateral aid in many ways (Bayram & Graham, 2017). Is it possible that our findings are driven by the bilateralisation of multilateral aid?
We further show that our main impact of staff representation remains robust even if we focus on the earmarked or non-earmarked Voluntary Contribution instead of the total Voluntary Contribution. Staff representation fully mediates the impact of non-earmarked voluntary contribution on procurement, and partially mediates that impact of the earmarked voluntary contribution. This shows that the earmarked voluntary contributions affect the procurement allocation in a slightly more direct way compared to the non-earmarked contributions, which largely affect procurement through staff representation. Details can be found in our supplementary materials. The bilateralisation of multilateral aid does not have a huge direct impact on procurement because the multilateral bureaucracies keep the institutional power over procurement decisions. For example, the World Bank forbids the donors to tie earmarked donations to procurement decisions (Reinsberg, Michaelowa, & Knack, 2017).

Discussions
Why does national staff representation affect procurement allocation? Our interviews with UN staff implies the importance of information. Mid-level UN officials responsible for the execution of specific development projects make key procurement decisions. UN officials might spread the official procurement through their personal networks, attracting more bidders from their home countries. Alternatively, bidders from the countries that have more staff in the UN system are more likely to track the official procurement announcements in a more efficient way. In both cases, the countries with more UN staff tend to have more procurement bidders.
In the screening stage, many bidders are eliminated simply because of technical or procedural problems. Bidders who are familiar with basic procurement procedures are far more likely to survive the screening. It is plausible that the countries having contributed a lot to UN staff tend to know more about procurement rules and applications.
The information channels, in addition to the subtle personal preferences (Clark & Dolan, 2021) channel, may explain the correlation between staff representation and procurement bias. They are legitimate in the UN legal framework. There is no qualitative evidence to support the view that mid-level UN officials cultivate corrupted personal connections with potential bidders from their home countries.

Conclusion
This paper presents one of the first studies on donor state advantages in multilateral procurement. Using comprehensive data from the UN system, where UN bureaucrats control Notes: ACME stands for average causal mediation effects. ADE stands for average direct effect. The bootstrapped unstandardised indirect effect (ACME) was 0.088 and the 95 confidence interval ranged from 0.041 to 0.130 for each 1000 bootstrapped samples. Thus, the indirect effect was statistically significant. Standard errors are clustered by country and organisation. Control variables contain GDP per Capita (ln), local IO activities (ln), level of democracy, level of corruption, population(ln), GDP growth (annual %), trade openness, and average years of education. Fixed effects at the country, UN organisation, and year level are included. procurement decisions, this paper argues that procurement deviates from the UN regulations and Handbook and is biased in favour of donor states. We identify the representation of the donor state in the organisation staff as the key pathway through which donor states benefit from procurement bias. We also find that the country of origin of the head of a UN organisation barely influences staff representation or procurement decisions. This paper indicates that the independence of IO bureaucrats may be a double-edged sword. It may infringe on the impartiality of the UN system, compromise the principles of fairness and transparency, and contravene the rights and benefits of developing countries. The established formal procurement rules could not effectively prevent donor countries from pursuing private benefits. Oath of allegiance to the international community is hardly sufficient to prevent IO bureaucrats who may remain loyal to their countries of origin, prefer products and services from home countries or are influenced by other subtle factors, at the potential expense of developing countries. Here bureaucracy independence and informal norms combined appear to be the root of the problem. On the other hand, bureaucracy independence could prevent the head of bureaucracy from enjoying deep, penetrating top-down power, as the leadership positions appear to have little power over routinised decisions such as lower-level recruitment and procurement. Here the benefits of bureaucracy independence appear to derive from the formal procurement rules.
Our paper has implications for UN system reform. Given that developed countries represent the majority of staff in UN organisations (our calculation based on official UN data), we suggest that future UN reform efforts should focus on bureaucratic representation rather than independence. These efforts should focus particularly on bureaucratic representation in resource allocation and the informal norms and rules that drive resource allocation decisions. However, we do not recommend relaxing the rule-based decision-making process inside the UN bureaucracy. Without such rules, UN bureaucracy could become excessively politicised and turn into a battlefield of big powers. Our paper should not be interpreted as an allegation against the UN bureaucrats. The UN bureaucrats may be unable to resist influence from their home countries, given the UN's chronic underfunding and the numerous methods that powerful nations try to maintain control over their citizens. The UN reform should think about effective methods of shielding UN bureaucrats from external influences.
Our paper also suggests several lines of future research. First, further research is needed to clarify how financial contributions improve a donor country's bureaucratic representation, specifically whether financial contributions transform the IO's recruiting procedures or preferences and whether countries that provide more donations also actively supply more applicants to the IOs than other countries. Second, future studies should explore how the citizenship of an IO bureaucrat affects his or her routine decisions, and whether this potential influence is intentional or unintentional. Third, future scholars can explore how such analysis might be expanded from procurement to other forms of IO bureaucratic decision-making. We believe that these lines of research will help us to better understand how international agencies fulfil their mandates in a complex environment. Notes 1. Voluntary contribution includes two types of contributions: specific contributions and non-specific contributions.
In this research, we sum these types to determine the voluntary contribution and do not distinguish the types in our data because they are highly correlated. 2. We did not use the dataset from Novosad and Werker (2019), which is only updated to 2013. 3. Trade openness is 'the sum of exports and imports of goods and services measured as a share of gross domestic product'. 4. Local IO activity is defined as the size of the locally hired general services staff (for example, drivers and secretaries).