Differences in Aid Allocation Motives of
Official Bilateral, Multilateral, and NGO Aid
Master Thesis for the Double Degree Programme between University of Göttingen and University of Groningen
Master of Arts: International Economics
Master of Science: International Economics and Business
First Supervisor: Dr. Anna Minasyan (University of Groningen) Second Supervisor: Prof. Dr. Stephan Klasen (University of Göttingen)
Countries allocate foreign aid through three main channels: Official Development Assis-tance (ODA), Non-Governmental Organisations (NGO) and multilateral organisations. This analysis investigates the aid allocation from Sweden and Switzerland to 151 recipient countries between 2000 and 2014 with respect to four donor motives: Need-based, Strategic, Merit-based and Mimic motive. Tobit panel regressions weakly confirm a need-based motive. Whereas Swedish ODA is found to follow economic motives, NGOs and Swiss ODA allocate altruistic. Moreover, ODA focuses on politically difficult environments. Surprisingly, NGO allocation is not only shaped by ODA, but by private donors and multilateral organisations, as well. Especially the allocation pattern of the United Nations is correlated with NGO aid on a similar level as ODA.
Keywords: Foreign aid allocation, donation motives, principal-agent model
Adrian Bernhard Monninger
Contents
1 Introduction 1 2 Theoretical Background 3 2.1 Need-based motive . . . 3 2.2 Strategic motive . . . 4 2.3 Merit-based motive . . . 6 2.4 Mimic motive . . . 73 Data and descriptive statistics 8 3.1 Aid Allocation . . . 9
3.2 Explanatory variables . . . 13
4 Methodology 16 5 Results 18 5.1 Need-based, Strategic and Merit-based donor motives . . . 18
5.2 Mimic donor motive . . . 23
5.3 Robustness tests . . . 30
6 Conclusion and Policy Implications 34
7 Bibliography 36
1
Introduction
Foreign aid is used for multiple purposes. It can rebuild societies after conflicts and/or meet humanitarian needs. The common view is that foreign aid can reduce poverty and stimulate economic growth in less developed countries (Burnside and Dollar, 2000; Collier and Dollar, 2002). Nevertheless, there are as manifold donors of development aid as there are reasons for it. Additionally, no clear international guideline exists which donor should allocate how much aid to each recipient country. The freedom of decision making makes aid heterogeneous. Different actors can allocate aid with a different focus, based on their agenda. Hence, the pattern may vary across foreign aid channels. Apart from private aid, the three main channels are: Official Development Assistance (ODA), Non-Governmental Organisations (NGO) and multilateral organisations.
ODA is bilateral government to government aid and aims at increasing the economic development of developing countries (OECD, 2018b). The incumbent government can decide which projects in which countries get funded by taxpayer money. The aid allocation process in NGOs is less centralised. Every NGO executes projects based on their individual agenda. Therefore, NGOs use own resources, mostly private donations, and official aid funds (Dreher et al., 2010). Multilateral organisations receive their budget from member countries. Some donations are bound to specific projects or areas, whereas with others the organisations can decide independently which projects are promoted (OECD, 2015).
Since the 1980s the argument from Tendler (1982) that NGO aid is the superior channel of aid allocation found much support on a theoretical basis. It was argued that NGOs are more flexible and innovative than other channels which could result in a more efficient poverty alleviation (Fruttero and Gauri, 2005). Moreover, NGOs may be less distorted from strategic donor motives due to their political independence. Unfortunately, the empirical evidence since the 2000s shows ambiguous results depending on the donor country and questions the status of a superior aid channel.
In this paper the three channels are compared in terms of need-based, strategic, merit-based and mimic aid allocation motives. For this purpose, three studies which investigate the differences between ODA and NGO aid in Sweden (Dreher et al., 2010) and Switzerland (Dreher et al., 2012; Nunnenkamp et al., 2009) are replicated and extended. The scholars use a similar methodology and investigate country specific characteristics. In line with them, the research question of the paper is: Do the motives for foreign aid allocation differ across various channels?
The answer to this question can have an important political impact. If all channels support different projects due to different motives, society has to choose which channel reflects their preferences best. This channel needs then to be promoted and the others need to change. Society finances official and multilateral aid directly through taxes and NGOs through donations. Hence, donation motives should match.
countries official numbers are given for a long time period which makes the comparison across channels possible. Second, the papers find differences in donor motives between the channels of ODA and NGO aid in the countries for the period from 2001/02 until 2005/06 (Dreher et al., 2010, 2012; Nunnenkamp et al., 2009). Third, previous literature supports contradicting results. While Neumayer (2005) and Berthélemy (2006) find that Swedish and Swiss ODA is altruistic, the key papers identify political strategic motives.
All in all, the paper aims at five different aspects. First, an extended time period functions as an additional robustness test. Current data for Sweden and Switzerland can be used to verify the previous results. Moreover, the time period includes multiple government changes which may have affected the foreign aid allocation.
Second, a different methodology is used. The key papers use averages of aid allocation from 2001/02 until 2005/06 and estimate their regressions with a cross-sectional data set. The extended time period allows for a panel data analysis. This offers several methodological advantages (see section 4).
Third, previous analyses have only focused on the overall aid allocation pattern and assumed homogeneity across regions. While allowing for heterogeneity and investigating donor motives separately for each region, different foci can be investigated. This gives a deeper understanding of donor behaviour.
Fourth, ODA and NGO aid allocation across the two donor countries can be compared directly. While using the exact same methodology, structural differences or behavioural patterns can be examined. Especially the comparison between NGOs is interesting, because previous literature regarding this is scarce and the data sets differ in their budget funding. While Swedish NGOs are mainly state funded, Swiss NGOs receive only private donations. Lastly, the principle-agent model of Fruttero and Gauri (2005) is extended by allowing for multilateral agencies and private donors as additional principals. This gives a deeper understanding of herding behaviour of aid channels and can have an influence on government spending. When the mimicking aid behaviour of other channels replaces donor motives and prevents efficient aid allocation, the dependency of NGOs may need to change and other funding options may be necessary.
by ODA, but by private donors and multilateral organisations as well. Especially the United Nations seem to serve as an orientation for NGOs, since the allocation pattern of the latter is very similar to that of the UN.
The remainder of the paper is structured as follows. First, a detailed overview of the literature of the four donor motives is given. Previous results lead to the hypotheses. Afterwards, the data set and methodology are explained in detail, followed by the results and robustness tests. Finally, the conclusion consists of a brief discussion with policy implications, limitations and future research.
2
Theoretical Background
The conceptual framework of this thesis is based on three key papers which investigate donor motives of official bilateral and NGO aid. Dreher et al. (2010) investigate Sweden, whereas Nunnenkamp et al. (2009) and Dreher et al. (2012) focus on Switzerland. This paper will replicate and extend their findings. The target of Dreher et al. (2010) and Nunnenkamp et al. (2009) is aimed at three main motives of donor behaviour.
1. Need-based motive: The poorer the country, the more aid it receives respectively. 2. Strategic motive: Individual economic and political donor interests can drive foreign
aid. If not aid is labeled as ’altruistic’.
3. Merit-based motive: Countries with a more difficult political environment receive more aid.
The last motive below is inspired by Dreher et al. (2012) who investigate the correlation between ODA and NGO aid in order to test whether both channels allocate in a similar way or not. Additionally, the data set allows to compare NGO with multilateral aid, as well. Therefore, motive four states:
4. Mimic motive: Donor agencies act similar to each other.
In the next chapters the intuition behind the motives is explained and previous findings guide towards the hypotheses.
2.1
Need-based motive
to allocate most of their aid to middle income countries, whereas Germany, France, the Netherlands and Japan have no clear favourite (Baulch, 2006). Dreher et al. (2010) and Nunnenkamp et al. (2009) find for Sweden and Switzerland that in general more bilateral aid is allocated to recipient countries with a lower GDP per capita. However, whether low or middle income countries are more focal is not investigated.
NGOs are seen to have a stronger poverty focus than official donors due to the direct contact with target groups (Riddell et al., 1995). Moreover, their small scale, lower cost and flexibility compared to ODA may make them more efficient in aid targeting (Dreher et al., 2010). However, the empirical results are ambiguous. On the one hand, Nancy and Yontcheva (2006) analyse European NGOs and find that the national poverty level is the strongest motive and Koch et al. (2009) and Büthe et al. (2012) identify a higher NGO aid allocation to poorer countries for international and US NGOs. On the other hand, Edwards and Hulme (1996) and Koch et al. (2007) question the effectiveness of NGOs while arguing that NGOs behave more similar to the donor agencies whose policy agenda they have to implement. Therefore, they argue that the need-based motive is not the most driving factor of aid allocation.
For the focal countries of this study, Nunnenkamp et al. (2009) and Dreher et al. (2012) find that NGOs have no stronger poverty focus than ODA in Switzerland, nevertheless, they allocate more to poorer countries. Swedish NGOs, in contrast, are not driven by this motive which is an interesting result, because ODA is found to be shaped by GDP per capita. Hence, Swedish official bilateral aid is superior in the need-based motive compared to NGO aid.
On the multilateral level, in general the highest aid share flows to the least developed countries (Maizels and Nissanke, 1984), but the empirical evidence differs again. Scholars find that the World Bank seems to be influenced by this motive (Frey and Schneider, 1986), where two-thirds of all foreign aid flows to low income countries (Baulch, 2006). The UN gives a similar share, but at the same time a large disbursement to few richer countries as well, whereas the EU prefers middle income countries (Baulch, 2006). Other studies support these findings (Canavire et al., 2006; Neumayer, 2003a). Nevertheless, whether multilateral aid is better targeted towards the poor than ODA is still in question. Dollar and Levin (2006) find support for this and Canavire et al. (2006) do not.
Accordingly, hypothesis one states:
1. Need-based motive: The poorer the country, the more aid it gets respectively. The extent varies across channels and countries.
2.2
Strategic motive
in several ways. First, official aid can be used to promote exports to recipient countries (Berthélemy and Tichit, 2004; Canavire et al., 2006). A larger amount of foreign aid helps the economic development of the country which then increases imports from donor countries. Second, ODA can ensure that recipients meet debt obligations towards the lender (Cohen et al., 2007). Additional grants reduce the risk of default. Finally, countries with a larger resource endowment may receive more aid in order to assure a steady supply of resources (Dreher et al., 2010).
The debate whether various donors are altruistic or base their allocation on strategic motives is especially large when it comes to ODA. Alesina and Dollar (2000) observe that several official donors follow a strategic agenda. The extend differs across countries which is investigated by several scholars (e.g. Berthélemy, 2006; Dollar and Levin, 2006). Countries like Denmark, the Netherlands, Norway, and Sweden seem to be more altruistic (Neumayer, 2005). In contrast, the motives for countries like France or the US were found to be strategic (Dollar and Levin, 2006) and therefore, can be labeled as ’egoistic’ (Berthélemy, 2006). Nevertheless, several donors reduced their policy orientation over the
years (OECD, 2004) and the evidence for certain countries is ambiguous. For example, Germany is found to prefer trading partners (Berthélemy, 2006), whereas Nunnenkamp and Öhler (2011) reject an economic, but support a political motive.
Dreher et al. (2010) and Nunnenkamp et al. (2009) find support for a political motive in the cases of Sweden and Switzerland, by testing for similarities in the voting pattern in the UN General Assembly between the donor and the recipient countries. However, an economic motive was not found and a larger resource endowment in the recipient country even decreased the Swedish ODA share.
NGO aid, on the other hand, was believed to be less distorted by commercial and political interests (Tendler, 1982). However, empiric results go in both directions. Some scholars support this hypothesis regardless of government dependency (Büthe et al., 2012; Nancy and Yontcheva, 2006), whereas others find that especially financially dependent NGOs allocate strategically and in line with their official back donor (Fruttero and Gauri, 2005; Koch et al., 2007). Hence, the altruistic motive gets replaced.
Dreher et al. (2010) and Nunnenkamp et al. (2009) find an altruistic behaviour for Sweden and Switzerland, respectively. Neither possible political nor economic interests of the donor countries seem to shape the NGO aid allocation pattern which makes the NGO channel superior in this motive.
2013). Historically, the UK wanted to support its former colonies, while Mediterranean and Nordic countries supported a humanitarian approach, regardless of the recipient’s location. After the fall of the Soviet Union more aid for Central and Eastern Europe is in favour of all member countries (Schneider and Tobin, 2013). In contrast, some scholars find that neither the UN, nor World Bank or IMF seem to be influenced by their donor countries (Canavire et al., 2006; Neumayer, 2003a; Schneider and Tobin, 2013).
Hence, the second hypothesis states:
2. Strategic motive: ODA is shaped by political and economical interest, whereas NGO aid is altruistic.
2.3
Merit-based motive
Merit-based foreign aid allocation should focus on politically difficult environments. This covers political instability, institutional flaws or violent conflicts. Merit-based motives can influence donor behaviour in two directions. First, aid allocation can reward an improvement of institutions and governance. When a government is willing to change political structures, more aid can follow. In addition to that, especially NGOs have the incentive to change the governance climate within the country and hence, focus on countries with a difficult environment. Second, worse institutions can function as barriers. Accordingly, foreign aid is more expensive, because projects have to deal with the environment which results in higher costs. Therefore, it is argued that the overall level of aid effectiveness can be increased by allocating more to well-governed countries (Dollar and Levin, 2006; Dollar and Pritchett, 1998).
Several scholars have argued that governments do not allocate significantly more to countries, in which basic institutions and economic policies may show good local conditions (Burnside and Dollar, 2000; Collier and Dollar, 2002). Moreover, Alesina and Weder (2002) and Neumayer (2003c) argue against the argument that ODA donors reward good governance. They find that governments do not give more aid to less corrupt governments, except Scandinavian donors. Neumayer (2003b, c) investigates all OECD countries with different results. He finds that the focal countries (Sweden and Switzerland) seem to reward political conditions and provide more aid to democratic regimes. He argues that official aid may not work when recipient countries have bad institutional requirements.
have a comparative advantage, but are active in all types of countries, regardless of the governance structure. The lower cost argument is directly tested by Nunnenkamp and Öhler (2012) who find no empirical support by comparing project costs of US NGOs. Several scholars indicate that U.S. and Norwegian NGOs allocate more aid to politically difficult environments, whereas this is not the case for Germany, Sweden or Switzerland (Edwards and Hulme, 1996; Koch, 2009; Koch et al., 2007; Nunnenkamp et al., 2009).
Dreher et al. (2010) and Koch et al. (2007) argue that financially dependent NGOs try to provide aid in well-governed countries in order to find easier future funding prospects. Those projects have a higher chance of success and generate visible results. Other projects contain a higher risk of failure or nonrecognition.
Furthermore, 39 low and middle income countries implemented NGO restrictions between 1993 and 2012, because the governments feared that those organisations may support political opponents which undermines their power (Dupuy et al., 2016). Therefore, access was denied and NGO aid could not be allocated to those countries.
On the multilateral level, most scholars argue that good governance gets rewarded with higher shares of aid. Frey and Schneider (1986) find in an early approach that the World Bank allocates more aid to politically stable countries, whereas the statistical support for UN agencies in that matter is weak (Neumayer, 2003a). Whether bilateral or multilateral aid focuses more on good governed countries is still in question. Dollar and Levin (2006) show that between 2000 and 2003 the governance mode had a stronger effect on multilateral aid, whereas Canavire et al. (2006) find no support for that.
Accordingly, it is stated:
3. Merit motive: NGOs have a comparative advantage of working in difficult environ-ments. Official agencies allocate more to good governed countries.
2.4
Mimic motive
The first three hypotheses try to explain differences in aid allocation across the channels based on different donor motives. The hypothesis stated in this section tries to explain similarities.
results in the past as well (Rose-Ackerman, 1982). Only successful projects can promote a fund raising campaign. Second, multilateral agencies do also fund projects and cooperate with NGOs e.g. through consultancy (Department of Economic and Social Affairs, 2018). Therefore, NGOs are dependent on all three principals (ODA, private donor, multilateral agencies) who provide them with funds in the future.
A positive side effect from the mimic behaviour are possible synergy effects. When aid channels cooperate and work in the same countries, the aggregated impact of the projects can be larger than the sum of the individual ones (Koch, 2009). Moreover, in cases of emergencies like natural disasters or wars, herding behaviour increases aid effort from donors (Frot and Santiso, 2009).
A negative side effect is that uncoordinated projects decrease aid efficiency and prevent possible synergy effects (Frot and Santiso, 2009). In addition to that, state agencies in the recipient countries have to deal with many NGOs separately which imposes an unnecessary burden on them (Djankov et al., 2009). Furthermore, herding can cause volatility in aid flows with potentially high costs for recipient countries, as evaluated in Arellano et al. (2009).
In contrast, arguments about a comparative advantage of NGOs, especially in difficult environments, give support for the theory that a different allocation pattern is beneficial. Additionally, differences in the three aforementioned donor motives may also diverge the allocation pattern. The last point may be especially relevant when it comes to different multilateral organisations. Every agency has different guidelines and different foci where to allocate the money. Hence, the correlation between NGOs and each organisation may vary.
Therefore, hypothesis four states:
4. The motives of NGO aid differs from bilateral and multilateral behaviour. Never-theless, NGOs are influenced by all channels with a different magnitude. Private funded NGOs act risk avers as well.
3
Data and descriptive statistics
The full panel data set consists of foreign aid allocation from Swedish and Swiss govern-ment to governgovern-ment, NGO and multilateral aid from the United Nations, World Bank, International Monetary Fund and European Union to 151 recipient countries from 2000 until 2016/17.1 Table 1 shows the number of recipient countries in the data set by region and income categories.2 Most recipient countries in the data set lie in Europe and Central Asia which can be explained by the better data availability in this region. In contrast to
1Some variables are restricted for certain years. The UN Votes for Switzerland are only
inter-pretable after 2001, because of the Swiss observer role which excluded a voting mandate. Moreover, UN Votes are only given until 2014. Therefore, the estimation sample for Sweden ranges from 2000 until 2014 and for Switzerland from 2002 until 2014.
Table 1: Recipient Countries by Region and Income Category
Region Number Income Category Number
East Asia & Pacific (EAS) 18 High income 44
Europe & Central Asia (ECS) 46 Upper middle income 41
Latin America & Caribbean (LCN) 23 Lower middle income 41
Middle East & North Africa (MEA) 16 Low income 25
North America (NAC) 2
South Asia (SAS) 7
Sub-Saharan Africa (SAF) 39
previous literature, high income countries are not excluded, because first they still receive a small portion of foreign aid - from Swedish and Swiss NGOs, UN and EU - and second, they are included in the decision of each channel to execute projects. Hence, by neglecting them, the motive gets distorted and the interpretation does only cover low and middle income countries.3
The panel data set is restricted by two factors. First, due to the estimation approach of a Tobit model, only positive aid allocation is used and therefore, negative allocation is replaced by zero.4 Second, due to problems in data availability only countries are included
where GDP per capita, as well as resource depletion data was found.
3.1
Aid Allocation
The dependent variable is specified as the natural logarithm of 1 + the amount of aid allocation in million $US to the recipient country i in year j. With this, the minimum value of the variable is zero and the data is less skewed (e.g. Nunnenkamp et al, 2009).
In order to get data of Swedish NGOs, the Swedish International Development Coop-eration Agency (SIDA) records all NGO projects which they financially promote since 1998. A restriction is that the NGO needs at least 10% own resources and the projects should follow the overall goal ’to work for a viable and pluralistic civil society in developing countries that act from a rights perspective for improved living conditions and for people living in poverty in all its dimensions for greater respect for human rights and a global sustainable development’ (SIDA.se, 2018). The small amount of own resources necessary may raise the question how independently those NGOs can allocate their funds. Riddell et al. (1995) and Pratt et al. (2006) argue that SIDA does not fund each project separately, but allocates money to each organisation. In addition, neither the mandate nor the ability is given to influence the NGO’s behaviour. Both arguments are in favour of a limited
3In the robustness tests in section 5.3 it is shown that high income countries do not change the
results.
4Negative aid allocation was only seen in Swiss NGO aid to Somalia in 2016. This means that
dependency between the government and NGOs, but do not rule it out completely. As figures 1 and 2 display, Swedish ODA is more allocated in the region of Europe and Central Asia, whereas NGO aid is relatively focused on Latin America and the Caribbean. Nevertheless, the correlation matrix in table 3 shows a large overall similarity between the channels. The descriptive statistics in table 2 indicate a lower mean contribution compared to other channels which can be explained by the restricted budget of NGOs.
Data on Swiss NGO allocation was made available by the Swiss Agency for Development and Cooperation (SCD) and reports the contribution of privately financed NGOs to recipient countries from 2000 to 2016. Here, all aid is received by low and middle income countries with a focus on Sub-Saharan Africa, South Asia and Latin America and the Caribbean (see figures 1 and 2). Important is the fact that Swedish and Swiss NGOs differ in their independence. Whereas Swiss NGOs are solely private, Swedish NGOs are up to 90% state funded. Nevertheless, the correlation matrix (table 4) shows that the correlation between ODA and NGO aid is larger in Switzerland than in Sweden. This suggests that herding behaviour can be observed even for NGOs which are financially independent from official funding.
The allocation of ODA as well as from multilateral organisations is available by OECD statistics. For this paper, aid between 2000 and 2016 to all sectors is used. The OECD Development Assistance Committee defines ODA ’as government aid that promotes and specifically targets the economic development and welfare of developing countries’ (OECD, 2018b, p. 1). Therefore, resource flows are measured which follow two conditions. First, ODA needs to be implemented by official agencies and second, the character has to be concessional, such as soft loans and grants, with a focus on the economic welfare of developing countries (OECD, 2018b). The collected aid activities cover direct government to government aid, as well as projects channeled through NGOs. Additionally, donor costs linked to development are measured as well (OECD, 2018a). Therefore, it is important to emphasise that the NGO databases cover only NGO projects which are partly to fully private financed, whereas the OECD database includes NGOs which undertake projects financed by the government. The former NGOs are more independent than the latter.
Figure 1: Share of aid allocation by Region of Recipient Countries
Share of aid to each recipient region. Aid between 2000 and 2016 or 2017 (for Swedish NGO) EAS: East Asia & Pacific; ECS: Europe & Central Asia; LCN: Latin America & Caribbean; MEA: Middle East & North Africa; NAC: North America; SAS: South Asia; SAF: Sub-Saharan Africa
Figure 2: Share of aid allocation by Income category of Recipient Countries
Table 2: Summary statistics of aid allocation.
Mean SD Number Min Max Total
NGO Sweden 0.45 1.12 2421 0 16.70 1086.19 NGO Switzerland 1.78 3.14 2421 0 29.99 4313.37 ODA Sweden 7.55 20.95 2421 0 320.95 18268.36 ODA Switzerland 5.71 11.89 2421 0 163.70 13822.68 United Nations 12.82 24.01 2421 0 190.89 31044.53 World Bank 83.16 254.18 2421 0 3814.60 201338.17
International Monetary Fund 7.94 35.10 2421 0 543.64 19226.50
European Union 67.80 195.86 2421 0 4051.56 164136.16
Aid allocation from donor country to each recipient country per year (2000-2016/17). In million $US.
Table 3: Correlations of Aid Channels: Sweden between 2000 and 2016 Swedish Swedish United World International European
NGO ODA Nations Bank Monetary Union
Fund Swedish NGO 1.00 Swedish ODA 0.60 1.00 United Nations 0.51 0.62 1.00 World Bank 0.37 0.51 0.69 1.00 International 0.15 0.28 0.38 0.53 1.00 Monetary Fund European Union 0.40 0.62 0.72 0.50 0.32 1.00
Table 4: Correlations of Aid Channels: Switzerland between 2000 and 2016
Swiss Swiss United World International European
NGO ODA Nations Bank Monetary Union
Fund
Swiss NGO 1.00
Swiss ODA 0.70 1.00
United Nations 0.73 0.66 1.00
World Bank 0.57 0.57 0.69 1.00
International Monetary Fund 0.28 0.31 0.38 0.53 1.00
3.2
Explanatory variables
This section describes the variables which measure the different donor motives. The summary statistics are displayed in table 5, a more detailed description can be found in table A1. The expected direction of correlation according to the hypotheses are part of table 6.
Hypothesis one investigates the need-based motive of aid allocation. The literature commonly uses GDP per capita which functions as a proxy for standard of living. A country with a higher GDP is richer and needs less aid. Here, the World Bank indicator is used and logarithmised to reduce skewness. Nevertheless, this measurement is one dimensional and neglects other aspects of development. In order to account for this issue, alternative measurements are tested in the robustness section.
Additionally, the variable Disaster measures the number of affected people by natural disasters. This variable captures large parts of humanitarian aid. When natural disasters occur, normally foreign aid increases. The database emdat reports consequences of biological, climatological, extra-terrestrial, geophysical, hydrological and meteorological disasters and is frequently used in the literature. In order to reduce skewness, the variable is logarithmised.
For the second hypothesis the strategic motive of the donor is investigated. As described earlier, the motive can be of political or economic nature. The former is tested by the variable Votes. This covers the percentage of the similarity in the voting pattern between the donor and recipient country in the UN general assembly for each year between 2000 and 2014. If the relationship with aid allocation is positive, this does not clearly state that aid is ’bought’ by the recipient country, but gives an indication for promotion of strategic partners with similar political beliefs (Dreher et al., 2010). The data was extracted from dataverse.harvard.edu which records the voting behaviour of all countries in the UN general assembly between 1946 and 2014. It is worth noting that Switzerland had an observer role in the UN general assembly until 2002 which precludes a voting mandate.
including additional measurements and testing for a possible omitted variable bias.5 The economic interest of the donor is tested by two variables. One possibility is that economic connectedness is shown in a larger trade flow between the countries. The Exports between donor and recipient country per year from 2000 until 2017 are taken from the IMF database. Due to the non-normality distribution, the natural logarithm of the exports in million $US is used. The way exports are defined here differs from previous definitions. For instance, Dreher et al. (2010) use the share of exports. They argue that a higher share shows a larger economic interdependency, hence, it may be an economic motive to give more aid to economically important countries in order to secure and improve this relationship. However, many countries import only small amounts, especially low income countries. Therefore, an increase in the overall exports to high income countries, can decrease the share to low and middle income countries even though their imports of Swedish or Swiss products stayed the same or increased on a smaller rate. This would not reflect the trade relation. Nunnenkamp et al. (2009) divide the exports by the GDP of the donor country which follows Berthélemy (2006) and Canavire et al. (2006). This measurement is dependent on donor GDP changes which is not necessarily
related with economic connectedness. Given these shortcomings, the natural logarithm measures the direct bilateral exports best. Moreover, interdependency can be measured by a country’s Resoures. If a recipient country has more natural resources, a higher aid allocation towards this country can increase the relationship between both countries and assure cheap resource imports. Here, the World Bank indicators Adjusted savings: mineral depletion and Adjusted savings: energy depletion give a large data set until 2016 and capture the natural resources of each recipient country. The two variables are summed up and logarithmised to reduce skeweness.
The third hypothesis tests a merit-based motive. A difficult environment in terms of basic institutions and economic policies can reduce effectiveness of foreign aid (Dreher et al., 2010). This is tested by several governance indicators of the World Bank. These are available for a large set of countries and fit the purpose of this analysis perfectly. By measuring different aspects of governance, a broader view of political difficult environment can be drawn. The key papers use only one or two dimensions,6 whereas, in this thesis
five are investigated. First, Government Effectiveness captures the quality of public and civil services. Second, Political Stability measures the likelihood of political instability and politically-motivated violence. Third, Voice captures participation in politics and freedom of speech, associations, and media. Moreover, levels of Corruption and Rule of Law measure slightly different governance factors. All indicators are in units of a standard normal distribution ranging from approximately -2.5 to 2.5, where a higher score shows a better status of the country.
5Other papers focus on colonial ties between donor and recipient countries. This can not be
investi-gated here, because both countries never had any colonies.
6Dreher et al. (2010) uses an indication of democracy, and Dreher et al. (2012) and Nunnenkamp
Lastly, the country’s logged Population is used as a control variable in order to account for differences is country sizes. Together with year dummies which control for time specific characteristics and country dummies which do the same for countries, the effect of each variable of interest is less distorted by a possible omitted variable bias.
Table 5: Summary statistics of independent variables
Motive Variable Mean SD Number Min Max
Need GDP per capita (ln) 8.42 1.54 2421 5.27 11.60
Need GDP per capita (ln) squared 73.26 26.12 2421 27.74 134.45
Need Disaster (ln) 6.09 5.56 2421 0.00 19.66
Population (ln) 16.23 1.62 2421 10.94 21.04
Strategic Resources (ln) 19.35 3.54 2421 4.85 26.40
Strategic Exports (Sweden) (ln) 4.06 2.66 2421 -4.47 9.86
Strategic Exports (Switzerland) (ln) 4.09 2.76 2421 -4.34 10.77
Strategic Votes (Sweden) 0.63 0.20 2114 0.00 1.00
Strategic Votes (Switzerland) 0.56 0.28 2114 0.00 1.00
Strategic Leader Orientation (Sweden) 0.23 0.42 2226 0.00 1.00
Strategic Regime (Sweden) 0.21 0.41 2377 0.00 1.00
Strategic Regime (Switzerland) 0.21 0.41 2377 0.00 1.00
Merit Corruption -0.10 1.00 2276 -1.81 2.47
Merit Government Effectiveness -0.04 0.96 2275 -2.27 2.35
Merit Political Stability -0.19 0.96 2274 -3.18 1.94
Merit Rule of Law -0.13 0.99 2275 -2.01 2.10
Merit Voice -0.15 0.99 2275 -2.26 1.80
Table 6: Expected signs of the Variables of Interest.
Hypothesis Variable NGO ODA
H1: Need-based motive GDP per capita − −
H1: Need-based motive Disaster + +
H2: Strategic motive Exports − +
H2: Strategic motive Resources − +
H2: Strategic motive Votes − +
H2: Strategic motive Regime − +
H2: Strategic motive Leader Orientation − +
H3: Merit motive Government Effectiveness − +
H3: Merit motive Political Stability − +
H3: Merit motive Voice − +
H3: Merit motive Rule of Law − +
4
Methodology
The summary statistics of the dependent variable (table 2) show that in more than 50% of the cases the aid allocation is zero. The reason for this is the large number of possible recipient countries and the long period of time. Some small countries do not receive aid from every channel or every year. Moreover, aid is clustered in specific regions or income categories as figures 1 and 2 show. Therefore, the data is censored with the limit of zero which appears to be the case for a large part of the observations. Hence, an OLS model would not suit this adequately, because it is not consistent (Dreher et al., 2012) and gives too much emphasis to the zero values.7 Therefore, two options are possible: the Heckman
selection model and Tobit model (Neumayer, 2002). The Heckman model assumes that the allocation process is divided into two parts. First, the donor decides if the country receives any money at all and in the second step, the extend of the amount. While it can be argued that this assumption holds, the method requires exclusive restrictions which are not obvious. It has to be assumed that some variables determine the first step, but not the second one. The Tobit model does not require such a restriction and is therefore the preferred option.
The Tobit method differentiates between two types of data. First, the dependent variable is zero, and second it is above zero. Therefore, a non-linear relationship between independent and dependent variable is calculated (Hill et al., 2008a). One important note is that the coefficients cannot be interpreted directly. In contrast to an OLS model, a Tobit model needs to be transformed first. In order to receive the marginal effects of a coefficient, the ’McDonald-Moffit’ decomposition is used (Hill et al., 2008b):
∂E(y|x) ∂x = P rob(y > 0) ∂E(y|x, y > 0) ∂x + E(y|x, y > 0) ∂P rob(y > 0) ∂x (1)
The first part measures the marginal effect of a change in x, when y is not zero, whereas the second part measures the change of the y population which becomes larger than zero due to a change in x (Hill et al., 2008a). The regression tables in section 5 are not transformed, but in order to interpret the coefficients, the sample mean of the other variables is used to calculate the marginal effect.
In contrast to the previous literature a panel data approach can be used due to the longer time period. This has several advantages compared to a cross-sectional data set. First, efficiency increases due to an expansion in observations. Not only the degrees of freedom rise, but, in addition, collinearity among explanatory variables decreases. Furthermore, this approach can test time trends. Especially a political and economic strategy can change when the government is replaced. Second, the omitted variable bias is reduced. Time invariant characteristics which correlate with the explanatory variables can be included and do not distort the estimators anymore. In this case, country characteristics such as language, culture, climate, or religion do not need to be added (Hill et al., 2008a).
In practice, there are two possible models: the fixed-effects and the random-effects model. On the one hand, the fixed-effect model examines the variation within one individual and includes individual-specific and time invariant characteristics in the intercept. On the other hand, a random effects model assumes that the differences between individuals are random. Here all individual differences are captured by the intercept and time invariant variables can be included in the model.
The main difference between fixed and random effect estimators is the critical assump-tion in the random effects model that the error term is uncorrelated with the explanatory variables. In case of a linear regression design, a Hausman test would be conducted. If the assumption holds, both models can be used, otherwise only the fixed effects model is consistent. Unfortunately, in a Tobit design, a fixed-effects model is generally biased8 (Dreher et al., 2012). Hence, a random effects model is used. This makes it necessary to include dummy variables for each recipient country and each year in order to control for country and year specific events. Those may affect the allocation behaviour of the donor countries, such as the financial crisis of 2008.
In this analysis a panel data set with two different estimation approaches is used. The first one estimates the aid allocation of NGOs and ODA separately as in Dreher et al. (2010) and Nunnenkamp and Öhler (2011) and tests hypotheses 1-3. The aid allocation from either NGO or government aid to recipient country i in year j is the dependent variable. This approach makes it possible to investigate differences between the two channels. The explanatory variables estimate the influence of donor motives on aid allocation and the variables population as well as the country and year dummies function as control variables. Therefore, the first Tobit random effects regression equation is:
Aidij =α1+ 3
X
n=2
αnN eedM otiveij + α4P opulationij+ 9 X n=5 αnStrategicM otivesij+ 14 X n=10
αnM eritM otivesij + Countryi+ Y earj + ij
(2)
The second approach is based on Dreher et al. (2012), Nancy and Yontcheva (2006) and Koch et al. (2009). Here, only aid allocation of NGOs to recipient country i in year j is the dependent variable. The aid allocation of the government as well as the multilateral organisations United Nations, World Bank, International Monetary Fund and European Union are included in the regression as explanatory variables. Therefore, the correlation of the allocation between government and multilateral aid and NGO aid can be observed (Hypothesis 4). Either a mimic behaviour or differences between NGO activities and the channels can be the result as explained earlier. Moreover, while controlling for aid allocation of various channels, the effects of the variables of interest show what drives the NGO aid allocation in specific.
Hence, the second Tobit random effects regression looks as follows: N GOAidij =α1+ α2ODAij + 6 X n=3 αnM ultilateralAidij + 8 X n=7 αnN eedM otiveij+ α9P opulationij + 14 X n=10 αnStrategicM otivesij + 19 X n=15 αnM eritM otivesij+ Countryi+ Y earj + ij (3) The second approach differs from the previous literature in three aspects. First, a panel data set is used differently. Koch et al. (2009) uses a cross-section data set, whereas Dreher et al. (2012) and Nancy and Yontcheva (2006) use aid from each NGO to each recipient country as their dependent variable. This is not possible in this data set for Switzerland, because only aggregated NGO data is available. Second, a longer and more current time period allows to investigate differences across time periods. Third, not only official bilateral, but in addition, multilateral aid is included in the regression.
5
Results
This section is divided into two parts. Subsection 5.1 investigates hypotheses 1-3 based on equation 2, whereas subsection 5.2 uses equation 3 and tests hypothesis 4.
5.1
Need-based, Strategic and Merit-based donor motives
Table 7 shows the regression output for the first estimation approach. Here, the allocation pattern of ODA and NGO aid can be interpreted separately. Interestingly, the coefficient Population differs from the previous findings of Dreher et al. (2010), Nunnenkamp et al. (2009) and Dreher et al. (2012) for Sweden and Switzerland, where a positive significant
relationship between aid and country size is estimated. In this sample, the coefficient is only negative significant for Swiss ODA, and hence, the coefficient has the opposite sign than the key papers found. However, when no country dummies are used as in table A2 the coefficient is positively significant for all channels. This allocation behaviour differs from other countries. For example, Alesina and Dollar (2000) find a bias towards less populated countries for various donor countries. A reason for this could be that smaller additional amounts of aid can have a larger impact on smaller countries than on larger ones and therefore, the impact is larger compared to populous countries (Neumayer, 2003a). The results now show that there are other country specifics which shape the aid allocation pattern and the population size is not a driving factor.
the recipient country and its squared version. This approach is in line with Alesina and Dollar (2000) and differs from the key papers which only allow for a linear relationship. Table 7 shows a concave relationship for the Swedish NGO and the Swiss channels. Hence, higher GDP per capita increases aid, but on a diminishing level. For Swedish ODA only a linear relationship between GDP per capita and foreign aid is given. Hence, the need-based motive is stronger for Swedish ODA than for the other channels. In marginal terms, when GDP increases by 1% official Swedish aid decreases on average by 0.59%, ceteris paribus.
Even, by including the variable Disaster which captures parts of aid based on humani-tarian needs, the results hold for all channels. Table A3 shows that the number of people affected by natural disasters is not significantly correlated with foreign aid. Only Swiss ODA is negatively correlated. This contradicts previous results. Dreher et al. (2010) and Nunnenkamp et al. (2009) find a positive and significant relationship with Swedish and Swiss NGO aid, whereas the correlation for ODA is insignificant.
For the other three channels, a bias towards middle income countries is observed which can also be seen in tables A4 and A5 where aid allocation to low- and middle income countries is reported respectively. Only for middle income countries this concave relationship is given which shows that the need-based motive does not hold for the poorest countries.9 Moreover, already the summary statistics in figure 2 indicated that middle income countries are the main recipients which confirms the results of previous studies, e.g. Alesina and Dollar (2000). In contrast, GDP per capita is not correlated with Swedish ODA in each income group, but in the combined data set. This shows that aid is generally targeted towards poorer countries, but not significantly differentiated across low and middle income countries.
The key papers for Switzerland and Sweden only use a linear term without allowing for a quadratic relationship. In Sweden, Dreher et al. (2010) find that ODA has a stronger influence of the need-based motive than NGO aid which is found to be not driven significantly by GDP per capita. The current analysis confirms these results. In Switzerland, Nunnenkamp et al. (2009) find that ODA is stronger determined by GDP per capita than NGO aid. This analysis finds the opposite. The relationship between aid and GDP is shaped in a concave way in both channels, but the maximum is lower for Swiss
NGOs than for ODA.10
To sum up, the analysis finds weak support of the need-based motive (hypothesis 1). While allowing exclusively for a linear relationship, the hypothesis holds only for Swedish ODA. The reason for this is that Swedish NGO and Swiss agencies follow a concave relationship. The latter result holds only for middle income countries. All in all, NGOs and Swiss ODA can target the poor better.
9A linear relationship is also not given.
Table 7: Hypotheses 1-3: Total sample of Swedish and Swiss NGO aid and ODA: Panel Tobit Results
Variable SWE NGO SWE ODA CHE NGO CHE ODA
GDP per capita (ln) 2.63*** -0.59** 2.30*** 4.64***
GDP per capita (ln) squared -0.19*** -0.18*** -0.32***
Population (ln) 0.60+ -0.22 0.14 -2.22*** Resources (ln) -0.00 0.02 0.01 -0.00 Exports (ln) 0.02 0.10** -0.02 -0.06+ Votes (ln) -0.36* 0.40 0.08 0.04 Leader Orientation 0.04 0.06 Regime -0.20+ -0.27 -0.26** 0.11 Government Effectiveness 0.07 -0.54** -0.05 -0.41** Political Stability 0.01 -0.48*** -0.08* -0.20** Voice 0.20* 0.68*** 0.15* 0.60*** Rule of Law -0.08 0.24 0.16+ -0.14 Corruption 0.00 0.39* 0.06 0.15 Constant -18.44*** 10.06 -8.39* 23.50** Number of Observations 1934 1934 1809 1809 Years 2000-’14 2000-’14 2002-’14 2002-’14
Country Dummies yes yes yes yes
Year Dummies yes yes yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
Tobit regression to investigate hypotheses 1-3. The dependent variable is defined as foreign aid to each country i in year j, measured in ln(1 + aid in million $US). Need-based motive is measured by GDP per capita, strategic motive by Resources, Exports, Votes, Leader Orientation, Regime and merit-based by Government Effectiveness, Political Stability, Voice, Rule of Law and Corruption.
Hypothesis two refers to a possible strategic motive in economic and political terms. First, the economic motives are investigated. The basic model (table 7) shows that only Swedish government aid is determined by exports. Here, strategic partners which buy more from Swedish producers receive more aid. In marginal terms, when exports increase by 1% official foreign aid increases on average by 0.10%, ceteris paribus. The economic motive is especially shaped in Latin America and the Caribbean (see table A6). Interestingly, this is only the case for Swedish bilateral aid and not for any other channel.
(2010) who find a negative relationship with resource endowment, this analysis calculates no significant coefficients for any channel at any conventional significance level.
Contrary to the aforementioned papers, political interest is measured by several coefficients. Not only the voting pattern in the UN General Assembly (Votes), but Leader Orientation (for Sweden) and Regime may indicate possible strategic partners as well. In the total sample, no significant relationship of ODA is found. In contrast, Swedish NGO aid is even negatively correlated with UN Votes and a country with a different Regime receives more NGO aid from both countries. However, investigating different regions shows that Middle Eastern and Northern African countries receive more Swedish official aid, when they have a similar voting pattern with Sweden (see table A7), whereas this is the case for Swiss ODA with Latin American countries (table A6). Hence, for some countries a similar voting pattern is rewarded by higher official foreign aid.
This data set does not verify the results by Dreher et al. (2010) and Nunnenkamp et al. (2009) who find support for a general correlation of UN voting pattern and official bilateral
aid from Sweden and Switzerland. Only in regional sub-samples a positive relationship is given for Swedish and Swiss ODA. A reason why the results differ from previous analyses could be the correlation of other variables which measure similar political beliefs. Not only the variables Regime and Leader Orientation, but variables for the third hypothesis (merit-based motive) may be correlated with the voting pattern as well. Therefore, the government in the recipient country does not vote based on possible additional aid from Switzerland, but maybe due to similar beliefs. When testing this without Leader Orientation and Regime (table A8) and by allowing only GDP, Population, Exports and Votes (table A9), a significant positive correlation for Swedish ODA is estimated at a 5-% significance level for the second approach. Hence, the findings of Dreher et al. (2010) and Nunnenkamp et al. (2009) that official aid is generally shaped by UN Votes cannot be verified and may hint at a possible omitted variable bias. This means that the governance indicators do not only have an impact on foreign aid, but on the voting pattern as well.
In summary, the results are ambiguous. In the case of Swedish ODA evidence for a strategic motive is found. Especially in the region Latin America and the Caribbean aid is allocated relatively more to countries which buy more Swedish products. Moreover, Swedish ODA favours countries with a similar voting pattern in the UN General Assembly in Middle East and North Africa and Swiss ODA in Latin America and the Caribbean. Nevertheless, the NGO channel of both countries do not follow such a motive and can be declared as altruistic in the sense of Berthélemy (2006).
government aid focuses on politically stable countries. However, it could be argued that NGO aid is more effective in those countries and needs less resources than bilateral projects and therefore, the large amounts of government aid may be a waste of resources due to inefficient spending. The data set does not give the possibility to confirm or reject this, because no cost-structure of the agencies is given.
Second, Voice is positively correlated with aid allocation of all channels, where the effect on official aid is larger than on NGO aid. This could indicate a reward scheme for ODA. Countries which improve democratic structures and strengthen rights of the media and protesters receive more ODA. Further tests give more insights on this matter. First, the changes of all explanatory variables are used to test if an improvement in a governance indicator is rewarded by additional foreign aid. Table A10 shows that a change in Voice is positively rewarded by more aid in the case of Swiss ODA. Second, an argument which may be very important for NGOs is that a certain level of participation is a necessary condition to work efficiently. This is tested by two sub-samples where recipient countries are split at the mean of the Voice indicator. If the coefficient is positive and significant in both samples, then a better status in this category is always connected with more foreign aid regardless of the overall level. The results show no significant estimates for Swedish aid and Swiss ODA (see table A11) for the sample with countries which perform above the mean, but do for countries which perform below the mean (table A12). Hence, this indicator is important in the early stage, but after a certain threshold it loses importance. Only Swiss NGO aid is shaped the other way around. This may lead to the conclusion that Swiss NGO aid is channeled more intensively to countries where a higher political participation is possible.
Interestingly, Corruption is also positively correlated with bilateral aid of Sweden, but not with NGO aid. This means that lower levels of corruption are correlated with more bilateral foreign aid. Again, this can be a reward for political actions. The results show (table A10) that Sweden does not reward an improvement in corruption, as found by
Alesina and Weder (2002) and Neumayer (2003c).
5.2
Mimic donor motive
For the fourth hypothesis equation 3 is used which includes the aid allocation of official bilateral and multilateral channels to test for a mimic motive. As a recap, this estimation wants to investigate the correlation between the aid channels. NGOs can follow other agencies and allocate in the same countries, or support projects in different ones. The results of the second estimation approach can be found in table 8.
The principal-agent model of Fruttero and Gauri (2005) finds support in both countries. ODA and NGOs allocate aid in similar countries. In marginal terms: an increase of ODA by 1%, increases Swedish NGO aid on average, ceteris paribus, by 0.058% and Swiss NGO aid by 0.080% for each recipient country. This does not seem to be much, but when someone accounts for the budget differences it gains importance. It was already seen in table 5 that the sum of ODA exceeds the budget for NGOs by a factor of 14 in Sweden and 3 in Switzerland. Interestingly, the correlation is larger in Switzerland than in Sweden. As mentioned before, the NGOs in this data set are up to 90% state funded in Sweden and 100% private funded in Switzerland. Nevertheless, the correlation with ODA is larger in Switzerland. This could support the argumentation by Riddell et al. (1995) and Pratt et al. (2006) that SIDA has restricted influence on the allocation process and the NGOs allocate
independently. On the other hand, it may proof that private donors act as principals in the sense of Fruttero and Gauri (2005) as well. This would imply that private donors condition future funding on previous success as well. Therefore, NGOs act risk averse and mimic the donation pattern of other channels rather than deviating from them.
Furthermore, in hypothesis 2 differences in strategic motives were shown in Sweden and Switzerland for two regions: Latin America and the Caribbean and Middle East and North Africa, where economic and political interests shaped the ODA allocation. When the second estimation approach is also differentiated across regions, tables 9 and 10 show that ODA is still significantly correlated with NGO aid. This means, even when strategic motives play a role for the back donor’s aid allocation, NGOs follow significantly.
After all, some differences are still visible between NGO, official bilateral and multilat-eral aid, otherwise, the other aid channels would perfectly predict NGO behaviour. In both countries, the need-based motive finds again weak support with the concave relationship. Moreover, the coefficient Voice shows how important this indicator is for NGOs to operate efficiently. Lastly, a different Regime has a priority for both countries.
In addition to that, table 8 includes multilateral organisations as well. Interestingly, NGO aid from both countries is positively correlated with UN aid. Surprisingly, the magnitude does not differ significantly from ODA.11 Furthermore, World Bank aid is
negatively correlated with Swedish NGOs and the IMF aid with Swiss NGOs, whereas EU aid is positively correlated with Swiss NGO aid. Hence, arguing that only the official back donor, which is the government, has an influence on NGO aid, misses the importance of
signifi-multilateral organisations. The underlying driving factors are investigated from different perspectives. First, the guidelines from each organisation and their allocation share across regions (figure 1) give an indication for donor motivation and regional focus. Additionally, the regressions between NGOs and agencies are conducted for each region (tables 9 and 10) and the need and merit-based motives of each organisation are reported in table 11.
The aid allocation of the UN consists of projects from different UN agencies, or the UN family, which have individual targets. While every agency has its own focus and guideline, some general tendencies can be seen in the UN Charter, which is the legal framework, as well as the expenditure by sector, which is the actual spending behaviour. The UN Charter states in Article 1 that a major priority is to ’achieve international co-operation in solving international problems of an economic, social, cultural, or humanitarian character and in promoting and encouraging respect for human rights and for fundamental freedoms for all without distinction as to race, sex, language, or religion’ (UN, 1945). Hence, the topics of humanitarian aid, sustainable development and human rights play a big role in the guidelines. Accordingly, the budget allocation is largest for humanitarian and sustainable development aid12 (UN System, 2018). Therefore, an overlap in aims between NGOs and
the UN is visible. An additional sign for the link between the UN family and NGOs is the possibility of getting consultancy which is also part of the UN Charter in Article 7113 (UN, 1945). This makes an active influence on NGO behaviour possible. In 2016 more than 5000 NGOs used UN consultancy (Department of Economic and Social Affairs, 2018).
The World Bank group focuses more on poverty reduction and improving the standard of living which is a clear sign for the need-based motive. For this, loans, credits and grants are provided with a focus on education, health, infrastructure and communication (UN, 2018). Hence, an overlap in motives can be found in the guidelines. Nevertheless, the aid allocation of NGOs does not seem to mimic World Bank behaviour. One reason is the different regional focus. Whereas the World Bank allocates 31% of all funds in South Asia, Swedish NGOs do only half of it. On the other hand, Latin American countries receive 16% of Swedish NGO and less than 2% of World Bank money (see figure 1). While this cannot explain the negative correlation, it gives an indication of different regional foci. Moreover, the correlation and hence the influence of the World Bank on NGOs differs across regions. In the region Middle East and North Africa, even a positive correlation is given for Swedish NGOs. Lastly, table 11 shows a concave relationship between GDP and aid and a significant role for the Voice indicator. While this would speak for similar donor motives, the large population focus of the World Bank does not.
Fostering economic growth and employment is the main focus of the International Monetary Fund (UN, 2018). Therefore, grants should help to resolve balance of payment
12The costs of peacekeeping activities are large as well, but due to reporting standards of the OECD
which does not measure military and security expenses this can be neglected (OECD, 2018b)
13’The Economic and Social Council may make suitable arrangements for consultation with
Table 8: Hypothesis 4: Total sample of Swedish and Swiss NGO aid: Correlation of aid channels: Panel Tobit Results
Variable SWE NGO CHE NGO
ODA (ln) 0.06*** 0.08***
World Bank (ln) -0.01* 0.01
United Nations (ln) 0.03+ 0.05***
International Monetary Fund (ln) -0.00 -0.02**
European Union (ln) -0.01 0.02***
GDP per capita (ln) 1.62*** 1.41***
GDP per capita (ln) squared -0.10*** -0.10***
Population (ln) 0.16 0.13 Resources (ln) -0.01 0.01+ Exports (ln) 0.02+ -0.03* Votes (ln) -0.21* 0.14 Leader Orientation 0.02 Regime -0.15* -0.22** Government Effectiveness -0.04 -0.04 Political Stability 0.04+ -0.03 Voice 0.14** 0.11* Rule of Law -0.04 0.05 Corruption -0.03 0.04 Constant -8.11*** -6.29** Number of Observations 1920 1809 Years 2000-’14 2002-’14
Country Dummies yes yes
Year Dummies yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
Tobit regression to investigate hypotheses 4. The dependent variable is defined as foreign aid to each country i in year j, measured in ln(1 + aid in million $US). The correlation between aid channels shows that the allocation pattern of the official back donor within the country ODA and in multilateral organisations UN, EU is similar to the NGO behaviour.
deficits which are conditioned on project specific actions (IMF, 2002). As a consequence, donor motives differ from NGO or ODA ones. Moreover, only lower middle and low income countries receive grants (see figure 2) which results in a large need-based motive (table 11) and does not match NGO behaviour.
The European Union has its own department called ’Humanitarian Aid and Civil Protection Department’ (ECHO). It focuses on humanitarian aspects like health, nutrition and sanitation14 (ECHO, 2018). Figure 1 has already shown that NGOs and the EU
have a different focus of regions. Whereas the EU allocated more than a quarter of the budget in the Europe and Central Asia region, Swedish (6%) and Swiss (7%) NGOs have other geographical preferences. Moreover, the EU prefers small countries (table 11). Even though, the correlation between NGO and EU aid across regions is either positive or significant (tables 9 and 10).
Table 9: Hypothesis 4: Swedish NGO aid across Regions: Panel Tobit Results
Variable NGO NGO NGO NGO NGO NGO
(EAS) (ECS) (LCN) (MEA) (SAS) (SAF)
ODA (ln) 0.10*** 0.02 0.16*** 0.14*** -0.05 0.05+ United Nations (ln) 0.08* 0.00 0.02 -0.01 -0.28* 0.00 World Bank (ln) -0.05** -0.01+ 0.01 0.04* -0.08* -0.03* International Mon-etary Fund (ln) 0.01 -0.00 0.04 -0.11*** -0.05 0.01 European Union (ln) -0.03 + 0.02+ 0.00 0.04* 0.06 -0.01 GDP per capita (ln) 1.43 + 1.35*** 10.81*** 1.84** 2.63 0.54 GDP per capita (ln) squared -0.07 -0.09*** -0.61*** -0.09** -0.28 -0.04 Population (ln) 1.08 -0.23 0.95 -0.06 -4.33+ -0.68 Resources (ln) -0.01 0.01 -0.01 -0.00 -0.10 -0.00 Exports (ln) -0.03 -0.02 0.04 0.06* 0.11 0.02 Votes (ln) -0.18 0.09 0.09 -0.40* -0.43 -0.09 Leader Orientation 0.06 0.02 0.05 0.10* 0.60** -0.06 Regime -0.06 0.06 . . -0.56** . Government Effec-tiveness 0.12 -0.08 + -0.16 -0.01 0.33 0.14 Political Stability -0.11 -0.02 0.05 -0.02 -0.06 0.06 Voice 0.11 0.01 0.18 -0.18*** 0.34 0.23+ Rule of Law -0.16 0.16* -0.00 0.01 -0.42 -0.02 Corruption 0.12 0.10* 0.04 -0.04 -0.99* -0.26* Constant -24.77* -2.01 -64.53*** -8.04* 68.83+ 9.71 Number of Obser-vations 234 543 313 221 90 491 Years 2000-’14 2000-’14 2000-’14 2000-’14 2000-’14 2000-’14
Country Dummies yes yes yes yes yes yes
Year Dummies yes yes yes yes yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
Table 10: Hypothesis 4: Swiss NGO aid across Regions: Panel Tobit Results
Variable NGO NGO NGO NGO NGO NGO
(EAS) (ECS) (LCN) (MEA) (SAS) (SAF)
ODA (ln) -0.01 0.03 0.04+ 0.14*** 0.06 0.07** United Nations (ln) -0.05 -0.03 -0.01 0.07 + 0.24* 0.02 World Bank (ln) 0.00 0.03* 0.01 -0.13*** -0.02 0.01 International Mon-etary Fund (ln) 0.01 -0.00 0.03* -0.05 -0.07** 0.00 European Union (ln) 0.06** 0.02* 0.00 0.13*** -0.01 0.01 GDP per capita (ln) 3.52** 3.72*** 3.86** -2.86 + 1.89 2.16* GDP per capita (ln) squared -0.19** -0.21*** -0.21** 0.13 -0.24 -0.21*** Population (ln) 0.64 -0.75** 0.26 0.01 -3.85 -1.11* Resources (ln) -0.00 0.00 0.00 0.01 -0.09 0.01 Exports (ln) -0.06+ -0.01 0.02 0.27*** -0.01 -0.02 Votes (ln) -0.53 -0.03 0.10 -0.35 -1.13 0.15 Regime . . -0.11 -0.87*** . -0.17+ Government Effec-tiveness -0.05 -0.07 0.09 0.04 0.64 + -0.01 Political Stability 0.32*** -0.03 0.13** -0.12 -0.57*** -0.05 Voice -0.10 0.18* 0.13 0.26+ 0.37 -0.07 Rule of Law -0.32+ 0.19* 0.21** 0.32+ -0.02 -0.13 Corruption 0.17 -0.09 0.07 -0.15 -0.08 0.22* Constant -25.86+ -4.23 -21.41** 14.42+ 65.12* 14.74+ Number of Obser-vations 218 526 292 205 85 457 Years 2002-’14 2002-’14 2002-’14 2002-’14 2002-’14 2002-’14
Country Dummies yes yes yes yes yes yes
Year Dummies yes yes yes yes yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
Table 11: Hypothesis 4: Multilateral Aid: Panel Tobit Results
Variable United
Nations World bank
International Monetary Fund European Union GDP per capita (ln) 2.30*** 20.37*** -4.87*** 3.76* GDP per capita (ln) squared -0.16*** -1.43*** -0.23* Population (ln) 0.74** 3.38** 3.18 -1.71** Resources (ln) -0.01 -0.02 -0.08+ 0.03 Government Effectiveness -0.25** -0.02 -1.16 -0.07 Political Stabil-ity 0.01 0.04 0.28 -0.55*** Voice 0.06 2.00*** 1.92** 0.59** Rule of Law 0.23* -0.63 -0.11 0.58* Corruption 0.15+ 0.25 0.50 0.17 Constant -17.14*** -121.68*** -20.76 20.25+ Number of Ob-servations 2274 2274 2274 2274 Years 2000-’15 2000-’15 2000-’15 2002-’15 Country
Dum-mies yes yes yes yes
Year Dummies yes yes yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
Tobit regression to investigate hypothesis 4. The dependent variable is defined as foreign aid to each country i in year j, measured in ln(1 + aid in million $US). Aid allocation of the multilateral channels (ODA, UN, World Bank, IMF, EU ) is regressed on indicators for a need-based and merit motive. It shows the different focus of each institution.
Moreover, the similarity between ODA and NGO aid increases over time in both countries. In table 12 the panel is divided into state legislative periods.15 Every government
can change the allocation behaviour. For Sweden the impact of ODA increased vast from about 0.052% to almost 0.216% and in Switzerland from about 0.022% to 0.124%. Hence, two possible explanations can be derived. First, the motives of both donor channels converted over time. Therefore, the same focus of recipient countries is the explanation of this trend. The second option is that the impact of the principal agent model increased. Hence, the rising influence of the official back donor drives the mimic behaviour.
15General Elections took place in Sweden in 2002, 2006, 2010 and 2014 and in Switzerland in 2003,
Table 12: Hypothesis 4: Swedish and Swiss NGO aid across legislation periods: Panel Tobit Results
Variable SWE SWE SWE CHE CHE CHE
(03-06) (07-10) (11-14) (04-07) (08-11) (12-14) ODA (ln) 0.05* 0.04 0.22*** 0.02 0.04* 0.12*** United Nations (ln) -0.06 + -0.04 -0.07 0.03 0.03 -0.06+ World Bank (ln) -0.03 -0.01 -0.02 -0.02 0.01 0.00 International Mon-etary Fund (ln) 0.04* -0.00 0.04 -0.04* 0.00 0.02 European Union (ln) -0.02 -0.03 0.04 0.02 0.02 + 0.01 GDP per capita (ln) 1.94 6.01 + -4.82 3.90* 3.25* -1.07 GDP per capita (ln) squared -0.15 -0.35 0.33 -0.29* -0.23** -0.03 Population (ln) 1.22 0.41 0.99 -1.72 -0.70 2.76* Resources (ln) 0.02+ -0.03* -0.02 0.01 0.00 0.00 Exports (ln) 0.03 0.03 0.07 -0.02 0.02 0.03 Votes (ln) 0.12 -0.35 -0.10 0.06 0.17 0.38+ Leader Orientation 0.19+ 0.35 -0.16 Regime . 0.11 0.46 -0.64* 0.12 . Government Effec-tiveness 0.38* -0.15 0.12 0.21 0.18 -0.12 Political Stability 0.04 0.27* -0.18 -0.03 -0.20** 0.00 Voice 0.17 0.20 -0.29 0.16 0.08 -0.04 Rule of Law -0.20 0.25 1.00* 0.18 0.25 0.28 Corruption -0.41* -0.25 -0.07 0.09 -0.19 -0.33+ Constant -26.51 -27.52 1.33 18.30 1.91 -38.74+ Number of Obser-vations 549 554 550 556 556 415 Years 2003-’06 2007-’10 2011-’14 2004-’07 2008-’11 2012-’14
Country Dummies yes yes yes yes yes yes
Year Dummies yes yes yes yes yes yes
+p < 0.10,*p < 0.05, **p < 0.01,***p < 0.001
The first explanation can be evaluated by testing for a time trend in the three afore-mentioned donor motives. By using the same time periods, tables A13 and A14 do not support a convergence across time. The second possibility argues that the independence of NGOs with their back donors decreases. NGOs mimic the behaviour of other channels, because they have to show visible results in order to secure future funding. An example for the increase in dependency would be a larger competition in the NGO market which would make visible results more important to fund future projects. Nunnenkamp and Öhler (2012) analyse increases in competition for US NGOs, where fundraising expenditure did not increase, but NGO efficiency did. Nevertheless, whether the pressure to generate visible results increases is not answered. Another explanation could be that possible synergy effects (Koch, 2009) are the driving factor, meaning that NGOs and other channels operate in herding behaviour to increase the overall impact of each project. Unfortunately, this analysis cannot give evidence for either argument.
In summary, this analysis accepts hypothesis four: official bilateral aid shapes NGO aid. This relationship even holds for privately funded NGOs and increases over time. Moreover, especially the aid allocation of the United Nations correlates positively with the NGO pattern in both countries and European Aid with Swiss NGOs. This shows that two important principles are missing by neglecting private donors and multilateral organisations in the principle-agent model of Fruttero and Gauri (2005).
5.3
Robustness tests
In the previous sections, different sub samples and methods were already applied in order to find the driving factors of each effect. In this part, four additional approaches are used. First, alternative measurements replace the variable used in the section before. Second, lagged variables control for reversed causality. Third, high income countries are excluded. Lastly, a different estimation approach investigates whether the method changes the relationship.
