Beware of the Dragon’s Gifts:
Assessing the Determinants of China’s
Aid Allocation to Africa
MASTER THESIS
Name: Julian Allan Watkinson
Student number: S3214753
Address: Prinsesseweg 5A, 9717BA Groningen
Mail: j.a.watkinson@student.rug.nl
Study program: MSc International Economics and Business
Faculty: Faculty of Economics and Business
University: University of Groningen
Supervisor: Dr. Tarek Harchaoui
Co-Assessor: Dr. Robbert Maseland
II
Abstract
The Western countries have long been the unchallenged “sponsor” of the third world by providing countries in need with development aid. However, the recent rise of China as a large donor country has triggered an alteration of the existing status quo, which raises the question: What are the fundamentals behind China’s growing aid allocation and do they contrast with those of the Western donors? This study applies a Tobit model in order to analyse the determinants of China’s aid allocation to the African continent during 2000 to 2013. The study is novel both in terms of its contribution to current debates and technique, as it considers the special nature of the censored data, which enables us to analyse the different types of China’s development aid separately. The results show that China emphasizes less on institutional qualities, which stands in contrast to traditional donors, but more on its own economic interests. Furthermore, China is highly selective in terms of its aid allocation, as it uses two types of flows for different types of countries. Close trade partners and autocratic regimes receive relatively more development-oriented flows (ODA), while natural resource rich and democratic nations obtain commercial-oriented forms of aid (OOF). Recipients’ needs play a less dominant role, as ODA flows tend to flow to comparatively wealthier countries.
III
Table of contents
List of figures ... IV List of tables ... V List of abbreviations ... VI 1 Introduction ... 12 Foreign Aid: Forms, origins and trends ... 2
3 Literature Review ... 6
3.1 The Various Contributions ... 6
3.2 Take-Away Points and Our Working Hypotheses ... 9
4 Quantitative Analysis ... 12
4.1 The Model ... 12
4.2 The Source Data ... 13
4.3 Econometric Results ... 19
4.4 Robustness ... 27
5 Conclusion and Implications ... 29
5.1 Conclusion ... 29
5.2 Policy implications ... 30
Literature ... 32
IV
List of figures
V
List of tables
Table 1: The allocation of bilateral DAC and multilateral ODA in 2014 ... 4
Table 2: The Chinese financial development flows to Africa from 2000-2013 ... 14
Table 3: Comparison of the ODA allocation to Africa between 2000-2013 by donor country ... 20
Table 4: Chinese ODA flows to Africa between 2000-2013 ... 22
Table 5: Chinese OOF flows to Africa between 2000-2013 ... 24
Table 6: Split-up of Chinese aid flows to Africa between 2000-2013 ... 26
Table A1: The allocation of Chinese aid by flow type and sector between 2000-2013 ... 38
Table A2: Data description and sources ... 38
Table A3: Summary statistics ... 40
Table A4: Split-up of Chinese aid flows to Africa between 2000-2013 robustness check ... 41
Table A5: Split-up of Chinese aid flows to Africa between 2000-2013 with count data ... 41
VI
List of abbreviations
DAC FDI GDP HIV NGO ODA OECD OLS OOF UN USD RMBDevelopment Assistance Committee Foreign Direct Investment
Gross Domestic Product
Human Immunodeficiency Virus Non-Governmental Organisation Official Development Assistance
Organisation for Economic Co-operation and Development Ordinary Least Squares
Other Official Flows United Nations United States Dollar
1
1 Introduction
The traditional Western donor countries, such as the U.S. and Germany, have long been shaping the global foreign aid landscape. While their relative dominant position as the “sponsors” of the Third World has remained unchallenged, the recent rise of China has triggered an alteration of the existing status quo, both in terms of order of magnitude of aid and its underlying fundamentals.
The magnitude of aid disbursements over the period from 2000-2013 to the African continent illustrates this. Despite being a developing country itself, China transferred development aid equal to 9.5% of the Development Assistance Committee (DAC) donor countries combined aid budget during the pre-Great Recession period. The share of Germany, as the second largest DAC member, was in comparison only 8.5% during the same time. While the DAC donors decreased their development aid to the African continent in the aftermath of the Global Financial Crisis in 2007, China took the opportunity and scaled its aid flows up. By doing so China’s aid reached an average aid level equal to 16.3% of the level of all DAC donor countries in the post-Great Recession period. The fact that China’s standard of living merely stands at 18% of that in Germany during the period under observation but nevertheless significantly outweighs the aid flows of Germany and other traditional Western donor countries is first, a remarkable occurrence and second, raises the following quintessential question: What are the fundamentals behind China’s growing aid allocation and do they contrast with those of the most important DAC donors such as the U.S., Germany and Japan? Traditional donors have been known to use their aid distribution to enforce conditions on the aid recipient countries, such as the implementation of policies towards democracy or free market principles. If the aid allocation determinants of those new donors significantly differ from the traditional donors, then the influence and power of the traditional donors to enforce those conditions is highly weakened.
2 We explore our research questions on a panel dataset of 50 African economies over the period from 2000-2013. Africa constitutes thereby a clean case to test the fundamentals behind China’s foreign policy as well as the way it contrasts with that of competing longstanding Western donors. Africa alone, received nearly 1/3 of the $81.6 billion of total bilateral DAC ODA disbursements in 2014.1 Similar, of the $138 billion in constant prices of China’s total development finance to Africa over the 2000-2013 period, 5/8 geared to the energy, natural resources and infrastructure sector. This is remarkable, as China’s intensification of its activities on the continent coincides with the recent resurgence of the African economy that began in the new Millennium following decades of economic stagnation.
To take the censored data of our dependent variable into account, we apply the random-effect Tobit estimator to analyse our research questions in a quantitative way. By doing so we introduce controls for the unobserved heterogeneity of the recipient countries as well as for time specific effects that may influence the aid allocation process. The results of our study are thereby robust to a variety of checks, including a Poisson count data as well as a two-stage model. Overall, the impact of the recipients’ needs show a mixed picture. Although countries with poor public health conditions receive overall larger aid shares, short term needs measured on the number of people affected by disasters decrease the aid inflows from China. In addition, the income level of African countries also has a negative impact on the Chinese aid allocation. Countries with a lower income level are discriminated and receive less ODA flows from China. In terms of merit, ODA flows favour autocratic and corrupt countries, while democratic countries receive in contrast more OOF flows. Furthermore, the selfish motivation behind China’s aid allocation supports the fact that China uses its two aid types for different country types. Countries with high levels of bilateral trade with China but with fewer natural resources receive more ODA flows, while natural resource-rich countries receive higher shares of the commercial orientated OOF flows, independent from their bilateral trade volume.
2 Foreign Aid: Forms, origins and trends
Global development finance can flow in a variety of ways and sources and can be defined in a broader and narrower sense. The origin is thereby either of official or private nature, with the latter surpassing the former in terms of volume. With $637.6 billion, FDI is the largest form of development finance followed by remittances with $396.2 billion. ODA, the closest form of the commonly known foreign aid follows far behind with $128.3 billion (World Bank 2016a). Figure 1 suggests that there is more to the structure of global development finance than FDI, remittances and ODA. While private development finance flows are of crucial significance for the development path of developing nations, this study will focus solely on the official forms of
3 development finance. Any reference to ‘aid’ will therefore refer to official flows of aid and in particular to ODA-like flows.2
Figure 1: The classification of foreign aid (based on Brautigam 2011; Strange et al. 2013)
The 30 DAC donor countries allocate their ODA flows to nearly every developing country on the world with a few exceptions such as North Korea. Table 1 provides the geographical breakdown of this item for 2014. With $19.0 billion, defined as net OOF flows and $81.6 billion ODA flows in 2014, the DAC donors clearly focus on bilateral disbursements of pure development aid. Multilateral donors, such as the United Nations and the World Bank, distribute with 47.5% the lion’s share of their multilateral ODA flows to the African continent. The DAC members likewise allocate their highest continental share to Africa. However, with only 31.1% of their total ODA disbursements, Africa possesses a smaller relevance in the bilateral context (OECD 2017).
The allocation of aid is often conditional. The DAC donors expect the recipient countries to improve their nations in a variety of areas according to their own ideals. The commitment for democracy, good governance, market based economies, openness and compliance to law are thereby essential (OECD 2008b). In terms of ODA disbursement, the recipient’s willingness to increase governance, peace and inclusion is at the centre of the DAC aid allocation. A disbursement by sector analysis of the DAC showed that the governance and peace sector, with respectively 15.7%, received the highest sector allocable share (OECD 2014). However, in recent times
2
4 corruption is just as relevant for the DAC donors. According to the OECD, corruption undercuts trust, democracy, decentralisation and the quality of public services while emphasizing egocentrism. It therefore detains the countries from making significant development progress, and hinders the donors to use their aid effectively (OECD 2016; Klitgaard 2015; Marquette 2014). Nevertheless, the complete elimination of poverty remains the top priority of the committee, as declared during the Paris Accra Agenda for Action in 2008 (OECD 2008b).
Table 1: The allocation of bilateral DAC and multilateral ODA in 2014 (based on data from OECD
2017)
2011 USD billion Africa Europe North America South America Asia Oceania Unspecified Total
DAC bilateral 25.3 2.8 2.3 2.8 21.6 1.3 25.5 81.6 Multilaterals 17.3 4.1 1.4 0.7 9.5 0.3 3.1 36.4 DAC bilateral % 31% 3% 3% 3% 26% 2% 31% 100% Multilaterals % 48% 11% 4% 2% 26% 1% 9% 100% Total 42.6 6.9 3.7 3.5 31.1 1.5 28.6 118
China stands in a sharp contrast to the traditional donor country such as the DAC members. Although actively engaged for quite some time now as a donor country itself, China received more total ODA funds than any other country on the African continent since the Millennium (OECD 2017). The early beginnings of the Chinese foreign aid program can be traced back to 1950, only one year after the foundation of the People’s Republic of China. China began its aid programme by providing material assistance to its bordering socialist countries, respectively the Democratic People’s Republic of Korea and Vietnam (State Council 2011). In their comprehensive historical review of China’s aid global allocation during the half of a century that started in 1956, Dreher and Fuchs (2015) identified five phases of evolution.
The first phase, ranging from 1956-1969, featured aid to non-socialist countries on the African continent and the Arabian Peninsula. Towards the end of this phase, China formulated its eight core principles regulating its future economic aid and technical assistance. As a developing country itself, China placed the emphasis on an equal partnership while promoting mutual benefits and no strings attached to its aid. The second phase, which covered the 1970-1978 period, has seen China promoting the path towards the independence of African countries from their colonial powers while initiating its first aid projects in Latin America. Clearly, ideological aspects shaped the first two phases of the African-Chinese development aid relations thanks to which China recovered—with the help of African countries—its legal seat in the United Nations at the expense of Taiwan’s (Davies 2007; State Council 2011).
5 subsequent phases (1989-1995 and 1996-2006). These last two phases contributed to reshaping China’s aid, utilized to increase economic cooperation in the form of joint ventures with Chinese small and medium scale enterprises, cooperative projects, technical training and debt relief. The end result was diversification of China’s aid portfolio, which earlier consisted almost exclusively of grants and interest-free loans (Dreher & Fuchs 2015; State Council 2011).
Combined with its progress in the structural transformation of its economy, which contributed to turn China into an export powerhouse, this shift in the foreign aid has helped to considerably upgrade trading relations with Africa, which advanced at a relentless 31% annually since the new Millennium (see Figure 2). On the heels of this economic success, China pledged at the Beijing Summit of the Forum on China-Africa Cooperation to double its aid to China-Africa by 2009, having the aid amount in 2006 as a reference point (Huan & Qi 2012).3 According to Chinese officials, China will
keep increasing its aid to developing countries in the future, and in particular to the low-income countries in order to support the realization of the Millennium Development Goals of the United Nations (State Council 2014). In the meantime, with the gaining importance and complexity of China’s foreign aid, its organizational and governing structure has been redesigned with the goal to ensure its value for money.4
Figure 2: Sino-African trade in billions of constant 2011 USD (based on data from UN Comtrade)
3 Although the AidData set does not attest the fulfilment of this promise, a clear upward trend from
2006 on is noticeable.
4 The Ministry of Commerce is the major government body to formulate, implement and supervise the
foreign aid activities and is thereby responsible for grants and interest-free loans. The Ministry of Foreign Affairs supports the Ministry of Commerce, as it manages the diplomatic channels within bilateral aid projects. The Finance Ministry thereby provides the funds for the aid projects and the Chinese embassies around the world monitor the aid activities on the ground. Last of all, the Export-Import Bank of China provides concessional loans, which are defined as loans that have substantially more generous terms than market loans. However, the activities of the bank are mostly not to be considered ODA-like and are thus only classified as aid activities in the broader sense (Davies 2007; Dreher & Fuchs 2015).
6
3 Literature Review
The aid allocation of the traditional donor countries has been the focus of an impressive literature grounded on well-established econometric methods. In contrast, its counterpart on emerging donors, while bourgeoning, still remains in its infancy leaving, along the way, a few crying gaps some, of which are the focus of the present paper.
3.1 The Various Contributions
DAC Member Countries
Alesina and Dollar (2000) analyse the determinants of the allocation of ODA flows from DAC members over the 1970-94 period using OLS regressions. Their main finding is that significant donor heterogeneity exists. Specific donors favour certain recipient countries. The U.S. allocates high aid volumes to Egypt and Israel while France transfers large aid shares to its former colonies. Other countries, such as Japan, favour recipient countries that have similar voting patterns in the UN or are its larger trading partners. However, the recipient’s needs on the other hand play only a secondary role. Thus, they conclude that DAC members rather follow their own strategic interest in their allocation of aid than the diminishment of poverty or the support of democracy and good policies.
A more recent study by Clist (2011) examines the development of DAC ODA allocation determinants over time from 1982-2006 by using a two-stage model. It consists out of a Probit model in the first stage, testing the eligibility to receive aid, followed by an OLS regression excluding recipients that did not receive any aid in year t. The results are similar. The behaviour of donors differs considerably cross-sectional but only little over time. External events, such as the end of the cold war or 9/11, had only little impact on the aid allocation determinants of the DAC donors at best. Furthermore, the donor countries merely show a slow sensitivity to the relevance of poverty and good governance. Thus, they conclude that a shift from rather strategic interests during the cold war to placing emphasis on the policy advancements and the country’s needs has not taken place.
7 A study by Dietrich (2013) used this new information of the aid distribution channels in order to add to the extensive literature on corruption and DAC aid by using OLS and Probit regressions over the relatively short time period from 2005 to 2009. Dietrich finds that the reason why the previous literature failed to find higher aid shares for less corrupt nations is the fact that DAC donors are more likely to bypass governments, when the risk of aid capture by corrupt state institutions is high. Thus instead of government institutions, the aid is transferred to NGOs, which does not affect the overall aid allocation to the country. This could be an explanation for the failure of the previous literature to find significant aid selectivity for recipient characteristics such as corruption or good governance.
In terms of best practices in foreign aid, a study by Easterly and Pfutze (2008) assesses both bilateral and multilateral agencies. By doing so, they relied on a variety of sources, including the DAC database, the Credit Reporting System’s database on aid activities as well as personal responses from aid agencies. Based on this data they ranked the aid agencies in five dimensions, respectively in their specialisation, selectivity, ineffective aid channels, overhead costs as well as transparency. The overall results show that the aid sector is splintered in too many small agencies, which each of them being specialised in too many sectors and active in too many countries. Furthermore, too much aid is still allocated to corrupt and autocratic regimes with the least developed countries receiving too little. Although the transparency in the aid sector is disastrous, multilateral development banks are still the closest in terms of best practices in foreign aid. Bilateral agencies in contrast rank in the lower middle range and United Nations agencies perform the worst.
Overall the literature shows that DAC donors are rather not altruistic. Although a tendency to an aid allocation in accordance with the needs of the recipient countries is noticeable, the selfish incentives are predominant. Nonetheless, the DAC donors are by far the largest source of bilateral aid for developing countries. However, the rise of emerging donors such as India and China constitutes a disruptive factor to the traditional aid hierarchy. The next section will therefore summarise the research on the aid allocation of China as to date.
China
8 their aid funds to Africa. Immediately after China tried to take over the lead-partnership with Africa, they increased their aid volumes again, in particular to those countries to which they have strategic interests.
Dreher et al. (2016) examined the determinants of the Chinese aid allocation over the period from 2000-2012 while differentiating between the different financial flows from China. Their results suggest that ODA flows from China are closely linked to foreign policy interests, but do not depend on the institutional quality of the recipient’s country nor their natural resource wealth. OOF flows in contrast appear to be conditional on the bilateral trade volume and the natural resource wealth of the recipient’s country. Furthermore, the coefficient for corrupt nations is positive and significant while it is also positive for ODA flows but not significant at the 10% level. The introduction of fixed effects, however, results into a drastic reduction of significant results at the 10% level overall. The focus on foreign policy interests remains the only significant coefficient.
Dreher and Fuchs (2015), also engaged in the previous study, analyse in collaboration with four further researchers two determinants of the Chinese aid allocation on a regional level in more detail. First, they test whether a disproportionate share of Chinese aid is allocated to the birth regions of African leaders. Second, they examine whether ethnic groups, which belong to the ethnic group of the leader, are favoured by Chinese aid flows. By using OLS regressions with 1650 Chinese development finance projects, covering 3,097 physical locations and 609 ethnic groups they come to the following results. As assumed, Chinese aid is allocated in particular to the birth regions of the African leaders at the cost of regions facing greater needs. Furthermore, they find similar but less statistically significant results for regions populated by their leader’s ethnic group.
Isaksson and Kotsadam (2016) perform another study on a regional level by focussing on the relationship of corruption and the location of Chinese development aid project sites. By using OLS and Probit models, they analyse whether individuals across 29 African countries and 8685 regional clusters experience higher levels of corruption when living next to one of the 227 analysed Chinese ODA project site. Although controlling for time-invariant effects, they find that Chinese project sites increase the local corruption level when being implemented, without having an impact on the economy. In contrast, World Bank projects do increase the economic activity without having any noticeable impact on the corruption level.
9 conclude that Western aid donors should thus not focus on political development but rather on a diversification of the economy, which would diminish the influence of Chinese aid, since it is focussed on natural resources.
Effectiveness
While the previously discussed literature purely focuses on the allocation of aid, an even larger number of studies analyses the effectiveness of foreign aid. Instead of citing a variety of studies, we chose to mainly refer to a well-known summary paper, which summarizes the findings of the last decades. Doucouliagos and Paldam (2011) compiled the results of 105 studies with 1217 estimates of the impact of foreign aid on the recipient’s economies. Although the literature keeps expanding, it continuously comes to the same results, respectively the ineffectiveness of aid. Their conclusion is in line with Edwards (2014), who discusses the effectiveness of aid from a historical perspective. According to Edward (2014), empirical analyses of panel data are just not suited for the complex ways and channels on which aid effects economies.
3.2 Take-Away Points and Our Working Hypotheses
We now critically summarize the main features of these empirical studies, identify the gaps, discuss how our contribution fits within the existing literature and outline, along the way, our working hypothesis. By doing so, we add further literature to support specific points. In particular, we frame our hypotheses along three main incentives to allocate aid: the need, the merit and self-interest.
Need
The State Council of China states in the introduction of its second official white paper on foreign aid that Chinese aids aim to reduce poverty and to improve the livelihood in developing countries and in particular within the least developed countries (State Council 2014). Davies (2007), however, points out that China is more interested in increasing its economic cooperation with African countries than focusing on diminishing their poverty. Copper (2016) in contrast argues that China, as a developing country itself, can access the needs of poor countries comparatively better than the traditional developed donor countries. Dreher et al. (2016) supports this line of arguing by providing empirical evidence for a focus of Chinese development aid flows on poorer countries. However, since the literature entertains conflicting hypothesis that have been tested, our study offers to fill this gap and advance the idea that poorer nations with higher needs receive comparatively more Chinese aid flows.
10
Merit
Nevertheless, aid does not only have to depend on the needs of a country but could also be granted to countries as a reward for showing improvements or goodwill in terms of their policies. In contrast to many DAC donors, China occupies the unenviable 141th position out of 159 in the 2016 edition of the Human Freedom Index (Vásquez & Pornik 2016). A country promoting civil liberties abroad while putting little emphasis on them at home seems to lack considerable credibility. This is supported by the fact that China is known for its non-interference policy which is also highlighted in its first official white paper concerning its development aid programs: “China never uses foreign aid as a means to interfere in recipient countries’ internal affairs or seek political privileges for itself.” (State Council 2011).
In contrast, a large strand of literature contends that China undermines the fight against corruption and autocracy and rather exploits its aid to bribe the decision-maker in favour for its own interests (see Condon 2012; Dreher et al. 2015; Curran 2016; Isaksson & Kotsadam 2016). Furthermore, DAC donors committed themselves to support countries promoting anti-corruption policies and democracy, thus leaving out corrupt and autocratic nations with less aid flows (ADB/OECD/Transparency International 2005). Thus, China sees itself as filling an important gap in terms of foreign aid in countries left out by the DAC donors. The emblematic case in point is Zimbabwe, an autocratic and corrupt regime ruled since the 1980s and without interruption by Robert Mugabe. After the DAC donors slowly cut off their aid year by year from 1996 till 2004 (OECD 2017), China came to the rescue of this regime and bailed the bankrupt country out in 2003 and 2004. The ODA amount that China used to do so exceeded the yearly ODA volume of all DAC donors combined (Strange et al. 2017). Interestingly, shortly after China increased its aid activities in Zimbabwe, the DAC reversed their drawback from Zimbabwe and increased their yearly ODA disbursements to levels even higher than before. Another example is the financial aid to the oil sector of the dictatorial regime in Sudan, allegedly responsible for the genocide in the Darfur region (Copper 2016; Transparency International 2016). In light of this anecdotal evidence, we posit that corrupt and autocratic regimes receive a disproportionately higher share of China’s aid.
H2: More corrupt and autocratic recipient countries receive more development aid.
Self-interest
11 with Chinese products sold below market price. The unattended effect of this strategy is the destruction of local businesses and a premature process of deindustrialization in Africa.
While foreign aid makes it possible for China’s products to aggressively penetrate Africa’s markets, it also facilitates the exploitation of Africa’s natural resources used to fuel China’s relentless economic growth. A report for the US Congressional Research Service comes to the conclusion that the Chinese growing market need for oil and minerals is the main driver of the Chinese aid activities on the continent (see Lum et al. 2009). The report points out that 70% of the total Chinese infrastructure financing of the continent over the 2002-2007 period went to Angola, Ethiopia, Nigeria and Sudan—all of which possess huge endowments in terms of oil. The main findings of Tseng and Krog (2016) are completely in line with this report, but fail to clearly differentiate between ODA and OOF flows.
Figure A1 in the appendix shows a striking coincidence between the allocations of China’s development aid flows to African countries and the corresponding distribution of the natural wealth. This figure supports the previously outlined results of the literature that countries that are growing or that maintain close economic relations with China, especially those with large natural resources endowments, have higher chances to receive Chinese foreign aid. In addition, based on this, we assume that the Chinese aid allocation is more self-interest founded than the aid allocation of DAC donors. Our working hypothesis is thus:
H3: China places more emphasis on self-interests than DAC donors and thus rewards recipient countries that are of economic strategic importance for China in terms of their trade, market potential and natural wealth with higher amounts of aid.
Flow types
The next hypothesis is not related to the incentives to allocate aid but to the choice of the aid flow type as illustrated in Figure 1. Recent literature, such as Dreher et al. (2016) and Tseng and Krog (2016) failed to analyse the flow types separately. Dreher et al. (2016) analysed ODA flows and the sum of OOF and Vague flows, with the latter flow type representing unspecified ODA and OOF flows. Tseng and Krog (2016) on the other hand analysed either ODA flows or the sum of ODA and OOF flows and thus were also not able to analyse OOF flows individually. We argue that this is closely related to the choice of the estimation method. By using linear regressions, they do not take the characteristics of their data into account, as it contains a large share of zeroes for their dependent variable. This method has two crucial shortcomings.
12 aid flows in the following year. With over 40% of ODA observations clustered at the zero bound, the OLS estimator leads to biased and inconsistent results. Neglecting all zero values to overcome this issue, however, would lead to a selectivity bias and is thus also not recommended (Wooldridge 2002; Hill et al. 2007).
Furthermore, this is also the reason why Dreher et al. (2016) and Tseng and Krog (2016) were not able to analyse OOF individually. Since over 70% of OOF observations are clustered at the zero bound, the inconsistency of linear estimators would be even higher. In order to overcome this issue both studies had to pool two flow types to decrease the inconsistency and bias of their estimators. In order to fill this gap of the existing literature, this study chooses an estimation method that allows an individual analysis of the OOF flows despite its zero bund clustering.
By doing so we are able to formulate another hypothesis concerning the difference of the two flow types. Based on the fact that ODA flows should inherit a large development objective while OOF flows are rather commercially oriented, we come to our following last hypothesis:
H4: ODA flows are more positively affected by country needs while OOF flows are more positively affected by self-interest determinants.
4 Quantitative Analysis
4.1 The Model
The aid allocation literature has not reached a consensus on the best estimation strategy and thus uses a variety of models. For example, the landmark contribution by Alesina and Dollar (2000) applies ordinary least squares (OLS) on the presumption that their dependent variable contains very few zero observations for most of their donors. Yet, they recognize that for a few donors with a larger share of zeroes in the dependent aid flow variable a Tobit procedure would be more appropriate. The comparison of the OLS and Tobit results for those donor countries shows that the results differ significantly, with the coefficients associated to income changing from a positive to a negative sign for Scandinavian countries. Other studies rely on either a Tobit model (Dollar & Levin 2006; Berthélemy & Tichit 2004), a Two-Stage model, consisting out of a Probit as a first stage followed by a simple linear regressions for only positive values5 (Hoeffler & Outram 2011; Clist 2011) or a Heckman procedure correcting for a potential selection bias (Lebovic & Voeten 2009; Berthélemy 2006).
Our baseline model rests on the panel Tobit, which characterizes the aid allocation process as follows:
5 By doing so, Clist (2011) assumes that both stages, with the first stage deciding whether a country
13 !!" = !! ! + !! ! + !!!!"# ! + !!" (1) with !!" = !!+ !!" (1.1) with y representing the aid flow of interest, r the recipient country, t the time period, X a set of explanatory variables and ! the error-term. The error term u is thereby composed of the recipient specific effects !! and the random noise of the data !!" (Hansen & Tarp 2000). However, taking the special nature of the dependent variable into account, a simple linear regression would be biased and inconsistent due to the large share of zeroes, which contradicts the assumption of a normal distribution (Hill et al. 2007). The logical solution to such an issue in the field of econometrics is the choice of a censored model, whereby the Tobit model recognizes that we have two sorts of data with
!!"= !! !" !!"∗ !" !!"∗ > ! !" ∗ ≤ ! (2.1) !!"∗ = !! ! + !! ! + !!!!"# ! + !!" !"#ℎ !!"~!(!, !!) (2.2)
The model, developed by Tobin (1958), assumes thereby a latent variable !!"∗ , whose
linearity depends on the set of explanatory variables !!" via the coefficient !! with a normally distributed error term. However, in reality not all the data of the latent variable is observed. As in 2.1 described, for values of !!"∗ smaller or equal to zero
!!" equals zero. Values of !!"∗ higher than zero, on contrary, are in !
!" observed. The
marginal effect used in the following regressions will illustrate how a change in the explanatory variable x affects the outcome of the dependent variable for a fictional country that possesses the mean country characteristics of the observed countries.
The results represented in the tables display the marginal effect of an increase in the explanatory variable on the aid dependent variable at the means of the explanatory variables. Interpreting the results is straightforward for log-log regressions, with the explanatory variable being log-transformed just as the dependent variable. A one-percentage increase in the explanatory variable leads to a ! percentage increase in the dependent variable. However, interpreting log-linear regressions is not as simple. When the explanatory variable is not log-transformed, a one-unit increase of the explanatory variable results into 100 ∗ !!− 1 percentage
change of the log transformed dependent variable (Kephardt 2013).
4.2 The Source Data
Overview
Our source data rests on the information compiled by AidData at the College of William & Mary and the Brigham Young University.6 This open source data tracks
14 China’s aid flows on a project-level basis from 2000 to 2013. We use their newest dataset at the time of writing this thesis, respectively version 1.2 (Strange et al. 2017). Compared to OECD’s International Development Statistics database widely utilized in the research on foreign aid (see Arndt et al. 2015 for example), which only reports China’s aid inflows but not its outflows, AidData fills an important gap by providing aid outflows information.7
The AidData methodology to track unreported financial flows consist out of a two-stage process. The first two-stage is based on a general search through the Factiva media database on a donor-recipient basis, which draws information from 28,000 media sources in 23 languages from all over the world. The information gathered is mostly based on newspaper, television and radio transcripts. In addition, research assistants collect data from government websites, so-called aid information management systems that provide information on country specific aid inflows and from academic articles. The second stage has the aim to enhance the quality of the data collected in the first stage. Research assistants evaluate every project individually and by doing so either confirm or disconfirm the existence of projects. Furthermore, if confirmed, they increase the quality of the record through various search engines and information repositories and especially countercheck the project’s value. As a result, the dataset of version 1.2 includes data on ODA-like, OOF-like, Vague and Total Official Investment-like flows with both the USD value as well as the number of projects.8
In addition to this, we use a variety of additional data sources for our explanatory variables. These include among others the Penn World Table version 9, the United Nations Comtrade database and the World Bank database. A full list of the variables, including their definition and sources used in this thesis, is to be found in Table A2 of the appendix.
Trend Analysis
Table 2: The summed Chinese financial development flows to Africa from 2000-2013 (based on
data from AidData 2017)
USD 2011 billion ODA OOF Vague Official Investment Total aid
USD value 38,185 51,602 40,570 7,574 137,932 Number of projects 1,774 350 484 39 2,647
Our work uses a panel dataset represented by 50 African economies tracked over the 2000-2013 period. The dataset covers over 2,647 development aid projects with
7
As a non OECD DAC member country, China is not compelled to publish detailed information about its aid outflow activities.
8 AidData defines Vague (Official Finance) flows as such flows that are either ODA-like or OOF-like,
15 a total value of $137,9 billion in constant 2011 USD prices.9 It covers the four types of
official development finance flows, as outlined in Figure 1. With 1,774 projects, ODA flows are clearly the largest type of financial flows in terms of the volume of development aid projects. However, with $38,2 billion, they are only ranked third in terms of value in constant prices.
Figures 3 A and B complement Table 2 by offering a perspective on both trend and level displayed by total aid and ODA-like flows to Africa.10 Except for 2007, when
these Chinese flows skyrocketed to outbalance the decrease of DAC flows immediately after the Great Recession (OECD 2017), the data reports a steady trend with the result that in 2013 the pre-recession level has been completely restored, if not considerably surpassed. Figures 3 C and D further show that a substantial share of aid flows is allocated to the infrastructure sector, an area in which Africa reveals a considerable deficit that contributes to the hold-back of its economy (A more detailed illustration of the flows by sector can be found in Table A1 in the appendix).
Figure 3: The development of Chinese aid flows in total and ODA-like volumes (based on data
from Strange et al. 2017)
9 However, only 1517 projects include a USD value with 1010 of those being ODA-like projects and
201 OOF-like.
10 The Chinese State Council published its first official paper with detailed information about its aid
flows, revealing that China spend 37,7 billion USD in constant prices on total aid flows to developing countries by the end of 2009 (State Council 2011). On the basis of the AidData data, 39,5% of this budget was spent alone on the African continent in the 2000-2009 period on pure ODA-like flows.
0 7,500 15,000 22,500 30,000 2000 2004 2008 2012
A: Total aid flows in millions of constant 2011 USD 0 2,500 5,000 7,500 10,000 2000 2004 2008 2012
B: ODA-like flows in millions of constant 2011 USD 0 5,000 10,000 15,000 20,000 2000 2004 2008 2012
C: Total aid flows related to infrastructure in millions of constant 2011 USD
0 1,000 2,000 3,000 4,000 2000 2004 2008 2012
16
The Variables
The set of variables retained in our model are listed in Table 2A and 3A in the appendix, along with their descriptive statistics which do not report any abnormal pattern. Similarly, diagnostic checks have been performed to ascertain that the variables comply up to a reasonable degree of confidence with the assumptions underlying the model. All in all, the explanatory variables do not report any evidence of multicollinearity. The literature emphasizes the simultaneity issue in the field (Brückner 2011; Clemens et al. 2012). The disbursement of aid has an impact on a variety of factors that are used in order to determine the decision of the aid allocation in the first place. For example, significant aid flows in year t could increase the growth rate in the same year or the corruption level, when transferred to the wrong authorities.
The literature disagrees on whether to use total aid or aid per capita as the dependent variable. The latter has the advantage as it controls for the fact that countries differ significantly in their population size, while the former takes into account that countries commit their aid allocation in absolute terms. In the absence of a consensus, we pursue two options: first, we use for most of our regressions the aid per capita option. Second, following Clist (2011), this study chooses to include aid as a percentage of the total Chinese aid flow in year t to recipient r. This has the advantage that the dependent variable is not influenced by fluctuations in the yearly aid budget of the donor11 and still considers the total amount spent by country. The
log of the population size is added to control for population differences for the latter variable option. Due to the wide scattering of the dependent variable’s observations, the dependent variables are transformed in a monotonic fashion following Dollar and Levin (2006) so that the two types of the dependent variable become log (!"#$%&'!$"(!!"+ 1; !""∗!"#!"
!"!! + 1). By doing so, it is secured that the Tobit
estimator censors observations with zero aid.
The explanatory variables are grouped into the earlier mentioned three areas of aid allocation incentives: need, merit and self-interest. The most straightforward variable describing the need level of a country is the logarithm of its GDP per capita, measured on the output-side real GDP at chained power purchasing parities to include the difference in the cost-of-living between the countries. We use the Lorenz curve, a powerful method to analyse the wealth distribution by Max Lorenz in 1905 (Cowell 2014) to have a first pre-analysis of the cumulative distribution of Chinese ODA flows. We rank the countries by their total ODA-like aid inflows per capita between 2000 and 2013 in an ascending order.12 If China would focus its ODA
11 Time dummies could diminish this influence, however, only to some extent (Clist 2011).
12 Each country point is thereby multiplied by its average population size of the period in order to
17 allocation on poorer countries, then the red continuous line of Figure 4 would be above the perfect equality line at all times. However, Figure 4 shows that the poorest 20% of the African population receive only 15% of the total Chinese ODA disbursements. Thus, Figure 4 suggests that China allocates higher shares to the medium-income countries, while neglecting the poorest countries. This is an interesting result, which we are going to test later on in the Tobit regressions. In contrast, the ODA allocation of the United States, represented by the continuous blue line in Figure 4, is above the equality line at all times, showing a redistributive allocation and thus a focus of the US on poorer countries.
Figure 4: The cumulative distribution of GDP and ODA flows from China as well as the USA
(authors own illustration based on data from Feenstra et al. 2015 and Strange et al. 2017)
To capture not only the need in terms of differences in economic circumstances but also differences in health conditions, we initially considered the use of the prevalence of HIV as a proxy for critical health conditions in African countries. However, Fan et al. (2014) argues that China, in contrast to the DAC donors, does not target specific diseases but rather focuses on the projects related to the general public health system. Life expectancy at birth in years is therefore chosen as the second variable regarding the need of the recipients. Countries with a weaker public health system, and hence with a lower life expectancy at birth, are therefore expected to be more eligible to be a recipient and to receive higher shares of aid. The third need related explanatory variable, the logarithm of number of people affected by disasters, is used to capture the short-term need of a recipient, and is thus also expected to be positive.
According to the no interference policy of China, policy factors, such as the institutional quality or civil liberties, should not play a vital role in China’s aid allocation. However, based on the results from Isaksson and Kotsadam (2016) as well as Tseng and Krog (2016), the corruption level and the political system do seem to be important. The former is included by the Transparency International corruption indicator, which measures the public sector corruption level. For the sake of (F(y1)=0.2) would also have 20% of the complete aid (Φ(y1)=0.2). This would be illustrated by a line
with a 45° ascending line on the graph, also known as the perfect equality line (Cowell 2014).
0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
ODA-flows from China GDP Perfect equality line
0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
18 simplicity, this variable is called logarithm of Anticorruption, as the variable ranges from 0 equalling a highly corrupt to 100 being a very clean country. For the second merit variable, this study uses a rather modest definition of democracy. Due to the comparatively low level of democracy in Africa (The Economist Intelligence Unit 2016), a country is assumed to be rather democratic, if elections are used to determine the chief executive of the nation (Przeworski 2013). The dummy variable takes the value 0 for no elections and 1 for elections. Following the above-mentioned literature, we assume both variables to be negatively influencing the aid allocation.
In terms of selfish incentives, probably the most important variable is the logarithm of the bilateral trade volume. The variable is based on the commodity trade volume between the recipient country and China and is expected to be positive. Secondly, to test the imputation of the Chinese natural resource craving, the logarithm of the oil wealth of African nations is used.13 If the imputations are correct,
the coefficients of the variable will be positive. The third self-interest indicator is economic growth. While a variety of empirical studies in this field, including those by Hoeffler & Outram (2011) and Berthélemy & Tichit (2004), use the variable as a proxy for economic policies, this study uses it as a proxy for growing markets with a higher potential for China’s products. Thus, it is expected to be positively influencing the amount of aid flows a country receives.
In addition, year dummy variables are introduced to the regressions and most of them appear to be significant. Although the use of time dummy variables is widespread in applied econometrics, this practice is not widely accepted in the literature of the Tobit model. For example, Berthélemy and Tichit (2004) argue that introducing fixed-effect to a non-linear Tobit model leads to biased and inconsistent results. However, Greene (2004) shows that while there is no bias in the slop estimators when including fixed effects, the variance is subject to a bias. The order-of-magnitude of this bias is negligible if t is five or higher, a requirement met by our panel with 14 years. In any event, given this absence of consensus, we decided to interpret conservatively the results related to the country dummy variables.
Considering all the variables discussed above, equation (2.2) looks like this: !"#!" = !!+ !!!"#$%&!"!!+ !!!"#$$%&!"!"!!+ !!!"#$%&%'$(%!"
+ !!!"#"$%&'(()*$%'"!"!!+ !!!"#$%&'()!"
+ !!!"!"#$%!"!!+ !!!"#$!!"+ !!!"#$%&!"!!+ !!"
(5)
Note that for the benefit of space dummy country and dummy trend variables have been suppressed from equation (5). While we lagged explanatory variables to avoid the potential risk of endogeneity, two of them remained intact. These are the disaster and democracy variables. The number of people affected by disasters is not lagged,
13 The oil wealth is thereby calculated on the present value of rents form the extraction of oil in 2005
19 as it displays the urgent short-term need of a country. The election variable is not lagged, as China negotiates its aid programs directly with the heads of state (Copper 2016). The political system type in year t is therefore essential for the aid allocation in year t as the personal decision of the current head of state depends on it.
4.3 Econometric Results
Aid is not just aid. We begin with the factors that drive the Chinese ODA allocation and analyse whether a discernible different pattern exists between China and the main DAC donors. We then delve into the different types of Chinese aid flows to assess our working hypotheses. We first focus on the narrow definition of aid, respectively ODA flows, followed by the second detailed analysis of the more commercial orientated OOF flows.
China vs. ODA Donor Countries
The goal here is to tackle the first research question—whether there are differences in the fundamentals behind China’s aid and those of the most important DAC donors such as the U.S., Germany and Japan. Table 3 reports the econometric results of the Tobit model with ODA per capita as the dependent variable as well as the inclusion of time dummies. Another important feature of the model is to refrain from pooling the DAC ODA flows to control for heterogeneity in accordance with the literature on the DAC aid allocation (Alesina & Dollar 2000; Clist 2011). All regressions of Table 4 passed the log-likelihood test, comparing it with a pooled version of the Tobit model. This clearly shows that our dataset possesses panel-level effects and that the random-effect Tobit estimator therefore is the estimator of preference. Furthermore, Likelihood Ratio (LR) Chi-Square test of all regressions shows that at least one of the predictors is always not equal to zero.14
The model uses the overall quality of public sector institutions instead of the corruption variable as an explanatory variable to evaluate whether the DAC donors meet their own obligations to reward countries that emphasize on good governance. The institutional variable rates factors such as the enforcement of property rights, the quality of the public sector management or the corruption level on a scale from 1=low to 6=high (Independent Evaluation Group & World Bank 2009). The overall picture that emerges is that while China emphasizes on a variety of determinants, the DAC donors mainly focus on two country characteristics, namely the institutional quality and the political system. Of the two considerations, the U.S. tends to give precedence to the former so that a 0.01 increase in the institutional quality scale translates into a significant 3.88% increase in the ODA per capita inflows. The elections effects do not play any role. Germany differs from the U.S. on two important
14 Tobit regressions have the disadvantage that no equivalent to the commonly in OLS regressions
20 grounds: a larger effect of institutions with 4.18% and an important consideration given to countries having elections with rewarding countries that change from being autocratic to having elections with an increase in ODA flows by 17.7%.15
In contrast, Japan rewards recipient countries the least relative to its counterpart DAC donors from the standpoint of quality of institutions and places more emphasis on the political system. With Japan, African economies that hold elections increase significantly ODA per capita inflows by 56.2%. These results stand in clear contrast to the Chinese aid allocation, which puts no emphasis on the institutional quality and rather favours autocratic regimes.
Table 1: Comparison of the ODA allocation to Africa between 2000-2013 by donor country
(1) ODA pc China (2) ODA pc Germany (3) ODA pc USA (4) ODA pc Japan
ln GDP pc (t-1) 0.266** 0.019 -0.077 0.043 (0.123) (0.082) (0.097) (0.078) Life Expectancy (t-1) -0.021* -0.007 -0.008 0.005 (0.012) (0.011) (0.017) (0.009) ln Disaster -0.025** -0.009 -0.002 -0.007 (0.011) (0.006) (0.006) (0.007) Institutions 0.059 1.646*** 1.586** 0.597* (0.504) (0.475) (0.697) (0.342) Elections -0.556*** 0.163** -0.120 0.446*** (0.195) (0.078) (0.087) (0.084) ln Trade (t-1) 0.074*** -0.066* 0.031 0.035 (0.025) (0.037) (0.036) (0.023) ln Oil wealth -0.021** 0.010 -0.005 -0.016*** (0.009) (0.008) (0.012) (0.005) Growth (t-1) -0.432 0.277 0.215 0.003 (0.334) (0.174) (0.190) (0.208) Observations 462 458 460 439 Countries 33 33 33 32 Notes: This was estimated using a random-effect Tobit model with the coefficients representing the marginal effect at the mean values of the explanatory variables. Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. The dependent variable is the logarithm of the ODA per capita amount in USD for the respective countries with (1) being China, (2) Germany, (3) USA and (4) Japan. All regressions passed the test for random-effect specification at the 1% level (not shown here), which indicates that panel-level effects exist.
Turning to the self-interest aspect of allocation aid, Table 3 supports the fact that DAC donors do not use aid for selfish interests, at least in economic terms. Germany even favours countries with a lower bilateral trade volume and Japan supports countries with fewer natural resources, though the effect offers a great deal of uncertainty. Furthermore, the U.S., often considered to have a geopolitical strategy with the view to secure the access to natural resources, do not focus their ODA allocation on natural resource rich countries. However, Table 3 also shows that this study cannot find any support for a need focus of the DAC donors, which should be the main focus of any aid allocation at all. Table 3 cannot find statistical evidence that
15 Calculating the effect of a 0.01 increase in elections is irrational, as countries either have elections
21 countries with a low-income level, a low life expectancy or countries affected by disasters receive significantly more aid. This sharply contradicts the self-imposed DAC standard for the allocation of aid that emphasises the relevance of allocating aid to where it is needed the most (OECD 2008b).
The results out Table 3 support the fifth hypothesis that China emphasises more on its economic interests, in particular its trade volume. In contrast, the DAC donors focus on institutional and political factors that are at best neglected by China. However, while China transfers its aid flows directly to the official institutions, DAC donors often bypass these institutions in order to penalize governments that do not implement conditions such as good governance (see Dietrich 2013). These actions have the consequence that their aid allocation as seen in Table 3, can appear blurrier than it is. Governments with autocratic systems thus could receive significantly lower shares of DAC aid flows to governmental institutions due to the fact that large parts of their DAC aid inflows are bypassed to NGOs. The regressions of Table 3, however, are not able to capture this process.
ODA Flows
Table 4 shows that the ODA allocation of China is not just random but rather highly selective. Yet, when any possible time variant influences and country differences are taken into account, most determinants stay significant. The composition of Table 4 is as follows: The first two columns exclude both time and country dummies, while column 3 and 4 control for time variant factors by introducing year dummies. Column 5 and 6 extend this by including country dummies to control for time-invariant recipient factors. Odd column numbers, respectively column 1, 3 and 5 show the results for the country specific share of each year’s total aid allocation to the African continent. The remaining columns, respectively 2,4 and 6 show the results for the ODA per capita dependent variable.
22 significant at the 10% level.16 However, controlling for country differences in column 5
and 6 changes the picture. The GDP per capita variable becomes non-significant while the Life Expectancy significance increases and its marginal effect enlarges by nearly ten-fold.
Table 4: Chinese ODA flows to Africa between 2000-2013
(1) ODA share (2) ODA per capita (3) ODA share (4) ODA per capita (5) ODA share (6) ODA per capita
ln GDP pc (t-1) 0.210** 0.327*** 0.218** 0.270** 0.153 0.195 (0.084) (0.111) (0.086) (0.112) (0.141) (0.197) Life Expectancy (t-1) -0.016** -0.022** -0.017** -0.020* -0.144*** -0.092*** (0.008) (0.011) (0.008) (0.011) (0.031) (0.019) ln Disaster -0.008 -0.016 -0.012* -0.020** -0.014* -0.017 (0.007) (0.010) (0.007) (0.010) (0.008) (0.010) ln Anticorruption (t-1) -0.315* -0.275 -0.293 -0.244 -0.450** -0.357 (0.183) (0.252) (0.185) (0.254) (0.229) (0.315) Elections -0.261** -0.321* -0.309** -0.383** -0.425*** -0.461** (0.128) (0.173) (0.130) (0.175) (0.144) (0.189) ln Trade (t-1) 0.040*** 0.085*** 0.043*** 0.061*** 0.041** 0.061*** (0.014) (0.018) (0.016) (0.021) (0.017) (0.023) ln Oil wealth -0.026*** -0.038*** -0.027*** -0.034*** (0.007) (0.008) (0.007) (0.008) Growth (t-1) -0.185 -0.221 -0.262 -0.421 -0.311 -0.502 (0.235) (0.327) (0.237) (0.327) (0.236) (0.327) ln Population (t-1) 0.127** 0.139*** 2.132** (0.052) (0.053) (0.837)
Year dummies No No Yes Yes Yes Yes
Country dummies No No No No Yes Yes
Observations 588 588 588 588 588 588 Countries 42 42 42 42 42 42 Notes: This was estimated using a random-effect Tobit model with the coefficients representing the marginal effect at the mean values of the explanatory variables. Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. The dependent variable in column (1), (3) and (5) is the logarithm of the share a country receives of the total Chinese ODA allocation in year t. The dependent variable for the remaining columns is the logarithm of the ODA per capita amount in USD. All regressions passed the test for random-effect specification at the 1% level (not shown here), which indicates that panel-level effects exist.
Overall, Table 4 shows that China does take the needs of the recipient country into account, however, mostly negatively. Poorer countries and countries with disasters receive less ODA flows. The positive impact of GDP per capita stands in contrast to the findings of Dreher et al. (2016), who find a rather negative impact. However, the significance of the results decreases noticeable, when using GDP per capita values that do not incorporate the purchasing power of the citizen.17 Furthermore, the results of this study give a reasonable explanation for the mixed
16 We have to recognize, however, that the inclusion of fixed-effects in the non-linear Tobit estimator is
a controversial issue and the results should be interpreted with caution in terms of their significance (Greene 2004).
17 Dreher et al. (2016) do not declare which type of GDP per capita they use. This study assumes that
23 picture. Only in terms of its ODA allocation, China prefers autocratic countries with a slight tendency for corruption. Eric Uslaner (2010) shows that autocratic regimes are known for their high corruption levels that are based upon a foundation of unequal distributed resources, or in short based on inequality. Thus, it seems reasonable to suppose that China allocates its ODA flows to countries with higher inequality, that appear to be wealthier on paper, while having the majority of its population living in poverty. As a consequence, these countries also exhibit lower life expectancies. The results, nonetheless, speak unfavourably for the first hypothesis and suggest that a decrease in the GDP per capita leads to a decrease of the ODA flows from China. Although countries with worse public health conditions are supported by China and receive more ODA flows, the aid allocation of China is definitely not primarily focussed on diminishing the needs of its recipients.
The impact of corruption is generally small and its significance varies according to the nature of the dependent variable. A one percentage increase in the corruption level of the recipient country increases the aid share by only 0.3-0.45%, though the estimate is subject to a moderate level of uncertainty. In contrast, the effect of the Elections variable remains robustly significant. Having all other variables held fixed, a country that holds elections receives around 23.0-36.9% less ODA depending on the inclusion of time and country fixed effects. This is a remarkable impact and shows a strong favour of rather autocratic regimes just like China is one itself. Our results conflict with those obtained by Dreher et al. (2016) featuring a lack of a robust statistical significant evidence for the autocratic regime effect on ODA allocation. Our findings thus support the second hypothesis, stating that China favours corrupt and autocratic recipients.
Turning to the analysis of self-interest, the results show a mixed picture. The trade variable shows a positive impact of the bilateral trade, albeit with a varying degree of uncertainty. A one-percentage increase in the trade volume leads to a 0.04% increase in the ODA aid share, which is robust to the inclusion of time and country dummies. A higher oil wealth of a recipient has in contrast to the assumptions of the third hypothesis a negative impact on the ODA aid inflows. A one-percentage increase in the oil wealth leads to a reduction of the ODA share by approximately 0.03-0.04%.18 Replacing the oil wealth variable with other natural resources, such as natural gas, coal or minerals leads to similar or insignificant results. Although having a systematic negative coefficient, the growth variable remains insignificant across the regressions impact on the aid flows. The third hypothesis can therefore be partially rejected, as the natural wealth and the market potential, measured in terms of the growth rate, have no positive impact on the aid allocation.
24
OOF Flows
Turning to the more commercial orientated aid flow type OOF, Table 5 shows a less selective picture in comparison to Table 4, as more aid allocation determinants remain insignificant. The structure of Table 5 is thereby similar to the structure of Table 4.
Both, disasters and the income level of recipient countries have no significant impact on the allocation of OOF flows by China. The only variable that proxies the need of the recipient which is significant is the life expectancy, ranging between a 1.5-1.6% decrease of the OOF flow share for an additional year of life expectancy at birth. Including fixed effects, however, removes the impact nearly completely. OOF flows have therefore a slight tendency to be directed towards needier recipients, a result that partly supports the first hypothesis.
Table 5: Chinese OOF flows to Africa between 2000-2013
(1) OOF share (2) OOF per capita (3) OOF share (4) OOF per capita (5) OOF share (6) OOF per capita
ln GDP pc (t-1) 0.022 0.010 0.032 0.006 0.000 -0.001 (0.055) (0.059) (0.053) (0.056) (0.016) (0.126) Life Expectancy (t-1) -0.014** -0.017** -0.014** -0.016** -0.000 -0.002 (0.006) (0.007) (0.005) (0.006) (0.102) (0.212) ln Disaster 0.003 0.007 0.003 0.005 0.000 0.000 (0.006) (0.006) (0.005) (0.006) (0.007) (0.004) ln Anticorruption (t-1) -0.055 -0.061 -0.043 -0.043 0.001 -0.000 (0.127) (0.142) (0.119) (0.131) (0.096) (0.019) Elections 0.153*** 0.180*** 0.130*** 0.146*** 0.004 0.004 (0.050) (0.055) (0.047) (0.052) (0.520) (0.455) ln Trade (t-1) 0.002 0.014 0.001 0.006 -0.001 -0.000 (0.009) (0.011) (0.010) (0.012) (0.064) (0.042) ln Oil wealth 0.005 0.007* 0.004 0.008* (0.004) (0.004) (0.004) (0.004) ln Mineral wealth 0.009* 0.012** 0.008* 0.012** (0.005) (0.005) (0.004) (0.005) Growth (t-1) -0.318* -0.307 -0.318* -0.307* -0.016 -0.013 (0.181) (0.198) (0.169) (0.177) (1.826) (1.469) ln Population (t-1) 0.074* 0.076** -0.046 (0.038) (0.036) (5.219) Observations 588 588 588 588 588 588 Countries 42 42 42 42 42 42 Notes: This was estimated using a random-effect Tobit model with the coefficients representing the marginal effect at the mean values of the explanatory variables. Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively. The dependent variable in column (1), (3) and (5) is the logarithm of the share a country receives of the total Chinese OOF allocation in year t. The dependent variable for the remaining columns is the logarithm of the OOF per capita amount in USD. All regressions passed the test for random-effect specification at the 1% level (not shown here), which indicates that panel-level effects exist.
