Abstract
Why do developing countries give foreign aid? Although emerging donors are gaining importance in development finance, lack of systematic data on their aid allocation limits our understanding of their motives. We address this gap using detailed data on a major new donor, Turkey, since 1992. We show that domestic politics has had a large impact on Turkey’s priorities in giving aid. Turkish aid used to be determined by international alignments and coethnicity, but after the Islamic AKP (Adalet ve Kalkınma Partisi) took power, political ties lost importance. Turkey began to give more economic aid to trade partners and more humanitarian aid to Muslim nations. While this new focus on trade ties makes Turkey more similar to traditional donors, the growing role of cultural ties sets Turkey apart. The broader lesson of this study on Turkey is that government change can significantly influence the way emerging donors give aid and these changes can vary in predictable ways across different types of aid.
Introduction
The growing importance of developing countries as aid donors is a significant change in development finance. Until recently, these “emerging” (or new) donors have been primarily recipients, but today they give more than 10 percent of global aid (Dreher, Fuchs, and Nunnenkamp 2013; Walz and Ramachandran 2011). Moreover, this trend is likely to continue as developing countries become richer, which has led one observer to argue that a “silent revolution” is underway (Woods 2008). These developments may have positive or negative consequences for recipient states depending on how and why emerging donors give aid. Broadly, there are two views on emerging donors’ motives. Those who suspect that new donors are mainly motivated by self-interest warn developing countries of “donors who are more interested in extracting [their] resources than in building [their] capacity” and “new colonialism.” 1 Others claim that emerging donors are motivated by solidarity with other developing economies and have the potential to promote growth. 2
The scarcity of reliable data on new donors remains a major obstacle to studying their motives systematically. Walz and Ramachandran (2011, 6) note that there is a conspicuous lack of clear data from some of the largest and most rapidly evolving aid donors such as China, India and Brazil. Many countries have no standard system for reporting [official development aid], or even definitions of what qualifies as development assistance.
For this reason, although existing studies on emerging donors have made important contributions, they are limited to a small number of donor countries and years (Dreher and Fuchs 2015; Dreher, Nunnenkamp, and Thiele 2011; Fuchs and Vadlamannati 2013).
In this paper, we contribute to the literature on foreign aid by analyzing the bilateral aid allocation of a major new donor, Turkey, since 1992. The data come from the Organisation for Economic Co-operation and Development–Development Assistance Committee (OECD DAC) database and have a number of advantageous features that allow a detailed analysis of this emerging donor. Turkey is not a member of the OECD DAC but has reported to the DAC database since 1992. Turkish bilateral aid has been recorded according to the guidelines and definitions of the OECD, which facilitates comparison with the traditional OECD donors. 3 Moreover, our data are sector level, which allows us to disaggregate total aid into its economic and humanitarian components and test whether Turkey gives different types of aid for different reasons. Using this long series of data, we can study new questions such as the effect of domestic political and economic change on emerging donor aid. In addition to data availability, Turkey merits study due to its size. Turkey is among the world’s 20 largest economies and since 2003 has given more than $600 million dollars of aid annually. According to an OECD estimate, Turkey was the second largest emerging aid donor after China in 2008 (Smith, Fordelone, and Zimmermann 2010, 2–3), and in 2011, Turkey’s aid budget exceeded one billion dollars (author’s calculation based on OECD data). 4 In short, this paper analyzes the aid allocation of a major new donor using reliable data over several years.
Our main finding is that the Islamist AKP (Adalet ve Kalkınma Partisi), which came to power in 2003, has made Turkey’s aid program much larger and changed its priorities. Before the AKP, the two main determinants of Turkish aid, aside from recipient need, were international alignments and ethnic ties. These factors lost their importance when the AKP came to power; Turkey began to give more economic aid to its trade partners and more humanitarian aid to Muslim countries. Comparing Turkey with traditional donors, the growing importance of trade ties has made Turkey more similar to traditional donors, whereas the growing role of cultural ties in humanitarian aid makes Turkey different from them.
To our knowledge, this is the first paper to explore the role of domestic politics in emerging donor aid even though there are separate literatures on new donors and domestic–international linkages, which are cited below. We show that domestic politics can change an emerging donor’s priorities and that these changes may vary across different types of aid. We propose that one reason for such variation is that different types of aid have different domestic political consequences. For instance, humanitarian crises are often newsworthy, which motivates politicians to distribute humanitarian aid in ways that will appeal to the general public, whereas economic aid projects are less visible, which means weaker incentives to use economic aid for populist goals. Similar to Milner and Tingley (2015), who argue that foreign policy tools such as foreign aid and economic sanctions need to be understood differently because they have different domestic consequences, our findings suggest that it may be beneficial to analyze different types of aid separately for they can also have different domestic consequences. Of course, this paper is limited to the analysis of one country, which means that our findings cannot be generalized to all emerging donors. We hope that our work will encourage other scholars to study the effects of domestic politics on other emerging donors as more data become available.
We proceed as follows. In the next section, we briefly review Turkish foreign policy in the period under study. The “Theoretical Framework” and the “Data and Method” sections present our theoretical framework and research design. In the “Results” section, we analyze the determinants of Turkish aid in different time periods. Then in the “Is Turkey Different Than Traditional Donors?” section, we report on the differences and similarities between Turkey and traditional OECD donors. The final section concludes by discussing the broader implications and limitations of our paper.
The Impact of the 2002 Election on Turkish Politics
The purpose of this section is to briefly review Turkish politics in the post–Cold War era and highlight the significance of the 2002 election. This discussion will motivate the hypotheses we test below.
Turkey, a reliable ally of the West during the Cold War, maintained its Western focus in the 1990s. Between 1990 and 2002, all Turkish governments (with a brief exception) ideologically belonged to “Kemalism,” the pro-Western, nationalist, and secular founding ideology of the country. Turkey continued its efforts to become a full member of the European Union but remained uninterested in its eastern neighbors (Özel 2003). One active area of Turkish foreign policy was the newly independent Turkic nations in Central Asia (Azerbaijan, Turkmenistan, Kazakhstan, Uzbekistan, and Kyrgyzstan). Turkey launched a broad effort to form closer links with these republics and in 1992 established the main coordinating agency of Turkish foreign aid, TIKA (Türk İşbirliği ve Koordinasyon Ajansı Başkanlığı), with the express goal of providing aid to coethnic nations (Fidan and Nurdun 2008). However, Turkey suffered from significant economic and political instability in the 1990s and had limited ability to give aid broadly (Özel 2003).
In 2002, Turkish politics experienced a fundamental break. A severe economic crisis in 2001 caused the economy to contract by almost 10 percent. In the 2002 election, the Turkish party system collapsed. The big winner of the election was the newly founded AKP, which won two-thirds of the seats in the Parliament and formed the first single-party government in Turkey in more than a decade. Since 2002, the AKP has won four consecutive elections by providing economic stability and increasing the role of Islam in public life (Müftüler-Baç and Keyman 2012). Virtually all observers of Turkish politics agree that the AKP victory in 2002 was a momentous change that brought to power a new set of conservative and ethnically diverse interests (see, for example, Aktürk 2012; Öniş and Keyman 2003; Yavuz 2009). The rise of the AKP gave a new group of foreign policy elites the opportunity to implement their vision (Ipek 2015). The AKP’s foreign policy is based on the view that Turkey’s strictly pro-Western foreign policy has become inadequate, and Turkey, given its historical legacy and geographical position, should be more active, especially in the Middle East (Davutoğlu 2001).
Turkish foreign policy has undergone a number of changes since 2002. In general, AKP governments have pursued a more active foreign policy than their predecessors (Bayer and Keyman 2012). Although there have been setbacks recently, Turkey’s relations with its neighbors improved significantly between 2002 and 2010, partially as a result of the “zero problems with neighbours” policy, which sought to improve Turkey’s relations with its neighbors (Özel and Özcan 2011). Another source of rapprochement was the AKP’s Islamist ideology. In contrast to previous Turkish governments, the AKP comes from the Islamist tradition, which has emphasized the worldwide Muslim community (ummah) and attacked Turkish nationalists as “ethnicist” (kavmiyetci) (Aktürk 2012, 177–79). The Islamic ideology of the AKP, combined with the AKP’s support for Islamist groups in Syria and Egypt and falling out with Israel, has led observers such as former U.S. diplomat Peter Galbraith to conclude that “Turkey used to identify itself as a nationalist Kemalist regime; today it is Sunni . . . With the AKP, being a Sunni has become the common denominator and the AKP chose to emphasize this identity” (Basaran 2012). Based on these observations, we test below whether Muslim countries receive more aid from Turkey in the AKP era than before.
A second important development has been the increasing role of economic factors in foreign policy. Turkey’s foreign trade has grown from 23 percent of its gross domestic product (GDP) in 1995 to almost 40 percent by 2005, and domestic interest groups who rely on this trade have become influential (Kirişçi 2009). According to Kirişçi, before the AKP came to power Turkey’s foreign policy was driven by security considerations. However, under the AKP Turkey has become a “trading state,” whose primary foreign policy goals include seeking markets and economic benefits (Kirişçi and Kaptanoğlu 2011, 701). This account leads us to the hypothesis that perhaps AKP governments have used foreign aid more strongly to gain economic advantages such as access to new export markets.
To summarize, there is a consensus that Turkish politics experienced a sudden and dramatic change when the AKP came into office in 2002. A new set of decision makers with different preferences and constituency came to power which led to significant shifts in domestic and foreign policy. This sudden change in Turkish politics gives us an ideal case to observe the effects of domestic political change on the aid policies of a major new donor.
Theoretical Framework
This section proposes a theoretical framework to analyze Turkish aid that is flexible, incorporates several recipient characteristics, distinguishes between different types of aid, and allows for the possibility that different governments have used aid to pursue different foreign policy goals.
Domestic Political Change and Foreign Aid Allocation
Scholars have shown that there are strong links between domestic politics and interstate relations. For example, especially in less developed states, government turnover is often followed by changes in trade relations (McGillivray and Smith 2004) and how the country votes at the United Nations (Dreher and Jensen 2013). In the U.S. context, domestic actors are heavily involved in foreign policy making to shape its consequences (Milner and Tingley 2015). Studying OECD DAC members, Therien and Noel (2000) and Tingley (2010) show that government turnover in a donor country affects how that country distributes aid. In this paper, we extend this line of research to Turkey.
We expect a relationship between domestic politics and foreign policy, because different governments often rely on the support of different constituencies. Domestic groups may have different policy preferences due to conflicting economic interests or divisions along religious, ethnic, and ideological lines (Leeds, Mattes, and Vogel 2009). We expect differences in group priorities to be reflected in the policies that leaders make, because each leader needs to satisfy its support base to remain in power (Bueno de Mesquita et al. 2003). Importantly, we do not predict every instance of government turnover to lead to a significant change in policy; we only expect to observe it when the new government represents a different core group than its predecessor (Leeds, Mattes, and Vogel 2009).
Based on the scholarly consensus that the AKP’s victory in 2002 brought to power a new set of interests, we hypothesize that this significant break in Turkish politics will be reflected in Turkish foreign aid policies. Strong economic performance has allowed AKP elites to increase the total amount of aid Turkey gives (Ipek 2015). Our contribution is to analyze the distribution of this aid in detail and explore the changes in Turkey’s priorities.
Disaggregating Sectors of Aid
Motivated by recent work on foreign aid, we argue that economic and humanitarian Turkish aid are driven by different factors, which means that they need to be analyzed separately (Nielsen 2013). First of all, economic and humanitarian aid respond to different types of recipient needs. Economic aid helps support a recipient’s long-term developmental goals through education and technical assistance. Examples of Turkey’s economic aid include funding new schools and training state officials in the areas of banking and statistics in recipient countries (Özkan and Demirtepe 2012, 659). In contrast, humanitarian aid is more appropriate for immediate disaster relief and aims to provide basic necessities for victims of natural and man-made disasters. For instance, Turkish humanitarian aid to Pakistan after the 2010 floods consisted mainly of food, medicine, tents, and medical personnel to work at temporary hospitals (Republic of Turkey, Ministry of Foreign Affairs 2015). 5
Second, relative to long-standing problems of underdevelopment in other countries, humanitarian crises attract more attention from the public and the media (Shoemaker and Reese 2011). If people pay closer attention to where the government sends humanitarian aid, then the government has incentives to give humanitarian aid in ways that please the general public and gain popularity (Strömberg 2007). A good example is then-Prime Minister Erdoğan’s visit to Somalia (a majority Muslim country) in 2011 during the holy month of Ramadan; his visit was highly publicized and accompanied by Turkey’s highest humanitarian aid package (77 million dollars) to the country. In contrast to humanitarian aid, economic aid, which is less monitored by the public, can be more easily used to benefit special interests. Such differences matter, because politicians use foreign policy tools with different domestic consequences in different ways (Milner and Tingley 2015). Consequently, we can expect Turkish governments to use humanitarian and economic aid in different ways that advance the government’s international as well as domestic political goals.
In Figure 1, we take a look at the broad trends in Turkish aid. It shows that the initial increase after 2002 was driven completely by the increase in economic aid. Turkish humanitarian aid increased only after 2010 and this increase was due to the humanitarian crisis in Syria and the aid Turkey gave to Syrians. When we exclude aid to Syria, the increase in Turkish humanitarian aid disappears. 6 Figure 1 lends initial support to our hypothesis that Turkey distributes economic and humanitarian aid for different purposes. 7
Amount of Turkish aid by type.
The Role of Recipient Characteristics
We include a set of recipient characteristics in our baseline models and test several additional covariates in robustness checks. First, recipients’ need for aid is naturally an important factor in Turkey’s aid allocation. We use multiple indicators of long- and short-term needs.
Second, we expect to see a significant relationship between Turkish aid and indicators of recipients’ economic and political value, because previous literature has consistently found that donor interest impacts aid allocation (Alesina and Dollar 2000). Moreover, we expect the relationship between Turkish aid and a recipient’s commercial importance to be stronger in the AKP era given AKP governments’ efforts to gain prestige and export markets abroad (Kirişçi and Kaptanoğlu 2011).
Third, good governance has become an increasingly important part of the debate on foreign aid. Traditional donors have pledged to reward well-governed countries with more aid, because aid given to corrupt authoritarian regimes may be stolen or used to buttress the regime rather than promote development (Easterly 2006). Anecdotal evidence suggests that Turkey prioritizes economic ties over human rights in its dealings with repressive regimes (Özel and Özcan 2011, 128). Here, we conduct the first systematic test of the link between governance and Turkish aid.
Another potentially significant factor is cultural affinity. Donors may give more aid to culturally similar recipients for instrumental or altruistic purposes (Neumayer 2003). Such altruism stems from demands by donor country citizens that their state help coethnic and co-religious nations. Elected officials, who need to respond to the public’s demands to stay in power, are likely to distribute aid accordingly, especially in more dramatic cases (such as the aftermath of a natural disaster) that attract more attention from citizens.
We measure cultural similarity in ethnic and religious terms. Importantly, we allow the weight of cultural factors to change over time depending on the ideology of the government ruling Turkey.
As Turkey’s aid program was founded in the 1990s to support newly independent Turkic nations, we expect to find a strong relationship between ethnic ties and Turkish aid in the pre-AKP era. However, we expect the importance of ethnic ties to diminish in the AKP period as a result of Sunni-Muslim identity replacing language as the core cultural feature of Turkishness (Saraçoğlu and Demirkol 2015). There is some evidence of such change in the aggregate aid share of the five Turkic nations in Central Asia. Between 1992 and 2002, Turkic nations received almost 50 percent of all Turkish foreign aid, but in the AKP era their share fell to 35 percent on average, which is still significant and yet lower.
Religion is another important source of group identity (Neumayer 2003). We expect the impact of Muslim solidarity to increase under the AKP governments because of the greater role of religion in the AKP’s ideology. Given the greater visibility of humanitarian aid, one can expect a more pronounced increase in the importance of religion in humanitarian aid.
Data and Method
In this section, we describe our empirical research design in detail. Our unit of analysis is a donor–recipient–year. In the first set of analyses, we compare the aid allocations of different Turkish governments (pre-AKP and AKP era) and the only donor in our sample is Turkey. In the second set of analyses, we contrast Turkey’s aid allocation to that of the OECD DAC donors. The universe of recipients is all low- and middle-income countries based on World Bank criteria.
Dependent Variables
Our dependent variable is the logged amount of aid (in constant U.S. dollars) given to a recipient in a year. 8 Data on aid disbursement come from the OECD DAC, 9 and we disaggregate a donor’s total aid into “humanitarian” and “economic” aid using the OECD categorization. 10 Our results are robust to using as an alternative dependent variable, the percentage of a donor’s aid given to a recipient.
Independent Variables
We create a dichotomous variable AKP to indicate the years in which Turkey is governed by the AKP. Because the AKP came to power in November 2002, this variable takes the value of 0 until 2002, and 1 for the years 2003 and onward.
Our first indicator of recipient need is Infant Mortality, which is defined as the number of infants dying before reaching one year of age (per 1,000 live births) in recipient country (World Bank 2015). Infant mortality captures the level of abject poverty. Second, we control for the recipient’s GDP Per Capita (World Bank 2015). This variable is related to, but distinct from, infant mortality. Although many countries in the world have achieved low infant mortality rates, they continue to vary greatly in wealth. For instance, in former communist countries such as Moldova and Serbia, infant mortality is as low as Saudi Arabia even though the former have much lower GDP per capita. These two variables are lagged by one year to avoid endogeneity. Our third indicator, Disaster Victims, is the number of people affected by disasters (natural and man-made) in a given year and measures the need for emergency relief (Guha-Sapir, Below, and Hoyois 2016). All three variables are logged.
We measure a recipient’s economic importance by its level of imports from Turkey (Barbieri and Keshk 2012). This variable, Imports from Donor, is logged and lagged by one year. 11 We measure the political relations between Turkey and other recipients by the similarity of their votes in the United Nations General Assembly (UNGA; Strezhnev, Voeten, and Bailey 2009). UN Voting Similarity is also lagged and varies between −1 and 1. 12
We measure cultural similarity between Turkey and recipient nations along ethnic and religious lines. Our indicators of cultural similarity are continuous and measure the percentage of people in a recipient country that are similar to the population of Turkey, which is majority Turkish and Muslim. Common Language Group is the percentage recipient population that speaks a Turkic language (Fearon 2003). Common Religion is the percentage of Muslim people in a recipient country (Barro and McCleary 2003). Our results are robust to using dichotomous measures of cultural similarity based on the largest ethnic and religious groups in a recipient nation.
Our two indicators of good governance are a recipient’s level of democracy and respect for human rights. Democracy is the recipient’s yearly democracy index score (Polity2) from the Polity IV data set and varies between −10 and 10 (Jaggers and Gurr 1995). Political Terror Scale measures violations of physical and personal integrity rights committed by a recipient state (Wood and Gibney 2010). We lag these variables by one year.
Control variables include recipient’s population (World Bank 2015) and the geographical distance between Turkey and recipient (Mayer and Zignago 2011). Both variables are logged. Last, we capture Turkey’s ability to give aid by controlling for Turkey’s GDP per capita, GDP growth rate (World Bank 2015), and number of disaster victims in a given year (Guha-Sapir, Below, and Hoyois 2016).
As robustness checks, we explore the effects of several other variables, including denominational differences among Muslim recipients (Shia vs. Sunni), recipients’ historical ties to the Ottoman Empire, and the ideological positions of Turkish governments on the left–right scale. Including additional controls does not have an impact on our main findings. These robustness checks and summary statistics are presented in the online appendix.
Statistical Model
We analyze our data using the Tobit estimator, which offers significant advantages over alternative methods (Wooldridge 2001). Turkish aid data have a large cluster of observations at zero and only take nonnegative values. The simple ordinary least squares (OLS) would produce biased estimates in this case. Another alternative, the Heckman selection model, requires us to include at least one variable that influences the selection (“gate-keeping”) stage but not the outcome (“allocation”) stage. Several scholars have noted the difficulty of finding such a variable in the foreign aid context and thus used the Tobit estimator, which deals with nonnegative data by estimating in a single equation whether an observation will receive aid and the amount of this aid (Dreher, Fuchs, and Nunnenkamp 2013; Nielsen 2013). Although our preferred estimator is Tobit, our findings are robust to Heckman and OLS estimators as well. 13 In all models, robust standard errors are clustered by recipient.
Our preferred specification is a nested model that includes each explanatory variable of interest and its interaction with the AKP indicator. The nested regression model is preferable to running two separate regressions on two subsamples (pre-AKP and AKP eras), because the interaction terms can directly tell us the size and statistical significance of any change in Turkish aid after 2003. 14 We took care to ensure that multicollinearity does not inflate our standard errors and make it difficult to detect statistically significant effects. Specifically, we used model comparison tests to check that regressions including these interaction terms fit the data better than other specifications. To deal with autocorrelation, we include a lagged dependent variable in every model.
As Tobit is a nonlinear estimator, regression coefficients are not informative about substantive effects. More importantly, we cannot directly interpret the statistical or substantive significance of our interaction terms (Ai and Norton 2003). For this reason, we follow the example of Dreher, Nunnenkamp, and Thiele (2011) and in the main text present the marginal effect of each independent variable. We provide regression coefficient tables in the online appendix. Although researchers often use sample means when estimating marginal effects, these may be substantively unrepresentative. Therefore, we average over the observed values in our sample as recommended by Hanmer and Ozan Kalkan (2013). 15 For each explanatory variable, we show the marginal effect and the corresponding standard errors (in parentheses), evaluated for the AKP indicator equal to zero and, respectively, equal to one, at the mean of the other explanatory variables. We also conducted Wald tests for equality for each pair of marginal effects, and the p values of these tests are shown in squared brackets. For example, the marginal effect for Common Language Group in the pre-AKP era in column 2 is 0.83 (SE = 0.22), and its marginal effect in the AKP era is 0.40 (SE = 0.22). We can reject the null hypothesis of no difference between the two effects at the .05 level.
Results
Table 1 compares Turkish aid before and after the AKP came to power. At first glance, we can make a number of observations. First, several of the interactions with AKP are statistically significant, which implies that the determinants of Turkish aid have changed when the AKP came to power. Second, the estimates for total and economic aid are quite similar, which is not surprising, because economic aid makes up a bigger portion of Turkey’s total aid than does humanitarian aid. To avoid repetition, we will not discuss estimates for total aid in the remainder of the text. Third, changes in economic and humanitarian aid are different. Economic aid, once strongly related to international alignment and ethnic ties, became focused on Turkey’s commercial interests. In contrast, in humanitarian aid, shared religion gained importance.
Marginal Effects for Comparison of Turkish Governments.
This table reports marginal effects of the independent variables. Brackets report p tests for Wald tests of equality between the marginal effects of variables in different eras. Standard errors of the marginal effects are in parentheses. Regression table is in the online appendix. Estimator: Tobit. Dependent variable: Amount of Aid. DV = dependent variable; AKP = Adalet ve Kalkınma Partisi; UN = United Nations; GDP = gross domestic product.
p < .1. **p < .05. ***p < .01.
It is clear from Table 1 that Turkey gives more aid to Turkic nations, but this effect was significantly larger before the AKP came to power. Before 2003, the level of economic aid going to a recipient that is 90 percent Turkic would be on average 110 percent higher than a recipient with no Turkic population. After 2003, this number fell by more than half to 43 percent. This drop in the effect of Turkish identity supports the argument that under the AKP governments, the importance of ethnic ties to other countries has diminished (Saraçoğlu and Demirkol 2015).
What is the effect of religious similarity? AKP has raised the level of aid to Muslim nations, but only humanitarian aid. The effect size is relatively large; under the AKP governments, a 100 percent Muslim recipient receives 50 percent more humanitarian aid than a comparable non-Muslim recipient. One explanation for the AKP’s selective favoritism is the higher visibility of humanitarian crises. If AKP supporters follow crises in Muslim nations closely, their attention provides the government with an opportunity to portray itself as a benefactor to fellow Muslims and gain popularity. However, the public pays less attention to the distribution of economic aid, which reduces the government’s incentives to deviate from commercial interests.
Commercial ties gained and political ties lost importance after the AKP came to power. Before 2003, Turkey did not give significantly more aid to its trade partners, but after 2003, a country that goes from zero imports from Turkey to the sample average is expected to receive 120 percent more economic aid. Note that the size of this increase is comparable with effect of being a coethnic recipient. This finding is robust to using bilateral trade instead of Turkish exports and consistent with the account of Kirişçi and Kaptanoğlu (2011). In contrast, political ties became statistically and substantively insignificant after 2003. This change reflects the AKP’s efforts to engage neighbors, some of which are international pariahs. As Turkey began to give aid to countries that vote differently from the Western bloc, UN Voting Similarity became a less useful predictor of Turkish aid.
Can the 2001 financial crisis that preceded the AKP governments provide a deeper explanation for the growing importance of trade ties? 16 Perhaps after such a severe crisis, a non-AKP government would also use economic aid to increase Turkish exports and trade, which means that government turnover is not critical. This alternative explanation wrongly assumes that all postcrisis governments pursue similar policies. While the AKP sought to diversify and expand Turkey’s trade portfolio after 2003, it never pursued the mercantilist goal of increasing exports and limiting imports. In fact, under AKP, Turkey’s imports have grown faster than its exports (Öniş and Güven 2011). In contrast, often governments that come to power during or soon after an economic crisis follow beggar-thy-neighbor policies. Most possibly in 2003, there were some non-AKP parties that would use aid to increase trade but also others that would cut foreign aid and adopt protectionism. For this reason, we find the claim that any government following a crisis would allocate aid to promote aid unconvincing.
Recipient needs matter for Turkey’s aid allocation. Turkey gives more economic aid to poorer countries and more humanitarian aid to those that suffered deadlier disasters. Infant Mortality is negatively and significantly related to humanitarian aid.
Institutional quality has some effect on Turkish aid. Democracy is not a significant predictor of economic aid, but in the AKP period, Turkey gives more humanitarian aid to more democratic countries. On the contrary, Turkey gives more humanitarian aid to countries with worse human rights records. This negative correlation between aid and human rights could be due to repressive countries having greater humanitarian need that is not captured by other measures.
As expected, Turkey gives more economic aid when Turkish economy is doing better, but humanitarian aid is not affected by the economy or disasters in Turkey. Turkey gives more aid to more populous and geographically closer countries, but the effect of Distance has weakened since 2003. We attribute these changes to the more active foreign policy Turkey has pursued under the AKP.
Sensitivity Analyses
We conduct several sensitivity checks for each type of aid. To conserve space, these analyses are presented in the online appendix. Our results are broadly robust to using the Heckman and OLS estimators. Using alternative versions of our key variables produces similar findings: changing the dependent variable to the percentage of Turkish aid or measuring cultural similarity with a dichotomous variable does not change the results.
We expand our models by including other control variables that capture religious, political, and historical recipient characteristics: whether a Muslim country has a Sunni or Shia-majority, and recipient’s former status as an Ottoman territory (or colony). 17 In addition, we control for the Turkish government’s left–right ideology in a given year to show that our findings are robust to controlling for changes in government other than the AKP’s rise to power. We construct this variable, Donor Government’s L-R Ideology, using a methodology similar to Huber, Ragin, and Stephens (1993). Briefly, for each year we obtained data from the Comparative Party Manifesto Project (Klingemann et al. 2006) on the ideological positions of parties in the government (on a left–right scale) and then weighted each party’s position by its share of cabinet seats.
Including those additional controls does not affect our original findings. We find that under AKP governments, Turkey began to favor former Ottoman territories, which is consistent with the idea that the AKP has pursued a neo-Ottoman foreign policy. Turning to Government’s L-R Ideology, before the AKP, leftist governments in Turkey gave more aid, but after 2003 a government’s left–right ideology ceased to have an impact.
Syria is an outlier among Turkey’s humanitarian aid recipients since 2011, but excluding Syria does not change the results. We also consider that Turkish aid may have evolved across AKP governments. We reran the analyses excluding the years since 2007, which marks the AKP’s second election victory. Our results remain similar in this shorter panel.
We have dealt with endogeneity so far by lagging time-variant variables, and without good instrumental variables, a better solution is difficult to find. However, as Turkey gives on average less than 5 million dollars to a recipient, Turkish aid is unlikely to change recipient characteristics like infant mortality or UN voting patterns. One exception is Imports from Donor. The positive association between Turkish exports and aid in the AKP era could be due to exports leading to more Turkish aid (i.e., Turkey rewards importers with more aid) or aid leading to greater exports (i.e., Turkey uses aid to open up markets). However, regardless of the direction of causality, our basic finding—that economically important countries receive more aid from Turkey after 2003—holds.
Is Turkey Different than Traditional Donors?
In this section, we compare OECD DAC donors to Turkey in the pre-AKP and AKP eras. We show that Turkey and traditional donors have become more similar after 2003 as Turkish economic aid became primarily focused on trade partners and other factors lost their importance. Meanwhile, the growing importance of cultural similarity in Turkey’s humanitarian aid makes it different from traditional donors.
Our research design is similar to the previous section. We run nested Tobit regressions to test for differences between traditional donors and Turkey. 18 The unit of analysis is a donor–recipient–year. The dependent variable is the amount of aid given to a recipient. We again study economic and humanitarian aid separately to capture sector-specific dynamics. As estimates of total and economic aid are sufficiently similar, we do not present models of total aid. We interact our independent variables with a DAC indicator that is coded 1 for the OECD DAC members and 0 for Turkey. Our models include the same set of explanatory variables as before. As we now have several donors besides Turkey, dyadic variables (e.g., Distance) are calculated for the relevant donor–recipient dyad. We measure ethnic/religious similarity between donors and recipients using the percentage of people in a recipient who belong to the largest ethnic/religious group in donor country.
Results
Table 2 presents our findings on the similarities and differences between Turkey and other donors. We start with cultural ties. Both ethnic and religious similarity have an effect on DAC donors. However, the effects are stronger for Turkish aid. Relative to DAC donors, Turkey gave more economic aid to culturally similar recipients before 2003, but now prefers to give them more humanitarian aid. This shift has made Turkish humanitarian aid different from DAC aid.
Marginal Effects for Comparison between Turkey and OECD DAC Members.
This table reports marginal effects of the independent variables. Brackets report p tests for Wald tests of equality between the marginal effects of variables for different donors. Standard errors of the marginal effects are in parentheses. Regression table is in the online appendix. Estimator: Tobit. Dependent Variable: Amount of Aid. OECD DAC = Organisation for Economic Co-operation and Development–Development Assistance Committee; UN = United Nations; GDP = gross domestic product.
p < .1. **p < .05. ***p < .01.
Could higher Turkish aid to Muslims be due to the War on Terror, which has led all donors to engage Muslim countries more after 2003? To test this alternative explanation, we replace Common Religion by three variables that measure the percentage of Christians, Muslims, and followers of other religions in a recipient nation. We continue to find that (1) Turkey began to give more humanitarian aid to Muslim nations after 2003 than before, and (2) in the period after 2003, when giving humanitarian aid, Turkey shows a stronger bias toward Muslim nations than traditional donors do. This result strengthens our claim that the change in Turkish humanitarian aid was primarily due to political change in Turkey.
We now turn to donors’ political and economic interests. In the current era, both Turkey and traditional donors give more aid to their trade partners, but political similarity (as measured by UNGA voting) does not have an effect on either. Political similarity mattered strongly for Turkish aid before 2003, but it lost its importance in the AKP period.
Both traditional donors and Turkey respond to recipient need. They give more aid to countries that have low GDP per capita or suffered deadlier disasters. Turkey seems more responsive to both measures than traditional donors are. However, in some models, Infant Mortality is negatively related to foreign aid, and the effect is even stronger for Turkish aid. This finding may be due to the low capacity of some recipients to spend aid effectively.
More democratic countries get more economic aid from traditional donors and also more humanitarian aid from AKP-era Turkey. Regarding the effect of human rights, both Turkey and traditional donors seem to give more aid to countries with more state repression, and the effect sizes are not statistically different from each other. Overall, OECD DAC donors do not seem to pay more attention to institutional quality than Turkey does, especially after 2003.
As Turkish foreign policy became more ambitious under the AKP, the impact of geographic proximity diminished and Turkey became more similar to other donors. Last, Turkey gives more economic aid to recipients with larger populations.
Sensitivity Analyses
To ensure that our findings are robust, we first rerun our analyses using the Heckman and OLS estimators. Then, we operationalize our key variables in alternative ways. We use a recipient’s share of aid as our dependent variable and dichotomous measures of religious and ethnic similarity. Third, we control for additional political variables including colonial ties and government ideology in donor country. These changes do not affect our main findings.
Next, we relax the assumption that the OECD DAC is a homogeneous group of donors. Following Fuchs and Vadlamannati (2013), we categorize DAC donors into three groups: major powers (the United States, the United Kingdom, France, and Germany), OECD donors with a reputation for giving generous and high-quality aid (Canada, the Netherlands, Denmark, Sweden, and Norway), and the remaining DAC members. 19 We find that most of our findings regarding what separates Turkey from traditional donors hold for each subset, too. One interesting exception is that “generous donors” resemble Turkey more than the other DAC members and do not give more humanitarian aid to their trade partners. As in the previous section, we checked that our results are robust to excluding the post-2007 period.
Conclusion
This paper contributes to the debates on emerging aid donors using detailed data on one important new donor, Turkey. We first showed that Turkish aid policy changed significantly when the Islamist AKP came to power, but different types of aid changed in different ways. Turkey’s economic aid program greatly expanded soon after 2003, but there was not a significant expansion of humanitarian aid until the Syrian War. Relative to previous Turkish governments, AKP gives less weight to Turkey’s international alignments and ethnic ties in allocating economic aid; it instead gives more to Turkey’s trade partners. On the contrary, in humanitarian aid, ethnic and, for the first time, religious ties became important under AKP governments. AKP-era Turkey’s focus on trade has made it more similar to OECD DAC donors in economic aid, but the growing importance of ethnic and religious identity in humanitarian aid allocation has made Turkey different than other donors.
Our results are based on data from one donor country, which limits their generalizability. More research is necessary before we can say with some confidence whether domestic political change in general impacts emerging donors’ aid patterns. We can offer at least two conditions under which domestic political change will not matter as much as it did in the Turkish case. First, in countries where political actors are not polarized and hold similar preferences, government turnover will matter less. The election of AKP in 2002 was a significant event, because it brought to power new cadres who wanted to take Turkey in a new direction. If an incoming government agrees with its predecessor’s foreign policy goals, then we will probably not observe any differences between them. Second, if a new government is ideologically fractured, then it may not be able to change policy significantly as internal factions block each other’s initiatives. AKP has always ruled as a single party; if it had been part of a coalition, it would probably have less control over foreign policy and make less of a difference. In short, we should only expect government turnover to matter in cases where the incoming government has both the willingness and the ability to make changes to the foreign aid program.
Despite its limitations, our study can offer some suggestions for the broader literature on foreign aid. Our results suggest that we can improve our understanding of emerging donors by exploring how their motives change over time with government turnover. In addition, it can be fruitful to disaggregate a country’s total aid into different types and study whether emerging donor motives vary by aid types. One promising starting point would be the work of Milner and Tingley (2015) who recommend researchers to consider the different domestic political consequences of different policy tools. From this perspective, we can think of foreign aid as a bundle of policy tools (rather than a single tool) that a government can use to advance its domestic and international goals.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
