Development Aid and Exchange Rate Movements

(1)

University of Groningen and Georg-August Univesität Göttingen

Master Thesis

Development Aid and Exchange Rate Movements

Can aid function as a curse?

June 2015

Author:

Lauren van Straalen

Student number: S1891766 (Groningen) Student number: 11403932 (Göttingen) Email: l.c.van.straalen@student.rug.nl

Supervisor:

Dr. G.J. De Vries

Faculty of Economics and Business, University of Groningen

Co-Assessor:

Dr. Felicitas Nowak-Lehmann Danzinger

(2)

ii Abstract

This thesis investigates if development aid can result in economic effects associated with the “Dutch disease”, such as an appreciation of the real effective exchange rate and a loss of competitiveness. Where the aid effectiveness literature predicts that aid inflows result in an appreciation of the exchange, I find that aid inflows tend to depreciate the real effective exchange rate. Additionally, I find that aid inflows tend to increase tradable-non-tradable output ratio, something that is also associated with a gain in competitiveness. A closer look at the sectorial output composition shows that aid increases agriculture and services output, but that it decreases output in the manufacturing sector. This indicates that the relation between aid and the “Dutch disease” is less straight forward than suggested by the literature.

(3)

iii

The origins of aid can be traced back to the beginning of the nineteenth century, where it mostly occurred in the form of food transfers. Although food transfers are still an important source of aid, they comprise only a small percentage of total aid flows. In 2013, almost 1.2 billion US Dollar was spent on food aid, while total aid flows equaled 135 billion US Dollar (OECD, Detailed Aid Statistics). Money transfers for development assistance began to emerge after the Second World War, when the United States provided European countries with large sums of money for the reconstruction of their economy (Szirmai, 2005). After a period of reconstruction, the Commonwealth countries were the first who signed a binding commitment to provide development contributions to former British colonies. This example was later followed by other European nations, which also put the focus on their own former colonies. By the 1960’s multilateral institutions emerged as major actors in the development programs. When the external conditions of the 1970’s, such as the oil crisis and the recession in Western countries, led to a deterioration of government’s budgets, multilateral institutions such as the World Bank and the International Monetary Fund (IMF) began to expand their influence with the Structural Adjustment Programs (SAP) in the beginning of 1980’s (Szirmai, 2005).

(7)

2

Official Development Assistance (ODA) flows have been steadily rising since 1960; increasing from 4,7 billion current USD in 1960 to 135,2 billion current USD in 2014, which sums to around 2,6 trillion USD over 50 years (Figure 1). The recent financial crisis and European debt crisis have put pressure on government budgets. The Netherlands was for example always one of the few countries in the Organization for Economic Cooperation and Development (OECD) that pledged 0,7 % percent of its gross national income (GNI) to developing countries, but was unable to reach this goal in 2013, for the first time since 1974. In 2014, only Sweden, Luxembourg, Norway, Denmark and the United Kingdom reached the 0,7% target set by the United Nations (OECD, 2015).

As governments face more budget constraints, the debate about the effectiveness of aid has received a lot of attention again. While influential people like Bill Gates try to reason with governments to stick to the 0.7% of GNI per year, others are not convinced about the effectiveness of development assistance. Although aid most likely has a positive influence on socio-economic indicators such as education and health, its influence on economic growth is not clear. One would expect that aid would increase the state of welfare in the recipient country since a country receives more income to invest, educate and provide better health facilities. When investment and education improve, one would also expect that aid is growth enhancing. However, if we look at the unconditional correlation between aid flows and economic growth in Figure 2, it seems that aid is growth reducing rather than growth enhancing.

(8)

3

Figure 2. Apparent relationship Aid inflows and Economic Growth

(9)

4

In this thesis, I will investigate whether aid inflows can result in the “Dutch disease”, and more specifically, whether large aid inflows result in an appreciation of the real exchange rate and/or a loss of competitiveness of the manufacturing sector. This effect is tested by a regression that has the real effective exchange rate and the tradable-non-tradable ratio as a dependent variable. For several reasons this investigation might prove interesting. First of all, not many scholars have empirically investigated this mechanism, although this effect was identified as a potential negative influence. Among the few who do investigate this relation, Sackey (2001) runs his investigation only in Ghana, Nyoni (1998) only in Tanzania and Rajan and Subramanian (2011) limited their investigation to the 1980’s and 1990’s. On top of that, Rajan and Subramanian (2011) only include countries that receive aid inflows greater than 1% of GDP, which results in a sample of only 32 countries for the 1980’s, and 15 for the 1990’s. In this thesis, I will use an initial sample of 134 developing countries from the years 1970-2013, which could provide a more comprehensive result for a possible “Dutch disease” effect. Secondly, when aid inflows have a negative influence on the competitiveness of a country, it may provide an explanation for the lack of evidence of a positive relation between aid and economic growth. This thesis may therefore provide some answers to the aid effectiveness debate.

This thesis is structured as follows; in the first part, an overview of the aid effectiveness literature since the 1960’s is provided. I focus on the article of Doucouliagos and Paldam (2009) who review previous results in the literature and identify three main groups of study. Additionally, they identify a fourth channel that might provide an explanation for the unobserved positive effect of aid. For completeness, I briefly discuss the effects of aid on socio-economic indicators such as health and education, and their effect on economic performance1. The second part of the thesis explains the methodology and elaborates on the variables and data that are used. The third part describes the data and discusses potential problems associated with regressions. The fourth section gives the regression results. The last section

(10)

5

discusses the limitations of the model and provides some recommendations for further research. Lastly, I provide a short conclusion of my findings and discuss how these contribute to the literature.

2. Literature Review

2.1 The Aid Effectiveness Literature

Although there has been a great deal of research on the effectiveness of aid, it is hard to make one simple statement about the economic consequences. Some studies find a positive effect of aid on economic growth (Lof et al. 2014; Hanson and Tarp, 2001), others find no effect (Boone, 1996; Rajan and Subramanian, 2007) and some even find a negative effect (Nowak-Lehman et al. 2012; Moyo, 2009). One observation that can be made is that both positive and negative effects are marginal, and Doucouliagos and Paldam (2009) find that a correlation of zero is reached between aid inflows and economic growth as the number of observations and years increases. In “The aid

effectiveness literature: The sad results of 40 years of research”, Doucouliagos and

(11)

6

(2009). This additional influence of aid on economic growth runs along socio-economic indicators such as health and education, and influences socio-economic growth through higher productivity and human capital accumulation.

Figure 3. Expected effects of aid on economic growth.

(12)

7

effect on economic growth; besides the fact that investment in health saves human lives, it also provides people with more opportunities for the future, so that there might be a way out of poverty. Health problems create both direct (treatment and medicine) and indirect costs (human capital and income loss). Antle and Pingali (1994) show that health affects productivity and that a loss in productivity results in lower income generating activities, and thus higher levels of poverty. An important consequence of a loss in income is that it results in human capital deprivation. For example, Bell et al. (2006) investigate the effects of HIV/AIDS on economic development in South Africa, and find that as most HIV/AIDS deaths fall into the 15-49-age category they leave behind orphaned children who lack the means to finish education.

Since the creation of many global funds to fight communicable diseases, such as UNAIDS and the Bill and Melinda Gates Foundation, malaria prevalence has been reduced with 26% (WHO, 2014); HIV/Aids infections have fallen with 38% since 2001. The number of people that have access to antiretroviral treatment rose from 5,2 million in 2009 to 12,9 million in 2013 (UNAIDS, 2014). In my opinion, development aid clearly has a positive effect on socio-economic indicators such as education and health. It provides medicine for the sick, and is able to provide children with a brighter perspective for the future.

Although aid can have a positive influence on socio-economic indicators, what is the influence of large aid flows on economic growth? One would expect that a country that has more aid inflows is able to invest more, educate more students and provide food aid to the ones that are in need, which results in higher levels of welfare. However, this picture does not emerge in the literature.

(13)

8

consumption, they also lack the means to save part of their income for investment. This theory was therefore the basis of the first aid flows, such that aid could function as the bridge between savings and investment, resulting in increased investment and capital accumulation in the receiving country. Higher capital accumulation should lead to an increase in income, and therefore also to a rise in domestic savings. Weiskopf (1972) however, argues that domestic savings behavior is independent of foreign capital inflow, and that aid therefore does not result in an increase in investment. Griffin and Enos (1970) argue along the same line, and find that foreign savings tend to displace rather than to increase domestic savings. An additional inflow of capital will more likely result in an increase in consumption than in an increase in domestic savings. Griffin and Enos (1970) find that an extra dollar of aid is associated with a 75 cents increase in consumption and a 25 cents increase in investment, Weiskopf (1972) finds that foreign capital inflows have a negative effect on domestic saving in developing countries and Boone (1996) obtains no evidence that aid significantly increases investment. These authors argue that aid inflows reduce the need to finance additional consumption by an increase in taxes, and that in extreme cases aid can become a substitute for tax reforms. The neoclassical theory of capital accumulation (and therefore economic growth) is thus not very applicable to developing countries, and foreign capital inflows can lead to a crowding out of domestic savings via increased consumption. Although there are various scholars who show a marginally positive effect of aid on investment (Grifin and Enos, 1970; Weiskopf 1972; Boone, 1996) there are also studies that find a stronger positive relation between the two variables (Hanson and Tarp, 2001; Papenek, 1973). Papenek (1973) disaggregates foreign inflows into their principle components, and finds that aid does not affect domestic savings, but that it does fill the savings gap (unlike foreign private flows and other flows). The expected relation that is denoted in figure 3, thus cannot be uniformly confirmed in the literature.

(14)

9

of all these studies, which show both positive and negative outcomes of aid on economic growth, and they conclude that the variation between the results is falling over time and with sample size. Moreover, the average effect is steadily decreasing and converging to zero (it is now on average small and positive with a 0.02% increase in growth with a 1% increase in aid). The extent to which aid is effective, depends highly on the choice of estimator and the choice of control variables (Hanson and Tarp, 2001). Hanson and Tarp (2001) combine the outcomes from studies that found decreasing marginal returns to aid with the ‘conditionality’ literature (described below). When they controlled for investment and human capital, they found that aid did not have a positive effect on economic growth, and was characterized by decreasing marginal returns. Rajan and Subramanian (2008) use a cross sectional panel data for the period 1960-2000 to estimate the effects of aid on economic growth. After controlling for several variables such as initial income, ethnic fractionalization, trade policy, geography and institutional quality, they do not find a positive (or negative) relationship between aid and economic performance.

(15)

10

Collier and Dollar (2002) suggest that the marginal effect of aid depends on the quality of policies, but they also add that the amount of aid that a country is receiving has a significant influence too.

However, this body of literature is also not very consistent. Easterly, Levine and Roodman (2003) found that the model of Burnside and Dollar (1997) loses its significant results when additional periods and countries are added to the data. They therefore conclude that the results of Burnside and Dollar (1997) are not as sound as they present them to be, and that one cannot claim that good policy environments increase the effectiveness of aid. Hansen and Tarp (2001) also make use of a larger sample size than Burnside and Dollar (1997) and state that aid is effective regardless of ‘good’ policies. Lensink and White (2001) in turn, test whether aid is only effective in good policy environments and whether it starts to have negative returns above a certain threshold. After controlling for aid/GDP, aid/GDP squared and other variables such as secondary school enrolment, debt-to-GDP ratio, and initial income, they do

not find evidence for the good policy model, but they do find that aid has negative

returns above a certain threshold.

(16)

11 2.2 Resource Curse Literature

Although the link of resource abundance and slow economic growth emerged in the 1980’s, Richard Auty first named the term resource curse in 1994 (Auty, 1994). The apparent negative relation between natural resource abundance and poor economic performance has led to a large body of literature. Even after controlling for a number of variables such as initial income, investment, trade-openness and geography, several authors have provided strong evidence for the existence of such a curse (Sachs and Warner, 1995; Collier and Gorderis, 2008; Humphreys et al, 2007). Natural resource abundance leads to slower economic growth (Sachs and Warner, 1995, 2001) and Collier and Goderis (2008) have associated a boom in commodity prices with a short-run increase in GDP growth, but with a negative effect on long term GDP growth. Natural resources can work as a “curse” on economic performance via exchange rate appreciation, an increase in violent conflicts and/or government malfunctioning. The first influence of the resource curse, also known as “Dutch disease”, was introduced by the magazine The Economist in an article describing the decline of the manufacturing sector in the Netherlands in relation to the discovery of a large natural gas field in Groningen. Natural resource abundance can cause a decline of the tradable sector via two ways: First, it will lead to a focus on the export of these resources, and an increase in these exports will lead to a real exchange rate appreciation, as there is a large demand for the exporter’s currency. This increases the price of domestic manufacturing products that are exported and thus leads to a loss in competitiveness of the tradable goods sector. Second, a focus on natural resource exports will draw more capital and labor into this sector, which would otherwise have been employed in the manufacturing sector (Sachs and Warner, 1995).

(17)

12

be drawn into this sector compared to the tradable sector (White, 1992). Higher labor demand will result in higher wages, and as part of these higher wages is spent on the non-traded sector, this will increase the price of non-traded goods. In turn, higher wages in the non-tradable sector will also exercise an upward pressure on wages in the tradable sector, reducing competitiveness and resulting in a decline in exports (Rajan and Subramanian, 2011). Rajan and Subramanian (2011) examine the effects of aid on the growth of the manufacturing sector. They find that aid has adverse effects on the recipient’s competitiveness, which is reflected in a lower growth of the export sector of an aid recipient country relative to a donor country. This effect arises via an appreciation of the exchange rate, as the authors argue that aid inflows can lead to spending pattern that is biased towards the non-tradable sector. However, aid can also have a positive effect on a country’s competitiveness, when aid programs are focused more on technology transfers, capital accumulation and trade reforms.

(18)

13

access for rebels to use these natural resources as a financing source for their rebellion. Another part of the conflict research is the effect of resource abundance on the duration of the conflict. Although some studies suggest that natural resources are correlated only with a subset of conflicts, Doyle and Sambanis (2000) conducted a study of 124 wars/conflicts and found that primary commodity exports have a negative effect on peace building efforts. Especially oil seems to be correlated with the likelihood of conflict and civil wars (Collier and Hoeffler, 2002; de Soysa, 2002). Collier and Hoeffler (1998) argue that most conflicts are either driven by greed or grievance (problems with the government) and that the control over natural resources is an essential source of financing rebellions. Other explanations could be that resource exporters are more vulnerable to trade shocks due to price volatility, which can lead to unrest in times of bad shocks (Humphreys, 2005) or that natural resource abundance leads to weak states, which in turn increases the chance on a civil war (Fearon and Laitin, 2003).

In my opinion, the relation between natural resource abundance and conflicts is not something that we can observe with aid flows. Although aid flows also exist of free rents, it is not easy to gain control over these resources as a rebellion group. Aid flows are often ceased or postponed in times of conflict, and therefore are not likely play a role in the decision for rebels to start a conflict. De Ree and Nillesen (2009), find that aid inflows reduce the duration of civil conflict, but they fail to find any significant relation between aid inflows and the beginning of a conflict.

The argument that resource abundance leads to weak institutions is a last channel that is mentioned in the literature as an indication of the resource curse. The weak institutions manifest themselves by high levels of corruption and authoritarian governments. When governments receive rents from natural resource exports, they are less dependent on taxes, and therefore tend to tax the population less heavily. As the population does not contribute much to the government budget, this results in less pressure and lobbying, and thus less incentive for democracy (Moyo, 2009; Ross, 2004a).

(19)

14

structural changes (the structural adjustment programs in the 1980’s). However, it can also erode the quality of the government by creating dependency, encouraging rent-seeking, increasing corruption and weakening accountability (Moyo, 2009; Knack, 2001). Although many authors found a link between aid effectiveness and policy environment (Boone, 1996; Burnside and Dollar, 1997), they failed to make conclusions about the reversed causality; they were unable to prove that aid can influence the functioning of governments.

In sum, because aid inflows are unlikely to influence the number of conflicts in a country and as several authors already investigated the relationship between aid and government effectiveness, I will focus on the “Dutch disease”.

3. Methodology

To investigate whether an aid recipient country faces the threat of the “Dutch Disease”, I use the economic framework from Lartey, Mandelman and Acosta (2008). In their paper “Remittances, Exchange Rate Regimes, and the Dutch Disease: A Panel

Data Analysis”, they investigate via a dynamic panel analysis whether worker’s

remittances have an influence on the exchange rate and competitive position of a country. The authors find that remittance inflows tend to increase the share of services in total output, and that it reduces manufacturing output, which is an indication of the “Dutch disease”. Furthermore, a spending effect arises that increases the relative price of non-tradable goods and results in an exchange rate appreciation. This effect occurs as total aggregate demand increases with remittances transfers, and given that the prices of tradable goods are determined within the world market, the price of non-tradable goods rises relative to non-traded goods. Burstein et al. (2006) found that price movements of non-tradable goods relative to the price of traded goods account for 50% of exchange rate fluctuations. Therefore an increase in the price of non-tradable goods will result in an exchange rate appreciation. Moreover, the extra demand for non-tradable goods, increases relative wages, initiating a movement of resources from the tradable sector into the non-tradable sector.

(20)

15

sector, via an exchange rate appreciation and a loss in international competitiveness. There are several similarities between remittances and aid flows. Both capital inflows can be seen as free rents, where remittances are free rents for individuals, aid flows pose free rents for the public sector. The main problem with free rents is that it does not create any incentives for the one receiving the capital inflow, which can result in unproductive spending and investments (Gelp et al, 1991). Additionally, both aid and remittances have to be converted into local currency before governments or residents are able to spend the money, which results in an increase in the demand of the local currency.

Because of the similarities between aid inflows and remittances mentioned above, I use the framework proposed by Lartey et al. (2008) and add aid inflows to the independent variables. Lartey et al. (2008) use a generalized method of moments estimator (GMM), as the GMM model is tailored to deal with potential endogeneity and provides additional information about unknown variables. On top of that, it is robust to distributional assumption, and more specifically with heteroskedasticity. The GMM estimator uses lagged values as explanatory variables, thus assuming that all variables are endogenous. Despite all its advantages, the GMM model also has some drawbacks. For example, in the case of autocorrelation, internal lagged values might not be a useful instrument, because endogeneity will not disappear in that case. Additionally, the GMM is very complex and sensitive to the assumptions used, which can easily lead to misinterpretation of findings and it is hard to determine whether the instruments are valid (Roodman, 2009). I chose a fixed effects model after conducting a Hausman test, as it controls for differences between groups; each country may have individual characteristics (individual heterogeneity) that can influence the independent variables (such as culture, climate, etc.). The advantage of a fixed effects model is that it removes these time-invariant effects, such that the net effect can be assessed I use a fixed effects panel model with 134 developing countries for the years 1970-2013 in order to assess the effect of aid on the competitiveness of a country. The equation for the fixed effects model is

(21)

16

where 𝑦_𝑖𝑡 represents the Real Effective Exchange Rate (REER) or the Tradable-Non-Tradable (TNT) ratio for country i in period t: 𝑎𝑖𝑡 denotes aid inflows and is assumed to be exogenous. 𝑥𝑖𝑡 represents a set of control variables that can also influence the dependent variable (assumed to be exogenous), 𝛾𝑖 denotes country specific characteristics that do not vary with time (country fixed effect), 𝛿_𝑖 controls for time specific effects and 𝜖𝑖𝑡 is the error term that is assumed to be identically and independently distributed random variables.

I will run two separate regressions, first testing the influence of aid inflows on the real effective exchange rate and second, the effect of aid inflows on the tradable-non-tradable output ratio. Following the resource curse literature, my first hypothesis is that large aid inflows will lead to increased demand of the local currency, and therefore will lead to an increase of the exchange rate. Such an appreciation will lead to a decrease in exports as domestic products will become more expensive for foreign consumers, and will have a negative effect on economic performance. My second hypothesis is that aid inflows will result in increased spending in the non-tradable sector, and therefore increase output in the non-tradable sector rather than in the tradable sector, resulting in a decrease of manufacturing output.

Hypothesis 1 𝑅𝐸𝐸𝑅 = 𝑓(𝑎𝑖𝑑, 𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠, 𝐿𝑁( 𝐺𝐷𝑃 𝐶𝑎𝑝𝑖𝑡𝑎), 𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ, 𝑀2, 𝑇𝑜𝑇, 𝑡𝑟𝑎𝑑𝑒 𝑜𝑝𝑒𝑛𝑛𝑒𝑠𝑠) + + + - /+ + + + Hypothesis 2 𝑇𝑁𝑇 𝑅𝑎𝑡𝑖𝑜 = 𝑓(𝑎𝑖𝑑, 𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠, 𝐿𝑁( 𝐺𝐷𝑃 𝐶𝑎𝑝𝑖𝑡𝑎), 𝐺𝐷𝑃 𝑔𝑟𝑜𝑤𝑡ℎ, 𝑀2, 𝑇𝑜𝑇, 𝑡𝑟𝑎𝑑𝑒 𝑜𝑝𝑒𝑛𝑛𝑒𝑠𝑠) - - - - /+ - + +

(22)

non-17

tradable sector relative to the tradable sector, and thus a loss in the competitiveness of the manufacturing sector.

3.1 Dependent variables

The REER represents the nominal effective exchange rate adjusted for changes in consumer prices in both the home country and its trading partners, and valuates the performance of a country’s currency relative to that of others. As the REER is calculated from a base year, it also provides a comparison within a country, and thus can be used as an evaluation for a country’s changing international competitiveness. The REER data is taken from the study of Zsolt (2014), who calculates the REER as 𝑅𝐸𝐸𝑅𝑡 =𝑁𝐸𝐸𝑅_𝐶𝑃𝐼 𝑡∗𝐶𝑃𝐼𝑡

𝑡𝑓𝑜𝑟𝑒𝑖𝑔𝑛

. 𝐶𝑃𝐼𝑡 represents the consumer price index of the home country, the NEER is the nominal effective exchange rate of the home and represents the average of the nominal bilateral exchange rate between the home country and its trading partner i, measured as the foreign currency price of one unit of domestic currency2. 𝐶𝑃𝐼_𝑡𝑓𝑜𝑟𝑒𝑖𝑔𝑛 is the geometrically weighted average of the CPI indices of 67 trading partners. As the NEER is denoted in the foreign currency price of one unit of domestic currency, the NEER would drop with inflation in the home country, which signals a currency depreciation. However, as the REER is adjusted for inflation, a depreciation in the nominal effective exchange rate due to other factors than inflation, this is in turn reflected by a depreciation in the real effective exchange rate. When a currency is stable over time, this is reflected in a stable real effective exchange rate. An increase in the REER would indicate an appreciation of the local currency, signaling to a loss of competitiveness as the real effective exchange rate of the country is increasing compared to the world average (exports become more expensive). Lartey et al. (2008) defines the tradable-to-non-tradable ratio as the sum of agricultural and manufacturing output divided by services output. As a decrease in

2_{Let’s make a simple thought experiment, so that we can have a clear picture of the REER’s}

(23)

18

this ratio signals that output in the service sector increases relative to the agricultural and manufactures sector.

3.2 Independent Variables

Aid inflows are measured as net Official Development Assistance (ODA) received as percentage of gross national income (GNI), and therefore give a good indication how much of total GNI countries receive in aid, and can be seen as an indicator of aid dependence. As all other variables are taken as percentage of GDP, I will transform 𝑂𝐷𝐴

𝐺𝑁𝐼 by multiplying it with 𝐺𝑁𝐼

𝐺𝐷𝑃, so that the variable that is used in the regression is also denoted as a fraction of GDP. I use bilateral aid inflows (% of GDP) as an extra control variable. Whereas ODA includes both governmental aid as well as aid flows from international organizations, bilateral aid inflows only include governmental aid flows.

(24)

19

country might pursue a strategy of currency appreciation. Additionally, fast growing countries might be more likely to depreciate their currencies, so that they are able to keep exporting their products (correct for the large demand for their currency, see the example of China). The M2 variable is included as the quantity of money that is available in the short run; this influences the price level of an economy and serves as a proxy for domestic expenditures. Terms of trade are incorporated as it affects the imports and exports of a country, and therefore the real effective exchange rate. When a country faces an increase in the price of its imports or a decrease in the price of its export, its terms of trade deteriorate. Negative shocks to a country’s terms of trade would result in a contraction in the export sector and therefore would depreciate a country’s currency (less demand for the domestic currency). The variable trade openness serves as a proxy for government policies, as a less restrictive trade environment would result in more imports and exports. Trade openness thus investigates how trade policies affect the price level of the economy (the effect on non-tradable goods), and therefore the real exchange rate. Remittances are included as an additional control variable, as this is the main variable of interest in the study of Lartey et al. (2008) and because Lartey et al. (2008) and Amuedo-Dorantes and Pozo (2004) found that these inflows also have an effect on the exchange rate and the tradable-non-tradable output ratio.

4. Data

(25)

20

The REER data is taken from the most complete database about real effective exchange rates, calculated by Zsolt Darvas (2014), who computed this index for 178 countries3.

Table 1. Composition of the Sectors

United Nations Category ISIC Division Sector Agriculture, Forestry,

Fishing and Hunting A-B Primary

Mining, Manufacturing and Utilities

C-E Primary/

Secondary

Manufacturing D Secondary

Wholesale, Retail Trade, Restaurants and Hotels

G-H Tertiary

Construction F Tertiary

Transport, Storage and Communication

I Tertiary

Other Activities J Tertiary

Mining* C,E Primary

* Calculated as Mining, Manufacturing and Utilities minus Manufacturing.

The World Bank Development Indicators also have extended data availability, and there I obtained the data about aid inflows, primary school enrollment, remittances and the M2. All data is in constant 2005 local currency units or in percentage of GDP, such that inflation is accounted for4. The data observations include 134 developing countries5 for the period 1970-2013. I only take into account developing countries as they receive aid inflows on a steady basis. As there was no data available on aid inflows for Puerto Rico and Myanmar, I left these countries out of my analysis. Some observations are missing, especially for the former Soviet countries, which only have data from 1990 onwards. On top of that, there were no values available for Sudan and South Sudan in the United Nations Statistics Division, but they only provided values for former Sudan.

3_{Although the database from Zsolt (2014) is the most complete available, it misses values for Cuba,} Micronesia, Kiribati, Kosovo, Marshall Islands, Montenegro, Palau, Solomon Islands, Somalia, Timor-Leste, Tuvalu and Zimbabwe

4_{Additional information about the variables can be found in Appendix 1. Appendix 1 describes how} the variables are calculated and from which database they come from. Moreover, a description of outliers is provided.

(26)

21 4.1 Data Analysis and outliers

Outliers are observations that differ substantially from the other observations, and they can distort the coefficients of the regression results when they are not removed from the sample. Outliers can be the result of model specifications or errors in the data, and exclusion of these outliers will provide a better estimation of the regression. Below the outliers in the dataset are examined, and although there is missing and incomplete data in the sample, for some outliers an explanation can be provided in terms of economic of political events.

Table 2. Summary Statistics Initial Sample

(1) (2) (3) (4) (5)

VARIABLES N Mean Standard

Deviation Min Max REER 4,836 153.4 829.4 0.381 56,666 Tradable-Non-Tradable Ratio 5,494 1.051 1.417 0.0554 26.84 Aid Inflows (% of GDP) 4,520 8.670 12.05 -2.321 207.2

Bilateral Aid Inflows (% of GDP)

5,100 6.445 12.75 -2.766 473.8

Remittances (% of GDP)

3,492 4.948 9.119 3.90e-05 106.4

GDP/Capita 5,494 961,464 3.032e+06 56.14 3.220e+07

Ln[GDP/Capita] 5,494 10.85 2.563 4.028 17.29 GDP Growth 5,360 3.871 7.779 -66.12 130.8 M2 4,544 41.63 318.0 -99.79 12,513 Terms of Trade 4,381 115.9 162.4 1.780 3,984 Trade Openness 5,491 72.46 42.37 0.0268 354.1 Investment 5,453 35.75 147.1 -8.235 4,765 Government Expenditures 5,453 22.47 63.10 0.0113 1,407 Primary School Enrollment 4,462 94.75 27.30 7.863 222.0 Agriculture (% of GDP) 5,494 19.85 13.99 1.138 85.51 Primary Sector (% of GDP) 5,494 31.78 17.96 2.412 99.59 Secondary Sector (% of GDP) 5,459 11.76 7.477 0 73.67 Tertiary Sector (% of GDP 5,494 56.25 16.66 3.606 94.77

(27)

22

from the United Nations Statistics Division, the number of observations differs among the variables. There are a few things that can be noticed here. The highest values of the real effective exchange rate are obtained by Nicaragua for the period 1982-1990. Compared to 2005, Nicaragua had a large loss of competitiveness in this period, which is not strange as the country was during that time characterized by large rates of inflation. Other countries with large values for the real effective exchange rate are Uzbekistan, Ghana and Guinea, but all for the period before 2005. Countries with low REER values were mostly former Soviet countries, which obtained low values for the beginning of the 1990’s. Uganda also has very low values for the REER at the start of the 1970’s.

Countries with a very high tradable-non-tradable output ratio are countries that typically depend very much on tradable goods, and more specifically, on primary products. Among these countries are Equatorial-Guinea, Iraq, Libya, Timor-Leste and Afghanistan. Countries where services dominate economic activity are Seychelles, Palau, Barbados, Marshall Islands and the Maldives, which is not strange if you know that these countries are tropical islands, where tourism is the main source of income. Aid inflows and bilateral aid inflows can be negative as developing countries might also provide aid to others.

The large values for aid inflows are obtained by Tuvalu, who received aid inflows as much as 473,8% of GDP, when the United Kingdom, Australia and New Zealand set up the Tuvalu Trust Fund in 1987. In 2004 Tuvalu again received large aid inflows, but this time from the United States, who provided the country with 700 million dollars of foreign aid. Lesotho is the country that receives most of the remittances, ranging from 20%-106% of domestic GDP.

The values for GDP Growth might also look extreme at first sight. However, these large fluctuations can often be explained. For example, Iraq and Libya faced large contractions in GDP growth during the Iraq war and the Libyan Civil war. Surges in economic growth can often be explained by political stability, the discovery of natural resources and more surprisingly, war activity. For example, Timor-Leste had a growth in GDP in 2004 of 130%, after large oil fields were found in the country (Cotton, 2005).

(28)

23

period where money growth was explosive, provides an explanation for these large values. Iraq, Turkmenistan, Panama and Armenia are among the countries that show the highest value for trade openness, and Togo, Somalia, Morocco, Democratic Republic of Congo and Sudan show the lowest values.

Countries with the lowest values of terms of trade are Equatorial Guinea, Guinea-Bissau, Cabo Verde, Venezuela and Nigeria, but as all values were for the period before 2005, it means that their export sector has been increasing, as their terms of trade improved over the years. Eritrea also has very high values for terms of trade in the years 2011, 2012 and 2013, which also indicates that their exports have increased. Countries with high values, who experienced a contraction in their export sector were Albania, Bangladesh, Burundi, Malawi, Niger and Kazakhstan. These countries had very high values of terms of trade for the period before 2005, indicating that their terms of trade deteriorated.

Iraq had the highest investment over the whole period, ranging from 4765% to 762% of GDP, followed by Cote d’Ivore with values ranging from 196% to 85% of GDP. Countries with negative investment are Nicaragua in 1978 with a divestment of 8% and Azerbaijan in 1992 with divestment of 0.42%.

Iraq has the largest government expenditures relative to GDP over the whole period, ranging from 283% of GDP to 1407 % of GDP. Tuvalu follows second with high government expenditures for the whole period, ranging from 50% of GDP to a maximum 147% of GDP in 2000. Also Timor-Leste reaches high government expenditures for the years 1992-2003, reaching values as high as 137% of GDP.

The maximum value for primary school enrollment ratio exceeds the 100, because I used the gross enrollment ratio. This implies that they count all the children who are enrolled, and they do not take into account age. The ratio therefore might exceed 100%, as older children who actually belonged in another class are also taken into account. I chose to use the gross enrollment ratio as it had much more observations compared to the net enrollment ratio.

(29)

24

from the sample, so that regression results are not influenced by the outliers. The resulting variables are described in Table 3.

Table 3. Summary Statistics Without Outliers

(1) (2) (3) (4) (5)

VARIABLES N Mean Standard

Deviation Min Max REER 4,222 132.1 68.09 32.62 585.1 Tradable-Non-Tradable Ratio 4,686 0.915 0.699 0.108 5.820 Aid Inflows (% of GDP) 3,943 7.869 9.602 -0.000516 54.98

Bilateral Aid Inflows (% of GDP)

4,365 5.475 7.367 -0.00550 49.72

Remittances (% of GDP)

3,134 4.497 6.343 0.00305 42.33

GDP/Capita 4,686 906,585 2.912e+06 56.14 3.046e+07

GDP Growth 4,576 3.959 5.025 -17.60 24.30 Ln(GDP/Capita) 4,686 10.80 2.501 4.028 17.23 M2 4,003 21.56 25.64 -15.80 353.5 Trade Openness 4,683 70.90 38.40 2.022 190.3 Terms of Trade 3,816 106.1 37.20 19.27 316.7 Investment (% of GDP) 4,645 23.69 12.21 1.186 129.6 Government Expenditures (% of GDP) 4,645 16.71 10.34 0.670 147.1 Primary School Enrollment 3,867 95.26 27.10 7.863 215.4 Agriculture (% of GDP) 4,686 19.76 13.51 1.138 76.64 Primary Sector (% of GDP) 4,686 30.69 16.46 2.687 90.53 Secondary Sector (% of GDP) 4,651 12.09 7.116 0 50.89 Tertiary Sector (% of GDP) 4,686 57.10 15.32 10.97 93.04

4.2 Endogeneity Issues, Multicollinearity and Heteroskedasticity

(30)

25

that a country receives, or when the tradable-non-tradable output ratio is not only dependent on aid, but also determines the level of aid for the incoming country. Reversed causality can occur with economic growth as dependent variable, as countries that grow slow might be more likely to get development aid. However, I do not think it is likely that exchange rate movements or sectorial composition of the recipient country determine the level of aid, as these factors do not. Regardless, it is unlikely that the distribution of aid is completely random. As mentioned above, aid is likely to depend on economic growth of a country, sudden events such as natural disasters and the relation between the recipient and the donor country.

Multicollinearity occurs when two explanatory variables are correlated with each other, and thus do not explain separately an effect that is tested in the regression. When two explanatory variables are highly correlated, one might be left out of the regression, as it does not pose any extra influence on the dependent variable. To look for multicollinearity, a correlation table is presented in Appendix 4, from which we can see that the GDP/Capita and the LN(GDP/Capita) are highly correlated, which is not a problem as these variables are not used simultaneously in the regressions. Another large correlation coefficient is observed between aid and bilateral aid inflows, but again these correlations are not used simultaneously as they serve as mere alternative variables for each other. Something that also can be noticed from the correlation table is that government expenditures are correlated with both indicators of net aid inflows, which may indicate that government expenditures in the countries under study are dependent of foreign assistance. The variation inflation test (VIF) is another way to measure multicollinearity in the model. When dummies are excluded, the value for the VIF is very low (1.19) but when country and year dummies are added to the regression, the VIF rises to values ranging from 4.34-5.70. This is still well below the VIF cutoff value of 10 (Craney and Surley, 2002), indicating that there is no multicollinearity in the model.

(31)

26

be seen in the Appendix 4, adding the assumption of clusters and robust errors, does not change the data in a significant manner; all coefficients remain the same, but for some variables the standard errors change. This indicates that the regressions are not prone to biases caused by heteroskedasticity.

5. Results

5.1 Aid inflows and the Real Effective Exchange Rate

(32)

27

Table 4: Regression results, REER as dependent variable

(1) (2) (3) (4) (5) (6)

VARIABLES FE FE FE FE FE FE

ODA Inflows (% of GDP) -1.243*** -1.261*** -1.233*** -1.252*** -1.248*** (0.195) (0.192) (0.196) (0.196) (0.197)

Bilateral Aid Inflows (% of GDP) -1.979***

(0.304) Remittances (% of GDP) 1.009*** 1.075*** 1.104*** 1.010*** 1.079*** 1.506*** (0.276) (0.272) (0.273) (0.276) (0.273) (0.262) LN(GDP/Capita) 27.04*** 26.80*** 27.87*** 25.93*** 27.17*** 14.09*** (4.380) (4.356) (4.337) (4.698) (4.672) (4.591) GDP Growth -0.592*** -0.653*** -0.616*** -0.582*** -0.662*** -0.794*** (0.194) (0.196) (0.195) (0.195) (0.197) (0.194) M2 Growth -0.309*** -0.314*** -0.312*** -0.310*** -0.313*** -0.291*** (0.0376) (0.0373) (0.0374) (0.0377) (0.0374) (0.0362) Terms of Trade 0.0435 0.0507 0.0552 0.0427 0.0517 0.0872*** (0.0340) (0.0337) (0.0338) (0.0340) (0.0339) (0.0337) Trade Openness -0.128** -0.144*** -0.135** -0.127** -0.145*** -0.0736 (0.0555) (0.0550) (0.0546) (0.0555) (0.0551) (0.0531) Domestic Investment (% of GDP) 0.168 0.167 0.114 (0.110) (0.112) (0.109) Government Expenditures (% of GDP) -0.124 -0.0747 -0.0197 (0.263) (0.269) (0.262)

Primary School Enrollment -0.121 0.0266 -0.230

(0.185) (0.190) (0.184) Constant -100.2** -102.4** -109.5** -84.31 -106.4* 17.92 (49.57) (49.18) (49.04) (55.17) (54.96) (54.34) Observations 2,471 2,440 2,440 2,471 2,440 2,530 R-squared 0.315 0.322 0.321 0.315 0.322 0.302 Number of country 101 100 100 101 100 100

Country fixed effects YES YES YES YES YES YES

Year fixed effects YES YES YES YES YES YES

(33)

28

theory, which states that fast growing countries are net borrowers, and they try to decrease their debt burden by allowing inflation so that they have to pay less for their loans (when they are denoted in local currency). As the M2 growth variable serves as a proxy for domestic expenditures, the variable should have a positive effect on the real effective exchange rate. However, the regression denotes a negative relation between M2 growth and the real effective exchange rate. Albeit this is counterintuitive, the observed effect may be explained by the fact that fast money growth in the economy leads to higher levels of inflation. When a country experiences high levels of inflation, the domestic currency becomes less valuable compared to others, which signals to depreciation. Terms of trade shows the expected coefficients, which indicates that when a country faces a positive shock to a country’s terms of trade, it would expand the country’s export sector and would therefore appreciate the exchange rate (more demand for the domestic currency). The results of the trade openness variable are counterintuitive, as they point to negative results in most regressions. As this variable serves as a proxy for trade restrictions, it means that less restrictive and more open countries have lower real effective exchange rates. Although it is not in line with my initial hypothesis, Nyoni (1998) also finds that openness tends to depreciate the exchange rate in Tanzania. He states that trade restrictions serve as additional costs for doing business, and that a reduction in trade and exchange controls therefore can improve the competitiveness of a country. The control variables are insignificant.

The reader should bear in mind that the regression in Table 4 also takes into account countries that do not have a flexible exchange rate or that use another currency. For example, Ecuador and El Salvador use the US dollar and many African countries share a common currency.6 As it is unlikely that the power of these currencies will be affected when the country experiences a surge in development assistance, I remove these countries from the sample for an additional check. Table 5 presents these regression results, and one can see that not much changes in the regression results, but most coefficients become larger in their influence. Although still insignificant, government expenditures become positive and the domestic investment variable becomes significant. The latter might occur as foreign investors face more risk when a

(34)

29

Table 5: Regression results without Fixed Exchange Regimes, REER as dependent variable

(1) (2) (3) (4) (5) (6)

ODA Inflows (% of GDP) -1.519*** -1.521*** -1.533*** -1.528*** -1.542*** (0.248) (0.244) (0.252) (0.249) (0.251)

Bilateral Aid Inflows (% of GDP) -2.066***

(0.380) Remittances (% of GDP) 1.150*** 1.337*** 1.277*** 1.157*** 1.319*** 2.025*** (0.345) (0.340) (0.341) (0.346) (0.340) (0.327) LN(GDP/Capita) 38.38*** 38.27*** 39.23*** 37.42*** 40.05*** 23.49*** (5.330) (5.238) (5.269) (5.801) (5.739) (5.584) GDP Growth -0.713*** -0.879*** -0.735*** -0.702*** -0.891*** -1.043*** (0.239) (0.240) (0.240) (0.241) (0.243) (0.240) M2 Growth -0.342*** -0.353*** -0.347*** -0.343*** -0.352*** -0.321*** (0.0427) (0.0423) (0.0425) (0.0429) (0.0424) (0.0411) Terms of Trade 0.0257 0.0280 0.0373 0.0257 0.0271 0.0726* (0.0436) (0.0432) (0.0434) (0.0436) (0.0433) (0.0430) Trade Openness -0.174*** -0.239*** -0.178*** -0.173*** -0.240*** -0.142** (0.0655) (0.0657) (0.0645) (0.0655) (0.0658) (0.0639) Domestic Investment (% of GDP) 0.684*** 0.713*** 0.746*** (0.164) (0.167) (0.168) Government Expenditures (% of GDP) 0.0200 0.239 0.204 (0.341) (0.348) (0.337)

Primary School Enrollment -0.101 0.211 -0.127

(0.242) (0.247) (0.238) Constant -209.2*** -223.0*** -221.4*** -195.8*** -251.8*** -94.87 (59.34) (58.47) (58.75) (67.45) (67.48) (66.42) Observations 1,955 1,924 1,924 1,955 1,924 2,001 R-squared 0.320 0.334 0.328 0.320 0.335 0.308 Number of country 78 77 77 78 77 77

(35)

30

country has a flexible exchange rate. This may cause countries with flexible exchange rates to be more dependent on domestic investments compared to countries that use a foreign currency. Domestic investment results in an appreciation of the exchange rate, which can be explained in an increase demand of the currency. The influence of primary school enrollment becomes a little bit smaller, but remains insignificant.

5.2 Aid Inflows and Tradable-to-Non-Tradable Ratio

Table 6 denotes the results for a country fixed effects regression with the tradable-to-non-tradable output ratio as the dependent variable, where column 1 denotes the base model, column 2-4 provide regression results with an additional variable, column 5 provides the regression results for all variables, and column 6 shows the regression with bilateral aid flows as the independent variable. As the tradable-non-tradable output ratio is defined as the sum of agriculture, mining and manufacturing output divided by services, a decrease in this ratio signals that output in the service sector increases relative to the tradable sectors.

(36)

31

Table 6: Regression results, TNT as dependent variable

(1) (2) (3) (4) (5) (6)

ODA Inflows (% of GDP) -0.00152 -0.00143 0.000103 0.00136* 0.00160** (0.00114) (0.00113) (0.00116) (0.000720) (0.000745)

Bilateral Aid Inflows (% of GDP) 0.000396

(0.00114) Remittances (% of GDP) -0.000558 -0.000307 -0.000212 -0.000735 -0.000565 -0.00166 (0.00167) (0.00167) (0.00166) (0.00106) (0.00107) (0.00104) LN(GDP/Capita) -0.192*** -0.179*** -0.183*** 0.194*** 0.201*** 0.219*** (0.0264) (0.0266) (0.0264) (0.0179) (0.0182) (0.0181) GDP Growth 0.000746 0.00169 -0.000304 -0.00282*** -0.00287*** -0.00209*** (0.00116) (0.00118) (0.00117) (0.000737) (0.000763) (0.000759) M2 Growth -0.000350 -0.000262 -0.000240 0.000162 0.000150 8.17e-05 (0.000227) (0.000228) (0.000227) (0.000144) (0.000146) (0.000143) Terms of Trade -0.000588*** -0.000556*** -0.000528*** -0.000312** -0.000317** -0.000386*** (0.000203) (0.000204) (0.000204) (0.000129) (0.000131) (0.000132) Trade Openness 0.00127*** 0.00156*** 0.00124*** 0.00104*** 0.00106*** 0.000827*** (0.000335) (0.000336) (0.000333) (0.000212) (0.000215) (0.000211) Domestic Investment (% of GDP) -0.00383*** -0.000397 -0.00114*** (0.000670) (0.000437) (0.000434) Government Expenditures (% of GDP) -0.0103*** -0.000902 -0.000815 (0.00157) (0.00103) (0.00101)

Primary School Enrollment 0.0420*** 0.0420*** 0.0411***

(0.000707) (0.000741) (0.000730) Constant 2.991*** 2.927*** 3.016*** -2.495*** -2.557*** -2.650*** (0.299) (0.299) (0.298) (0.210) (0.214) (0.214) Observations 2,501 2,470 2,470 2,501 2,470 2,560 R-squared 0.104 0.121 0.125 0.642 0.640 0.630 Number of country 103 102 102 103 102 102

(37)

32

this regression. In section 5.4, we will test how remittances influence the real effective exchange rate and the tradable-non-tradable output ratio when the aid variables are left out of the regression. The coefficients for Ln(GDP/Capita) change from positive to negative when primary school enrollment is added to the regression, which points to a non-linear relationship between GDP/Capita and the tradable-non-tradable sectors. Countries with a lower GDP/Capita first experience a fall in the tradable-non-tradable sector, indicating that primary products become less important compared to non-tradable goods. When GDP/Capita reaches a certain threshold, the tradable-non-tradable output ratio increases again. This is in contrast to the neoclassical theory, which predicts that countries should have an inverted u-shaped tradable-non-tradable relation in the development process. The effect of GDP/Capita on the specific sectors will also be further discussed in section 5.3. The significant results for GDP growth show a negative correlation with the tradable-non-tradable ratio. This indicates that fast growing countries have a decrease in the tradable output and/or an increase in the non-tradable output. Better terms of trade exhibit negative influence on the tradable-non-tradable sector, which is strange, as a positive shock in the terms of trade expand the export sector. Trade openness has a positive influence on the tradable sector, which is expected, as a less restrictive country will trade more, and therefore the tradable sector will increase. Domestic investment reduces the TNT ratio, indicating that this investment is biased towards the non-tradable sector rather than the export sector. An increase in government expenditures also seems to reduce the tradable sector, which is not strange as government expenditure might also be biased towards the service sector.

5.3 Aid flows and Sectorial Output Decomposition

(38)

33

on the agricultural is larger alone, than when the mining sector is also taken into account.

We can see from table 7 that aid inflows seem to decrease output in the primary and in the secondary sector, but that it raises output in agriculture and service sector, which is an indication of the “Dutch disease”. Rajan and Subramanian (2011) examine the effects of aid on the growth of the manufacturing sector, and they find that aid has adverse effects on the recipient’s competitiveness. The loss of competitiveness is reflected in a lower growth of the manufacturing export sector because aid inflows can lead to spending patterns that are biased towards the non-tradable sector. Aid is relatively more spent on food, schooling and health developments rather than on manufactures, and thus biased towards the service sector. Table 7 also provides additional evidence that remittances tend to decrease the output in the manufacturing sector and that it increases output in the services sector. This finding confirms the effect that already has been found by Lartey et al. (2008) and Ameudo-Dorantes and Pozo (2004). These authors identified that remittances inflows lead to spending patterns that are biased towards the non-tradable sector, reducing the competitiveness of the tradable sector.

Table 7: Sectorial Output Composition

(1) (2) (3) (4)

VARIABLES Agriculture Agriculture & Mining Manufactures Services

ODA Inflows 0.0141 -0.0657*** -0.0129 0.0639*** (0.0127) (0.0205) (0.00996) (0.0187) Remittances 0.00928 -0.0259 -0.0910*** 0.116*** (0.0183) (0.0294) (0.0142) (0.0269) LN (GDP) -8.566*** -10.20*** 0.156 8.859*** (0.236) (0.379) (0.192) (0.347) GDP Growth 0.0711*** 0.0907*** -0.0172* -0.0795*** (0.0128) (0.0206) (0.00999) (0.0189) M2 -0.00678*** -0.0114*** 0.00378* 0.00616* (0.00254) (0.00408) (0.00198) (0.00374) Terms of Trade 0.000377 -0.00406 -0.00467*** 0.00724** (0.00228) (0.00367) (0.00178) (0.00336) Trade Openness -0.0273*** 0.00358 0.0238*** -0.0221*** (0.00379) (0.00609) (0.00295) (0.00557) Constant 113.3*** 141.7*** 10.63*** -39.24*** (2.553) (4.107) (2.097) (3.759) Observations 2,501 2,501 2,478 2,501 R-squared 0.450 0.261 0.044 0.236 Number of country 103 103 103 103

(39)

34

The LN(GDP/Capita) variable is able to shed more light on the non-linear relationship found in Table 6. The variable in this regression shows the expected effect that is predicted by neoclassical theory, which states that countries will face a transition from agriculture to manufacturing, and eventually to services as income rises. Higher ratios of GDP per capita, are indeed associated with lower agricultural output shares, and vice versa with higher shares of manufacturing and services. However, the GDP growth variable seems to paint another picture: this variable seems to reduce manufacturing and services output. One explanation might be that fast growing countries are very dependent on natural resources, such as Timor-Leste, which experienced a GDP growth rate of 130% in 2004, after large oil fields were found in the country (Cotton, 2005). Liberia’s growth increased with 106% in 1997 after the end of the war in 1996. Although this reason was due to political stability, Liberia’s exports are heavily dependent on agricultural products. The terms of trade variable seems to also have a different influence than expected, where a positive shock in the terms of trade should lead to an increase in exports, it seems to reduce the manufacturing sector and increasing the services sector. Trade openness has the expected value for manufacturing output, stimulating manufacturing output as the country opens up for trade. However, it has a negative influence on the agricultural and mining sector and a positive influence on the service sector.

5.4 Comparison to Lartey et al. (2008).

(40)

tradable-35

non-tradable output ratio are indicators of the competitiveness, the results prove that remittances can lead to effects that are associated with the “Dutch disease”. Although many variables have the same coefficient sign as Lartey et al. (2008), in the REER regression, trade openness shows a negative influence, while Lartey et al. (2008) find a positive relation. For the regression with the tradable-non-tradable ratio, more variables show an opposite relation compared to the findings from Lartey et al. (2008), which is something that I also observed in the previous regression.

Table 8: Lartey et al. (2008) copy

(1) (2) (3) (4)

VARIABLES REER REER TNT TNT

Remittances 1.023*** 1.123*** -0.00139 -0.00469*** (0.251) (0.249) (0.00151) (0.00101) GDP/Capita 3.51e-07 3.79e-07 2.01e-08*** 1.17e-08***

(7.81e-07) (7.72e-07) (4.69e-09) (3.13e-09) GDP Growth -0.736*** -0.776*** 0.000979 -0.00146* (0.188) (0.192) (0.00112) (0.000767) M2 Growth -0.287*** -0.291*** -0.000384* 1.55e-05 (0.0363) (0.0361) (0.000217) (0.000146) Terms of Trade 0.0909*** 0.0939*** -0.000983*** -0.000360*** (0.0331) (0.0331) (0.000197) (0.000133) Trade Openness -0.0237 -0.0388 0.000673** 0.00104*** (0.0519) (0.0516) (0.000311) (0.000209) Domestic Investment 0.107 -0.000796* (0.108) (0.000439) Government Expenditures -0.273 -0.00107 (0.256) (0.00101)

Primary School Enrollment -0.447*** 0.0373***

(0.171) (0.000692) Constant 158.0*** 175.1*** 1.140*** -0.222*** (16.06) (17.75) (0.0963) (0.0719) Observations 2,655 2,624 2,686 2,655 R-squared 0.270 0.277 0.103 0.606 Number of country 101 100 103 102

Country fixed effects YES YES YES YES

Year fixed effects YES YES YES YES

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(41)

36

where I use a simple fixed effects model. However, as Lartey et al. (2008) argue that their results are significant and robust, they should also appear in a simple model. Another explanation for the differences in results can be found in the number of observations and the period they use. Where Lartey et al. (2008) work with a set of 109 developing countries for the period 1990-2003, I use a sample of 134 countries for the period 1970-2013. Even replicating the regression only with the years 1990-2003, does not change the values of the coefficient by much.‡‡ Additionally, the variables that Lartey et al. (2008) use are denoted in US dollars, which implies that already some exchange rate conversion has been applied to the data. For example, Lartey et al. (2008) use constant US dollars as an indicator for GDP per capita, but the remittances per capita are denoted in current US dollars.

6. Limitations and Recommendations

Although the real effective exchange rate is a popular measure to analyze competitiveness of a country, it also has its limitations. For example; consumption patterns can change more rapidly than market baskets construct. This means that the REER measure is slow to adjust the weights of products in the basket to account for changing consumption patterns, and might therefore not adequately reflect the relative price level. Second, it is not taken into account that higher productivity countries have lower production costs and therefore have lower prices. Due to international competition, lower international prices will lead to lower prices in all tradable goods, reducing the price ratio of tradable to non-tradable goods, something that can have an appreciating effect on the REER. One limitation of the aid variables is that they only take into account flows from the Development Assistance Committee (DAC), and thus ignore flows from developing countries to other developing countries. Consequently, net donors might emerge as aid recipients in this dataset, and total aid flows might be undervalued for other countries. As we have seen in the literature review, Hanson and Tarp (2001) stress that regressions are very sensitive to the number of observations and the assumptions that have been made by the author. This is also likely to be of influence in my regressions. For example, I make the assumption that all my variables are exogenous. However, it is unlikely that remittances do not depend on the exchange rate, as workers are more likely to be

(42)

37

reluctant to send money when the exchange rate is unfavorable. In this case, they may wait when transfers are not urgent, so that they receive more return on their earnings. On top of that, aid flows are likely to be dependent on the relationship between the donor country and the recipient or certain events such as war and natural disasters. Political motivations for giving aid can be of influence on the allocation of aid, and how it is spend. This research has been conducted in order to shed some light about the unobservable effect of development aid on economic performance. As this thesis does not find any evidence for the “resource curse of aid”, there might be another effect at work that is responsible for the marginal effect of development aid on economic growth. This is something that might be interesting for a new field of research in the aid effectiveness literature.

7. Conclusion

The aid effectiveness literature has trouble coming to a consensus because these analyses are very sensitive to the variables that are used and the assumptions that are made. This thesis tries to provide more clarity about the found irregularities in the literature, by investigating whether aid flows could result in an appreciation of the currency. However, instead of an appreciation, I find that aid inflows tend to

depreciate the real effective exchange rate. A 1 percentage point increase in aid

Development Aid and Exchange Rate Movements

University of Groningen and Georg-August Univesität Göttingen

Master Thesis