Does Aid for Trade enhances the trade performance of
recipient countries?
Examining the 2005 Aid for Trade initiative
Name: Menouschka Claudine Plugge Studentnumber: 10364757 Faculty: Economics and Business
Master: Economics Field: Development Economics
Supervisor: Prof. E.J.S. Plug Academic Year 2016/2017
Abstract
This study examines whether the trade performance of recipient countries of Aid for Trade are affected by the Aid for Trade initiative in 2005. I do so by estimating a difference-in-differences estimation using the 2005 Aid for Trade initiative as the con-sidered treatment and compare countries receiving yearly Aid for Trade to countries not receiving Aid for Trade over the period 1995-2015. The results of the difference-in-differences estimation suggest that exports increased 33% more and import increased 24% more in countries receiving yearly Aid for Trade, post-Aid for Trade initiative rel-ative to pre-initirel-ative, than in countries not receiving Aid for Trade. Further, a Two Stage Least Squares estimation is used to estimate the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of the recipient. These results suggest that a 1% increase of Aid for Trade received increased exports with 1.06% and imports with 1.15%.
This document is written by Student Menouschka Claudine Plugge who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Contents
1 Introduction 3 2 Context 7 3 Data 9 3.1 Sample Selection . . . 9 3.2 Data Explained . . . 9 3.3 Descriptive Statistics . . . 10 4 Empirical Method 15 4.1 Difference-in-differences . . . 15 4.2 Two Stage Least Squares . . . 185 Results 23
5.1 Difference-in-Differences . . . 23 5.2 Two Stage Least Squares . . . 27
6 Conclusion 32
7 References 34
1
Introduction
The main purpose of this study is to evaluate the relationship between Aid for Trade and the trade performance of recipient countries by examining the 2005 Aid for Trade initiative. The Aid for Trade initiative was launched at the Ministerial Conference of the World Trade Organization (WTO) in Hong Kong in 2005. It was agreed to expand aid to support developing countries in increasing exports of goods and services, and benefiting from free trade and increased market access (OECD, 2006). Even if trade-related Aid has always existed as part of Official Development Assistance (ODA) flows, the official creation of Aid for Trade has put a new light on these specific Aid flows and launched discussions and debates on their effectiveness (Vijil et al., 2011). This setting creates an environment to examine whether the trade performance of recipient countries of trade-related Aid is affected by the Aid for Trade initiative in 2005. I do so by estimating a difference-in-differences estimation using the 2005 Aid for Trade initiative as the considered treatment and compare countries receiving yearly Aid for Trade to countries not receiving Aid for Trade over the period 1995-2015. Further, a Two Stage Least Squares estimation is used to estimate the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of recipient countries.
Nowadays, there is still no consensus among economists if more trade will lead to more economic growth and development. However, Rodrik et al. (2004) stated that “Trade can be an underlying source of growth once certain institutional pre-requisite have been fulfilled”. In other words, trade can be an opportunity for economic growth, economic development, and poverty elevation if certain other factors have been realized (Stiglitz & Charlton, 2006). Building on this, outward-oriented growth has been a pop-ular development strategy in low-income countries, resulting in non-existing or very low trade controls (Wade, 1992). However, the share of the poorest developing countries in global trade flows has not increased and only a few of these countries effectively suc-ceeded in reducing poverty. In fact, the share in world exports of the Least Developed Countries has constantly decreased in the last three decades even when enjoying pref-erential market access for goods and services in most high-income countries (UNCTAD, 2006). It seems that trade liberalisation and increased market access is not sufficient for some countries, especially the least developed countries, to improve the economic development and poverty reduction prospects. This is mainly due to internal obstacles to trade that most developing countries face, such as lack of knowledge, excessive red tape, insufficient financing and poor infrastructure (Vijil & Wagner, 2012). The solution for these internal obstacles to trade could be country specific trade facilitation reforms, but these are difficult to do and cost money.
Building on this outlook, the Aid for Trade initiative was launched at the Hong Kong Ministerial World Trade Organization (WTO) Conference in December 2005. The WTO calls for the expansion of Aid for Trade to help developing countries benefit from free trade agreements and more broadly expand their trade. The short-term objective is helping developing countries overcome the internal obstacles to trade and the long-term
objective is a more equitable distribution of global benefits across and within developing countries (OECD, 2006). Expected is that Aid for Trade enhances the trade perfor-mance and through trade perforperfor-mance, increased economic growth, increased income and reduced poverty.
The literature on the impact of Aid on economic development and poverty reductions has so far failed to provide strong and convincing results. This is due to difficulties in capturing the causal effect of Aid on economic development. Focusing on more specific outcome variables is a solution for addressing the difficulty in capturing the causal effect. In this setting, focusing on the causal effect of Aid for Trade on the trade outcomes of recipient can give perspective. However, the evidence of whether Aid for Trade is effective in enhancing the trade outcomes is still surprisingly limited. There are only a few empirical studies that assess the effectiveness of Aid for Trade on the trade flows of recipient countries. I will give a brief overview of the four main articles written, emphasising on the method used and results gained.
Cali and Te Velde (2009) assess the extent to which Aid for Trade have helped recip-ient countries trade outcomes. Based on an augmented export demand equation, they estimate the direct effect of Aid for Trade on total exports. The paper distinguishes between two main categories of Aid for Trade: aid to economic infrastructure and aid to productive capacity. They use a panel data fixed effect estimation for 120 recipients developing countries, controlling for total population, market potential measure, govern-ment effectiveness and the Consumer Price Index. Solving for the endogeneity problem, the paper uses instruments for aid to economic infrastructure and for aid to productive capacity. The instruments used are the degree of respect for civil liberties, as mea-sured by Freedom House (2009), and the affinity of nations index proposed by Gartzke (2009). The empirical results show a positive impact of aid to economic infrastructure on exports, while aid to productive capacity does not seem to have any significant effect on exports. In addition, they include an analysis at the sectoral level using four main tradable sectors: food production, manufacturing, mineral extraction, and tourism. The results show a positive and significant impact of aid to economic infrastructure and aid to productive capacity building on sectoral exports.
Helble et al. (2009) relates recipient countries exports to changes in Aid for Trade. The paper distinguishes between two main categories of Aid for Trade: i) narrow trade facilitation, including aid to trade policies and regulations; ii) broad trade facilitation, including aid to productive capacity and aid to economic infrastructure. Using a gravity model with fixed effects, they find strong empirical evidence that Aid for Trade has a small, but significant and positive relationship to trade flows. Suggesting that a 1% increase in Aid for Trade could generate an increase in global trade of about USD 415 million. In addition, the narrow trade facilitation, is stronger both in robustness and magnitude. A one percent increase in this aid category could generate an increase in global trade of about USD 697 million.
Vijil and Wagner (2012) address the question of the effectiveness of aid for trade on trade performance using a two-step empirical analysis. The first empirical step tests whether institutions and infrastructure are significant determinants of export
perfor-mance. From theoretical model, institutions and infrastructure can be considered as potential channels of transmission for the impact of Aid for Trade. The second em-pirical step tests whether Aid for Trade sectoral flows are significant determinants of infrastructure and institutions. Just like the previous mentioned paper, both empiri-cal steps are empiri-calculated using an aggregation of gravity equations of trade flows. With controlling for countrys size, international market access, volatility of exchange rate, a proxy for infrastructure quantity, the quality of institutions, and the trade restrictive-ness imposed on its imports from the rest of the world. To address the endogeneity problems, they instrument infrastructure and institutions variables. Infrastructure is in-strumented by a variable reflecting internal geography and institutions is inin-strumented by the numbers of documents needed to export. The first step empirical result sug-gests that infrastructure has a highly significant positive impact on developing countries export performance, whereas institutions have limited impact. The second empirical results suggest that aid for infrastructure has a strong and positive impact on the in-frastructure level. Suggesting that a 10% increase in aid for inin-frastructure commitments leads to an average increase in the exports over GDP ratio of 2.34%.
H¨uhne et al. (2014) test the hypothesis that Aid for Trade is as much in the self-interest of donor countries as it may have promoted the exports of recipient countries. They simultaneously estimate and compare the effects of Aid for Trade on recipient and donor countries export. Again, like previous mentioned papers, it uses an aggregated gravity model controlling for GDP, population and distance between trading partners. The study deals with the endogeneity concerns by introducing lags and by performing separate estimations for donors perceived altruistic. They find that Aid for Trade in-creases recipient exports to donors as well as recipient imports from donors. The first effect tends to rule the latter, which contradicts the sceptical view that donors grant Aid for Trade primarily to promote their own export interests.
The above-described studies all provide empirical evidence that Aid for Trade, or at least some categories of Aid for Trade, do positively affect the export performances of recipient countries. However, this thesis tries to complement previous studies by emphasizing on the 2005 Aid for Trade initiative. This thesis is the first, to the best of my knowledge, to fill in the evaluation gap by exploring a difference-in-differences estimation using the 2005 Aid for Trade initiative as the considered treatment. The findings of the difference-in-differences estimation suggest that exports increased 33% more in countries receiving yearly Aid for Trade, post-Aid for Trade initiative relative to pre-initiative, than exports in countries not receiving Aid for Trade. For imports, I find a 24% more increase for countries receiving yearly Aid for Trade flows than countries not receiving Aid for Trade, post-initiative relative to pre-initiative. The thesis also finds that the rate of growth of imports increased more in countries receiving yearly Aid for Trade, post-initiative relative to pre-initiative, than countries not receiving Aid for Trade. To complement the difference-in-differences estimation, this thesis uses a Two Stage Least Squares estimation to estimate the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of recipient countries. I find strong empirical evidence that a 1% increase in the amount of Aid for Trade received increases exports
of recipient by 1.06% and imports of recipient by 1.15%.
This thesis proceeds as follows. The next section describes the context of Aid for Trade. Section 3 describes the data. Section 4 provides details of the empirical strategy that is employed. Section 5 presents and discusses the empirical findings. Section 6 summarizes and concludes.
2
Context
According to the WTO (2006) Aid for Trade refers to a subset of development assistance designed to help developing countries address the internal barriers to trade and boost their capacity to take advantage of expanded trade opportunities. Aid for Trade is a form of development assistance with an aim of building supply-side capacity and trade-related infrastructure with the expectation that this will enhance economic growth and reduce poverty in developing countries. The structure of Aid for Trade has been broadly identified, but the scope is differently refined by various institutions and organizations. The scope has been categorized to range from a narrow to a wide definition. As starting point, there are four main categories provided by the Aid for Trade Task Force. This Task Force is composed of 14 WTO member countries, the coordinators of the African, Caribbean and Pacific Group of States, the African Group and the LDC Group (WTO, 2006). The categories are:
1. Aid for Trade Policy and Regulations is according to the OECD (2006):“to help countries negotiate, reform and prepare for closer integration in the world trad-ing system; it covers activities such as analysis and implementation of multilateral trade agreements, trade policy mainstreaming and technical standards, trade facil-itation including tariff structures and customs regimes, support to regional trade arrangements and human resources development in trade”.
2. Aid for Economic Infrastructure addresses the internal barriers resulting from in-adequate infrastructure. As illustration Stiglitz and Charlton (2006) report:“poor transport infrastructure can prevent local farmers from accessing large domestic markets and international ports; poor storage facilities can increase inventory costs; and bad energy and water supplies can disrupt production or increase costs”. 3. Aid for Building Productive Capacity is to help enterprises engage in trade and
improve the business climate, access to trade finance and trade promotion in the productive sectors (agriculture, forestry, fishing, industry, mining, tourism and ser-vices). Thus, help strengthening local economic sectors to increase competitiveness in export markets (OECD, 2006).
4. Aid for Trade-related Adjustment reflects the fact that trade liberalization may impose costs as resources move from one sector to another in process of the re-form. In other words, adjustment costs are the price to be paid for the benefits of multilateral tariff reduction. As example, Stiglitz and Charlton (2006) state: “transitional unemployment, additional risks from transitioning to an open econ-omy, and changes in relative prices of factors of production”.
The narrowest definition includes only category 1. of Aid for Trade Policy and Regula-tions, while the broadest definition includes all four categories of Aid for Trade, resulting in the differently redefined scope of Aid for Trade (Vijil et al., 2011). Figure 2.1 shows
the possible elements in the Aid for Trade agenda that are taken into account by the Task Force. Starting with the narrowest definition of Aid for Trade represented by Cat-egory 1 and finally the broadest definition of Aid for Trade represented by CatCat-egory 1, 2, and 3 plus the adjustment costs.
Figure 2.1: Aid for Trade and the Expanding Agenda
Note: Source OECD (2006)
Trade-related Aid has always existed as part of the Official Development Assistance (ODA) flows. However, the official creation of Aid for Trade at the Ministerial Conference of the WTO in 2005 has put a new perspective on these specific Aid flows and launched discussions on their effectiveness. Before the Aid for Trade initiative in 2005, assistance was mostly targeted to the narrow Aid for Trade agenda. In 2004 this included a total of nearly USD 23 billion and a combined share of over 24% of total ODA. The total aid volume was not the issue, but there where issues surrounding the effectiveness of Aid for Trade on the trade outcomes of recipient countries. In fact, the share in world exports of the Least Developed Countries has constantly decreased in the last three decades. The problem is that the process of building internationally competitive economies takes time and is highly country specific. The inclusion of Aid for Trade in the Hong Kong Declaration offers the aid and trade communities the opportunity to establish a framework to deliver aid for trade-related assistance (OECD, 2006). In addition, the Aid for Trade definition has been broadened to include support to productive capacity, trade-related infrastructures and trade-related adjustment (Vijil et al., 2011). Hopefully resulting in a more effective enhancement of Aid for Trade on recipient countries trade outcomes.
3
Data
3.1
Sample Selection
The sample includes data of 124 developing countries from all over the world. This thesis defines an observation as any point in time for which data on export or import and Aid for Trade exists. All countries used in this thesis have relatively open economies and are exporting and importing countries. The included sample countries and the number of observations is based on the availability of yearly export, import and Aid for Trade flows. For some countries, data on the export and import flows are not complete. This results in an unbalanced panel dataset including a total of 124 developing countries. This thesis uses data from the pre-initiative years 1995-2004 and the post-initiative years 2005-2015. The treatment group used in this thesis consists of the countries receiving yearly Aid for Trade flows. This group consists of 118 developing countries and are selected in the sample because they received yearly Aid for Trade flows and there is data available on exports and imports of most years in the period 1995-2015. The full list of countries receiving yearly Aid for Trade used in this thesis can be found in table A.1 in the Appendix. The control group used in this thesis consists of 6 developing countries not receiving Aid for Trade flows and with data available for exports and imports flows for most years in the period 1995-2015. The countries in the control group are Aruba, Bahrain, French Polynesia, Malta, New Caledonia, and S˜ao Tom´e and Pr´ıncipe. See table A.2 in the Appendix for a table with the list of countries used for the control group.
3.2
Data Explained
The data are drawn from existing databases and publicly available information websites. Data for export and import flows of the recipient countries are employed from the World Banks World Integrated Trade Solutions (WITS) database. The export and import flows are recorded at current value USD.
The data on Aid for Trade come from the OEDC/DAC Creditor Reporting System (CRS) database on aid disbursements. The CRS is a database covering around 90% of all ODA and was acknowledged as the best available data source for tracking global Aid for Trade flows (OECD, 2016). This study follows a data driven definition of Aid for Trade by the explanatory note of Aid for Trade data by the OECD (2016). This study uses three categories of Aid for Trade: i) Aid for Trade Policies and Regulations and Trade-related Adjustment; ii) Aid for Economic Infrastructure; iii) Aid for Building Productive Capacity. Total Aid for Trade is calculated by adding up the three categories. It should be kept in mind that the CRS does not provide data that exactly match these Aid for Trade categories, see the explanatory note on Aid for Trade by the OECD (2006) for the full list of CRS codes. The Aid for Trade flows recorded are net disbursements of ODA at constant prices in 2015 USD.
Also, the study controls for a number of factors that may affect exports and imports. First, the study controls for total population, other things being equal a larger population is associated with a larger flow of exports and imports. Second, government effectiveness is included to control for the institutional strength of the country, which is likely to influence export and import capacity. Third, countries income level is included, which is likely linked to better institutions and rules in a country. Fourth, area size of country is included, which negatively impacts trading costs. Final, the study controls for whether a country is considered landlocked or not. A landlocked country is a country entirely enclosed by land, or whose only coastlines lie on closed seas. Other things being equal being landlocked increases costs for taking goods to the port of departure (Cali & te Velde, 2009). Data on GDP, area size and population size are employed from World Banks World Development Indicators (WDI) databank. The government effectiveness indicator is from The Worldwide Governance Indicators (WGI) project. Information on landlocked countries are retrieved from Vijil et al. (2011), see table A.3 in the Appendix for the full list of landlocked countries.
3.3
Descriptive Statistics
Table 3.1: Descriptive Statistics on the 124 developing countries included in the sample Mean Std. Dev. N Source
(1) (2) (3) (4)
Outcome Variables
Export 28368.77 130248.4 2203 World Bank WITS Log(Export) 21.5529 2.4781 2203 World Bank WITS Import 26896.21 112116.2 2214 World Bank WITS Log(Import) 22.1187 1.9624 2123 World Bank WITS Treatment variable
Total Aid for Trade 189.9439 416.7182 2604 OECD CRS
Log(AfT) 17.6040 2.1254 2392 OECD CRS
Control variables
Area Size (squared km) 0.5808 1.2475 2583 World Bank WDI Landlocked 0.2097 0.4072 2604 Vijil et al. (2011)
GDP 126.1267 502.5964 2526 World Bank WDI
Population 40.4245 156.356 2604 World Bank WDI Government effectiveness -0.2954 0.6361 2557 WGI
Note: Yearly data on the outcome, treatment and control variables over the period 1995-2015 included in the thesis. Column (1) shows the mean, (2) shows the standard deviation, (3) total number of observations and (4) the sources used to obtain the data. GDP is in billions of USD. Export, Import,
total Aid for Trade are in millions of USD. Population and Area Size are in millions.
Table 3.1 reports descriptive statistics on the 124 countries included in the sample. The large standard deviations of the control variables highlight the large differences
be-tween the countries included in the sample. Using the difference-in-differences estimation will control for these differences between the control and treatment group. Furthermore, in this thesis the logarithms of the statistics of the outcome and treatment variables are used. Logarithms can be used because the share of countries receiving yearly Aid for Trade compared to countries not receiving Aid for Trade remained relatively stable over time. See Figure A.1 in Appendix A for further explanation. Table 3.2 compares the descriptive statistics of the treatment and control group. With the treatment group receiving yearly Aid for Trade flows and the control group receiving zero Aid for Trade flows over the full period 1995-2015. There are large differences between the treatment and control group for all the variables included in this thesis. When the difference-in-differences estimation is implemented, these difference-in-differences are controlled for.
Table 3.2: Descriptive Statistics on the treatment and control group used in the sample
Treatment Group Treatment Group Control Group Control Group Mean Std. Dev. Mean Std. Dev.
(1) (2) (3) (4) Outcome Variables Export 29638.04 133371.4 3258.876 4796.113 Log(Export) 21.5932 2.5059 20.757 1.6643 Import 28042.55 114806 4322.852 4595.949 Log(Import) 22.1373 1.9998 21.7523 0.8981 Treatment variable
Total Aid for Trade 199.602 424.9222 0 0
Log(AfT) 17.6040 2.1254 -
-Control variables
Area Size (squared km) 0.6052 1.2679 0.0046 0.0070
Landlocked .2096774 0.4087 0.1667 0.3741
GDP 128.4455 507.4868 13.5935 7.3189
Population 42.3923 160.0319 1.7254 3.1448
Government effectiveness -0.3278 0.6054 0.7199 0.7372
Note: Descriptive Statistics for the treatment and control group on the outcome, treatment and control variables over the period 1995-2015 included in the thesis. Treatment group is group of countries receiving yearly Aid for Trade and the control group is the group of countries not receiving Aid for
Trade. Column (1) and (2) shows the mean and standard deviation for the treatment group and Column (3) and (4) shows the mean and standard deviation for the control group. GDP is in billions of
USD. Export, Import, total Aid for Trade, Population and Area Size are in millions.
Figure 3.1 graphically shows the treatment and outcomes variables of the group of countries receiving yearly Aid for Trade and the group of countries not receiving Aid for Trade included in this sample over the period 1995-2015. In this setting, the control group are the countries not receiving Aid for Trade over the full period and the treatment group are the countries receiving yearly Aid for Trade flows. In this thesis, there are 118 countries included in the treatment group and 6 countries included in the control
group. See Table A.1 and A.2 in the Appendix for an overview of the countries included in the treatment and control group. Looking at the figure, note that after the 2005 Aid for Trade initiative, Aid for Trade levels increased sharply for the treatment group. The Aid for Trade levels in the post-intervention years almost doubled compared to the pre-intervention years. Exports of the treatment group shows a similar pattern as the Aid for Trade levels for the treatment group: a sharp increase in exports post- 2005 Aid for Trade initiative. While the treatment group shows an upward trend in exports, the exports of the control group remained relatively stable. Examining the import levels of the control and treatment group again shows this similar pattern. After the 2005 Aid for Trade initiative there appears to be an upward trend in the imports of the treatment group, while the imports of the control group remained relatively stable. Figure 3.1 could suggest that Aid for Trade post-initiative effectively enhanced the exports and imports of the recipient countries, while the exports and imports of the control group remained relatively stable.
Figure 3.2 and 3.3 presents the logarithm of exports and imports for the control and treatment group. It suggests, without controlling for any other characteristics, that prior to 2005 export grew at roughly the same rate across the group of countries receiving yearly Aid for Trade and the group of countries not receiving Aid for Trade. Suggesting that the control and treatment group experienced common trends in exports prior to the initiative. After the 2005 Aid for Trade initiative, there appears to be some divergence, with the gap opening more as time progresses. This could be explained by the time lag in the effectiveness of Aid. It is increasingly recognized that aid may not become effective immediately, but only after a certain period of time (Helble et al, 2012). Import shows a similar pattern. Again, it suggests, without controlling for any other characteristics, that prior to 2005 import grew at roughly the same rate across the treatment group and the control group. Suggesting that, prior to the initiative, imports of the control and treatment group experienced a common trend. After the 2005 initiative, there appears to be some divergence, with the gap opening more as time progresses. Also, the graphs suggest that not only the export and import levels were affected by the Aid for Trade initiative, but that the rate of the change in those levels was also affected. The graphs in Figure 3.2 and 3.3 suggest that the rate of growth in exports and imports increased after the 2005 Aid for Trade initiative. The difference-in-differences analysis that follows is designed to quantify the divergence that appears in Figure 3.2 and 3.3 and to determine if it is statistically significant after controlling for country characteristics.
Figure 3.1: Log(Aid for Trade), Log(Exports), and Log(Import) for treatment and con-trol group over the period 1995-2015.
Note: Control group is the group of countries not receiving Aid for Trade and the treatment group is the group of countries receiving yearly Aid for Trade.
Figure 3.2: Log(Export) of treatment and control group over the period 1995-2015
4
Empirical Method
4.1
Difference-in-differences
To identify a causal impact of Aid for Trade on export and import of recipient countries the difference-in-differences (DID) approach is undertaken. The idea behind the DID approach is to determine whether some statistic of interest (e.g., import and export) changed more for one group of observations after some event than for another group of observations. The DID compares changes over time in a group of countries that are affected (treated) by a policy change, to a control group that is not affected. Basically, estimating the true counter-factual. By analysing changes in outcome over time, the estimation allows for controlling for observed and time-invariant characteristics as well as time-varying factors.
In this thesis, the 2005 Aid for Trade initiative is the considered policy change and data on export and import are the statistics of interest. The treatment group considered is the group of countries receiving yearly Aid for Trade flows over the period 1995-2015. The control group is the group of countries not receiving Aid for Trade flows over the full period. The standard DID estimation focuses on differences in export and import levels and includes dummy variables in a simple Ordinary Least Squares (OLS) regression indicating whether the time period is pre- or post-Aid for Trade initiative, whether the country is receiving yearly Aid for Trade, and an interaction of these two indicators (Hotchkiss et al, 2002). The dummy variable indicating pre- or post-initiative captures possible factors that would cause changes in exports and imports even in the absence of the Aid for Trade initiative. The dummy variable indicating treated or control captures the possible differences between the group of countries receiving yearly Aid for Trade and the group of countries not receiving Aid for Trade prior to the initiative. The interaction term will be the coefficient of interest and shows the treatment effect (NBER, 2007).
log(Export)it= α0+ α1T reatedi+ α2P ostt+ α3(T reatedi∗ P ostt) + α4Xit+ ξit
with 1=1,...,N t=1995,...,2015 (4.1)
log(Import)it= β0+ β1T reatedi+ β2P ostt+ β3(T reatedi∗ P ostt) + β4Xit+ ξit
with 1=1,...,N t=1995,...,2015 (4.2)
Where log(Export)it and log(Import)it are the variables of interest and are
respec-tively the logarithm of yearly exports in country i in year t and the logarithm of yearly imports in country i in year t. The variable T reatedi is the binary treatment variable.
This dummy equals one if the country received yearly Aid for Trade and is zero other-wise. It captures the possible differences between treatment and control group prior to the policy change. The variable P ostt is the binary time variable. The dummy equals
one if the year is post-Aid for Trade initiative and is zero if the year is pre-initiative. This variable captures the possible changes in exports or imports in absence of the policy change. The interaction term (T reatedi∗P ostt) multiplies the dummy variable T reatedt
and P ostt, which is the same as a dummy variable equal to one for those countries
re-ceiving yearly Aid for Trade in the period following the 2005 Aid for Trade initiative and is zero otherwise. The variable Xitcontains a set of variables to control for cross country
differences. The coefficients of interest are α3 and β3 and measure the level of export
and import in countries receiving Aid for Trade, relative to countries not receiving Aid for Trade, after the 2005 Aid for Trade initiative, relative to before the 2005 Aid for Trade initiative.
In addition, this thesis also examines the change in the rate at which exports and imports is growing pre- and post-Aid for Trade initiative. The graphs in Figure 2 suggest that not only the level of exports and imports looks changed pre- and post-initiative, but the rate at which exports and imports was growing looks changed as well. Therefore, a modified DID specification will be explored in addition to the standard one. Specifically, the thesis will explore whether there was a change in the export and import trend pre-versus post- initiative. The modified DID specifications take the following form:
log(Export)it = γ0+ γ1t + γ2T reatedi+ γ3t ∗ (T reatedi∗ P ostt) + γ4Xit+ ξit
with 1=1,...,N t=1995,...,2015 (4.3)
log(Import)it= δ0+ δ1t + δ2T reatedi+ δ3t ∗ (T reatedi∗ P ostt) + δ4Xit+ ξit
with 1=1,...,N t=1995,...,2015 (4.4)
Where log(Export)it and log(Import)it are the variables of interest and are
respec-tively the logarithm of yearly exports in country i in year t and the logarithm of yearly imports in country i in year t. The variable t is a time trend variable incremented by 1 in year 2005 for the first year following the initiative, 2 in year 2006 for the second year after the initiative, etc. Same for the years pre-initiative: t is incremented by -1 in year 2004 for the first year pre-initiative, -2 in year 2003 for the second year pre-initiative, etc. The variable T reatedi is the binary treatment variable. This dummy equals one
if the country received yearly Aid for Trade and is zero otherwise. The variable P ostt
is the binary time variable. The dummy equals one if the year is post-initiative and is zero if the year is pre-initiative. The variable Xit contains a set of variables to control
for cross country differences. The impact of treatment and time is allowed to show up through the change in the trend t. The coefficients of interest are γ3 and δ3 and measure
the change in export and import growth trends in countries receiving Aid for Trade, relative to countries not receiving Aid for Trade, after the 2005 WTO initiative, relative to before.
The key assumption behind the DID analysis is the parallel trend assumption. Mean-ing that in absence of the treatment, the average change in response variable would have been the same for both treatment and control groups (Abadie, 2005). Thus, in absence of the Aid for Trade initiative, the average change in exports and imports would have been the same for the group of countries receiving yearly Aid for Trade and the group of countries not receiving Aid for Trade. The parallel trend assumption is tested by estimating the following equation for the pre-initiative years 1995-2004:
log(T radeOutcome)it= δ0+ δ1t + δ2T reatedi+ δ4t ∗ T reatedi+ ξit
with 1=1,...,N t=1995,...,2004 (4.5)
Where (T radeOutcome)it is either the exports or imports in country i in year t.
Again, the variable t is a time trend variable and the variable T reatedi is the binary
treatment variable. If the parallel trend assumption holds, than the (t ∗ T reatedi)
coeffi-cient should be insignificant. Table 4.1 shows the test for the parallel trend assumption. The (t∗T reatedi) coefficients for log(Export) and log(Import) are insignificant. Meaning
that the parallel trend assumption holds and the DID analysis can be used. Table 4.1: Test for Parallel Trend Assumption
log(Export) log(Import) (1) (2) t 0.0837 0.0736 (0.00)*** (0.00)*** Treated 0.3244 -0.0000 (0.75) (1.00) t*Treated -0.0239 -0.0238 (0.32) (0.97)
Note: Table shows the test for parallel trend assumption in the years before the Aid for Trade initiative: 1995-2004. With (t*Treated) as the coefficient of interest. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.
As final notice, the control group consists of only 6 countries while the treatment group consists of 118 countries. A more extensive control group would be preferred in the DID estimation. However, there are only 6 developing countries not receiving yearly Aid for Trade with yearly export and import data available over the full period. Including developed countries would give difficulties for the parallel trend assumption. Thus, the thesis continues with using the control group including only developing countries.
To complement previous estimation, this thesis will try to measure the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of the recipient
country. If the difference-in-differences estimation suggests that Aid for Trade enhances the trade outcomes of recipient countries, interesting to know is by how much the trade outcomes would change if Aid for Trade is increased by one percent. The group of countries not receiving Aid for Trade will be dropped for these estimations and only the group of countries receiving yearly Aid for Trade will be used. I will estimate the DID of equations (4.1) and (4.2) with the newly specified treatment and control groups. In this setting, the treatment group will be the group of countries receiving large amounts of Aid for Trade and the control group will be the group of countries receiving small, but positive amounts of Aid for Trade. See Table A.4 in Appendix A for the list of countries classified by small or large amounts of Aid for Trade received and see Table A.5 for the descriptive statistics corresponding to the newly specified treatment and control group. Figure 4.1 and 4.2 show the same graphs as Figure 3.1 and 3.2 respectively, only using the newly defined control and treatment group with the control group receiving positive, but small amounts of Aid for Trade and the treatment group receiving large amounts of Aid for Trade. Noteworthy is the sharp increases in exports and imports after the 2005 Aid for Trade initiative for both the control group and the treatment group. If Aid for Trade really enhances the trade performance of recipient countries, expected is a more effective enhancement of trade performance for the treatment group in comparison of the control group. Examining the graphs does not suggest a more effective enhancement of trade performance for the treatment group. The differences between export and import levels of the treatment and control group remain relatively stable over time. However, using an empirical analysis gives a more reliable result than examining graphs, so the difference-in-differences analysis is estimated in Section 5.
4.2
Two Stage Least Squares
In addition to the DID estimation, an Ordinary Least Squares (OLS) estimation will be used to estimate the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of the recipient country. If the DID estimation suggests that Aid for Trade enhances the trade outcomes of recipient countries, interesting to know is by how much the trade outcomes would change if Aid for Trade is increased by one percent. Again, the group of countries not receiving Aid for Trade will be dropped for the TSLS estimation and only the group of countries receiving yearly Aid for Trade will be used. See Table A.1 in the Appendix for the full list of countries receiving yearly Aid for Trade. An important underlying assumption of performing an OLS regression is that the independent variables in the regression are exogenous. However, the Aid for Trade variable is possibly endogenous to exports and imports. The endogeneity problem is possibly due to reverse causality between Aid for Trade and the export and import of the recipient country. This is the case for example if faster performing countries tend to receive more Aid for Trade than others. As a result, the outcome of the regressions then generates an upward bias in the Aid for Trade coefficients (Cali & te Velde, 2011). This makes it difficult to isolate the effect of Aid for Trade on exports and imports of country recipient.
Figure 4.1: Log(Aid for Trade), Log(Exports), and Log(Import) for the new treatment and control group over the period 1995-2015.
Note: Control group is the group of countries receiving small, but positive amounts of Aid for Trade and the treatment group is the group of countries large amounts of Aid for Trade.
Figure 4.2: Log(Export) of treatment and control group over the period 1995-2015
Figure 4.3: Log(Import) of treatment and control group over the period 1995-2015
Note: Control group is the group of countries receiving small, but positive amounts of Aid for Trade and the treatment group is the group of countries large amounts of Aid for Trade.
A commonly used approach in order to address the reversed causality issue is to use an instrumental variable for the possibly endogenous Aid for Trade variable. Two assumptions must hold in order to use instrumental variables. First, the instrument chosen should be relevant. This entails that the instrument significantly influences the endogenous variable. This requirement can be tested in the first stage regression using a simple test of weakness of the instrument. Second, the instrument chosen should be exogenous. This entails that the instrument is uncorrelated with the error term. Usually, this second assumption cannot be formally tested. When the two assumptions hold a Two Stage Least Squares (TSLS) analysis can be employed. The TSLS model consists of two stages. In the first stage, the available instruments and the exogenous variables are regressed on the endogenous coefficient. This results in a predicted value for the endogenous coefficient. The predicted value estimated in the first stage is than used in the second stage to estimate the effect of the predicted value on the dependent variable (Stock & Watson, 2012).
In this thesis, I will instrument Aid for Trade by the interaction term of the post-Aid for Trade initiative dummy and the time trend variable t used in the DID trend-analyses estimated in equation (4.3) and (4.4). Subsequently, I proceed to estimate the effect of the predicted value of Aid for Trade on the trade outcomes in the TSLS approach. Formally, I will estimate the following equation:
log(T radeOutcome)it= πlog(Aidf orT rade)it+ ξit
with 1 = 1,...,N t = 1,..., T (4.6)
Where (T radeOutcome)it is the variable of interest and is either the logarithm of
yearly exports or the logarithm of yearly imports in country i in year t. The variable log(Aidf orT rade)it is the logarithm of yearly Aid for Trade in country i in year t. In
order to estimate equation (4.5) without endogeneity problems, log(Aidf orT rade)it will
be instrumented in the TSLS approach. The first stage is characterized as follows: log(Aidf orT rade)it= η1t + η2(P ostt∗ t) + η3Xit+ λi+ ξit
with i = 1,...I t = 1,..., T (4.7)
Equation (4.7) describes a fixed effect estimation where log(Aidf orT rade)it is the
dependent variable. The variable t is a time trend variable incremented by 1 in year 2005 for the first year following the initiative, 2 in year 2006 for the second year following the initiative, etc. Same for the years pre-initiative: t is incremented by -1 in year 2004 for the first year pre-initiative, -2 in year 2003 for the second year pre-initiative, etc. The variable P osttis the binary time variable. The dummy equals one if the year is post-Aid
for Trade initiative and is zero otherwise. The variable (P ostt∗ t) is the instrumental
variable for the logarithm of Aid for Trade. The variable Xit contains a set of variables
The reduced form is obtained by substituting the first stage equation (4.7) in the second stage equation (4.6). Hence, the reduced form is formally:
log(T radeOutcome)it= η1t + η2(P ostt∗ t) + η3Xit+ λi+ ξit
with i = 1,...,N t = 1,...,T (4.8)
In order to use the TSLS estimation, the relevance and exogenous assumption must hold. The first assumption is tested using the weakness of the instruments in the first stage. The first stage estimation will be discussed in Section 5.2. The second assumption entails that the instrument is uncorrelated with the error term. Since the post-Aid for Trade initiative dummy and the year dummies are exogenously determined, this is likely to be the case.
5
Results
5.1
Difference-in-Differences
Table 5.1 shows the DID results from equations (4.1) and (4.2). The coefficients of interest to this thesis are those corresponding to T reatedi ∗ P ostt. This coefficient
measures the level of exports or imports in countries receiving Aid for Trade, relative to countries not receiving Aid for Trade, after the 2005 Aid for Trade initiative, relative to before the 2005 Aid for Trade initiative. The coefficient for the logarithm of Export in column (1) indicates that, without controlling for country differences, exports increased 37 percent more in countries receiving yearly Aid for Trade, relative to countries not receiving Aid for Trade, after the Aid for Trade initiative, relate to before the initiative. This coefficient is strongly positive and highly significant. When controlling for country differences, the coefficient for the logarithm of export in column (2) shows a slightly lower, but still strongly positive and significant result. The coefficient now suggests that exports increased 33 percent more in countries receiving Aid for Trade, relative to countries not receiving Aid for Trade, after the initiative, relative to before the initiative. The coefficients for GDP, population and government effectiveness show the expected positive sign and are all highly significant, indicating that these coefficients positively effects the level of exports in country i. The coefficient on area size indicates an unexpected positive effect on exports, but this might be explained by the increasing amount of resources available in a larger country. The landlocked dummy shows the expected sign, but is not significant in this estimation.
Imports also experienced a boost in countries receiving yearly Aid for Trade. The coefficient in column (3) indicates that, without control for country differences, import increased 26 percent more in countries receiving Aid for Trade, relative to countries not receiving Aid for Trade, post-initiative compared to pre-initiative. This coefficient is highly significant. When controlling for country differences, the coefficient in column (4) indicates that import increased 24 percent more compared to countries not receiving Aid for trade, post-initiative, relative to pre-initiative. Again, this coefficient is highly significant. The control variables in column (4) show the same pattern as the control variables in column (2). In conclusion, table 3 finds strong empirical evidence that Aid for Trade increased the levels of exports and imports in countries receiving yearly Aid for Trade, relative to countries not receiving Aid for Trade, after the initiative, relative to before the initiative.
Table 5.2 contains the results for equations (4.3) and (4.4). The coefficients of the control variables included in column (2) and (4) follow the same patterns as the control variables in table 3. In these estimations, the impact of treatment and time is allowed to show up through the change in the trend t. The coefficients of interest is t ∗ T reatedi∗
P osttand measures the change in exports and imports growth rate in countries receiving
Aid for Trade, relative to countries not receiving Aid for Trade, after the 2005 WTO initiative, relative to before. Column (1) shows that, without controlling for country
Table 5.1: Difference-in-Differences regression using equations (4.1) and (4.2)
log(Export) log(Export) log(Import) log(Import)
(1) (2) (3) (4) Treated 0.3979 -0.0232 0.0830 -0.0798 (0.71) (0.99) (0.92) (0.93) Post 0.6527 0.6371 0.8091 0.7779 (0.00)*** (0.00)*** (0.00)*** (0.00)*** Treated*Post 0.3730 0.3345 0.2644 0.2416 (0.00)*** (0.03)** (0.00)*** (0.04)** GDP 0.0001 0.0001 (0.00)*** (0.00)*** Population 0.0047 0.0056 (0.00)*** (0.00)*** Government Effectiveness 0.2489 0.1984 (0.00)*** (0.08)*** Area Size 0.5172 0.2284 (0.00)*** (0.08)* Landlocked -0.4506 -0.5202 (0.38) (0.17) R2 (within) 0.5181 0.5519 0.6908 0.7212 N 2203 2135 2214 2145
Note: Treatment group when country receives Aid for Trade, Control group when country not receives Aid for Trade. GDP is in billions of USD, Population and Area Size are in millions. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1%
level
differences, the export growth rate of countries receiving yearly Aid for Trade increased 0.1546 percentage points per year more after the initiative, relative to before initiative, than countries not receiving Aid for Trade. However, when controlling for country differences the result is not significant and thus suggests that the growth rate in exports of countries receiving yearly Aid for Trade did not increase more than countries not receiving Aid for Trade, after the initiative, relative to before the initiative. Nevertheless, the growth rate in imports did experienced a boost. Without controlling for country differences, the rate of growth in imports in countries receiving yearly Aid for Trade became 0.0284 percentage points per year larger post-initiative, relative to pre-initiative, compared with countries not receiving Aid for Trade. With controlling for country differences there is a 0.0266 percentage points per year increase in the growth rate of imports for countries receiving Aid for Trade compared to countries not receiving Aid for Trade, post-2005 initiative, relative to pre-initiative. Concluding, table 4 finds strong empirical evidence that the growth rate of imports experienced a boost in countries receiving yearly Aid for Trade, relative to countries not receiving Aid for Trade, after the
2005 initiative, relative to before. While the growth rate of exports remained relatively the same for countries receiving yearly Aid for Trade, compared to countries not receiving Aid for Trade, after the initiative, relative to before the initiative.
Table 5.2: Difference-in-Differences trend regression using equations (4.3) and (4.4) log(Export) log(Export) log(Import) log(Import)
(1) (2) (3) (4) t 0.0792 0.0798 0.0767 0.0744 (0.00)*** (0.00)*** (0.00)*** (0.00)*** Treated 0.5787 0.1393 0.1575 -0.0652 (0.57) (0.91) (0.84) (0.94) t*Treated*Post 0.1546 0.0093 0.0284 0.0266 (0.01)*** (0.14) (0.00)*** (0.00)*** GDP 0.0001 0.1480 (0.01)*** (0.01)*** Population 0.0027 0.0036 (0.01)*** (0.00)*** Government Effectiveness 0.2053 0.1480 (0.00)*** (0.00)*** Area Size 0.7047 0.4160 (0.00)*** (0.01)*** Landlocked -0.5335 -0.6092 (0.27) (0.09)* R2 (within) 0.5983 0.6229 0.7920 0.8079 N 2203 2135 2214 2145
Note: Treatment group when country receives Aid for Trade, Control group when country not receives Aid for Trade. GDP is in billions of USD, Population and Area Size are in millions. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1%
level
Table 5.3 contains the results from the DID estimation using a different control and treatment group. The new control group is the group of countries receiving small, but positive amounts of Aid for Trade and the treatment group is the group of countries receiving large amounts of Aid for Trade. The coefficients of interest to this thesis are those corresponding to T reatedi∗ P ostt. This coefficient measures the level of exports
and imports in countries receiving large amounts of Aid for Trade, relative to countries receiving small amounts of Aid for Trade, after the Aid for Trade initiative, relative to before initiative. The coefficient for the logarithm of exports in column (1) indicates that, without controlling for country differences, exports increased 58 percent more in countries receiving high intensity Aid for Trade than exports in countries receiving small intensity Aid for Trade. When controlling for country differences, the coefficient for the logarithm of exports in column (2) shows a slightly lower, but still strongly positive and significant result. The coefficient now indicates that exports increased 52 percent more
Table 5.3: Difference-in-Differences regression using equations (4.1) and (4.2)
log(Export) log(Export) log(Import) log(Import)
(1) (2) (3) (4) Treated 1.8377 1.5978 1.4872 1.3340 (0.00)*** (0.00)*** (0.00)*** (0.00)*** Post 0.6498 0.6417 0.7952 0.7890 (0.00)*** (0.00)*** (0.00)*** (0.00)*** Treated*Post 0.5803 0.5271 0.4323 0.3716 (0.00)*** (0.03)** (0.00)*** (0.00)** GDP 0.0001 0.0001 (0.00)*** (0.00)*** Population 0.0032 0.0044 (0.00)*** (0.00)*** Government Effectiveness 0.2323 0.1857 (0.00)*** (0.00)*** Area Size 0.4820 0.1954 (0.00)*** (0.11) Landlocked -0.9921 -0.9593 (0.04)** (0.01)*** R2 (within) 0.5661 0.5815 0.7193 0.7389 N 2097 2097 2107 2107
Note: Control group is group of countries receiving positive, but small amounts of Aid for Trade and the treatment group is the group of countries receiving large amounts of Aid for Trade. GDP is in
billions of USD, Population and Area Size are in millions. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1% level
in countries receiving high intensity Aid for Trade, post-2005 relative to pre-2005, than exports in countries receiving low intensity Aid for Trade. The control variables show the same signs as in previous estimations, with as only difference the landlocked coefficient shows a significant result now. Again, imports also experienced a boost in countries receiving high intensity Aid for Trade. The coefficient for the logarithm of imports in column (3) indicates that, without control for country differences, import increased 43 percent more in countries receiving high intensity Aid for Trade, relative to countries receiving low intensity Aid for Trade, post-initiative compared to pre-initiative. When controlling for country differences, the coefficient in column 4 indicates that import increased 37 percent more compared to countries receiving low intensity Aid for trade, post-initiative relative to pre-initiative. Again, this coefficient is highly significant. The control variables in column 4 show the same pattern as the control variables in previous estimations. In conclusion, the results in table 5.3 are in line with the expectation that Aid for Trade enhances the trade outcomes of recipient countries. According to the results, receiving more Aid for Trade enhances the exports and imports of recipient
countries more effectively than receiving low Aid for Trade, post-initiative compared to pre-initiative.
5.2
Two Stage Least Squares
Table 5.4 and 5.5 report the results of the TSLS regression analyses presented in equation (4.6) to (4.8).
The first stage of the TSLS procedure with the logarithm of Aid for Trade as de-pendent variable is shown in table 5.4. The estimation provides information about the correlation between the instrument and the instrumented variable. The instrument is valid if it shows strong correlation with the endogenous regressor. There are two differ-ent first stages used: column (1) shows the first stage for estimating recipidiffer-ent exports, column (2) shows the first stage for estimating recipient imports. The results slightly differ, this is explained by the differences in observations between exports and imports. For some countries, data on the export and/or import flows are missing what resulted in an unbalanced panel data set. The instrument P ostt∗ t shows strong correlation with
the logarithm of Aid for Trade suggesting the instrument to be relevant in explaining Aid for Trade. The relevant assumption is formally tested using the F-statistic. As a common rule of thumb the F-statistic should be larger than 10 for the instrument to be relevant (Stock & Watson, 2012). The F-statistic corresponding to the first stages in table 5 shows that the instrument is strong in explaining the logarithm of Aid for Trade. Thus, the relevance assumption holds and the TSLS estimation can be used. As final notice, the control variables for area size and the landlocked dummy are deleted from the estimation. By performing the TSLS it was omitted because of multicollinearity.
Table 5.5 presents the effect of the logarithm of Aid for Trade on the logarithm of yearly exports or imports using P ostt∗ t as an instrument. The R2 is negative in the
TSLS regression and therefore not reported as it has no statistical meaning. Applying a TSLS estimation results in a substantial positive and statistically significant effect of Aid for Trade on the yearly exports and imports of recipient. The results in column (1) shows that a one percent increase in the amount of Aid for Trade received increases exports of the recipient by 1.06%. This result is significant on the 1% level. The result suggests that the elasticity of Aid for Trade on recipient exports is close to 1. The results in column (2) suggest that a one percent increase in the amount of Aid for Trade received increases imports of the recipient by 1.15%. Again, this result is significant at the 1% level. The control variables GDP and Government Effectiveness show the expected positive and significant effects. The control variable for Population shows an insignificant result. In conclusion, the TSLS results suggest Aid for Trade to have a positive, statistically significant effect on trade outcomes of the recipient countries with an elasticity bigger than 1. Suggesting that an extra dollar received of Aid for Trade results in more than a dollar increase of exports and imports of recipient country. This finding is in sharp agreement with previous literature and the expectations prior to the results.
coun-Table 5.4: First-stage regression of instruments and exogenous variables on the logarithm of Aid for Trade using equation (4.7)
log(Aid for Trade) log(Aid for Trade)
(1) (2) Instrumental Variable Post*t 0.1213 0.1201 (0.00)*** (0.00)*** Exogenous Variables GDP -0.0005 -0.0005 (0.00)*** (0.00)*** Population 0.0059 0.0061 (0.02)** (0.02)** Government Effectiveness 0.4877 (0.4719) (0.00)*** (0.00)***
Fixed Effect Yes Yes
R2 (within) 0.1557 0.1531
N 2027 2036
F-statistic 34.72 34.69
Note: Column (1) shows the first stage for estimating the effect of Aid for Trade on exports, column (2) shows the first stage for estimating the effect of Aid for Trade on imports. GDP is in billions of USD, Population and Area Size are in millions. Standard errors are in parentheses: *significant at the
10% level; **significant at the 5% level; ***significant at the 1% level
Table 5.5: Effect of the logarithm of Aid for Trade on the logarithm of yearly exports and imports using Post*t as an instrument using equation (4.6) and (4.7)
log(Export) log(Import)
(1) (2)
log(Aid for Trade) 1.0651 1.1591 (0.00)*** (0.00)*** GDP 0.0006 0.0006 (0.00)*** (0.00)*** Population -0.0006 -0.0006 (0.84) (0.83) Government Effectiveness -0.2723 -0.3766 (0.05)** (0.01)****
Fixed Effect Yes Yes
R2 (within)
N 2027 2036
Note: GDP is in billions of USD, Population and Area Size are in millions. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1% level
tries to assess whether total Aid for Trade is particularly effective where it may be needed the most. First, all recipient countries are classified per region. Interesting is evaluating the effects of Aid for Trade on the trade outcomes of recipient countries located south and north of the Sahara, assuming that many countries in these regions are still lagging behind in the global trade flows. Most Asian countries are well known to be closely integrated into the world market, same is for large parts of Latin America (H¨uhne et al, 2014). See Table A.6 in Appendix A for the list of countries classified by region. Second, all recipient countries are classified according to income group. This is classified according to the World Banks World Development Indicators (2016). Assuming that low income countries are particularly vulnerable for internal barriers to trade, I would expect that Aid for Trade is more effective if it benefited low income groups. See Table A.7 in Appendix A for the full list of countries classified by income.
Table 5.6 shows the results for the selected regions. The corresponding first stage F-statistic shows whether the instrument P ostt∗t is relevant in explaining Aid for Trade for
that particular sample. Only for countries located in the Oceania region the instrument is not relevant. Thus, the effect shown of Aid for Trade on the trade flows for the region Oceania can be biased. For all other regions, the F-statistic is larger than 10 and the results can be used. The effects of total Aid for Trade vary considerable between the different regions, although all results show a highly significant and positive effect. It seems that total Aid for Trade, with controlling for country differences and country fixed effects, is most effective in enhancing the exports of countries in East Asia, Europe and South America. For imports, countries located in East Asia, South America and South America show the most effective enhancement as result of Aid for Trade. Remarkable is that countries located North of the Sahara, the Middle East and South of the Sahara show the least effective enhancement of exports and imports as a result of Aid for Trade flows. Thus, the countries who are still lagging behind in global trade flows are the once who are benefiting least of an extra dollar of Aid for Trade received.
Table 5.7 shows the results for the four income groups. The effects of total Aid for Trade on exports and imports of recipient vary considerable between the income groups, although all results show a highly significant and positive effect. By interpreting the results for the low income group and the high income group, there is need for some caution. The first stage F-statistics for these groups are lower than 10, suggesting that the instrument used P ostt∗ t is not relevant in explaining Aid for Trade. Thus, the
effect shown of Aid for Trade on the trade flows for these income groups can be biased. Noteworthy is that the effect of total Aid for Trade on the exports and imports of the upper middle income group is considerably more effective than of the other income groups. This seems counter intuitive, assuming that low income countries are particularly vulnerable for internal barriers to trade and thus Aid for Trade would be expected to be more effective for those countries. However, this difference in effectiveness for exports could be explained by the difference in composition. Primary commodities rule the supply of the low income group, while manufactured goods rule the supply of the upper middle income group. Generally, manufactured goods respond faster to improved stimulus than primary commodities (H¨uhne et al, 2014). The difference in effectiveness
Table 5.6: Effect of the logarithm of Aid for Trade on the logarithm of yearly exports and imports using Post*t as an instrument: sub-samples of recipients by region
log(Export) First Stage log(Import) First Stage F-statistic F-statistic
(1) (2) (3) (4)
Europe 2.0408 14.27 1.6801 14.27
(0.23) (0.23)
North Sahara & Middle East 0.5846 22.84 0.8390 22.84
(0.00)*** (0.00)*** South Sahara 1.1180 30.64 1.0929 30.74 (0.00)*** (0.00)*** Central America 0.8887 13.71 1.3783 12.86 (0.00)*** (0.00)*** South America 2.0503 20.80 2.1645 21.16 (0.00)*** (0.00)*** East Asia 3.4563 21.47 3.6076 21.47 (0.20) (0.20) South Asia 1.9260 33.00 2.4301 33.00 (0.00)*** (0.00)*** Oceania 0.3209 1.43 0.5177 2.96 (0.00)*** (0.00)***
Control Variables Yes Yes Yes Yes
Country Fixed Effects Yes Yes Yes Yes
Note: The effect of Aid for Trade on the export and import of recipient, categorised by region. Column (1) shows the corresponding first stage F-statistic corresponding to log(export) for countries in that
particular region and column (2) shows the corresponding first stage F-statistic corresponding to log(import) for countries in that particular region. Standard errors are in parentheses: *significant at
the 10% level; **significant at the 5% level; ***significant at the 1% level
for imports could be explained by the higher demand followed by higher incomes in the income group.
Table 5.7: Effect of the logarithm of Aid for Trade on the logarithm of yearly exports and imports using Post*t as an instrument: sub-samples of recipients by income group
log(Export) First Stage log(Import) First Stage F-statistic F-statistic
(1) (2) (3) (4)
Low Income Group 0.6897 8.79 0.6922 8.79
(0.00)*** (0.00)***
Lower Middle Income Group 1.0295 36.11 1.0856 36.77
(0.00)*** (0.00)***
Upper Middle Income Group 1.3425 19.28 1.4779 19.98
(0.00)*** (0.00)***
High Income Group 0.5068 8.30 0.8651 7.77
(0.00)*** (0.03)**
Control Variables Yes Yes Yes Yes
Country Fixed Effects Yes Yes Yes Yes
Note: The effect of Aid for Trade on the export and import of recipient, categorised by income group. Column (1) shows the corresponding first stage F-statistic corresponding to log(export) for countries in
that particular income group and column (2) shows the corresponding first stage F-statistic corresponding to log(import) for countries in that particular income group. Standard errors are in parentheses: *significant at the 10% level; **significant at the 5% level; ***significant at the 1% level
6
Conclusion
The Aid for Trade initiative was launched at the Ministerial Conference of the World Trade Organization (WTO) in Hong Kong in 2005. It was agreed to expand aid to support developing countries in increasing exports of goods and services, and benefiting from free trade and increased market access. Even if trade-related Aid has always existed, the official creation of Aid for Trade has put a new light on these specific Aid flows and launched discussions and debates on their effectiveness. This thesis studies whether Aid for Trade enhances the trade outcomes of recipient countries by examining the 2005 Aid for Trade initiative.
This thesis is the first, to the best of my knowledge, to fill in this evaluation gap by exploring a difference-in-differences estimation using the 2005 Aid for Trade initiative as the considered treatment. The findings of the difference-in-differences estimation suggest that exports increased 33% more in countries receiving yearly Aid for Trade, post-Aid for Trade initiative relative to pre-initiative, than exports in countries not receiving Aid for Trade. Further, the results suggest that imports increased 24% more in countries receiving yearly Aid for Trade, post-initiative relative to pre-initiative, compared to countries not receiving Aid for Trade. The thesis also finds that the rate of growth of imports increased more in countries receiving yearly Aid for Trade, post-initiative relative to pre-initiative, than countries not receiving Aid for Trade. However, the control group used in this thesis consists of only 6 countries. A more extensive control group would be preferred.
To complement the difference-in-differences estimation, this thesis uses a Two Stage Least Squares method to estimate the monetary effect of an extra dollar of Aid for Trade received on the trade outcomes of recipient countries. Using an interaction term of the post-Aid for Trade initiative dummy and the time trend variable t as instrumental variable for Aid for Trade. The TSLS results suggest Aid for Trade to have a positive, statistically significant effect on trade outcomes of recipient countries. The findings suggest that an increase by one percent in Aid for Trade received increases yearly exports by 1.06% and yearly imports by 1.15% for the recipient. When re-estimating the TSLS for various sub-samples, the results suggest that the countries who are still lagging behind in global trade flows are the once who are benefiting least of an extra dollar of Aid for Trade received.
These findings point to some tentative policy implications. First, the difference-in-differences estimation suggests that post-Aid for Trade initiative the effectiveness of Aid for Trade on the trade outcomes of recipient has increased. Further research may provide additional insights on the change in effectiveness resulting from the Aid for Trade initiative. More importantly, is the change in effectiveness appointed by the expanding Aid for Trade flows or by changes in the redefined scope of Aid for Trade. Second, targeting Aid for Trade more strongly in line with regional needs may insure greater effectiveness of Aid for Trade on the trade performance of recipient countries. This is suggested by the TSLS estimation for different subgroups of Aid for Trade recipients.
Third, targeting Aid for Trade more strongly to the enhancement of primary commodities may insure greater effectiveness of Aid for Trade on the trade performance of the Least Developed Countries. Suggesting that the supply of Least Developed Countries are dominated by primary commodities.
7
References
Abadie, A. (2005). Semiparametric Difference-in-Differences Estimators. The Review of Economic Studies, 72(1), 1-19.
Cali, M. & te Velde, D. W. (2011). Does Aid for Trade Really Improve Trade Per-formance? World Development, 39(5), 725-740.
Freedom House (2009). Freedom in the world country rating (online database). Re-trieved from: https://freedomhouse.org/report/freedom-world
/freedom-world-2009
Gartzke, E. (2009). The affinity of Nation Index, 1946-2002, version 4.0 and related documents. Retrieved from: https://www.researchgate.net/
publication/254145743 The Affinity of Nations Index 1946-2002
Helble, M., Mann, C. L., & Wilson, J.S. (2012). Aid-for-Trade Facilitation. Review of World Economics, 148(2), 357-376.
Hotchkiss, J. L., Moore, E., & Zobay, S. M. (2002). The Impact of the 1996 Summer Olympic Games on Employment and Wages in Georgia. Southern Economic Journal, 69(January 2003), 691-704.
H¨uhne, P., Meyer, B., & Nunnenkamp, P. (2014). Who Benefits from Aid for Trade? Comparing the Effects of Recipient versus Donor Exports. The Journal of Development Studies, 50(9), 1275-1288.
NBER (the National Bureau of Economic Research). (2007). Difference-in-Differences Estimation. Imbens/Wooldridge, Lecture Notes 10, Summer 07, 1-19.
OECD (Organisation for Economic Co-operation and Development) (2006). Aid for Trade: Making It Effective. Paris: OECD, 1-94.
OECD (Organisation for Economic Co-operation and Development) (2016). Aid-for-trade data: Creditor Reporting System Explanatory Note. OECD, 1-8.
OECD/DAC (2017). Creditor Reporting System Database. Retrieved from: https://stats.oecd.org/Index.aspx?DataSetCode=CRS1
Rodrik, D., Subramanian, A., & Trebbi, F. (2004). Institutions rule: The primacy of institutions over geography and integration in economic development. Journal of Eco-nomic Growth, 9(2), 131165.
Stiglitz, J. E., & Charlton, A. (2006). Aid for Trade. International Journal of De-velopment Issues, 5(2), 1-41.
Stock, J. H., & Watson, M. (2012). Introduction to Econometrics. (3nd ed., pp. 354-382). Harlow: Pearson.
UNCTAD (2006). The Least Developed Countries Report: Developing Productive Capacities. UN: Geneva
Vijil, M., Huchet-Bourdon, M., & le Mou¨el, C. (2011). Aid for Trade: A Survey. French Agency for Development Working Paper, 110, 1-47.
Vijil, M. & Wagner, L. (2012). Does Aid for Trade enhance export performance? Investigating the Infrastructure channel. World Economy, 35, 838-886.
Wade, R. (1992). East Asias Economic Success: Conflicting Perspectives, Partial Insights, Shaky Evidence. World Politics, 44(1), 270-320.
Winters, A., McCulloch, N., & McKay, A. (2004). Trade Liberalization and Poverty: the Evidence so Far. Journal of Economic Literature, 42 (1), 72-115.
World Bank (2016). World Development Indicators. Retrieved from: http://data.worldbank.org/data-catalog/world-development-indicators
World Bank (2016). The Worldwide Governance Indicators. Retrieved from: http://info.worldbank.org/governance/wgi/home
World Bank (2016). World Integrated Trade Solutions. Retrieved from: http://wits.worldbank.org/
WTO (2006). Recommendations of the Task Force on Aid for Trade. Official publi-cation: WT/AFT/1. 27 July 2006.
