Foreign aid and tax transition in developing countries
Transcription
Foreign aid and tax transition in developing countries
Official Development Assistance and Tax Transition *** Gbewopo Attila† Gérard Chambas Jean-Louis Combes *** Clermont University, University of Clermont I, CERDI-UMR CNRS *** March 2009. Abstract: This paper examines the impact of Official Development Assistance (ODA) on tax transition in developing countries. In an initial approach we constructed qualitative indicators of tax transition and considered by what mechanisms ODA affects the adoption of tax transition reform, and then we used a non linear model of probability to evaluate the probability of the adoption of tax transition reform and extract the effect of ODA. Using this model on a sample of 106 developing countries over the period 1980-2005, it was demonstrated, after a check using structural factors and endogenous bias correction, that official development assistance significantly accelerates the transition. This result is robust to several specifications based on alternative measures of both foreign aid and tax transition. Résumé : Cet article analyse l’impact de l’aide publique au développement sur la transition fiscale dans les pays en développement. Après avoir construit des indicateurs qualitatifs de transition fiscale et envisagé les mécanismes par lesquels l’APD affecte l’adoption d’une réforme de transition fiscale, on utilise un modèle non linéaire de probabilité pour évaluer la probabilité d’adoption d’une réforme de transition fiscale et dégager l’effet de l’APD. De ce modèle estimé sur un échantillon de 106 pays en développement sur la période 19802005, il ressort, après contrôle par des facteurs structurels et correction du biais d’endogénéité, que l’APD favorise le succès de la transition fiscale. Ce résultat est robuste à différentes spécifications : définition de l’aide et de la transition fiscale. Key words: Tax transition, official development assistance. JEL classification : H20, H30, O19. ² † Clermont University, University of Auvergne, CERDI UMR CNRS 6589. This study has had the support of the National Research Agency (ANR). The authors alone are responsible for the contents of this study. For all correspondence, write to Gbewopo Attila, CERDI, 65 Boulevard François Mitterrand 63000 Clermont Ferrand, France. E-mail : [email protected] . Tel: + 33 6 67 32 73 13; Fax: + 33 4 73 17 74 28. 1 Introduction In order to reduce the economic distortions brought about by fiscal deductions, most developing countries set up during the seventies tariff reduction policies as essential components for reform of trade liberalization. These tariff reduction policies resulted in major tax losses. Insofar as the State in developing countries succeeds in providing public goods reasonably efficiently but insufficiently abundantly, it is essential that the change in taxation structure is compatible with the stabilisation or even the reinforcing of public finance. The aim must be, during a phase of tax transition characterised by a drastic change in the structure of taxation, not to diminish the supply of public goods (the institutional framework, public infrastructure, education and health services…) indispensable for economic development (Lucas, 1988; Romer, 1988; Easterly, 2003) and for the countries concerned to attain to the MDG (Millennium Development Goals). In a nutshell, it involves setting up fiscal reform within the Development and the Reduction of Poverty Strategies. In order to compensate for the loss of revenue from tariffs and to stabilise the overall level of their public revenue, tax transition policies aim at reducing the contribution to public finance of taxes on foreign trade at the origin of major distortions (Berg and Krueger, 2003) and replacing them by domestic tax revenue. Tax transition reforms thus are based in particular upon a reinforcement of indirect domestic taxation, especially VAT from now on installed in practically all developing countries (Ebrill et alii, 2001). In recent years, a good deal of literature has covered the effects of aid, in particular macroeconomic ones: effect on growth, on savings, on investment and even on public spending. Quite recently, research has been carried out on the effect of aid on the level of public taxation and the behaviour of States in matters of fiscal operations (Heller, 1975; White, 1992; McGillivray, 2000; Franco-Rodriguez, 2000; Mavrotas, 2002, Ouattara, 2006)1. However, a gap can be observed: to our knowledge no analysis has covered the effects of aid on tax transition. Now, it is true that tax transition reforms constituted a major phenomenon in the economic policies of developing countries towards the end of the twentieth century because they were the result of a gradual renunciation of economic protection policies, which until then were widely practised, and conversely of a choice in favour of a free trade policy. In addition, Official Development Assistance appears liable to have a major influence upon tax transition in three different ways. The awarding of certain flows of aid can be conditional upon tax transition reform. Such reform consisting of a coherent package of coordinated measures aimed at ensuring the freeing up of tariffs and concomitantly compensating for the loss of revenue thus occasioned. Quite frequently, free-trade policy and tax reform programmes are implemented separately; in this instance, aid may buttress in different ways one or another 1 For all these contributions see Brun et alii, 2007. 2 of the components of tax transition reforms and thus promote tax transition in an unequal manner. ODA can also facilitate tax transition reform by compensating for certain costs stemming from tax transition reform; these costs may be budgetary (in particular the loss of tariff revenue), economic (the transformation of economic activity) or social (the social cost brought about by the loss of jobs). However, one cannot exclude an adverse effect of aid on tax transition reform insofar as aid might lead to a status quo by compensating for the loss of collective welfare in the absence of reform. Finally, technical assistance through ODA and leverage provided by ODA for promoting reform can make it easier for putting reforms into action. Indeed, the risk, or at least its perception by the authorities, making an obstacle for reform, can be considerably reduced by technical assistance. Thus the objective of this paper is to fill a crucial gap both in aid policies and in economic policies of developing countries by highlighting the impact of ODA on tax transition. In order to achieve this objective, we must first determine the criteria characteristic of tax transition (a reduction in taxes on foreign trade in favour of domestic taxes and an evolution of the fiscal behaviour of State apprehended by an indicator of fiscal effort). The various channels of transmission of the effects of ODA on tax transition are subsequently highlighted. Finally the different hypotheses relative to the effects of aid on tax transition are submitted to an econometric test. 3 Section I- Tax transition in Developing Countries: an evaluation 1.1- What choice of tax transition indicators? Constructing a tax transition indicator requires taking into consideration criteria relative both to the structure of taxation and the level of public taxation. Structural criteria of public taxation Tax transition consists of the reduction of tariff revenue and its substitution by domestic fiscal revenue insofar as, in relation to other taxes (domestic direct or indirect taxation), tariffs are at the origin of the major fiscal distortions2. In a general way, tax transition in Developing Countries has taken place broadly in the form of domestic indirect taxation, in particular VAT. It should be noted that the introduction of VAT and the substitution3 of the latter for former turnover taxes has improved the economic neutrality of domestic indirect taxation, doing so progressively, step by step as VAT legislation has been modernised (broadening of the base, generalisation of the VAT mechanism). The improvement in the administration of VAT has certainly also constituted a factor in favour of a more neutral tax (Ebrill et al., 2001; Keen and Lockwood, 2007). In some countries with intermediate income such as Morocco, tax transition has, however, been implemented by direct taxation through an increase in company taxes (Brun et al, 2007). Criteria of the level of public taxation Tax transition criteria can be brought to light based on the evolution of the level of effective public taxation. However, this kind of criterion has the disadvantage of being a subjective reference, the former level of public taxation corresponding to low taxation or, on the contrary high taxation, depending upon the characteristics of the country. For example, it is difficult to consider that a country going through a phase of particularly low public revenue succeeds its tax transition because it manages to modify the structure of taxation while at the same time maintaining an abnormally low level of public finance. 2 Because of the lack of data, it is not possible to take into account the transformation of quantitative restrictions in tariffs occurring in the first phases of liberalisation and which put into practice a trade freeing up despite an increase in tariff revenue. 3 This substitution began in 1960 for the Ivory Coast and has continued ever since. 4 In order to overcome this difficulty, the standard of the level of taxation retained is the potential of public resources determined by the structural characteristics of each country4. Indicators of transition In order to be in a position to discern transitions essentially taking place towards VAT compared to transitions taking place towards other categories of public revenue, the process of tax transition is identified through two main criteria. According to the first indicator (indicator 1), during the course of a year, a country is considered in tax transition if two conditions A and B are satisfied conjointly. Condition A is fulfilled in one of two possibilities. The ratio of tariff revenue to GDP drops over a period of 5 years (the fall in this revenue is measured by the level of growth calculated from an exponential adjustment). Tariff revenue represents at the most 10%5 of total public revenue. Condition B: The level of public revenue of a country approaches its structural potential of revenue: public revenue used over the period representing at least 90% of the fiscal potential. The second indicator (indicator 2) is more restrictive than the first one since it makes it possible to identify, in countries in tax transition meeting indicator 1, those whose tax transition is taking place towards VAT. That is to say that to conditions A and B, a condition C is added: the ratio of revenue from VAT to GDP increases over 5 years (this increase is measured through the level of growth calculated from an exponential adjustment). 1.2- Tax transition reform: an observation Applying indicator 1 indicates a frequency of occurrence of tax transition of 75% of observations for all developing countries over the period 1980-2005. This frequency varies according to the groups of countries and reflects various situations with respect to the implementation of tax transition reform (table 2)6. 4 The fiscal potential is calculated as a predicted value of the estimated equation of public taxation on structural variables (see Brun, Chambas, Combes, 2006): the real product per capita as a variable capturing to the level of development, the trade openness and the share of exports of mining and petroleum products in total exports of goods as variables of the international trade, the population as a variable of size and agriculture value added as a variable of the structure of the economy. A positive impact is expected for the first four variables and negative for the last one. The equation is estimated on a sample of 106 countries on annual data over the 1980-2005 period by the method of least ordinary squares with individual fixed effects. The fixed effects represent nonobserved structural factors, constant in time and which determine the fiscal potential. Table A2 in the Annexes presents results conforming to the predictions except for the non-significant variable of population. 5 The initial ratio of tariff revenue compared to total revenue was about 25% before the trade liberalization; the threshold of 10% retained thus corresponds to a country well advanced in free trade policies. 6 Descriptive statistics relative to indicator 2 are shown in table A3 in the Annexes 5 Table 2: Frequency of occurrence of the transition per group of countries (period 1980-2005) Region Eastern Asia and the Pacific Eastern Europe and Central Asia Latin America and the Caribbean North Africa and the Middle East Southern Asia Sub-Saharan Africa All developing countries 1980-1995 Transition Transition completed not completed 31 9 77,5% 22,5% 22 2 91,67% 8,33% 129 35 78,66% 21,34% 61 25 70,93% 29,07% 32 10 76,19% 23,81% 81 63 56,25% 43,75% 356 144 71,2% 28,8% 1995-2005 Transition Transition completed Not completed 24 4 85,71% 14,29% 64 12 84,21% 15,79% 73 15 82,95% 17,05% 39 6 86,67% 13,33% 16 5 76,19% 23,81% 100 42 70,42% 29,58% 316 84 79% 21% 1980-2005 Transition Transition completed Not completed 55 13 80.9% 19.1% 86 14 86.0% 14.0% 202 50 80.2% 19.8% 100 31 76.3% 23.7% 48 15 76.2% 23.8% 181 105 63.3% 36.7% 672 228 74.7% 25.3% NB: The first statistic refers to the number of countries and the second to the frequency of the occurrence. Source: the authors. Table 2 shows for the period 1980-2005 a lesser frequency of occurrence of transition in the sub-Saharan African countries. This situation can be observed particularly in 19801995. It is less great over the second period.7 As for all reform of economic policy, an essential question is to appreciate the degree of fragility of the process of reform; in other words, what risk is there of a return to the former situation when tax transition is put into practice? In order to evaluate the more or less reversible aspect of the tax transition reform, we shall evaluate the conditional probabilities on the initiation situation (table 3). 7 The statistics stemming from indicator 2 (table A3 in the Annexes) show little difference compared to indicator 1. This observation corroborates the idea according to which VAT forms an integral part of a group of measures aiming at stabilising and making fiscal administration more efficient. However, an increase in VAT revenue seems necessary to fully ensure the success of transition. Thus, some countries have experienced a slowing down of reform which can be explained in part by an insufficient increase in VAT revenue. The following are involved, for example, South Africa, Kenya and Algeria over the period 1990-1995 or even Mexico or Chile during the 1980s. 6 Table 3: Conditional probability of transition Eastern Asia and the Pacific Transition in t Transition in t given absence given transition in t-1 of transition in t-1e 28.6% 94.7% Eastern Europe and Central Asia 94.6% 53.8% Latin American and the Caribbean 88.8% 37.9% North Africa and the Middle East 89.7% 38.5% Southern Asia 85.5% 36.4% Sub-Saharan Africa 81.8% 32.2% 86.4% 34.4% All Developing Countries NB: The calculation is based on indicator 1 According to table 3, most countries in transition in year t-1 are also in transition in year t. The process of transition thus appears as broadly irreversible. However, the irreversibility is less in Sub-Saharan Africa and in Southern Asia which include most countries with a low income. In these two regions, one third of the countries only which were not in transition in a particular year became so the following year: it is more difficult to make reform effective when it did not take place in the previous year. This situation can be put down to the greater cost of tax transition reform and also a lesser capacity to make effective a complex reform whose efficiency requires in particular good administrative capacity and a favourable environment. 7 Section II- Development aid and tax transition In order to better elucidate the role of aid in the process of tax transition we shall first of all examine the obstacles which reform in developing countries comes up against. Then we shall take a look at the effect of aid when confronted with these obstacles. 2.1- The obstacles to the reform of tax transition Tax transition is a large scale structural reform; it is confronted with four categories of obstacles. A first obstacle stems from the uncertainty in the evaluation of the costs and advantages of reform. Just as any major reform, tax transition reform is at the heart of a climate of uncertainty. It is complex to describe the modalities of tax transition reform and depending upon the fundamental approach various choices are possible (timing, coordination of the measures…). This uncertainty is all the greater when reform is undertaken in vulnerable countries, subject to shocks and where forecasting is chancy. Uncertainty tends to create resistance by decision-makers, actors of the economy and, in a more general way, citizens as a whole and consumers in particular. Because of uncertainly, this resistance can also concern groups which expect to benefit from the reform. Thus, experience shows that actors of the economy benefitting from the reform often take a wait and see stance or even associate themselves with those losing out from the reform (Drazen, 1996). Uncertainty concerning the spread of costs and benefits of the reform thus constitutes a factor reinforcing opposition (Fernandez, Rodrik, 1991) and the maintenance of the status quo (Laban, Sturzenegger, 1994). A second category of obstacles is due to the institutional characteristics of the country. The quality of institutions determines in part a capacity to implement tax transition reform which demands a constant effort over a number of years. A third category of obstacles stems from the opposition of special interest groups liable to be negatively affected by the reform. This is true especially concerning costs borne because of trade liberalization by actors of the economy benefitting from income resulting from protection (Krueger, 1993). Such groups are prone to federate opposition: they can team up with employees in de-protected sectors. A status quo may be extended despite the fact that it is generally disadvantageous: from the theoretical point of view, optimal nature of the reform does not necessarily guarantee its adoption in the absence of a credible mechanism of compensation for losses by benefits to the groups gaining from it. This situation can be described as the result of a war of attrition (Alesina and Drazen, 1991; Drazen and Grilli, 1993) between two protagonists with divergent interests. This model is based on several hypotheses. The achievement of the reform demands unanimous agreement, i.e. each group possesses a right of veto. The cost of the absence of reform increases with time but this constitutes private information (cf. also Laban and Sturzenegger, 1994). In other words, members of a group know for themselves the cost of waiting but do not know what the cost will be for the other group for the non application of reform. Finally, the group which first accepts the reform bears the cost (or most of it), this 8 rule being exogenous and common knowledge.8 The result is a delay in reform: as time passes the weakest group is shown up (Alesina, Ardagna and Trebbi, 2006). Finally, the fourth category of obstacles concerns the risk of a reduction in public resources affecting more especially countries with low revenue (Baunsgaardt and Keen, 2005). Thus a policy of trade liberalization can lead to an overall reduction in public revenue failing a successful compensation policy through internal fiscal revenue. 2.2- The effect of aid on the obstacles Development aid is likely to have an influence on the different obstacles to tax transition reform. Although most often aid seems to work towards a reduction in the obstacles to reform, it is not possible to deduce firm conclusions concerning its effect. Reduction in uncertainty: The contribution of external technical expertise, more especially the missions of the IMF in the Department of Public Finance reduces uncertainty by allowing the proponents of reform to gather knowledge of the experience of other countries. The contribution of this technical assistance is particularly useful for the poorest countries often without the necessary technical knowledge for the conception and promotion of tax transition reform which is a complex task. For countries with a greater capacity, the contribution of external technical expertise often makes it possible to validate a programme of reform initially conceived chiefly from internal resources (the case of Morocco, Brun et al, 2007). Discontinuity and fragility in promoting reform: The assistance of a donor country over an extended period can encourage the national authorities of a country to become involved in reform even if some instability affects the centres of decision making. In addition, support in reinforcing institutions in particular of those administrations most concerned by tax transition reform (financial administrations) can also facilitate the implementation of reform. Insofar as aid increases the efficiency of public action through the technical assistance provided or by the encouragement to carry out valid policies (Collier, 2006), it reduces the cost of reform. Appropriate external aid in the framework of agreements made at a political level can also, through the conditions required, help in overcoming internal resistance to the administration (Brun et al., 2007). Opposition by special interest groups: ODA can have an ambiguous impact on the cost of non reform. In this way external financing can reduce the cost of the absence of reform and thus delay it. In other words, donors might be liable to encourage « the tyranny of status quo ». Conversely aid can increase the cost of the absence of reform and thus reduce waiting time. This kind of situation appears when the allocation of aid is conditional upon an undertaking to reform. 8 Formally the optimal time of reform is obtained by equalizing the marginal cost of non reform increasing in the waiting time with marginal benefit decreasing with the waiting time. The latter is equal to the product of the probability that the other player will give way next time multiplied by the benefit of not giving way first. Waiting time is all the longer as the cost of the absence of reform is slow and the distribution of benefits and costs between the two groups is unequal. 9 Aid is also liable to affect the distribution of the cost and benefits of reform. Thus, at least from the theoretical point of view, it is possible to envisage targeting aid in favour of groups losing out. Thus, for example, in Ethiopia, the introduction of VAT in 2003 increased the burden of tax of the poorest households, the effective rate of VAT being much higher than the effective tax rate on sales borne by households before reform (Munoz and Sang-Wook Cho, 2003)9. An identical situation was observed in Ghana in 1995 (Osei, 2000). When this happens, aid targeted to the poorest households affected by reform could facilitate the implementation of such a reform. Where companies are concerned, a good deal of opposition can be expected from the protected sector. Aid can encourage the adaptation of de-protected companies by compensating for the cost of adjustment bound up with the modernisation of firms as well as financing the reallocation of resources for the benefit of competitive companies or sectors. Risk of budgetary loss: Development aid awarded to countries in tax transition can make it possible to reduce reluctance bound up with the risk of falling public resources. More especially external financing can be aimed at compensating for the temporary loss of public resources. Section III- Econometric analysis of the effects of ODA on tax transition Because of the ambiguous nature of the effect of aid on tax transition reform an econometric analysis is called for. 3.1- Econometric methodology We shall work on the basis of a model where the explained variable is a dummy variable with a value of 1 if the country is in transition and 0 otherwise. Thus a non linear Probit model is involved. Regional heterogeneity is controlled by including dummy variables10. The variable of interest is ODA. It is measured by the ratio of ODA to the product. The data were obtained from the Development Assistance Committee (DAC) of the OECD. We consider the aid really paid out by all bilateral and multilateral donors. It is equal to the sum of the sum of grants, capital subscriptions and and net payments of loans (overall payments of loans less repayments of capital on the basis of earlier loans and offsetting entries for debt relief). The impact of aid is controlled by the factors characteristic of the economic, political and social environment. It involves in particular the level of development which should facilitate tax transition. Indeed the higher the per capita income, the more the State can increase its administrative capacity thanks to economies of scale in the administration of tax for implementing tax transition. The increase in the product also has a favourable action on the ability to contribute. The State can then put to greater use resources favourable to transition. Ethno-linguistic fractionalisation can negatively affect tax transition. Indeed it 9 According to Munoz and Sang-Wook Cho (2003) the introduction of VAT only has a very weak impact on poor households. The latter gain from reform if additional fiscal income from VAT is allocated to spending on health and education. 10 The introduction of temporal dummy variables to test common shocks likely to affect every country does not modify the results. 10 renders the emergence of a social consensus towards reform more complicated. The burden of debt service is liable to affect the application of the tax transition reform. Indeed, on the one hand, the burden of the indebtedness can stimulate countries much in debt to accelerate reform to make available additional resources, in particular to compensate for the loss of tariff revenue. But, on the other hand, countries much in debt may hesitate to take the risk associated with tax transition reform of an overall loss in public revenue. The quality of bureaucracy is also tested in the hypothesis where tax transition is a complex process requiring efficient administration. Finally macro-economic shocks, in particular those stemming from currency exchange fluctuations, can affect the process of tax transition. But the impact is broadly indeterminate insofar as reform can be accelerated by an unfavourable evolution of prices or by severe instability, in such a way as to render tax revenue less dependent upon the external situation. On the contrary, the occurrence of shocks makes the benefits of reform more chancy and reform can then be delayed. 3.2- On the endogeneity of ODA The ODA variable must be considered potentially endogenous and for two reasons. On the one hand, tax transition can be a condition of the award of aid (bias of simultaneity). On the other hand, it is possible that tax transition and the importance of the ODA received are simultaneously determined by unobservable characteristics of the recipient countries (bias of omitted variables). Two groups of instrumental variables are thus constructed. The first group exploits the variations in overall aid of the various donor countries by weighting them in a specific manner for each country depending on its proximity to the various donors. This approach is suggested by Tavarès (2003) and relies upon two hypotheses: aid is correlated positively with proximity and the latter only modifies tax transition indirectly, through its influence over aid. For each recipient country five main donors are selected each year to ensure a high variability of the instrumental variables. The proximity can be geographical, linguistic or religious. Geographical proximity is measured by the reverse of the bilateral distance between the receiving country and Washington (for Canada and the USA), Brussels (for European countries, Tokyo (for Japan) and Canberra (for Australia and New Zealand). The instrumental variable is then obtained by weighting the distance reversed by the overall aid provided by each donor country. Linguistic proximity is represented by a dummy variable indicating the existence of a common official language. Religious proximity is represented by a dummy variable taking the value of 1 for a dominant shared religion. The second group of variables relates to the budgetary situation of donor countries (Brun, Chambas and Guerineau, 2008). It is assumed that the budgetary situation of donor countries affects the payment of aid independently of the situation of recipient countries. The conventional deficit and the amount of debt compared to the GDP of donor countries are weighted by the reverse of the bilateral distance between the receiving country and the donor country. 11 3.3- Econometric results and comments The results obtained based on indicator 1 are presented in table 4. Wald’s test makes it possible to reject the hypothesis of exogeneity of the ODA variable. The test of superidentification does not reject the hypothesis of the validity of the instruments. We can therefore interpret the results (marginal effects) in terms of a causal relationship. It is evident that, whatever the test variables retained, ODA exercises a favourable effect on tax transition. An increase of the ODA ratio to the product increases the probability of success of the transition by 0.05 (column 3). The increase by 4 points of the aid ratio observed between the beginning of the 90s (6%) and the middle of the 2000s (10%) for the Sub-Saharan African countries thus corresponds to an increase in the probability of transition by 0.20. The effect is thus very significant. Amongst other determining elements of tax transition, a positive impact of the growth of product per capita is highlighted. A higher level of development of countries makes it possible to improve the efficiency of their fiscal policies. In addition, the success of tax transition depends to a large extent on the quality of their public institutions. In this instance, the countries which have high quality bureaucracies have a higher probability of success in their transition. Contrary to the previous factors whose impact is positive, the one concerning the debt service appears negative and significant: countries heavily indebted are likely to show reticence in taking the risk of tax transition reform. Finally, the other factors do not seem to significantly affect the probability of success of the transition: it is a matter of rates of exchange and their instability and the ethnolinguistic fractionalisation. 12 Table 4: Variable explained: Indicator of transition 1; marginal effects Foreign aid over GDP Log of real GDP per capita Terms of trade Instability of the terms of trade Debt service/GNP Ethno-linguistic fractionalisation Bureaucratic quality (1) 0.044*** (3.18) 0.264*** (5.12) 0.0009 (0.96) -0.0064 (-1.41) -0.020*** (-3.44) 0.007 (0.09) 0.093*** (3.85) (2) 0.040*** (2.84) 0.245*** (4.76) (3) 0.054*** (2.87) 0.303*** (4.15) -0.018*** (-2.88) -0.021*** (-2.88) 0.086*** (3.38) 0.074*** (3.22) -0.132* (-1.69) -0.089 (-1.26) 0.215*** (3.99) 606 51.2% 6.331 (0.012) 3.293 (0.51) 606 50.2% 4.566 (0.033) 6.153 (0.18) Sub-Saharan Africa dummy Latin America dummy Southern Asia dummy Number of observations Proportion correctly predicted Chi-squared test of exogeneity Exogeneity test Wald P-value Test of super-identification (P-value) 564 52.7% 7.841 (0.005) 3.302 (0.50) Transition indicator 1 : potential fiscal criterion RT>=0.90, fall of TD and TD <= 10% RT Threshold of predicted proportion 0.75.Robust statistics z to the heteroscedasticity in brackets; * significant at the threshold of 10%; ** significant at the threshold of 5%; *** significant at the threshold of 1% Table 5 presents the results of estimations based on tax transition indicator 2 taking into account more specifically the growth of VAT. These results are concordant with those obtained with indicator 1. In particular foreign aid significantly stimulates tax transition. Its marginal impact is of the same importance11. 11 A test of robustness with an alternative measurement of aid is present in Annex 5. The results are not noticeably different. 13 Table 5: Variable explained: Indicator 2, marginal effects Foreign aid over GDP Log of real GDP per capita Terms of trade Instability of the terms of trade Debt service /GNP Ethno-linguistic fractionalisation Bureaucratic quality (1) 0.043*** (3.17) 0.242*** (4.60) 0.0004 (0.49) -0.007 (-1.53) -0.020*** (-3.29) 0.007 (0.09) 0.091*** (3.55) (2) 0.038*** (2.68) 0.220*** (4.10) (3) 0.051*** (2.79) 0.278*** (3.66) -0.017*** (-2.74) -0.020*** (62.79) 0.080*** (3.00) 0.069*** (2.86) -0.141* (-1.82) -0.092 (-1.34) 0.221*** (4.03) 606 48.0% 4.318 (0.038) 4.72 (0.31) Sub-Saharan Africa dummy Latin America dummy Southern Asia dummy Number of observations Proportion correctly predicted Chi-squared test of exogeneity Exogeneity test Wald P-value Test of super-identification (P-value) 564 49.6% 7.314 (0.007) 2.261 (0.68) 606 47.5% 5.521 (0.019) 2.288 (0.68) Transition indicator 2 = Indicator 1 + condition C of the growth of VAT revenue (Transition 2) Threshold of predicted proportion 0.75.Robust statistics z to the heteroscedasticity in brackets; * significant at the threshold of 10%; ** significant at the threshold of 5%; *** significant at the threshold of 1% Conclusion The objective was to highlight the impact of Official Development Assistance on tax transition in developing countries. In order to do this, it was first of all necessary to extract criteria for identifying the phases of tax transition on a broad panel of Developing Countries: binary indicators of tax transition were constructed by combining information on the level and structure of taxation. Then, using available research, a survey was made of mechanisms through which aid influences tax transition. These mechanisms are diverse and examining them does not provide a definitive conclusion concerning the effects of aid on tax transition. Therefore, in order to remove uncertainty, a non linear model of probability of tax transition was estimated for ODA as a variable of interest. The major result obtained, robust at different specifications, is to bring to light a positive and significant impact of ODA on the real implementation of tax transition. It is to be underlined that this result was obtained by correcting the endogenous bias by a strategy of identification using the proximity between donor countries and recipient countries, and the changes in policy of donor countries bound up with their budgetary situation. 14 Some conclusions relative to the policy of donors can be deducted from this result. Whereas developing countries are strongly encouraged to undertake policies of trade liberalization in particular by making Economic Partnership Agreements, the analysis which has just been undertaken highlights the importance of the obstacles to the implementation of tax transition reform and the major role that external partners play in favour of reform as development aid exercises a favourable influence over the implementation of tax transition reform. This analysis has made it possible to identify the main obstacles that aid programmes can assist in overcoming: in order to contribute to promoting tax transition, aid programmes must aim to reduce uncertainty in particular through assistance in the form of technical aid. The continuity of the impetus of external partners is also liable to serve as a protection against an internal risk of discontinuity in the promotion of reform. The partners in development must also have as their aim to lighten or occasionally compensate for the cost of these reforms. The specific role of ODA is all the more important when reform of trade liberalization now concerns lesser developed countries with fragile structures. A possible extension of this research would be to identify the specific impact of the different components of ODA (technical assistance, loans and gifts). In other words, not only are the level of aid but also its structure liable to influence policies of tax transition. In the same way bilateral aid could have a different influence from multilateral aid, the objective function of donors being liable to be different. Bibliographical references Alesina, A. and A. Drazen (1991). “Why are stabilizations delayed?” American Economic Review vol. 81(5). pp.1170-1188. Alesina, A. and R. Perotti (1994). “The Political Economy of Growth: A Critical Survey of the Recent Literature.” The World Bank Economic Review vol. 8(3). pp.351-371. Alesina, A. and B. Weder (2002). “Do Corrupt Governments Receive Less Foreign Aid?” The American Economic Review, vol. 92, no. 4, pp.1126-1137. Alesina, A. F., S. Ardagna and F. Trebbi (2006). “Who Adjusts and When? On the Political Economy of Reforms”. NBER Working Paper 12049. Baunsgaard T., Keen M., (2005) « Tax Revenue and (or?) Trade Liberalization. » IMF, Working Paper WP/05/112. 31p. Berg, A. and A. O. Krueger, (2003). “Trade, Growth, and Poverty: A Selective Survey”, IMF Working Paper N°03/30. Brun Jean-François, G. Chambas and M. Laurent (2007). « Économie politique Économie politique de la réforme de transition fiscale: le cas du Maroc », Afrique Contemporaine, vol. 223, no. 3-4, pp.309-324. Brun, J. F., G. Chambas and J. L. Combes (2006). « Recettes publiques des pays en développement. Méthode d’évaluation », Revue STATECO, vol. 100. Casella, A. and B. Eichengreen (1996). “Can Foreign Aid Accelerate Stabilization?” The Economic Journal vol. 106(436). pp.605-619. Chambas, G. (2005). « Afrique au sud du Sahara - Mobiliser des ressources fiscales pour le développement », Economica. 15 Collier, P. (2006). “Is Aid Oil? An Analysis Of Whether Africa Can Absorb More Aid?” World Development vol. 34(9). pp.1482-1497. Djankov, S., J. G. Montalvo and M. Reynal-Querol (2006). “Does foreign aid help?” Cato Journal, vol. 26, no. 1, pp.1-28. Drazen, A. and V. Grilli (1993). “The Benefit of Crises for Economic Reforms.” The American Economic Review vol. 83(3). pp.598-607. Ebrill, L. P., M. Keen, J.-P. Bodin, et al. (2001). The modern VAT. International Monetary Fund. Gupta, S., B. Clements and A. Pivovarsky (2004). “Foreign Aid and Revenue Response: Does the Composition of Aid Matter?” IMF Working Paper N° 03/176. Gupta, K. L. (1997). “Public fiscal behaviour and foreign aid: Some model solutions”, Economic Modelling, vol. 14, no. 2, pp.203-214. Heller, P. S. (1975). “A model of public fiscal behavior in developing countries: aid, investment and taxation”, The American Economic Review, vol. 65, no. 3, pp.429-445. Keen, M. (2006). “VAT, Tariffs and Withholding: Border Taxes and Informality in Developing Countries”, mimeo, International Monetary Fund. Keen, M. (2007). “VAT attacks!” International Tax and Public Finance 14(4): pp.365-381. Keen, M. and B. Lockwood (2007). “The Value Added Tax: Its Causes and Consequences”, International Monetary Fund WP/07/183. Khan, H. A. and E. Hoshino (1992). “Impact of Foreign Aid on the Fiscal Behavior of LDC Governments”, World Development, vol. 20, no. 10, pp.1481-1488. Laban, R. and F. Sturzenegger (1994). “Distributional conflict, financial adaptation and delayed stabilizations.” Economics & Politics vol. 6(3). pp.257-276. Munoz, S. and S. S. W. Cho (2003). “Social Impact of a Tax Reform: Case of Ethiopia”. IMF Working Paper N° WP/03/232, International Monetary Fund. Osei, P. D. (2000). “Political liberalisation and the implementation of value added tax in Ghana.” The Journal of Modern African Studies vol. 38(02). pp.255-278. Ouattara, B. (2006). “Foreign aid and government fiscal behaviour in developing countries: Panel data evidence”, Economic Modelling, Vol. 23, No. 3. Roodman, David (2005). "An Index of Donor Performance," Working Paper 67, Center for Global Development, revised November 2006. Tavares, Jose. (2003). “Does Foreign Aid Corrupt?” Economics Letters, 79(1), pp. 99-106. 16 Annexes Annex A1 Definition and sources of data Variables Public revenue Trade tariffs Définition Tax revenue+ non tax revenue of central administration and subnational governments and social contributions of social security Tariff revenue from international trade VAT Value-added tax revenues Official development Assistance (ODA)/GDP Sum of grants, capital subscriptions and net loans (loans extended minus repayments of loan principal and offsetting entries for debt relief). Real GDP per capita in constant US dollars (base 2000) Overall expenses of final consumption compared to GDP: current spending of goods and services + spending on security and other expenses, military spending excluded Sum of exports and imports of goods and services divided by GDP Index of the price of a country's exports in terms of its imports base 2000. Gap between the terms of trade and an estimated trend (deterministic+stochastic) of terms of trade Varies from 0 to 4. High points are given to countries where the bureaucracy has the strength and expertise to govern without drastic changes in policy or interruptions in government services. Countries that lack the cushioning effect of a strong bureaucracy receive low points Probability that two randomly selected individuals will belong to different ethnic groups : Real GDP per capita Government consumption Trade openness Terms of trade Instability of terms of trade Bureaucracy quality Ethnic Fractionalization M FR AC =1- i= 1 Brun, Chambas et al (2006) brought up to date by the authors DAC(2008) WDI(2007) CNUCED (2007) CERDI ICRG(2006) Alesina et al; (2003) 2 n i Source ; M = n u m b er o f eth n ic g ro u p s, N = to tal N th p o p u latio n , n = n u m b er o f p eo p le b elo n g in g to th e i eth n ic g ro u p i 17 Table A2: Equation of public taxation Dependent variable: Ratio of taxation to GDP Log of real GDP per capita Trade openness Exports of minerals Agricultural value added Log of the Population Constant Number of observations Number of countries 2.848*** (3.359) 0.0317*** (3.489) 0.0227** (2.232) -0.178*** (-4.598) 0.0586 (0.159) 1.944*** (5.218) 1277 106 Student’s T (in brackets) robust to heteroscedasticity and to autocorrelation. * significant at a threshold of 10%; ** significant at a threshold of 5%; *** significant at a threshold of 1% Table A3: Frequency of occurrence of transition by group of countries (indicator 2) Region Eastern Asia and the Pacific Eastern Europe and Central Asia Latin America and the Caribbean North Africa and the Middle East Southern Asia Sub-Saharan Africa The full sample 1980-1995 Transition Transition Completed not completed 31 9 77,5% 22,5% 20 4 83,33% 16,67% 120 44 73,17% 26,83% 59 27 68,6% 31,4% 30 12 71,43% 28,57% 78 66 54,17% 45,83% 338 162 67,6% 32,4% 1995-2005 Transition Transition completed not completed 23 5 82,14% 17,86% 62 14 81,58% 18,42% 70 18 79,55% 20,45% 37 8 82,22% 17,78% 16 5 76,19% 23,81% 100 42 70,42 29,58 308 92 77% 23% Transition completed 54 79,41% 82 82% 190 75,4% 96 73,28% 46 73,02% 178 62,24% 646 71,78% 1980-2005 Transition not completed 14 20,59% 18 18% 62 24,6% 35 26,72% 17 26,98% 108 37,76% 254 28,22% 18 Table A4: Public aid for development by groups of countries and by period Eastern Asia and the Pacific 1980-1984 1985-1989 1990-1994 1995-1999 1.69 2.07 1.12 4.03 (1.14) (1.62) (1.14) (7.07) 2000-2005 4.55 (6.41) 1980-2005 3.26 (5.60) Eastern Europe and Central Asia 0.87 (0.38) 0.32 (0.13) 0.45 (0.42) 5.76 (5.87) 5.32 (4.10) 4.70 (5.03) Latin America and the Caribbean 1.71 (2.97) 1.60 (2.60) 2.22 (2.92) 2.43 (4.66) 2.44 (5.21) 2.17 (4.02) North Africa and the Middle East 6.89 (7.88) 3.19 (2.70) 5.84 (8.79) 1.93 (2.00) 1.88 (2.38) 3.56 (5.51) Southern Asia 5.94 (2.61) 5.72 (2.50) 8.92 (9.22) 2.86 (4.18) 2.04 (1.82) 5.06 (5.62) Sub-Saharan Africa 6.44 (4.60) 9.17 (6.83) 11.06 (9.78) 10.69 (9.25) 10.21 (10.68) 9.90 (9.12) All Developing Countries 4.44 (5.03) 4.02 (4.94) 5.58 (7.96) 5.88 (7.78) 5.47 (7.95) 5.32 (7.29) 19 Annex 5- Test of robustness with an alternative aid measurement In order to confirm the robustness of our results, a variable of net transfers of ODA (NAT) calculated by Roodman (2005) is introduced as a variable of interest. The use of Net Aid Transfer makes it possible to evaluate more accurately the impact of the aid component liable to be more easily affected when reforms are implemented. Indeed, contrary to the previously used variable, which is calculated net on repayment of capital only, NAT also excludes payment of interest. The results of the impact of net ODA transfers (compared to the GDP) on tax transition are presented in the table below. This impact is measured on both variables of tax transition. The results established previously are not invalidated even if the marginal impact of aid excluding payments of interest seems higher. Table A5: Test of robustness with a variable of net transfers; Marginal effects (1) (2) (3) Transition variable 1 0.047*** (3.31) 0.258*** (5.49) 0.060*** (3.49) 0.306*** (4.89) (4) (5) (6) Transition variable 2 0.050*** 0.045*** 0.057*** (3.41) (3.11) (3.28) Log of real GDP per capita 0.251*** 0.233*** 0.281*** (4.96) (4.70) (4.21) Terms of trade 0.0002 (0.27) Instability of the terms of trade -0.0059 (-1.35) Debt service/GNP -0.019*** -0.022*** 0.021*** 0.019*** 0.021*** (-3.62) (-3.12) (-3.24) (-3.46) (-2.97) (-3.10) Ethno-linguistic fractionalisation -0.007 -0.007 (-0.09) (-0.09) Bureaucratic quality 0.095*** 0.087*** 0.072** 0.092*** 0.081*** 0.066*** (4.01) (3.63) (3.24) (3.69) (3.23) (2.89) Sud-Saharan African dummy -0.137* -0.143* (-1.80) (-1.90) Latin American dummy -0.089 -0.091 (-1.37) (-1.43) Southern Asian dummy 0.215*** 0.220*** (4.16) (4.13) Observations 564 603 603 564 603 603 Proportion predicted correctly 52.1 49.9 49.3% 49.3% 47.6% 47.9% Wald chi-squared test of exogeneity 8.652 8.563 6.677 7.962 7.383 5.946 Wald-P value (0.003) (0.003) (0.010) (0.005) (0.006) (0.015) Super-identification test (P-value) 2.758 (0.59) 3.063 (0.54) 5.921 (0.20) 1.838 2.150 4.632 (0.76) (0.70) (0.32) Threshold of proportion predicted 0.75.Robust z statistics in brackets ; * significant at a threshold of 10%; ** significant at a threshold of 5%; *** significant at a threshold of 1% Roodman net transfers 0.051*** (3.45) 0.273*** (5.57) 0.0006 (0.71) -0.005 (-1.23) -0.022*** 20