82110 A Multi-Partner Evaluation of the Comprehensive Development Framework Assessing the Development Impact of CDF-like Experiences Ibrahim Elbadawi Development Economic Research Group, World Bank and John Randa Kenya Institute for Policy Research and Analysis (KIPRA) ii The findings, interpretations, and conclusions expressed here are those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank cannot guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply on the part of the World Bank any judgment of the legal status of any territory or the endorsement or acceptance of such boundaries. Contact: Operations Evaluation Department Partnerships & Knowledge Programs (OEDPK) e-mail: eline@worldbank.org Telephone: 202-458-4497 Facsimile: 202-522-3125 http:/www.worldbank.org/oed iii Abstract In 1999 the World Bank proposed the “Comprehensive Development Framework: CDF initiative.” The CDF vision is articulated around four major principles: long-term, holistic development framework; country ownership of development programs and policies; country-led partnership among various stakeholders; and, results orientation. Given its recent origin, a direct evaluation of the development impact of the CDF is not possible. However, the principles espoused by the CDF are not new. Moreover, they do suggest some explicit processes that can be approximated by quantitative indicators using available cross-country data. Guided by these processes, this paper develops quantitative indexes of CDF-like experiences and analyzes their development impact. There are three main findings and some policy implications to highlight. First, development strategies adopted by countries or the type of institutions they develop are endogenous to country-specific socio-political characteristics and initial conditions. Second, sustaining CDF-like development strategies is more challenging in countries with fractionalized societies and non-inclusive political regimes, especially when their economies are susceptible to external shocks. This suggests that more attention should be given to flexible and counter cyclical assistance programs to help aid-recipient countries smooth the impact of external shocks and that lending programs should provide more time and space for genuine national bargaining processes to evolve and mature, especially in fractionalized societies. Third, CDF-like development is associated with superior development outcomes, including that they promote aid effectiveness. However, when accounting for the CDF-like development effect institutions do not appear to have an independent effect on growth and aid effectiveness. The implication of this finding is that, whenever possible, directly promoting the right types of “deep” development processes is more effective than attempting to influence intermediate outcomes, such as institutions of development policy. iv TABLE OF CONTENTS FOREWORD ..................................................................................................................... V I. INTRODUCTION..................................................................................................... 1 2. Conceptual, Data and Methodology Issues................................................................ 3 2.1 The CDF Principles and Processes ................................................................. 3 2.2 Indexes of the CDF-like Principles................................................................. 4 2.3 Some Summary Statistics ................................................................................ 5 3. The Determinants of CDF-like Development Strategies .......................................... 9 4. The Marginal Contribution of CDF-like Development.......................................... 14 5. CDF-like Development and Aid Effectiveness ........................................................ 19 6. Conclusions and Policy Implications........................................................................ 23 v FOREWORD An earlier version was prepared as a background paper to a multi-stakeholder evaluation of the Comprehensive Development Framework (CDF). The views expressed in this paper do not necessarily represent Kenya Institute for Policy Research and Analysis (KIPRA) or the World Bank, its Board of Directors or other affiliated organizations. The authors would like to thank, without implication, Shanta Devarajan, David Dollar, Paul Collier, and especially John Eriksson for their comments on an earlier draft. Ibrahim Elbadawi would also like to acknowledge helpful suggestions by Ali A. Ali, Alan Gelb, Aart Kraay and Charles Soludo. “When I think of a development framework for a country and for regions, I think of a balance sheet with two sides. On the left is the macroeconomic presentation including the Article IV reports of the IMF, the National Income Accounts, the Balance of Payments and Trade Statistics, and all the other financial and economic analysis which are at the core of our current appraisal system… There is however, a clear need for a second side which reflects more adequately an analytical framework that presents the structural, social, and human aspects. It must go beyond the familiar statistics of infant and maternal mortality, unemployment and children in school, to address fundamental long-term issues of the structure, scope and substance of societal development.” James D. Wolfensohn (January 22, 1999: p. 3)1 “The supply response depends, at a minimum, on communicating reform and its objectives to farmers. Even if the government fully intends to make good on its promises, this is of little use if farmers are actually unaware of the regime shift in pricing. In the cashew case, it is astonishing how little communication there has been with farmers about the reforms in the cashew sector,” Margaret McMillan, Karen Horn and Dani Rodrik (2003: p. 27) on the liberalization of the cashew sector in Mozambique. 1 A Proposal for a Comprehensive Development Framework, A Discussion Draft, World Bank: January, 1999. 1 I. Introduction In 1999 World Bank President James D. Wolfensohn introduced the Comprehensive Development Framework (CDF) initiative. The framework is articulated around four major principles: long-term, holistic development framework; country ownership of development programs and policies; country-led partnership among various stakeholders; and, results orientation. The overarching objective of the CDF is achieving faster and sustainable reductions in poverty.2 To this end it puts forward a holistic approach to development, which seeks a better balance in policymaking and implementation by highlighting the interdependence of all elements of development—social, structural, human, governance, environmental, macroeconomic, and financial. This approach would, therefore, require a transition from donor-led to aid-recipient country-led development partnership. Needless to say, this transition must be underpinned by genuine leadership and ownership by government at all levels and vigorous participation by representative institutions, civil society and the private sector as well as active support by the international development community. The CDF initiative has, by and large, been met with positive reviews in the academic and policy circles. For example, in attempting to place the CDF in the context of the development literature, Ali and Disch (2002: p. 3) suggest that, “at the risk of simplification, the advent of the CDF can be taken as signifying a return to the application of the basic concepts of development economics, as distinct from the application of narrow neoclassical economic theory propositions, to developing countries.” 3 They go on to characterize this as representing, “a major shift in emphasis in the conduct of development business especially among the multilateral development finance institutions.” And that the central role in the CDF program of long-term holistic frameworks for the design and implementation of development policy is, “a major recognition of the complexity of the development process and its long-term nature” (p. 5). The development-effectiveness of CDF is, however, an open and untested question, given the very little time available since the launch of this initiative. It is, therefore, not possible to analyze its potential impact on intermediate outcomes, such as the policy and institutional environment, much less on such development goals as poverty reduction or gains in human development. Though the CDF, as an initiative by that name, is new, the principles upon which it is based are not. This suggests that the development impact of 2 The Millennium Development Goals (MDGs), most recently endorsed by the UN Millennium Summit in September 2000 represent the international consensus about how, at the global level, progress to that end can be achieved, and how to measure it. The goals have also been placed at the heart of the World Bank's approach to reducing poverty, as set out in its 2001 Poverty Reduction Strategy Process (PRSP). They are therefore indispensable, under CDF, for informing the choices that a country makes in setting out its long- term vision and medium-term strategies for reducing poverty. 3 See also the edited book by Meier (2001), which contains a comprehensive review of the state of development economics, including contributions by Adelman (2001), Yusuf and Stiglitz (2001) and Hoff and Stiglitz (2001). For example, Hoff and Stiglitz present a critical overview of neoclassical economic theory from a development perspective and concentrate on the influence of institutions, the distribution of wealth, history and the types of behavior that have spillover effects. All these influences lead to multiple equilibria, which poses serious problems for development policy design. 2 the CDF might be gauged by analyzing the experiences of countries that have adopted development strategies that approximate the CDF concept of development. We describe such development experiences as “CDF-like” development strategies. To give empirical content to this concept we develop parallel indicators of “CDF-like” principles, which in turn were aggregated to form an overall CDF-like index. This index was subsequently combined with other global development data to analyze the development impact of CDF-like development experiences. Needless to say, the insight from analyzing this type of data does not constitute an evaluation of the CDF initiative, as such. However, the analysis could provide useful lessons for a better design of future explicit CDF development strategies. Therefore, the key question to be addressed by this paper is whether CDF-like development strategies have actually contributed to better development outcomes, including better institutions; higher growth and lower poverty; enhanced human development as well as improved aid effectiveness. The rest of the paper is organized as follows: Section two briefly discusses conceptual, data and methodology issues related to the construction of the indexes of the CDF-like principles, leaving detailed description of the construction of the indexes to the Appendix. Section three examines cross-country evidence on the endogeniety of development strategies adopted by countries or the type of institutions they develop to deep structural factors and initial conditions, such as initial level of development and socio-political characteristics. In addition to analyzing these determinants, this section also examines the relationship between the CDF-like development processes and institutions, characterized as intermediate development outcomes. Section four contains analysis of the development impact of CDF-like development strategies, including on institutions, growth and illiteracy. This approach estimates the marginal contribution of CDF-like development strategy using a modified-control group-type model, which allows, among others, for the development processes to be endogenous to country- specific initial economic conditions and social characteristics. Section five assesses the aid effectiveness of CDF-like development strategy in a context of an empirical endogenous growth model. Section six concludes. 3 2. Conceptual, Data and Methodology Issues In this section we start by briefly stating the processes associated with the CDF principles, which provide the conceptual framework for developing the empirical indexes of the CDF-like principles. Then we provide a brief description of how these indexes are constructed, followed by a presentation of summary statistics of the CDF-like and other associated development variables. 2.1 The CDF Principles and Processes We state below the range of development processes envisaged under each principle, as outlined in the Design Paper for a Multi-Partner Evaluation of the Comprehensive Development Framework (CDF Evaluation Secretariat, World Bank, September 2001: pp. 10-11). i. Long-term, Holistic Development Framework’s Processes involve: 1) Identification of a 15-to-20 year vision statement containing monitorable development goals that: • take into account the broad aspirations of the population, and • include sustainable poverty reduction as an overarching goal and related sub-goals that are in the same areas as the MDGs (see footnote 1). 2) Formulation of a comprehensive yet realistic medium-term (3-to-5 year) strategy for making progress toward goals, specifically addressing the need for: • balance among macroeconomic and financial issues and structural and social concerns; • setting priorities in the face of capacity and hard budget constraints; • time-bound, concrete actions, with attention to phasing and sequencing. ii. Country Ownership’s Processes involve: 1) Identification of development goals and formulation of strategy by the country, not by development assistance agencies. 2) Regular and broad-based stakeholder participation, with evidence of sustained public support from top political leadership and intellectual conviction by key policymakers. iii. Country-Led Partnership’s Processes involve: 1) Government leadership in the management and coordination of aid resources, including: • analytical and diagnostic work; • aligning external support on the basis of the country’s development strategy and development agency comparative advantage; • harmonization of development agency procedures and practices, e.g. procurement, reporting, and evaluation. 4 2) Relations among government, development agencies, other stakeholders, marked by: • mutual trust, consultation, and transparency; • assumption of accountability for sound financial management and performance; • effective, demand-led support for strengthening government management and coordination capacity and not undermining it. iv. Results Orientation’s Processes involve: 1) Design of programs in support of the national development framework with clear and evaluable objectives that contribute to framework goals. 2) Monitoring and regular reporting and sharing of progress, with a focus on and accountability for results, including outcomes and goals, rather than only on inputs. 2.2 Indexes of the CDF-like Principles Guided by the above processes, we develop indexes of CDF-like principles that attempt to approximate the above processes, using available global databases. Starting with the two indexes of country ownership and result-orientation, which draw directly from available indicators of governance and institutional quality, we provide a brief description of the two indexes (a detailed description is provided in the appendix). The index of CDF-like ownership and participation principle is a weighted average of two components of a widely quoted index on the standard of democracy (Polity IV)4: on the processes governing the regulation and competition of political participation. The index for the CDF-like results orientation is a component of the “Country Policy and Institutional Assessment Index: CPIA”5 on the quality of poverty monitoring. Although the CPIA is an internal World Bank’s index, and is, therefore, subject to concerns about subjectivity and transparency, it is nevertheless a very comprehensive index and has recently been substantially enhanced in terms of criteria and transparency. As a recent report by the World Bank’s CDF Secretariat (2001: p. 25) observes, “the CPIA process has recently been strengthened by improving its clarity and making its criteria more explicit by requiring a written explanation of each country’s rating on each question, and implementing regular annual discussions with IDA recipient countries on the results for their country, thereby allowing the CPIA to be an input in the upstream dialogue with IDA countries.”6 The remaining two CDF-like indexes are not readily available and must be constructed from globally available data. The CDF-like approximation of the long-term holistic principle is associated with the ability of a country to maintain a balanced development policy across various sectors in the economy. This is measured by the dispersion in the quality of the CPIA ratings across three broad sectors: macro, social and structural.7 A 4 Source: Monty G. Marshall and Keith Jaggers (2000). 5 The Country Policy and Institutional Assessment (CPIA) is an internal World Bank index (see notes to Table 3 for a detailed description). 6 Moreover, the reports also notes that further work has begun on better linking of CPIA with CDF/PRSP processes, World Bank’s country assistance strategy (CAS) and other economic and sector work (ESW). 7 The fourth sector of public sector management, which was added to the CPIA categories in the 1990s, could not be included due to limited data availability. 5 large dispersion suggests failure to adopt long-term holistic development. Finally, The CDF-like country-led partnership is characterized as being reflected in the “quality” of aid. The overall quality of aid index, QA, is taken from Elbadawi and Randa (2003). Their index is a “principal components” weighted average of two characteristics of “aid delivery mechanisms”: “Aid Fragmentation” and “Excessive Technical Assistance,” with time-invariant indexes of aid dependency and debt overhang used as exponents in the ultimate QA formula (see appendix). First, Elbadawi and Randa’s concept of aid fragmentation accounts for the combined effects of the well known “donors fragmentation”8 and “sectoral concentration” of aid. The latter indicator is associated with a lop-sided allocation of aid across sectors, which suggests failure to pursue “holistic” development. Both indicators are measured by a Herfindahl- Hirschman index, which is a simple, yet sophisticated way of measuring fragmentation. In this study we measure aid fragmentation as one minus the ratio of donor to sectoral concentration raised by the power of (1+ period average aid dependency+ period average debt overhang). Second, “excessive” technical assistance is given by the excess of actual relative to an “optimum” level of technical assistance, where the latter is a decreasing function of the initial quality of institutions, as measured by an ICRG index.9 Again, period average debt overhang and aid dependency are used as exponents. Third, the period average debt overhang index is in turn constructed as a principal component-weighted average of three debt ratios (stock of debt to GDP, debt service to exports and debt service to government revenues), while the period average index of aid dependency is a weighted average of a qualitative variable reflecting the intensity of aid (as measured by the ratio of Aid/GDP) and the weights are the number of years corresponding to each indicator of aid intensity. Hence the QA index is directly influenced by the nature of the aid delivery mechanism, with debt overhang and the degree of aid dependency providing the magnification effects. The four indexes are normalized to fall between zero and 100, and the overall CDF-like index is obtained as a weighted average of the four indexes, where the weights are given by principal components. A detailed description of the construction of these indexes is contained in the Appendix. 2.3 Some Summary Statistics Summary statistics for the five CDF-like indexes for 92 countries spanning the 1980- 2000 period are described in Table 1.a, while Table 1.b presents the distribution of the CDF-like principles for the 92 countries over the 1980-2000 period. It is notable that the index of overall CDF-like principles (which will be the only index subsequently used in the empirical analysis) has an approximate bell-shaped distribution. Table 2 gives a summary of the CDF principles by region in 1998-2000. It is evident that Sub-Sahara Africa lags behind the sample median in all the four CDF-like principles as well as the overall CDF-like index, while East Asia dominates other region in terms of the overall 8 See for example, Knack and Rahman (2003), who developed an index of donor fragmentation based on a Herfindahl-Hirschman index. 9 ICRG is taken from the International Country Risk database, and is a global index composed of a set of indicators of institutional quality (see appendix for more details). 6 CDF index as well as on ownership and country-led partnership. Figure 1 depicts the evolution of overall CDF-like index by region over 1983-1985 to 1998-2000 periods. East Asia and Latin America and the Caribbean (LAC) region seem to have a higher median CDF-like index than other regions in the sample. The figure also shows the significant improvements in the median overall CDF-like indeed in Sub-Sahara Africa (SSA) from 13 in 1985-1988 to 57 in 1998-2000 period. There is also significant improvement in Middle East and North Africa (MENA) region from an overall CDF-like index of 26 in 1983-1985 to 57 in 1998-2000. However, South Asia and, especially, SSA and MENA continue to lag behind East Asia and LAC. Finally, and in anticipation of the subsequent analysis of the determinants of a CDF-like development strategy and of its potential development impact, Table 3 provides descriptive statistics for the key variables to be used in the regressions of the following sections. Table 1.a : Descriptive Statistics of the CDF-like Indexes (1980-2000) Mean Median Std. Deviation Min Max CDF –like (overall) 54 57 25 11 95 CDF-like long-term holistic principle 61 62 16 17 100 CDF-like county ownership 53 57 30 0 100 CDF-like country-led partnership 73 75 12 31 97 CDF-like results orientation 54 55 16 11 81 Notes to Table 1.a: 1. Of the 92 countries, 48 are classified as low-income countries, 29 as low middle-income, and 15 as high middle-income: SSA, Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Comoros, Congo Dem., Congo Rep., Cote d’Ivoire, Djibouti, Ethiopia, Gabon, Gambia, Ghana, Guinea, Lesotho, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Niger, Nigeria, Rwanda, Senegal, Sierra Leon, Sudan, Swaziland, Tanzania, Togo, Uganda, Zambia, Zimbabwe: MENA Algeria, Bhutan, Egypt, Iran, Jordan, Morocco, Tunisia, Turkey, Yemen, LAC, Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Trinidad and Tobago, Uruguay East Asia, Cambodia, China, Fiji, Indonesia, Malaysia, Papua New Guinea, Philippines, Thailand, Vietnam South Asia Bangladesh, India, Nepal, Pakistan, Sri Lanka Eastern Europe Azerbaijan, Bulgaria, Czech Republic, Georgia, Kazakhstan, Kyrgyz Republic, Latvia, Mongolia, Poland, Romania, Russia, Tajikistan. 2. The overall CDF-like index is a weighted average of the four CDF-like components, where the weights (0.0116 for long-term holistic, 0.7898 for ownership, 0.0666 for partnership and 0.132 for results orientation) are given by principal components, using the covariance approach (see the appendix). Since there is no wide variation in the variance of the CDF-like subcomponents, the covariance approach of the principal components analysis is the more appropriate than the standardized means approach. Moreover, the qualitative characteristics of the index do not change if a simple average index is used. 3. CDF-like indexes are normalized to range between 0 and 100. 7 Table 1.b: Distribution of CDF-like Principles (1980-2000) Score Long-term Country Country led Results Overall holistic ownership partnership orientation CDF-like 0-20 1 19 0 3 10 20-40 9 12 2 14 20 40-60 32 23 10 39 22 60-80 42 16 56 32 24 80-100 8 22 24 4 16 Table 2: CDF-like Principles by Region (1980-2000) Long-term Country Country led Results Overall Holistic Ownership partnership Orientation CDF-like Global sample 66 71 77 55 69 Sub-Sahara Africa 65 57 70 49 57 Middle East and North Africa 63 57 85 51 57 Latin America and the Caribbean 62 86 77 62 79 East Asia 65 86 80 69 82 South Asia 62 71 76 65 69 Eastern Europe 76 79 76 60 77 Notes to Table 2: Eastern European values are average for 1992-2000 Table 3: Descriptive Statistics of Relevant Development variables (1980-2000) Std. Mean Median Deviation Min Max GDP growth per capita 0.85 0.50 2.09 -4.67 8.27 Democracy 3.33 2.33 3.00 0.00 10.00 Shock -1.34 -1.01 2.80 -10.12 3.93 CPIA 2.95 2.92 0.52 1.63 4.24 ICRG 2.56 2.62 0.75 0.98 5.01 Illiteracy 38.5 37.4 23.6 3.08 88.4 Social Fractionalization 50.0 52.0 35.8 2.00 144.0 Ratio of Aid to GNP 8.67 5.63 9.25 0.05 50.5 GNP per Capita 1237 535 1551 104 8027 Notes to Table 3: 1. Democracy is the democracy index from Polity IV database discussed above. It varies between 1 and 10, with 10 being the highest value. 2. Social fractionalization is given by the index of ethnic heterogeneity, which is the sum of indices of racial division, linguistic division, and religious division. It was constructed by Vanhanen (1999) and ranges from zero (lowest heterogeneity) to 144 (highest heterogeneity). 3. The shock variable is calculated as: 8 Shock = Change in log (Terms of Trade) × lagged(Exports + Im ports ) / GDP . 4. The Country Policy and Institutional Assessment (CPIA) is an internal World Bank index. This index has 20 different components receiving equal weight in the overall index. World Bank country specialist rates each of the twenty components ordinally on a scale of 1-6, using standardized criteria. The components are grouped into four categories: (a) Macroeconomic management and sustainability of reforms (general macroeconomic performance, fiscal policy, management of external debt, macroeconomic management capacity, sustainability of structural reforms); (b) Structural policies for sustainable and equitable growth (trade policy, foreign exchange regime, financial stability and depth, banking sector efficiency and resource mobilization, property rights and rule based governance, competitive environment for the private sector, factor and product markets, environmental policies and regulations); (c) Policies for reducing inequalities (poverty monitoring and analysis, pro-poor targeting and programs, safety nets); (d) Public sector management (quality of budget and public investment process, efficiency and equality of revenue mobilization, efficiency and equality of public expenditures and accountability of the public survive). 5. ICRG – is constructed using measures of bureaucratic quality, Law and Order, and Corruption. Higher magnitude of these measures represent better institutions. We used principal components to construct our ICRG variables according to the following equation: ICRG = 0.28 Bureacracy + 0.39 Law & order + 0.33Corruption 6. Data on other standard variables are taken from the World Bank SIMA database. Figure 1: CDF-Like by Region 0.9 0.8 0.7 0.6 CDF-like index 0.5 0.4 0.3 0.2 0.1 0 1988 1991 1994 1997 2000 Period SSA MENA LAC EA SA 9 3. The Determinants of CDF-like Development Strategies Given that CDF-like development strategies constitute a set of development processes that determine development outcomes, including intermediate ones such as institutions, it can, therefore, be argued that countries adopting CDF-like development strategies should be more capable of developing better policies and institutions. On the other hand, it could also be argued that countries with high quality institutions and development- oriented policies should be more capable at implementing a CDF-like development strategy. A further, and in our view, a more plausible, argument would suggest that both good policy and institutional environment as well as the propensity of a country to adopt a CDF-like development strategy may both be driven by much deeper factors, such as initial level of development, the degree of social or economic fractionalization in society and the capacity of society to mediate potentially conflictive interests among different economic or social groups. It is, therefore, entirely possible that causation could go from CDF-like development strategy to good policy and institutions, or from the latter to the former or that no causation exists and that both are driven by the more deeper country- specific characteristics. Subscribing to the above argument, we estimate a model that assumes that CDF-like development strategy (as well as institutions, as measured by the overall institutional and policy environment: CPIA; or the overall quality of institutions: ICRG) to depend on the deep social and political characteristics of the country in question, as measured by (social fractionalization) and (democracy), as well as the country’s initial level of development (initial income per capita). In this context we would like to test the hypothesis that both the type of the development processes and the emerging institutions are driven by the same underlining factors. Moreover, we would also like to test whether or not these factors have independent influence on institutions once we acquire control for CDF-like development strategy. The justification for this hypothesis is that the influence of the country-specific characteristics on institutions are likely to be channeled through the range of processes that characterize the type of the development strategies in question. In other words, we hypothesize that while these processes are themselves endogenous to country characteristics, they are sufficient for shaping intermediate development outcomes, such as institutions. For completeness we also test for the hypothesis of whether country characteristics have independent effects on the choice of a CDF-like development strategy once we acquire control of institutions. Finally, we would also like to analyze the sustainability of CDF-like development strategies. In particular we would like to test the combined effects on the volatility of CDF-like index of exogenous shocks (such as terms of trade shocks), social fractionalization and the quality of political and governance institutions for mediating inter-social and economic group differences (e.g. as reflected by the presence of representative and accountable governance). Recent literature suggests that growth volatility is higher in countries characterized by “latent” social conflicts, weak or non- representative political institutions and economies that are susceptible to exogenous shocks. 10 Indeed, formal econometric results strongly support the above hypotheses. Given that the 10 See for example, Rodrik (1999) and Elbadawi (2002). 10 CDF-like and institution variables are positive by construction, a Tobit panel regression was used to estimate the models (pooled and random-effects Tobit regressions are reported in Tables 4.a & 4.b, respectively). The overall explanatory power of the estimated model is good taking into account the fact that we are dealing with panel data estimation. The likelihood ratios across the models show the overall significance of the estimated coefficient. The results are remarkably robust across the two sets of Table 4.a: Factors Influencing Development Strategies and Institutions (Pooled Tobit Regressions) Determinants of CDF-like Determinants of CPIA & ICRG Dependent CDF- Std CDF-like CDF-like CPIA CPIA ICRG ICRG Variable Deviation like of CDF-like (1' ) (1'' ) (3) (3') (4) (4') (1) (2) Log Initial GDP 0.0151 -0.0156 -0.0288 -0.0394 0.0670 0.0621 0.3315 0.3130 per capita (1.25) (-2.10)** (-1.41) (-1.57) (1.67)*** (1.47) (10.29)* (9.10)* Social -0.0033 0.0006 -0.0022 -0.0029 -0.0032 -0.0013 -0.0054 0.0025 Fractionalization (-8.92)* (2.37)** (-2.54)** (-4.97)* (-2.50)** (-0.67) (-3.05)* (0.99) Initial 0.0164 -0.0025ch 0.0002 0.0012 0.0314 0.0035 0.1017 0.0970 Democracy (4.42)* (-0.75) (0.03) (0.19) (2.58)** (0.20) (7.26)* (4.80)* Democracy x 0.0006 0.00006ch 0.0004 0.0006 0.0004** 0.0001 0.00003 -0.0014 Social (10.21)* (1.68)*** (3.41)* (8.94)* (2.16) (0.41) (0.11) (3.43)* Fractionalization Predicted 0.7941 1.5972 CDF-like (2.18)** (3.50)* Predicted CPIA 0.4403 (2.14)** Predicted ICRG 0.1231 (2.04)** Shock x Social 0.000002 Fractionalization (2.20)** x Autocracy Error-Correction -0.1047 Lagged (-2.50)** Constant 0.4440 0.1400 -0.5790 0.5178 2.6221 2.2931 0.4556 -0.4722 LR χ – Test 2 233.75 23.11 154.42 154.01 45.05 41.43 330.51 331.95 [0.0000] [0.0008] [0.0000] [0.0000] [0.0000] [0.0000] [0.0000] [0.0000] [P-value] # of Observation 326 84 231 231 304 298 371 367 Number of 93 56 80 80 82 89 102 101 Countries Pseudo R 2.6661 -0.1230 18.3493 3.7289 0.0732 0.0684 0.2753 0.2810 Notes: 1. Autocracy is given by 10 – Democracy, where Democracy is measured by Polity IV (see footnote of Table 3 for definition of other variables). 2. Predicted CPIA and ICRG variables are obtained from regressions 3 & 4, respectively. 3. The “Error-Correction” term is given by the residual of the regression of regression 1 (i.e. equal to: Predicted CDF-like – CDF-like). 11 4. t-values are reported in brackets, where robust standard errors are used. 5. *Significant at 1% level, ** significant at 5% level, *** significant at 10% level. 6. LR χ2 – Tests the null hypothesis that coefficients are NOT significant overall. Table 4.b: Factors Influencing Development Strategies and Institutions (Random Effects Tobit Regressions) Determinants of CDF-like Determinants of CPIA & ICRG Dependent CDF-like Std CDF-like CDF-like CPIA CPIA ICRG ICRG Deviation Variable of (1) CDF-like (1') (1'') (3) (3') (4) (4') (2) Log Initial GDP 0.0051 -0.0154 -0.0156 -0.0414 0.0388 0.0388 0.2268 0.2734 per capita (0.51) (-2.09)* (-1.24) (2.48)* (0.70) (0.65) (5.54)* (5.76)* Social -0.0047 0.0006 0.0028 -0.0041 -0.0012 0.0018 -0.0077 0.0049 Fractionalization (-14.76)* (2.42)* (-5.63)* (-8.26)* (-0.74) (0.69) (-3.15)* (1.97)** Initial 0.0087 -0.0026 -0.0032 -0.0042 0.0466 0.0206 0.1015 0.1379 Democracy (3.37)* (-0.78) (-0.90) (-1.13) (3.91)* (0.87) (8.11)* (5.90)* Democracy x 0.0009 0.00007 0.0007 0.0008 -0.00002 -0.0005 0.0004 -0.0017 Social (18.93)* (1.74)*** (10.30)* (12.86)* (-0.11) (-1.25) (1.67)*** (-4.63)* Fractionalization Predicted 0.9390 1.4202 CDF-like (2.96)* (3.90)* Predicted CPIA 0.2906 (2.68)* Predicted ICRG 0.1003 (2.92)* Shock x Social 0.000002 Fractionalization (2.23)** x Autocracy Error-Correction -0.1046 Lagged (-2.52)** Constant 0.5970 0.1388 -0.1672 0.6628 2.7376 2.2540 1.2202 -0.2544 257.7W 251.12 W LR χ – Test 2 632.71 27.22 620.35 240.39 22.66 17.21 [0.0000] [0.0001] [0.0000] [0.0000] [0.0001] [0.0041] [0.0000] [0.0000] [P-value] # of Observation 326 84 231 231 304 298 371 367 Number of 93 56 80 80 90 89 102 101 Countries Notes: 1. Autocracy is given by 10 – Democracy, where Democracy is measured by Polity IV (see footnote of Table 3 for definition of other variables). 2. Predicted CPIA and ICRG variables are obtained from regressions 3 & 4, respectively. 3. The “Error-Correction” term is given by the residual of the regression of regression 1 (i.e. equal to: Predicted CDF-like – CDF-like). 4. t-values are reported in brackets, where robust standard errors are used. 5. *Significant at 1% level, ** significant at 5% level, *** significant at 10% level. 6. LR χ2 – Tests the null hypothesis that coefficients are NOT significant overall. regressions. The results of pooled and random-effects Tobit regressions both strongly 12 corroborate the endogeniety of both CDF-like and ICRG indexes to the country-specific characteristics. According to these results countries with higher initial income, functioning democracy and less fractionalized societies are likely to adopt CDF-like development strategies (see regressions 1 of Tables 4.a & 4.b) as well as to have better institutions (regressions 3 & 4 of the two tables). Moreover, controlling for CPIA or ICRG in the CDF-like equation (regressions 1’ & 1’’ in both tables) at least some of these variables continue to be significant. On the other hand, when controlling for CDF-like development strategies in the CPIA regression the effects of the fundamental variables on CPIA disappear (regression 3’ in both tables and Figure 2). However, this finding does not carry over for the case of ICRG, where some of the fundamental variables remain significant even after controlling for the CDF-like effect (regressions 4’). The difference in the results could be explained by the fact that ICRG is a broader definition of institutions that also accounts for some of the CDF processes, while CPIA is confined to assessing institutions of direct relevance to economic policy. Figure 2: Index of CDF-like Fully Accounts for CPIA Residual Fitted l 2 Residual CPIA after controlling for CDF-like effect 1 0 -1 -2 2.4 2.6 2.8 3 Average of initial income, democracy, social fractionalization 13 Therefore, the evidence suggest that both the choice of a development strategy and the evolution of institutions are endogenous to “fundamental” country-specific factors. And that the latter tend to affect institutions, especially those directly relevant to economic policy, through the chosen development strategy. This latter and very interesting finding can be justified on the ground that CDF-like development strategy is about processes that should influence better intermediate outcomes, such as institutions. Hence, the initial economic, social and political characteristics of any given country must influence that country’s institutions through their influence on development processes and strategies. Finally, the volatility of CDF-like development strategy (measured by the standard deviation of the CDF-like index) is positively influenced by a composite term reflecting the combined effects of terms of trade shocks, social fractionalization, and autocracy (regression 2 of the two tables). This finding is consistent with the evidence from the growth literature referred to above. Moreover, our results confirm a concern expressed in the context of implementation of the current CDF initiative, where it has been noted that, “sustainability becomes a problem when countries are facing economic crisis. In such countries, macroeconomic issues dominate over social and structural ones in terms of policy priorities” (CDF Secretariat, 2001: p. 8). 14 4. The Marginal Contribution of CDF-like Development As we argued above the decision by countries to adopt a CDF-like or any other type of a development strategy is itself endogenous to country-specific socioeconomic and political characteristics as well as other factors exogenous to the country. These same factors are also likely to be important determinants of country economic performance, such as economic growth. This suggests that an adequate framework for estimating the marginal contribution of CDF-like development strategy should allow for correction of the “sample selectively” bias that results from the non-randomness of the decision to pursue a CDF-like development strategy. The problem of selectivity bias arises in evaluating the impact of CDF-like development strategy (or at a more narrow level economic reforms) on average economic performance (e.g. real GDP growth) when the average performance of the CDF-like countries would differ from that of non CDF-like countries even if the former do not pursue a CDF-like development strategy. This interpretation of the selection bias problem borrows from the literature on the impact of social programs,11 especially its application for assessing the impact of World Bank’s structural adjustment (e.g. Elbadawi, 1998). Drawing from this literature we estimate a modified control-group model, based on the following equations: (1) ′γ I (CDF − like) it = Cit ′ (2) g it = Fitg′ β F + Cit β S + I (CDF − like) it + vi + ε it for t>k ′ (3) g is = Fisg′ β F + Cis β S + I (CDF − like) is + vi + ε is for s