A Quality of Growth Index for Developing Countries: A Proposal

WP/14/172 A Quality of Growth Index for Developing Countries: A Proposal Montfort Mlachila, René Tapsoba, and Sampawende J. A. Tapsoba ©International Monetary Fund. Not for Redistribution © 2014 International Monetary Fund WP/14/172 IMF Working Paper African Department A Quality of Growth Index for Developing Countries: A Proposal Prepared by Montfort Mlachila, René Tapsoba, and Sampawende J. A. Tapsoba1 September 2014 This Working Paper should not be reported as representing the views of the IMF. The views expressed in this Working Paper are those of the authors and do not necessarily represent those of the IMF or IMF policy. Working Papers describe research in progress by the authors and are published to elicit comments and to further debate. Abstract This paper proposes a new quality of growth index (QGI) for developing countries. The index encompasses both the intrinsic nature and social dimensions of growth, and is computed for over 90 countries for the period 1990-2011. The approach is premised on the fact that not all growth is created equal in terms of social outcomes, and that it does matter how one reaches from one level of income to another for various theoretical and empirical reasons. The paper finds that the quality of growth has been improving in the vast majority of developing countries over the past two decades, although the rate of convergence is relatively slow. At the same time, there are considerable cross-country variations across income levels and regions. Finally, emprirical investigations point to the fact that main factors of the quality of growth are political stability, public pro-poor spending, macroeconomic stability, financial development, institutional quality and external factors such as FDI. JEL Classification Numbers: O40, O55, I10, I20, I32 Keywords: Quality of growth, social indicators. Authors’ E-Mail Addresses: [email protected]; [email protected]; [email protected] 1 We would like to thank, without implication, Tidiane Kinda, Samba Mbaye, Marco Pani, and attendants at an African Department seminar. A special thank you to Promise Kamanga for excellent research assistance. ©International Monetary Fund. Not for Redistribution 2 Contents Abstract ..................................................................................................................................... 2 I. Introduction ........................................................................................................................... 3 II. Methodology and Data ......................................................................................................... 4 A. Growth Fundamentals ...................................................................................................... 5 B. Social Dimensions of Growth .......................................................................................... 7 C. Construction of the Index ................................................................................................. 7 D. Data ................................................................................................................................ 10 III. Results ............................................................................................................................... 10 A. Some Stylized Facts ....................................................................................................... 10 B. Convergence Hypothesis ................................................................................................ 14 C. Putting the QGI into Perspective with Existing Development Indicators...................... 16 D. Drivers of the QGI: an Appraisal ................................................................................... 17 IV. Robustness of the QGI to Alternative Specifications ....................................................... 22 V. Conclusion ....................................................................................................................... 233 Tables 1. Full Sample QGI-based Ranking ........................................................................................ 11 2. Test of Convergence Hypothesis ........................................................................................ 15 3. Determinants of Quality of Growth over 1990–2011 ......................................................... 21 Figures 1. Conceptual Representation of the Quality of Growth Index ................................................ 6 2. QGI: Average Values .......................................................................................................... 13 3. QGI: Kernel Density ........................................................................................................... 14 4. Quality of Growth over Time ............................................................................................. 16 5. Correlation between the QGI and Existing Development Indicators ................................. 17 6. Quality of Growth Index: Key Correlations ...................................................................... .18 7. Social Dimension and QGI over Time: Convergence ........................................................ 23 Appendices Appendix 1. Data Sources and Definition .............................................................................. 25 Appendix 2. Dealing with Missing Observations in the Primary School Completion Rate ... 26 Appendix 3. Country Ranking using Geometric Mean-based QGI ........................................ 27 Appendix 4. QGI for the Full Sample over 1990–2011: Descriptive Statistics...................... 28 Appendix 5. Correlation Matrix of the Alternative QGIs ....................................................... 28 Appendix 6. Spearman’s Rank Order Correlation Test .......................................................... 28 Appendix 7. Spearman’s Rank Order Correlation Test .......................................................... 29 Appendix 8. Correlation Matrix between the Benchmark QGI and the Region-Specific QGI 29 Appendix 9. Determinants of Quality of Growth on Different Subsamples........................... 30 References ............................................................................................................................... 31 ©International Monetary Fund. Not for Redistribution 3 I. INTRODUCTION Recent history has shown that high growth on its own does not necessarily lead to good social outcomes. It matters if growth is inclusive or not. Thus, inclusiveness is an essential ingredient of any successful growth strategy. The concept has yet to be rigorously documented, as it is used to convey several aspects of growth. Numerous papers have proposed different definitions. For instance, the Commission on Growth and Development 2008 notes that inclusiveness of growth is associated with equity, equality of opportunity, and protection in market and employment transitions. In a similar vein, Ianchovichina and Gable (2012) define inclusive growth as rapid, broad-based across sectors and sustained growth that is inclusive of a large part of a country’s labor force. According to this definition, to be inclusive a growth path needs to be strong, pro-poor and redistributive, and further geared toward improving productive employment. Recently, Anand et al. (2013) refer to both the pace and distribution of economic growth to define inclusive growth, on the ground that for growth to be sustainable and effective in reducing poverty, it needs to be inclusive. This paper argues that all these aspects of inclusive growth have a common denominator, dubbed the “quality of growth”. A necessary condition to achieve all these different approaches of inclusive growth is “good quality growth”. Good quality growth is seen as high, durable, and socially-friendly growth. There is a consensus that high growth over the long run is necessary to achieve lasting improvements in social outcomes, but it is increasingly evident that high growth alone may not be sufficient in many cases. For instance, over the past few decades, many developing countries experienced strong growth episodes in the context of relative macroeconomic stability, sound policies, and strengthening institutions. However, relatively few posted significant declines in poverty, inequality and unemployment (see Dollar and Kraay, 2002; Dollar et al., 2013). It is therefore relevant for policy makers and academic professionals to assess whether the underlying “quality” of growth has been good. This paper builds on Martinez and Mlachila (2013) who explored the quality of the recent high-growth episode in sub-Saharan Africa. They delved into the concept of good quality growth, defining it as one that is strong, stable, sustainable, increases productivity and leads to socially desirable outcomes, like improved standards of living, especially in the reduction of poverty. The main objective of this paper is to introduce an index of the quality of growth. The proposed quality of growth index (QGI) encompasses both the intrinsic nature of growth and its social dimensions. Following Martinez and Mlachila (2013), a good quality of growth— more than just its (high) level—is important to enhance living standards and welfare, and to create opportunities for better employment. Thus how growth is generated is critical for its sustainability and for accelerating employment creation and poverty reduction. Our index attempts to capture the multidimensional features of growth. The concept of the QGI goes beyond the well-known Human Development Index (HDI) developed by the United Nations by concentrating not just on the levels of incomes, but the very nature of growth. We argue that it does matter how one reaches from level L1 to L2 of ©International Monetary Fund. Not for Redistribution 4 income for various theoretical and empirical reasons that are elaborated on below. Arguably, since it is income level-based, the HDI is the result of millennia of growth. On the other hand our index facilitates the assessment of the quality of various episodes of growth both within a country and across regions. There is ample evidence that not all growth is created equal: growth that is strong, stable, sustainable, increases total factor productivity, is broad-based sectorally, and export-oriented is likely to be more efficient in fostering socially desirable outcomes.2 Our QGI is also different from the recently developed Social Progress Index (Stern et al., 2014), as the latter focuses more on aspects that are close to the social dimension of the QGI, without accounting for the growth fundamental aspect, which is the core of the QGI.3 That said, it is worth mentioning that there is a dynamic and complementary relationship among social indicators such as education and health, and growth. Empirical evidence confirms this two-way relationship between investment in education and health, and growth (Bils and Peter, 2000). The paper’s key contribution is the rigorous development of the QGI, covering a wide panel of developing and emerging countries over 1990–2011. This allows us to explore how it has evolved over time and whether there are important regional variations in the quality of growth. It thus enables us to determine whether there has been some convergence in the quality of growth over time, or whether there exists a growth “quality trap”. The paper also explores whether the quality of growth is related to other development indicators identified in the literature. The paper’s main findings are four-fold. First, the quality of growth has been improving in the majority of countries over the past two decades. Second, the rate of convergence is relatively slow. Third, there are considerable cross-country variations across income levels and regions. Finally, empirical investigations show that political stability, public pro-poor spending, macroeconomic stability, financial development, institutional quality and external factors such as FDI, are associated with higher QGI. The paper is organized as follows. The next section describes in detail the steps involved in the creation of the index and introduces the dataset. Section III presents the computed index itself, shows some stylized facts and the country ranking, tests the presence of convergence in the QGI dynamics and explores the potential drivers of the QGI. Section IV checks the robustness of the index by examining various computation approaches. Section V presents concluding remarks. II. METHODOLOGY AND DATA In order to evaluate the evolution of the quality of economic growth across countries and over time, we build a Quality of Growth Index (QGI). This section first sets forth the 2 See Martinez and Mlachila (2013) for a detailed review of the literature on all these aspects. 3 The social progress index encompasses three main dimensions: (i) the basic human needs, (ii) the foundations of wellbeing, and (iii) opportunity. ©International Monetary Fund. Not for Redistribution 5 methodology used to derive the QGI along with its rationale, then discusses the sensitivity of the index to alternative assumptions, and introduces the dataset. The QGI is a composite index, resulting from the aggregation of two building blocks: the intrinsic nature of the growth sub-index (“growth fundamentals”) and the social dimension sub-index representing the desired social outputs from growth, as illustrated by Figure 1. A. Growth Fundamentals The sub-index for the intrinsic nature of growth encompasses four dimensions aiming at capturing the extent to which a given growth episode can be considered as of high quality, with regard to its (i) strength, (ii) stability, (iii) diversification of sources, and (iv) its outward-orientation. The strength of growth is measured by the annual change in real GDP per capita. We resort to GDP per capita instead of GDP, as the former is more in line with the concept of pro-poor growth which underlies the concept of quality of growth.4 The strength of growth is an important aspect of the quality of growth since high growth is a necessary ingredient to put a dent to poverty (Dollar and Kraay, 2002; Dollar et al., 2013). Accordingly, higher growth rate is expected to lead to substantial dent in poverty, and hence to a better QGI. Stability is measured by the inverse of the coefficient of variation (CV) of the level of growth as measured above. The CV is the ratio of standard deviation over the average. A five-year rolling window is used to derive time-varying CV. The CV offsets the apparent dispersion related to the level and allows a straight comparison of countries irrespective of the growth levels. It therefore allows smoothing out the influence of outliers such as small open economies that tend to be structurally more volatile or large countries with low growth that tend to be structurally less volatile. The higher the CV, the lower is the inverse of CV, and the less stable is the growth episode. However, growth instability is well-known to worsen poverty and equity, through a “hysteresis mechanism”. Indeed, swings in growth trajectories are particularly harmful to the poor, since the erosion of their human capital in “bad times” is not made up when the economy pulls out of the shock (Behrman et al., 1999; Ames et al., 2001; Guillaumont and Kpodar, 2006). Accordingly, a stable growth episode is expected to feed positively into the QGI. The diversification of sources of growth captures the extent to which growth is considered as generated by diversified sources. It is proxied by a diversification index computed as one minus a Herfindahl-Hirschman index (HHI) using exports data. The higher the index of diversification of export products, the more diversified are the sources of growth. The rationale of using the diversification of exports lies in the findings that that export diversification is associated with stronger growth and lower output volatility (Papageorgiou and Spatafora, 2012), which are both conducive to higher QGI, as argued above.5 4 Key trends and countries’ ranks are broadly robust to the direct use of GDP levels instead of GDP per capita. 5 A more intuitive indicator is the HHI of GDP value added by sector. However, widespread missing data (continued…) ©International Monetary Fund. Not for Redistribution 6 The degree of outward orientation of a growth momentum is proxied by the share of net external demand—in percentage of GDP,6 as opposed to the share of domestic demand. The rationale for this dimension builds on the fact that an outward orientation of growth is more likely to raise productivity growth through several mechanisms, including learning-by-doing processes, importation of more advanced technology, transfer of knowledge, the discipline of the world market, competition, and foreign direct investment (Diao et al., 2006). Note, however, that such an outward orientation of growth may increase the country’s vulnerability to external environment fluctuations and as such results in a more volatile and ultimately lower quality growth.7 Figure 1. Conceptual Representation of the Quality of Growth Index Quality of Growth Index Growth fundamentals (α) Social outcomes (β) Strength (γ1) Health (δ1) Volatility (γ2) Education (δ2) Sectoral composition (γ3) Demand composition(γ4) • Note:  and  represent the weights ascribed to growth fundamentals and social dimension in the QGI.  1 ,  2 ,  3 and  4 stand for the respective weights of the strength, stability, sectoral composition, and demand composition of growth in the growth fundamentals sub-component, while δ1 and δ2 are the weights assigned to health and education in the social dimension sub-component. B. Social Dimensions of Growth As mentioned above, a strong, stable, diversified, and outward oriented growth may prove insufficient in alleviating poverty substantially and improve living standards. This pro-poor prevent us from using such data. More decisively, output diversification is highly correlated with exports diversification (Papageorgiou and Spatafora, 2012). 6 Net external demand equals to the difference between exports and imports, both as percent of GDP. 7 This concern is somewhat addressed by accounting for the volatility of growth in the index. ©International Monetary Fund. Not for Redistribution 7 aspect of the quality of growth is factored in its social dimension sub-index, through indicators capturing two of the most basic dimensions of human capital building. These include (i) a long and healthy life, and (ii) an access to a decent education/knowledge8, which are both commonly acknowledged in the literature as key drivers of the changes in poverty levels (Schultz, 1999).9 The health component captures the extent to which a country’s population can enjoy a long and healthy life through the aggregation of two sub-components, namely: (i) the reverse of infant mortality rate; and (ii) life expectancy at birth. Both these health measures are considered as key poverty symptoms, consistently with Amartya Sen’s biological approach of measuring poverty (Sen, 2003). Education is captured by the primary school completion rate. The main motivation for using only this indicator is data availability. Several variables could also be good proxies for a country’s educational level, including inter alia, the average years of schooling and the net primary school enrollment rate. However, the lack of observations on these variables over a long period for many countries, either in the well-known Barro and Lee (2000) database or the WDI dataset, prevents us from employing such variables. It is worth noting that missing observations also do exist in the primary school completion rate variable, but we retain this variable, as a “lesser evil”. Nevertheless, to avoid reducing considerably the sample size owing to missing values from the primary school completion rate, we made some assumptions allowing us to generate and fill up these missing values, consistently with the approach set forth in Appendix 2. C. Construction of the Index The construction of the QGI follows a two-step approach: first the variables are standardized into indices of same scale, and then are aggregated into a single index using different weights. Standardization of the components The different variables presented above and representing the different components of the QGI are not expressed in the same unit, which makes their aggregation into a single index comparable to a “mixing apples and oranges” problem. Two main approaches allow us to deal with this issue, namely, the centered-reduced normalization or Z-score approach, and the Min-Max approach. The former consists of transforming a given variable X characterized by 8 Several other opportunity variables (such as employment, inequality or poverty itself) and socially-friendly policy measures (including public spending allocated to health and education) are relevant candidates for capturing the pro-poor dimension of growth but are not considered in the construction of the QGI because of data limitation. 9 Health and education are key components of the very well-known Human Development Index (Klugman et al., 2011). ©International Monetary Fund. Not for Redistribution 8 its mean µ and standard deviation σ, into an index or Z score expressed as follows: Z X    . If X is normally distributed, then Z follows a centered-reduced normal distribution, with a zero mean and a unity standard deviation. With this standardization, all variables are expressed in the same unit, namely the standard deviation, and can therefore be meaningfully aggregated into a single index. But one matter of concern related to this approach is the sensitivity of the transformed Z variable to the presence of outliers. For example, small open economies (predominant in our sample) have much more volatile growth, implying higher values for σ compared to the rest of countries. This leads to a highly dispersed distribution of Z-score (unbounded by definition) and renders the standardization strategy less appropriate to rank countries. The Min-Max approach also consists of transforming the variable X into an index Z’,  X  X min  , where Xmin and Xmax stand for the according to the following formula: Z '   X max  X min  minimum (min) and the maximum (max) of X, respectively. Unlike the aforementioned Z variable, Z’ is bounded, ranging from 0 to 1, and is consequently less likely to have a highly dispersed distribution, rendering it more fit for the country ranking perspective of this paper. But a key issue in building Z’ relates to the choice of the minimum (Xmin) and maximum (Xmax) of X. What should be taken as the maximum value of life expectancy at life for example? Should it be what ideally desired (positive argument) is or rather the highest value actually observed in the considered panel of countries (objective argument)? Given the potential controversy surrounding the subjective or positive-based choice of “ideal” max and min for X, we base our standardization on the max and min actually observed in the sample.10 Weighting Approach We assign equal weights (50 percent each) to the intrinsic nature of growth (α) and to the social dimension of growth sub-indices (β), respectively. Equal weight (γ1=γ2=γ3=γ4=25 percent) is also given to the four sub-components of the intrinsic nature sub-index, so is for the two sub-components of the social dimension of growth, namely health and education (δ1=δ2=50 percent). Equal weight (50 percent) is also assigned to the two sub-components of the health sub-index. 11 The main rationale for this weighting option, which is used in other well-known indices such as the Human Development Index (HDI) or the Economic Vulnerability Index (EVI), lies in its simplicity and transparency. But as well stressed by 10 Note, however, that this choice has consequences on the construction of the QGI, since the QGI for a given country will be heavily influenced by how far this country stands relatively to the sample’s max or min. The results therefore depend strongly on the retained sample. But this is somewhat mitigated by the fact that we include all developing countries for which data are available. 11 Given that the equal weights are somehow arbitrary, we conduct a sensitivity test by using alternative weights in the robustness section. ©International Monetary Fund. Not for Redistribution 9 Guillaumont (2009), this equal weighting does carry a dose of arbitrariness, since the weights are determined by the number of components, and hence depends heavily on the components retained themselves. Some alternative aggregation options (such as principal component analysis or regressionbased approach) exist in the literature but present several inconveniences for our purpose. The principal component analysis is difficult to apply when it comes to aggregating more than three variables, which is the case for the “growth fundamentals” index. Similarly, the validity of the regression-based approach depends heavily on the quality of the regressions, including notably issues related to the endogeneity of the regressors. That said, for the sake of robustness, we make use of alternative weights for the intrinsic nature of growth and the social dimension of growth sub-indices. Aggregation Approach The QGI is calculated as the arithmetic mean—with equal weighting, but with alternative weighting subsequently, for robustness purpose, as mentioned before—of the intrinsic nature of growth and the social dimension of growth sub-indices. The same averaging approach is applied for these sub-indices themselves.12 Simplicity and transparency are once again the main rationale for the choice of this aggregation strategy. But this strategy implicitly assumes the absence of interactions between the different components of the QGI, i.e., there is a substitutability relationship between the various components of the QGI. However, some complementarities may exist between the different components of the QGI, which would make the geometric averaging strategy more appropriate.13 Indeed, a country’s level of human capital— education and health—may influence the productivity of its economy, and hence its growth pace, and vice-versa. Such complementarities may also be at work between the sub-components of each of the two major sub-indices of the QGI. For the sake of ensuring that the construction of the QGI and its associated country ranking is not skewed by the chosen averaging approach, we resort, in addition to the arithmetic averaging, to a geometric approach. To sum up, the calculation of the QGI can be formally written as: QGI   Fundamenta ls    Social  (1) with “growth fundamentals” dimension defined as Fundamenta ls   1 Level   2 Stability   3 Diversific ation   4Orientatio n and the “social dimension” defined as Social  1School   2 Health. Under the geometric averaging strategy carried out for robustness purpose, the QGI is 12 This aggregating approach is also used for the construction of the EVI (Guillaumont, 2009). 13 This assumption underpins the construction of the popular HDI. ©International Monetary Fund. Not for Redistribution 10 defined as QGI  ( Fundamenta ls )1 ( Social ) 2 where Fundamenta ls  ( Level ) 1 ( Stability ) 2 ( Diversific ation ) 3 (Orientatio n ) 4 and Social  ( School )1 ( Health) 2 . D. Data For this paper, we use a panel data covering 93 developing countries between 1990 and 2011. The sample includes 57 middle-income countries and 36 low-income countries. To smoothen out the effects of short-term fluctuations on macroeconomic variables, each variable has been averaged over five-year (1990–94, 1995–99, 2000–04, and 2005–11). The variables used in this study are drawn upon from various databanks, including the IMF World Economic Outlook database, the World Bank’s World Development Indicators (WDI) database, COMTRADE, the International Country Risk Guide database, Barro and Lee (2010) and Xala-i-Martin (2006). Detailed sources and definitions of variables are provided in Appendix 1. Appendix 2 elaborates on the specific case of dealing with missing observations in the primary school completion rate. III. RESULTS In this section, we highlight our key findings. First, we present out some stylized facts of the QGI. Second, we rank the QGI and categorize countries based on their performances. We also assess the convergence hypothesis in the QGI. Third, we put the QGI into perspective with the existing development and living standard indicators, and explore the potential drivers of the QGI. A. Some stylized facts This section builds on the benchmark computed QGI to rank the full sample countries over 1990-2011. The ranking results are reported in Table 1 below. Over the most recent subperiod, namely 2005–2011, Bulgaria emerges as the top performer, with a QGI of 0.843, followed by China (0.842) and Argentina (0.830), while Chad (0.334), Central African Republic (0.402) and Niger (0.415) are the poorest performers, respectively. ©International Monetary Fund. Not for Redistribution 11 Table 1. Full Sample QGI-based Ranking Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 1990-94 country Malaysia China Thailand Argentina Chile Uruguay Poland Sri Lanka Indonesia Vietnam Panama Mexico Costa Rica Bulgaria Belarus Jordan Colombia Brazil Turkey Ecuador Philippines Cuba Syrian Arab Republic Romania Russian Federation Kazakhstan Venezuela Peru Kyrgyz Republic Tunisia Moldova Botswana Lithuania Egypt, Arab Rep. Armenia Albania Paraguay South Africa Kenya Honduras El Salvador Algeria Latvia Georgia India Iran, Islamic Rep. Namibia Mongolia Bolivia Azerbaijan Uzbekistan Swaziland Ghana Guatemala Morocco Tajikistan Bangladesh Nicaragua Gabon Pakistan Congo, Rep. Nepal Tanzania Senegal Lao PDR Cameroon Sudan Madagascar Côte d'Ivoire Lesotho Togo Yemen, Rep. Gambia, The Zambia Mauritania Djibouti Uganda Congo, Dem. Rep. Equatorial Guinea Nigeria Burundi Benin Mozambique Malawi Central African Rep. Sierra Leone Ethiopia Burkina Faso Rwanda Guinea Mali Chad Niger QGI Rank 0.811 1 0.772 2 0.754 3 0.750 4 0.748 5 0.746 6 0.742 7 0.733 8 0.725 9 0.721 10 0.719 11 0.712 12 0.707 13 0.703 14 0.692 15 0.689 16 0.685 17 0.685 18 0.684 19 0.678 20 0.676 21 0.674 22 0.673 23 0.673 24 0.672 25 0.668 26 0.667 27 0.661 28 0.661 29 0.656 30 0.655 31 0.651 32 0.649 33 0.630 34 0.630 35 0.628 36 0.625 37 0.622 38 0.619 39 0.618 40 0.615 41 0.612 42 0.609 43 0.604 44 0.596 45 0.594 46 0.588 47 0.587 48 0.586 49 0.577 50 0.572 51 0.569 52 0.566 53 0.563 54 0.552 55 0.521 56 0.515 57 0.513 58 0.511 59 0.507 60 0.498 61 0.491 62 0.482 63 0.480 64 0.478 65 0.451 66 0.447 67 0.439 68 0.435 69 0.433 70 0.430 71 0.429 72 0.428 73 0.406 74 0.404 75 0.400 76 0.394 77 0.392 78 0.386 79 0.373 80 0.373 81 0.362 82 0.346 83 0.340 84 0.338 85 0.328 86 0.327 87 0.324 88 0.320 89 0.308 90 0.287 91 0.286 92 0.258 93 1995-1999 country Malaysia Poland Vietnam China Chile Thailand Sri Lanka Uruguay Argentina Lithuania Mexico Indonesia Panama Brazil Albania Bulgaria Cuba Peru Costa Rica Tunisia Romania Egypt, Arab Rep. Colombia Jordan Bolivia Turkey Ecuador Latvia Armenia Russian Federation Georgia Syrian Arab Republic Kyrgyz Republic Kazakhstan Moldova Belarus Philippines Paraguay Venezuela South Africa El Salvador Algeria Iran, Islamic Rep. Botswana Uzbekistan Honduras Azerbaijan India Namibia Mongolia Kenya Nicaragua Guatemala Ghana Lao PDR Bangladesh Tajikistan Morocco Nepal Swaziland Gabon Pakistan Senegal Tanzania Togo Cameroon Sudan Côte d'Ivoire Equatorial Guinea Madagascar Zambia Congo, Rep. Gambia, The Malawi Mauritania Yemen, Rep. Uganda Lesotho Djibouti Benin Nigeria Mozambique Guinea Ethiopia Burkina Faso Central African Rep. Congo, Dem. Rep. Niger Rwanda Sierra Leone Mali Chad Burundi QGI Rank 0.809 1 0.790 2 0.784 3 0.784 4 0.764 5 0.754 6 0.753 7 0.749 8 0.742 9 0.740 10 0.736 11 0.732 12 0.727 13 0.726 14 0.726 15 0.724 16 0.721 17 0.719 18 0.716 19 0.711 20 0.706 21 0.706 22 0.705 23 0.699 24 0.698 25 0.697 26 0.696 27 0.696 28 0.696 29 0.694 30 0.688 31 0.687 32 0.685 33 0.682 34 0.679 35 0.676 36 0.671 37 0.661 38 0.660 39 0.654 40 0.654 41 0.649 42 0.641 43 0.639 44 0.639 45 0.637 46 0.635 47 0.630 48 0.623 49 0.615 50 0.603 51 0.599 52 0.592 53 0.588 54 0.587 55 0.584 56 0.580 57 0.571 58 0.557 59 0.548 60 0.541 61 0.519 62 0.502 63 0.499 64 0.499 65 0.497 66 0.492 67 0.471 68 0.467 69 0.462 70 0.461 71 0.460 72 0.452 73 0.451 74 0.449 75 0.443 76 0.441 77 0.429 78 0.410 79 0.407 80 0.376 81 0.371 82 0.371 83 0.363 84 0.355 85 0.346 86 0.343 87 0.330 88 0.327 89 0.320 90 0.311 91 0.298 92 0.294 93 2000-04 country China Latvia Vietnam Bulgaria Lithuania Poland Chile Sri Lanka Brazil Malaysia Mexico Argentina Cuba Albania Tunisia Thailand Panama Uzbekistan Romania Georgia Peru Costa Rica Armenia Russian Federation Belarus Uruguay Egypt, Arab Rep. Indonesia Turkey Jordan Colombia Ecuador Kazakhstan Moldova Kyrgyz Republic Philippines Syrian Arab Republic Bolivia Azerbaijan El Salvador Paraguay Mongolia Tajikistan Algeria South Africa Iran, Islamic Rep. Venezuela Namibia India Honduras Lao PDR Nicaragua Morocco Guatemala Bangladesh Kenya Nepal Ghana Botswana Gambia, The Tanzania Togo Cameroon Pakistan Gabon Swaziland Senegal Uganda Zambia Madagascar Congo, Rep. Benin Yemen, Rep. Côte d'Ivoire Nigeria Malawi Sudan Mauritania Djibouti Guinea Equatorial Guinea Ethiopia Lesotho Sierra Leone Burkina Faso Rwanda Niger Mali Congo, Dem. Rep. Mozambique Burundi Chad Central African Rep. QGI Rank 0.849 1 0.798 2 0.792 3 0.786 4 0.784 5 0.782 6 0.777 7 0.767 8 0.767 9 0.755 10 0.755 11 0.752 12 0.751 13 0.750 14 0.748 15 0.748 16 0.746 17 0.743 18 0.740 19 0.739 20 0.738 21 0.737 22 0.737 23 0.736 24 0.735 25 0.733 26 0.730 27 0.728 28 0.727 29 0.725 30 0.724 31 0.723 32 0.717 33 0.713 34 0.710 35 0.700 36 0.697 37 0.684 38 0.682 39 0.681 40 0.679 41 0.674 42 0.670 43 0.665 44 0.663 45 0.655 46 0.655 47 0.651 48 0.649 49 0.646 50 0.644 51 0.637 52 0.630 53 0.621 54 0.612 55 0.609 56 0.602 57 0.602 58 0.596 59 0.574 60 0.562 61 0.553 62 0.551 63 0.544 64 0.536 65 0.536 66 0.521 67 0.515 68 0.504 69 0.504 70 0.492 71 0.474 72 0.471 73 0.467 74 0.462 75 0.461 76 0.455 77 0.453 78 0.449 79 0.447 80 0.443 81 0.440 82 0.424 83 0.407 84 0.404 85 0.399 86 0.376 87 0.371 88 0.371 89 0.367 90 0.356 91 0.340 92 0.331 93 ©International Monetary Fund. Not for Redistribution 2005-11 country Bulgaria China Argentina Vietnam Indonesia Malaysia Uruguay Colombia Poland Panama Brazil Sri Lanka Peru Kazakhstan Chile Thailand Lithuania Mexico Belarus Romania Tunisia Turkey Cuba Jordan Syrian Arab Republic Albania Russian Federation Costa Rica Latvia Egypt, Arab Rep. Uzbekistan Armenia Georgia Ecuador Mongolia Lao PDR Moldova Paraguay India Philippines Bolivia Honduras Kyrgyz Republic El Salvador Morocco Algeria Iran, Islamic Rep. South Africa Tanzania Bangladesh Guatemala Nicaragua Venezuela Azerbaijan Tajikistan Namibia Kenya Ghana Nepal Botswana Zambia Pakistan Madagascar Gabon Ethiopia Swaziland Rwanda Togo Gambia, The Senegal Uganda Benin Cameroon Congo, Rep. Nigeria Mozambique Malawi Mauritania Guinea Djibouti Côte d'Ivoire Sierra Leone Congo, Dem. Rep. Yemen, Rep. Lesotho Sudan Equatorial Guinea Mali Burundi Burkina Faso Niger Central African Rep. Chad QGI 0.843 0.842 0.830 0.807 0.800 0.798 0.794 0.788 0.786 0.782 0.780 0.779 0.778 0.776 0.776 0.775 0.773 0.768 0.767 0.766 0.766 0.760 0.759 0.759 0.758 0.755 0.753 0.751 0.750 0.750 0.748 0.747 0.746 0.745 0.738 0.727 0.721 0.714 0.714 0.709 0.708 0.705 0.703 0.702 0.700 0.699 0.693 0.692 0.690 0.678 0.673 0.668 0.666 0.666 0.653 0.648 0.646 0.642 0.641 0.637 0.632 0.602 0.592 0.583 0.574 0.574 0.568 0.564 0.564 0.558 0.557 0.554 0.529 0.525 0.517 0.513 0.511 0.510 0.506 0.502 0.498 0.498 0.488 0.482 0.480 0.476 0.452 0.442 0.425 0.417 0.414 0.402 0.334 12 The QGI varies markedly across periods, countries, and income levels. The QGI has also improved over time (Figure 2a). The average value of the QGI stands at 0.604 (Figure 2). The minimum QGI is 0.258 for Niger over 1990–94 and the maximum of 0.849 for China over 2000–04. The QGI increases from 0.556 in 1990–94 to 0.656 in 2005–2011. Moreover, a density plot shows that the distribution of the QGI is shifting to the right over time (Figure 3a). At the same time, distributions have become narrower, denoting a certain level of convergence among countries over time. Moreover, there are significant differences across regions. LA countries exhibit the highest QGI scores whereas sub-Saharan Africa lags behind with the lowest QGI (Figure 2b). The density analysis shows that sub-Saharan African countries exhibit the flattest and leftmost density in their QGI distribution, with thick distribution tails (Figure 3b). This suggests that sub-Saharan Africa is the poorest performing country group in terms of quality of growth. In addition, the flatter density along with the thick tails tend to signal the presence of greater inequalities in the QGI scores, with a few countries performing quite well, namely above the full sample average score, while the bulk of sub-Saharan African countries—roughly more than 60 percent of observations—are left behind the full sample average score (0.604). Furthermore, the QGI also varies by income levels (Figure 2c).14 The QGI is positively correlated with countries’ income level. The upper-middle income countries record the highest QGI score, followed by the lower-middle income countries and the low-income countries, respectively. A density analysis suggests that the richer is a country group, the more in the right-hand side of the figure its density curve stands, confirming a positive association between countries’ level of development and their ability to draw upon a better quality of growth from their growth momentum (Figure 3c). We also focus on the fragility status and resource endowment.15 It emerges that fragile countries significantly underperformed the sample average by almost 16 percentage point (Figure 2d). This suggests that fragile countries face tougher structural impediments when it comes to achieving better quality of growth. Likewise, resource-rich countries have QGI scores standing slightly lower than their non-resource rich peers, which may fuel the natural resource curse debate (Sachs and Warner, 2001). From the density analysis (Figure 3d), it clearly transpires that most fragile countries are concentrated in the left side of the distribution, with around 25 percent of observations corresponding to a performance gap of as high as 0.2 point with regard to the full sample average, and around 40 percent of observations corresponding to a QGI score higher than that of the former group, but lower than the full sample average (0.604). This confirms the previously-underscored message from figure 2.d that fragility weighs severely on countries’ underperformance in terms of QGI. The distribution of QGI also shows that below the sample average, the density curve 14 The country sub-sampling in terms of income refers to the World Bank’s classification of countries. 15 The list of fragile countries is drawn upon from IMF (2011) which is based on the World Bank’s criteria of fragility while the list of resource-rich countries is extracted from IMF (2012). ©International Monetary Fund. Not for Redistribution 13 for resource-rich countries’ stands above the curve for non-resource rich countries, while the reverse is observed from the full sample average onwards. This therefore suggests that endowment in natural resources worked more as a curse rather than as a blessing for countries when it comes to achieving a better quality of growth. Figure 2. QGI: Average Values 2.a : QGI by period 2.b : QGI by region 0.68 0.8 0.66 0.7 0.64 0.6 0.62 0.5 0.60 0.4 0.58 0.3 0.56 0.54 0.2 0.52 0.1 0.50 0 1990-94 1995-99 2000-04 2005-11 SSA Full sample average MENA AP CEE LA Full sample average 2.c : QGI by level of income 2.d : QGI by fragility and resource-rich criteria 0.80 0.70 0.70 0.60 0.60 0.50 0.50 0.40 0.40 0.30 0.30 0.20 0.20 0.10 0.10 0.00 0.00 LIC LMIC Full sample average UMIC Fragile Non-Fragile Resource-rich Non-resource rich Full sample average Note: AP= Asia and Pacific; CEE= Central and Eastern Europe; LA= Latin America; MENA=Middle East and North Africa; SSA= Sub-Saharan Africa; LIC=Low-income countries; LMIC=Lower-Middle income countries; UMIC=UpperMiddle income countries. ©International Monetary Fund. Not for Redistribution 14 Figure 3. QGI: Kernel Density 3.b : by region 0 0 1 2 Density 2 Density 4 3 6 4 8 3.a : by period 0.2 0.2 0.4 0.6 QGI 0.8 0.4 0.6 0.8 QGI 1 Sub-Saharan Africa Asia and Pacific Latin America Central and Eastern Europe Middle East and North Africa 1990-94 1995-99 2000-04 2005-11 3.d : fragility and resource-rich criteria 0 0 2 1 Density 4 Density 2 6 3 8 4 3.c : by income level 0..2 0.2 0.4 0.6 0.4 0.8 QGI Low-Income Countries Lower-Middle Income Countries Upper-Middle Income Countries 0.6 0.8 QGI Fragile Countries Non-Fragile Countries Resource-rich Countries Non-resource rich Countries B. Convergence Hypothesis We first investigate the presence of convergence in the QGI process. To this end, we report in Table 2 below simple pooled OLS estimates linking the change in countries’ QGI to their past QGI. The results show that some convergence is at play in the QGI process. Past QGI performance, expressed either as the lagged (one period) value of the QGI or the initial (1990–94) value of the QGI, is found to be negatively associated with the growth rate of the QGI. This is reflected by the negative and significant estimated coefficient of the lagged QGI and the initial QGI. This result therefore suggests that the least performing countries tend to catch up the best performers over time. ©International Monetary Fund. Not for Redistribution 15 Second, building on this above-evidenced presence of convergence in the QGI process, we propose a categorization of countries, based on their QGI dynamics between the initial period (1990–94) and the final period (2005–2011), as reflected in Figure 4. The sample average QGI score is underlined by the red horizontal and vertical lines reprensenting the QGI values of the initial period and the final period, respectively. The dashed line represents the 45 degree line; the higher above this line, the greater the improvement in the level of the QGI. The intersection of these three lines yields 6 nonoverlapping regions. It is worth mentioning that virtually all the countries have improved their quality of growth over the past two decades. Region 2 includes countries whose QGI scores in the starting period as well as in the final period stand below the sample average, but did improve between the two periods. This group is made up of low-income and/or fragile countries, mostly from sub-Saharan Africa and incidentally from MENA, are labelled as the “hopefuls”, in that they have reasonably good prospects of converging progressively toward the sample average country . Region 3 includes countries that were able to improve their QGI from below to above the sample average between the starting and the final period, and as such are dubbed as the “contenders”, with a reference to the idea that they are contending to be among the high performers. This group includes mainly countries from Asia Pacific (for example Bangladesh, Laos and Nepal), from sub-Saharan Africa (for example Ghana,Tanzania and Zambia), from Middle East and North Africa (for example Algeria, Iran and Morocco), from Central and Eastern Europe (for example Azerbaijan, Tajikistan and Uzbekistan) and a few from Latin America (Guatemala and Nicaragua). Region 4 encompasses countries that not only recorded a QGI score superior to the sample average, in the initial as well as in the final period, but also experienced an improvement in their QGI between the two periods. This country group is labelled as the “club of best performers”, and includes chiefly upper-middle and lower-middle income countries. A noticeable finding is that a handful of sub-Saharan African countries belongs to this club of best performers, including notably Kenya, Namibia and South Africa. Finally, region 5 is chararcterized by countries that performed well above the sample average in terms of QGI in both the initial and the final periods, but have the particularity of having experienced a mild drop in their QGI between the two periods. This group comprises only two countries, namely Botswana and Malaysia, and is categorized as the “club of superior performers”. Table 2. Test of Convergence Hypothesis Dependent variable Lagged QGI (one period ) 1  QGI -0.066*** (0.0156) Initial QGI (1990-94) Observations R-squared 279 0.072 2  QGI -0.068*** (0.016) 279 0.074 Robust standard errors in brackets. *, **, and *** indicate the significance level of 10 percent, 5 percent, and 1 percent. Intercept included. ©International Monetary Fund. Not for Redistribution 16 Figure 4. Quality of Growth over Time Initial period vs. Final of period 0.8 0.4 QGI (Final period) 0.5 0.6 0.7 Contenders 3 ZMB PAK MDG GAB GMB RWA UGA TGO SEN BEN 2 MOZ CMR COG MWI NGADJI MRT GIN SLE CIV ZAR YEM SDN Hopefuls LSO GNQ MLI 1 BFA BDI NER CAF ETH BGR Best Performers CHN ARG VNM MYS KAZ COL IDN URY POL PAN BRA LKA PER CHL THA ALB LTU MEX BLR TUN ROM UZB GEO TUR CUB JOR SYR CRI LVAEGY RUS MNG ARM ECU 5 LAO MDA INDHND PRY PHL BOL KGZ SLV MAR DZA IRN ZAF TZA NIC BGD GTM AZE VEN TJK GHA Superior Performers NAMKEN NPL BWA 4 SWZ 6 0.3 TCD 0.3 0.4 0.5 0.6 QGI (Initial period) 0.7 0.8 Data source: Authors' calculations C. Putting the QGI into Perspective with Existing Development Indicators To have an idea about where our proposed QGI stands compared to existing development indicators, we display in Figure 5 correlations between the QGI and a selection of living standards variables. It appears that the QGI is positively correlated with the well-established United Nations (UN) HDI and real GDP per capita (albeit non-linearly), and negatively with the poverty rate and income inequality. These findings imply that the QGI could be another legitimate part of the toolkit available for gauging countries’ progress toward inclusvie growth. ©International Monetary Fund. Not for Redistribution 17 Figure 5. Correlation between the QGI and Existing Development Indicators Correlation between GDP per capita and Quality of Growth Correlation between Human Development Index and Quality of Growth 0.9 0.9 y = 0.7754x + 0.1993 R² = 0.7843 0.8 0.7 0.7 0.6 0.6 Quality of Growth Quality of Growth 0.8 0.5 0.4 0.3 0.5 0.4 0.3 0.2 0.2 0.1 0.1 y = 0.0019x + 0.5323 R² = 0.2003 0 0 0 0.2 0.4 0.6 0.8 0 1 2000 HDI 6000 8000 Correlation between Inequality and Quality of Growth Correlation between Poverty and Quality of Growth 0.9 0.9 0.8 0.8 y = -0.7372x + 0.6614 R² = 0.4271 0.7 Quality of Growth 0.7 Quality of Growth 4000 GDP per capita 0.6 0.5 0.4 0.3 0.6 0.5 0.4 0.3 0.2 0.2 0.1 0.1 y = -0.001x + 0.6762 R² = 0.0046 0 0 0 0.1 0.2 0.3 0.4 Poverty rate 0.5 0.6 0.7 20 30 40 50 60 70 Gini D. Drivers of the QGI: an Appraisal Pairwise correlation We investigate the key factors driving the QGI scores. We first adopt a pairwise correlation appraoch. We focus on living standards, politico-institutional indicators and external financing conditions. Figure ( 6.a ) points to a strong correlation between the QGI and politico-institutional factors. On the one hand, institutional quality, as measured by the quality of bureacracy, the rule of law or the control of corruption, is positively associated with the QGI, but the correlation is less marked with the latter. On the other hand, more political stability, as proxied by government stability go hand-in-hand with a higher QGI. A sound and stable macroeonomic environement as well as a better access to credits for financing good projects may be conducive to higher QGI scores. Indeed, Figure 6.b shows that the volatility of inflation and credit to the private sector are negatively and positively correlated with the QGI, repsectively. Figure 6.c points out a relevance of social spending in achieving a good quality