Information about Asia and the Pacific Asia y el Pacífico
Chapter

Chapter 4. Growth Slowdown: Are Frontier and Developing Asian Economies Different?

Author(s):
Alfred Schipke
Published Date:
April 2015
Share
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico

Related Material

Show Summary Details
Author(s)
Longmei Zhang and Damien Puy 

As is well known, generating growth in low-income countries with low capital stock, limited access to credit, and poor education systems is very challenging, and this phenomenon has been well studied in the “poverty trap” literature. By contrast, the problem of growth slowdowns in low-income countries has received less attention.

Over the past 40 years, several low-income countries have managed to achieve periods of rapid growth, but failed to sustain it. Most notably, several African countries suffered from slowdown and stagnation after an initially promising growth trajectory. A few Asian low-income countries have experienced sustained rapid growth and seem to be on a solid path to middle-income status. Figure 4.1 shows the evolution of log GDP per capita for a set of countries in Africa and Asia once they have reached the income level of $800. Note that the slopes of the lines can be read as growth rates. Interestingly, although Ghana, Malawi, and Mauritania reached that level earlier than Cambodia and Vietnam, income levels in the African countries have stagnated in the past four to five decades (with the exception of Ghana in recent years), despite an early start. In addition, their growth paths show much more volatility, with periods of rapid growth being reversed and followed by periods of long stagnation. In contrast, Cambodia and Vietnam have doubled their income levels within two decades, suggesting that success in Asian low-income countries can be translated into sustained rapid growth.

Figure 4.1Growth Trajectories in Low-Income Countries

Source: IMF staff calculations.

Note: t = 0 is defined as the year when GDP per capita for a particular country reached $800 in purchasing power parity terms or the earliest data available if the starting value is already higher than $800. The end period for Vietnam is when the GDP per capita reached $2,000 in purchasing power parity terms.

What explains the success in Asia compared with the slowdowns in Africa? Are slowdowns likely to strike low-income countries in Asia? And what are the potential policy measures to avoid such slowdowns?

This chapter addresses these questions by exploring the determinants of growth slowdowns in 58 low-income countries from 1960 to 2005. Building on Aiyar and others (2013), we first identify slowdown episodes in low-income countries using a conditional convergence framework, rather than relying on structural breaks in the time series patterns of growth (Berg, Ostry, and Zettelmeyer 2012).1 Having identified “slowdowns,” defined as substantial and protracted deviations from a predicted growth path, we then use standard probit analysis and a comprehensive set of explanatory variables to single out the determinants of these episodes. Two variations of Bayesian averaging modeling are then used to assess the robustness of our findings.

We find that the drivers of growth slowdowns in low-income countries are mainly of three types. The first is not economic in essence and involves exogenous natural disasters (droughts, floods) or political conflicts that do not have economic proximate causes. The second type of slowdown is directly related to economic structures. In most cases, low-income countries rely on exports of primary commodities, such as minerals, fuel, and food products, thereby making the economy highly sensitive to external demand and terms-of-trade shocks. And third, slowdowns can be induced by institutional deficiencies and poor macro policies. In this case, growth accelerations and sudden rises in income are followed by adverse institutional evolutions, such as rent seeking, civil wars, and excessive public sector growth, that make it impossible for countries to sustain this temporary growth or transform past income into productive investments.

These findings have important implications for policymaking in Asian low-income countries. Except for natural disasters, which are exogenous, most of the identified slowdown drivers can be influenced by appropriate policy interventions. In particular, our results suggest that some countries, such as Mongolia and Papua New Guinea, should diversify their economies and reduce their reliance on natural resource rents. On the other hand, Cambodia and Nepal do not seem to have immediate bottlenecks, but can still improve on most dimensions.

Before turning to the core of this chapter, it is important to note three distinctive features of our analysis with respect to other empirical contributions on growth dynamics in poor countries. First, we address growth slowdowns rather than poverty traps. Although the two phenomena could eventually overlap, we are interested in countries that experienced an abrupt reversal in their growth rate after experiencing a rapid growth period, rather than in countries that always stagnated at very low growth rates or experienced long periods of income stagnation. Therefore we differ from the very considerable literature initiated by Barro (1991), which explores the causes of low growth itself rather than growth reversals. In this respect, we are in line with Pritchett (2000), who called for more attention to “the hills, plateaus, mountains and plains” that are evident in growth records, but remain somewhat understudied, particularly in the case of low-income countries.

Second, the literature on growth slowdowns has mainly relied on statistical techniques to identify turning points in the growth series (Ben-David and Papell 1998; Berg, Ostry, and Zettelmeyer 2012) or applied intuitive rules of thumb (Hausmann, Rodriguez, and Wagner 2006; Reddy and Minoiu 2009; Eichengreen, Park, and Shin 2011). This study adopts an alternative approach that is better grounded in theory. As stated before, we define growth slowdowns as substantial and sustained deviations from the growth path predicted by the conditional convergence framework.

Thrid, although we rely on the same methodology, we differ from Aiyar and others (2013) in that we focus only on low-income countries and therefore restrict our sample to the determinants of slowdowns in countries with a purchasing power parity income per capita of less than $2,000 (measured in 2005 constant prices). We do so because we believe the determinants and nature of slowdowns in low-income countries may be very different from those affecting middle- or high-income economies. In fact, an important conclusion of Aiyar and others (2013) is that the determinants of slowdowns in middle-income countries are different from those identified when using the full sample of 138 countries.2

In the rest of this chapter, the first section explains the formal identification procedure of growth slowdowns and comments on the episodes that affected the low-income countries in our sample. The next section outlines the methodology for exploring the determinants of these slowdowns and discusses the key results. The section that follows derives policy implications from these findings, and the final section offers conclusions.

Identifying Growth Slowdowns

To formally identify the slowdown episodes, we operationalize predictions from conditional convergence theory and identify slowdowns, as noted earlier, as substantial and sustained deviations from the predicted growth path. To do so, we use annual data on income per capita in constant 2005 international dollars and compute a five-year panel of GDP per capita growth rates.3 The sample covers 138 countries over 12 periods (1955–2009).4 Our specification is parsimonious: GDP per capita growth is regressed on the lagged income level and standard measures of physical and human capital.5 For any country at any given point in time, the estimated relationship yields a predicted rate of growth conditional on its level of income and factor endowments.

We define residuals as actual rates of growth minus estimated rates of growth (thus a positive residual means that a country is growing faster than expected and a negative residual means a country is growing slower than expected). Country i experiences a growth slowdown in period t if the two following conditions hold:

Here p(0.20) denotes the 20th percentile of the empirical distribution of differences in residuals from one time period to another. Intuitively, condition (4.1) says that between period t–1 and t the country’s residual became much smaller (that is, its performance relative to the expected pattern deteriorated substantially). To be precise, the deterioration was sufficiently pronounced to place the country-period in the bottom quintile of changes in the residual between successive time periods. The second condition is meant to rule out episodes in which growth slows in the current period only to recover in the next by examining the difference in residuals between periods t–1 and t + 1; that is, over a 10-year period.6 For this chapter, we are more interested in countries that experience sustained slowdowns.

This methodology has at least three desirable characteristics. First, it emphasizes the relative nature of growth slowdowns. At different points in time, the neoclassical growth framework predicts different growth rates for different countries being conditional on world technology, current income, and factor endowments. By identifying growth slowdowns relative to these factors, and also relative to other economies, the methodology takes theory seriously. Second and relatedly, it clarifies what needs to be explained. A slowdown in the headline rate of growth could occur, for example, because a country has already attained a high level of income or because of a temporary shock. But neither of these phenomena needs explanation. Our proposed methodology demarcates countries that are growing slowly after accounting for expected income convergence and after accounting for short-lived shocks. We believe that our methodology passes the “smell test.” In particular, it captures the episodes that motivated this study; that is, substantial growth slowdown episodes in Ghana, Malawi, and Mauritania (Figure 4.1).

Table A4.1 (in the appendix) reports the full list of slowdown episodes identified by the criteria in equations (4.1) and (4.2), whereas Table 4.1 summarizes the slowdown variable restricting our sample to low-income countries and breaking down episodes by region. Both tables highlight important stylized facts. First, looking at Table A4.1, we find that growth slowdowns are quite numerous in middle-income countries. In fact, Aiyar and others (2013) show convincingly that middle-income countries are more prone to experience a slowdown than countries in other income classes, relating it to the so-called middle-income trap.7 However, Table 4.1 also shows that low-income countries are far from being immune to this phenomenon. We identify 32 episodes out of 348 observations8 so that 9 percent of our growth sample in low-income countries consists of growth collapses that are not accounted for by a convergence framework. Finally, we find that most of the slowdowns occurred in sub-Saharan Africa, with some countries, such as Burundi, Liberia, Malawi, and Zimbabwe, accounting for a third of all slowdown episodes. In fact, only five episodes took place outside Africa—namely in Afghanistan, Indonesia, Lao P.D.R., Mongolia, and Pakistan.

Table 4.1Slowdowns in Low-Income Countries—Regional Breakdown
Slowdown VariableEast Asia

and Pacific
Europe and

Central Asia
Latin America

and the

Caribbean
Middle East

and North

Africa
South AsiaSub-Saharan

Africa
Total
0 = no slow down47711949193316
1 = slow down3222532
Total507111151218348
Slowdown frequency6.0%18.2%4%11.5%9.2%

At first glance, this high representation of African countries is clearly reminiscent of Barro’s (1991) negative African dummy or the “Africa’s growth tragedy” highlighted in Easterly and Levine (1997). However, our diagnosis is somehow less pessimistic insofar as rather than highlighting an “unfulfilled potential, with disastrous consequences,” we find that many African countries have in fact often realized their potential but, for reasons that remain to be clarified, failed to sustain these growth spurts.9 We now turn to the investigation of the key determinants of slowdown episodes.

Determinants of Growth Slowdowns in Frontier Economies

This section turns to the identification of the drivers of slowdown in low-income countries. Following Eichengreen, Park, and Shin (2011), we estimate the impact of various determinants on the probability of a country experiencing a slowdown in a particular period using standard probit specifications. Because growth slowdowns can be generated by a host of factors, we consider a wide set of 40 potential explanatory variables, which we group into seven categories: economic structure, institutions, infrastructure, demography, trade, macroeconomic environment and policies, and others. Units and sources of these variables are detailed in Table A4.2 (in the appendix).

Note that the actual number of right-hand-side variables used is larger still because, as a general rule, we allow the data to speak to whether these variables influence slowdown probabilities in levels or differences.10 Because we are looking at the determinants of sustained slowdowns, we would expect the explanatory variables to matter mostly in differences, but in some cases the level may pick up important threshold effects. We emphasize that these variables have been selected because they all relate to major empirical or theoretical contributions. However, the growth literature from which these potential explanatory variables are taken is too large to be reviewed here. In presenting our results, we will attempt to set the stage by describing some of the intellectual precedents for our chosen variables in each category, but this will necessarily be in an illustrative rather than comprehensive nature.11

Our econometric exercise faces two important constraints that shape the empirical approach used in this chapter: data availability and model uncertainty. The first constraint is a recurrent problem in low-income studies (in particular those for African countries), whereby the lack of reliable data often restricts the analysis of qualitative or anecdotal evidence. In our setting, because we consider a vast number of potential regressors, the poor availability of data for low-income countries translates into inevitable data gaps. In particular, we observe a very poor overlap between the different data categories outlined above. For instance, if one were to use all the explanatory variables in a single estimation at the same time, the actual sample size would drop to less than 10 observations (and no slowdowns). To address this issue, we group the 40 variables in seven categories and estimate their impact on slowdowns separately, hoping that a larger sample size and a high overlap within categories will allow us to discriminate between alternative variables of a similar type.

The constraint of model uncertainty, on the other hand, is a standard issue in growth empirics in which ignorance of the “true” model tends to inflate the number of variables on the right-hand scale. When the sample size is limited—a rule rather than an exception in growth empirics—classical estimation methods are of limited use in sorting out robust correlates from irrelevant variables, and growth regressions tend to generate unstable and sometimes contradictory results (Durlauf, Kourtellos, and Tan 2008). Although the sample size considered in this paper is somewhat larger than in many contributions that rely on cross-sectional data, this issue remains relevant. Our approach to addressing model uncertainty is to employ Bayesian model averaging techniques. After every probit estimation, two Bayesian model averaging techniques are used to assess the robustness of the results: the weighted average least squares (WALS) methodology developed by Magnus, Powell, and Prüfer (2010) and the more standard Bayesian model averaging (BMA) developed by Leamer (1979) and popularized by Sala-i-Martin, Doppelhoffer, and Miller (2004).

To summarize, our empirical procedure follows two steps:

  • For each category, we start by running probit specifications with all possible explanatory variables within the specific economic category, both in level and in difference.12 We use both backward and forward selection procedures to identify a restricted set of robust regressors.

  • To assess the robustness of the preferred probit specification identified, Bayesian averaging techniques (BMA and WALS) are used over the full set of variables within the economic category of interest.

In line with these two steps, results are presented below for each category and provide (1) the best probit specification (that is, the probit model including variables selected in the first step); (2) the output of the second step under the form of individual post-inclusion probabilities (PIPs) for BMA and t-ratios for WALS. Note that Magnus, Powell, and Prüfer (2010) suggest a PIP threshold of 0.5 for inclusion of a variable, whereas in the case of WALS, a t-ratio with an absolute value of 1 or greater is typically recommended as a threshold for significance. We now turn to the results obtained under each category.

Economic Structure and Output Composition

This module starts by investigating the significance of four variables summarizing the nature and composition of countries’ output, namely the (1) industry share (in percent of value added), (2) services share (in percent of value added), (3) mining share (in percent of value added), and (4) a Theil index of output diversification computed by Papageorgiou and Spatafora (2012).

The inclusion of service and industry shares in the value added follows Kuznets’ pioneering work on structural transformation (Kuznets 1966). It was Kuznets who first explored what is now regarded as an inevitable accompaniment to modern economic growth: as an economy develops beyond its precapitalist stage, formal employment and output in the manufacturing sector expands, drawing labor from other parts of the economy, especially the initially dominant agricultural sector. The migration of labor from agriculture to manufacturing and the corresponding structural transformation of the economy came to be viewed as the engine of economic development and growth (Harris and Todaro 1970; Lewis 1980). A related aspect of structural transformation is the diversification of output across sectors. Papageorgiou and Spatafora (2012) document an inverse relationship between output diversification (across 12 economy-wide sectors) and real income for countries with a GDP per capita below $5,000. Imbs and Wacziarg (2003) argue that there is an inherent link between diversification of the product base and growth as poor countries diversify away from agriculture, although this relationship is nonlinear and may be reversed at higher levels of income. Finally, a vast number of empirical and theoretical contributions have documented the complex link between minerals—and natural resources in general—and economic growth. On the one hand, many studies have provided strong empirical support for the “resource curse” hypothesis, generating a second generation of studies that seek to explain the mechanisms through which this effect operates. This includes the so-called Dutch disease, market volatility, unsus-tainability or institutional side effects (rent seeking, corruption). On the other hand, Masanjala and Papageorgiou (2008) found that the share of mining was significantly and positively related to growth for African countries, suggesting that the presence of natural resources, in particular in Africa, also increased the trend of growth over 1960–2000.

Looking at Table 4.2, we find that the mining share is highly significant (at 1 percent) and associated with a positive coefficient, implying that a higher share of mining increases the probability of experiencing a slowdown. Using Papageorgiou and Spatafora’s (2012) index of output diversification covering 12 economy-wide sectors from 2000 onward, we find that sectoral diversification is associated with a lower probability of growth slowdowns. Note, however, that we tested the effect of diversification separately because the coverage is poor relative to the other variables in this module. In particular, we are only able to examine slowdowns over 2000–05, so that the regression collapses to a pure cross-section.

Table 4.2Economic Structure and Output Composition
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Mining share6.870.001
Pseudo R20.08
Observations181
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Mining2.520.96
Services share−0.120.070.310.07
Industry share−0.370.081.250.16
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

As we emphasized already, the significance of mining is not surprising per se insofar as many studies have highlighted a relationship between natural resources and growth dynamics. Even so, our results complement and qualify those of Masanjala and Papageorgiou (2008). In particular, although the presence of natural resources might increase the trend of mineral-based economies with respect to poorly endowed economies, our results suggest that it also increases the volatility of output. This “double-edged blessing” can be rationalized in several ways. First, countries that rely heavily on mining are typically subject to “mechanical” external demand shocks that can trigger sudden growth periods that are reversed when global prices fall again. For instance, high-performing African countries in the 1960s and 1970s, such as Zambia, experienced massive slowdowns after the collapse of the demand for copper.

A second interpretation, which is in line with Sala-i-Martin and Subramanian (2003) and illustrative of the governance view of the “natural resource curse” hypothesis, is that the reliance on natural resources is problematic because it generates adverse institutional evolutions, such as rent seeking, coups, or civil wars—all of them making it impossible for countries to sustain growth in income or transform past income into productive investments. Collier and Hoeffler (2001) found that the risk of civil war over a five-year period is 20 times higher for countries relying on primary commodity exports. Finally, it might well be a combination of these different effects. For instance, Schuknecht (1999) provides evidence that primary commodity booms tend to lure governments into unsustainable increases in spending that result in public bankruptcies once revenues fall again. In this case mechanical external shocks are combined with poor governance and macroeconomic management.

However, the relationship between diversification and growth slowdowns is clearly in line with Imbs and Wacziarg (2003). To the extent that sectoral shocks could lead to slowdown and stagnation in a concentrated economy, diversification is a form of insurance against idiosyncratic shocks to a particular sector and reduces the probability of such an event.

Interestingly, we find that these different interpretations are consistent with the slowdowns captured in our sample. On the one hand, several slowdowns are identified in rich countries that experienced massive civil conflicts with disruptive effects on institutions and investment. In this respect, the long slowdown identified in Liberia between 1980 and 1990 is illustrative. During the 1970s and 1980s, although strong growth was supported by iron mining, which accounted for more than half of Liberia’s export earnings, the first Liberian civil war in 1989–96 saw seven rival factions fighting for control of the country’s resources (mostly diamonds and iron) and ended up in the destruction of much of the economy and infrastructure.13 On the other hand, external shocks combined with poor diversification have also had very disruptive effects, without the interaction of internal civil conflicts. In the late 1970s, despite a growing economy throughout the decade, a global drop in the demand for flue-cured tobacco resulted in massive bankruptcies of tobacco estates in Malawi and a virtual doubling of the debt service ratio in only two years (1978–80).14

Institutions

It has been long acknowledged that institutions are important, indeed crucial, for growth, but recently much more attention has been paid to analyzing the role of different types of institutions. La Porta and others (1997, 1998) influentially argued that the quality of a country’s legal institutions—such as legal protection of outside investors—could affect the extent of rent seeking by corporate insiders and thereby promote financial development.15 Another strand of the literature has emphasized the advantages of limited government (Buchanan and Tullock 1962; North 1981, 1990; De Long and Shleifer 1993), continuing a tradition that stretches back to Montesquieu and Adam Smith. Mauro (1995) finds that corruption lowers investment, thereby retarding economic growth, although Mironov (1995) cautions that this is true of only certain kinds of corruption. Knack and Keefer (1997) provide evidence that formal institutions that promote property rights and contract enforcement help build social capital, which in turn is related to better economic performance. A large body of literature now exists on the relationship between financial openness and growth; for example, Grilli and Milesi-Feretti (1995); Quinn (1997); and Edwards (2001).

We use five institutional variables in this module. Four are drawn from the Economic Freedom of the World database compiled by the Simon Fraser Institute. The size of government index measures the extent of government involvement in an economy using general government consumption spending (as a percent of GDP). The rule of law index combines indicators of judicial independence, contract enforcement, military interference in the rule of law, protection of property rights, and regulatory restrictions on the sale of real property. The freedom to trade internationally index is constructed from measures of trade taxes, nontariff trade barriers, black market exchange rates, and international capital market controls. The regulation index is an average of subindices measuring credit market regulations, labor market regulations, and business regulations. All four indices are constructed so that a higher value of the index is more “desirable” (that is, a higher value indicates better rule of law, smaller government, more freedom to trade, and less regulation). The fifth variable used here is the Chinn-Ito Index of financial openness (Chinn and Ito 2006). This is based on binary dummy variables codifying the tabulation of restrictions on cross-border financial transactions reported in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions.

Table 4.3 shows that greater government involvement in an economy is strongly associated with a higher probability of slowdown. In fact, although the issue of the impact of government size on economic growth is still highly debated empirically, both BMA and WALS find that government size—as measured by the share of government consumption—is a very robust correlate of slowdowns. What explains this strong detrimental effect of government size in our sample? A theoretical interpretation—which is consistent with Guseh (1997)16—is that a large government size hinders economic growth because governments interfere with the efficient allocation of resources, thereby leading to a slowdown in economic growth (Scully 1988). More precisely, governments are (1) inefficient in the provision of Pigouvian goods and services, (2) engage in unproductive rentseeking activities, and (3) deter private enterprise and innovation. This interpretation turns out to be in line with the experience of many African countries that, for much of the postcolonial period, were governed by undemocratic governments with few agricultural or commercial interests. In particular, many sub-Saharan countries massively expanded the size of their public sector through public employment and inefficient government spending while imposing wideranging controls on private activity.17 As a result, Africa experienced a paradox of poor public services despite relatively high public expenditure and government size (Pradhan 1996).18

Table 4.3Institutions
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Size of government−0.370.001
Pseudo R20.16
Observations113
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Size of government−2.930.98−0.920.17
Rule of law−1.010.29−0.960.28
Freedom to trade−0.160.101.480.27
Regulation−0.120.090.800.11
Chinn-Ito index0.730.08−0.920.11
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

We find that rule of law is the regressor with the second highest significance, both in level and in difference, suggesting that a better rule of law decreases the likelihood of a slowdown. Unfortunately, the poor overlap between the seven subindices that constitute the final rule of law measure prevents us from using Bayesian averaging methods and determines which dimension of the rule of matters more. However, when running separate probit models on the subindices (using size of government as a control), we find that the extent of military interference in the rule of law and the political process are significant at a 5 percent level.19

Infrastructure

Infrastructure conveys beneficial externalities to a range of productive activities and, in many instances, has some characteristics of a public good (for example, a road network might be nonthreatening, at least up to some congestion threshold). It has been incontrovertibly viewed as positively related to economic growth and even as a necessary condition. Nonetheless, a survey by Romp and De Hann (2007) shows that the empirical literature has found mixed results, especially when proxies such as public investment are used to measure infrastructure development.20 In this section, we study four kinds of infrastructure development that have been viewed as important by the literature, using data taken from Calderon and Serven (2004) and the World Development Indicators database: (1) telephone lines is the log of telephone lines per 1,000 people, (2) power is the log of gigawatts of generating capacity per 1,000 people, (3) roads is the log of the length of a country’s road network per square meter, and (4) cell phones is the number of mobile phone subscriptions per 100 people.

As Table 4.4 shows, we are unable to obtain any significant variable in levels or differences. Given the preponderance of studies showing the importance of infrastructure to growth in the literature, one way to interpret these findings is to note the precise scope of the result—that poor infrastructure by itself is not responsible for sustained periods of growth slowdowns in our sample (and, conversely, good infrastructure is not sufficient to prevent slowdowns caused by other factors). Given the externalities between infrastructure and practically every other economic and social activity, it may be that infrastructure magnifies or retards the impact of other significant determinants of growth slowdowns. For now we leave this question to further research. Second, it is worth noting that Aiyar and others (2013) found that infrastructure matters only if they restrict the sample to middle-income countries, suggesting that the impact of infrastructure on the probability of slowdowns is sensitive to the stage of development of an economy, quite apart from any complementarities with other variables. In the case of low-income countries, it is likely that infrastructure development increases in importance only once the low-income stage of development has been crossed.

Table 4.4Infrastructure
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Pseudo R20.16
Observations113
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Telephone lines0.040.11−0.510.10
Power0.530.170.540.09
Road−0.530.09−0.290.09
Cell phones−0.180.08
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

Demography

Population growth subtracts from the rate of growth of output per capita in the Solow model, but the literature has found little systematic impact of population growth itself in cross-country settings. Instead, new research has focused on the age distribution of populations. Several papers document a positive impact on the working-age ratio on economic growth in a cross-section of countries (for example, Bloom and Williamson 1998, Bloom and Canning 2004). Others find that national saving rates are strongly connected to demographic structure (Higgins 1998; Kelley and Schmidt 1996; Bloom, Canning, and Malaney 2000; Mason 2001) and that East Asia’s economic “miracle” was associated with a major transition in age structure. Another demographic variable of interest is the sex ratio, a measure of gender bias. Sen (1992) and others have argued that the phenomenon of “missing women” reflects the cumulative effect of gender discrimination against all cohorts of females alive today. Gender bias could impact economic growth through higher child mortality, increased fertility rates, and greater malnutrition (Abu-Ghaida and Klasen 2004). In their study of Indian states, Aiyar and Mody (2011) find that a more equal sex ratio is robustly associated with higher economic growth.

We use five variables in this module—namely, annual population growth, fertility rate, old-age dependence ratio, sex ratio, and population density per square meter. Results are reported in Table 4.5. We find that only population density is significant at a 5 percent threshold, implying that higher population density is associated with a lower probability of slowdown. Interestingly, this result parallels those of important studies such as Hagen (1975) and Simon and Gobin (1980), who found a strong and positive association between density and growth using data on 54 countries in the period 1950–70. As argued by Boserup (1975), population density might be positively related to growth—and therefore negatively to slowdowns—because greater density implies different organizations of societies, with denser countries having greater specialization and higher investment levels. In particular, when population density is high, yields fall (as fields are fallow for less time), which drives people to develop techniques that enable more frequent cultivation. This requires more labor for regular farm work and for investments in land improvements, increasing output per unit of land (Tiffen 1995). Moreover, at higher densities, social structures and techniques get more complex, implying that the time needed to prepare children increases and education costs are higher. According to Caldwell (1976), this increase in education costs is often accompanied by changes in the nature of marriage, with the family becoming more nuclear. In other words, a higher population density accelerates the demographic transition, facilitating the escape from the Malthusian trap.21 In fact, this theory is supported by the negative correlation between population density and fertility rate in our sample (−0.25).

Table 4.5Demography
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Population density−0.0040.04
Pseudo R20.02
Observations232
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Population growth0.340.07−0.560.07
Fertility rate0.580.081.020.14
Old-age dependence ratio−0.980.15−0.620.09
Sex ratio−0.370.08−0.400.08
Population density−0.10.230.650.19
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

Trade

A vast literature has explored the importance of the trade structure of economies and its relevance to economic growth and resilience. In this section, we test the significance of five variables that map different facets of the trade structure. Trade openness, measured as the ratio of imports plus exports to GDP (at constant 2005 prices), reflects the degree of trade integration for countries in our sample. Fuel exports and food exports measure the composition of exports of each country and report, respectively, the share of these exports as a percent of GDP. Export diversification is measured by a Theil index calculated by Papageorgiou and Spatafora (2012) using product data at the four-digit Standard International Trade Classification level. Distance (GDP weighted) comes from the World Bank and sums, for every country i, the distance to every other country j in the world, weighting each distance by country j’s share of the world GDP. This captures the disadvantage of a country’s geographic location.

The use of an openness measure along with indicators of trade composition relates directly to the importance of the trade channel to explain growth reversals. Admittedly, countries with higher dependence on trade, in particular those relying on primary exports such as food or minerals, are more likely to suffer from global and external shocks. In fact, part of this channel has already been emphasized in an earlier section, which addressed the importance of output composition. Similarly, another strand of the literature looked at export diversification, which has generally been found to be favorably related to growth, especially at an early stage of development. For instance, Koren and Tenreyro (2007) show that economic diversification can increase the resilience of low-income countries to external shocks, and Agosin (2007) provides evidence that export diversification has a positive impact on growth in emerging market economies. Case studies like Gaertner and Papageorgiou (2011) provide similar evidence. Furthermore, distance from world and regional economic centers can be considered an important facet of a country’s endowments, with a more favorable geographical location being more conducive to growth through trade. Distance can directly raise transport costs and, by segmenting markets, may reduce scale economies for domestic firms. The work of Redding and Venables (2004) showing the association between distance metrics and income per capita has been replicated by other studies using different samples.22

Looking at Table 4.6, we find that both fuel exports and food exports are significant in the probit analysis, although the significance is higher for food exports. This implies that the reliance on primary commodity exports, such as fuel and food products, significantly increases the probability of a slowdown. These finding are clearly in line with those highlighted in the section above on Economic Structure and Output Composition insofar as they both emphasize the importance of international demand shocks for primary commodities23 in generating boom-bust cycles. On the other hand, the relatively low significance of export diversification is noteworthy and somehow contrasts with the results presented in the earlier section. For the record, using the index of output diversification, we found that sectoral diversification was associated with a lower probability of growth slowdowns. However, the poor sample coverage of the index relative to other variables prevented us from using other controls in the regression. In this module, we find that once we control for what countries specialize in—using fuel and food as a percent of exports along with a measure of diversification—the significance of export diversification disappears.24 One interpretation of this finding is that it is not the lack of diversification itself that is problematic but rather the specialization in volatile commodities.

Table 4.6Trade
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Fuel exports0.050.03
Food exports0.090.00
Pseudo R20.20
Observations101
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Trade openness0.040.100.190.11
Fuel exports0.800.35−0.160.18
Food exports2.690.981.440.32
Exports diversification0.600.130.410.12
Distance0.840.14
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

Macroeconomic Environment and Policies

A large variety of macroeconomic factors have been associated with economic growth and shocks to economic growth. First of all, although domestic investment is certainly crucial to economic growth, there is a long tradition in the literature pointing to the perils of overinvestment (Schumpeter 1912) and boom-bust cycles. For example Hori (2007) argues that the investment slump after the Asian financial crisis (1997–98) was at least partly due to overinvestment prior to the crisis. In turn, investment booms and credit bubbles have often been associated with excessive borrowing and rapid accumulation of public and/or external debt. Inflation has also been linked with negative growth outcomes (Fischer 1993), although Bruno and Easterly (1998) and subsequent contributions emphasize that the relationship is ambiguous when inflation is low to moderate. Finally, a considerable literature on the relationship between growth and competitiveness exists. For instance, Easterly and others (1993) and Mendoza (1997) find that terms-of-trade shocks can explain part of the variance in growth across countries. Such shocks could be particularly relevant for countries that are large importers or exporters of fuel and food. Relatedly, there is the concern that exporters specializing in natural resources could be subject to Dutch disease. Prasad, Rajan, and Subramanian (2007) find that exchange rate overvaluation may hinder growth in emerging market economies as manufacturing is crowded out by less productive sectors.

The names of the variables used in this module should be mostly selfexplanatory, so we only note the definitions of those that may not be for Table 4.7. “Banking crisis” is a dummy variable drawn from the database constructed and updated by Laeven and Valencia (2012), which takes the value 1 if the country experienced a banking crisis in any of the five years preceding the current year.

Table 4.7Macro
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
Banking crisis0.770.06
Public debt0.0060.05
Terms of trade0.020.03−0.010.03
Pseudo R20.20
Observations117
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Inflation−0.320.090.250.09
Public debt1.270.17
Terms of trade1.640.26−1.460.19
Real exchange rate−0.280.100.120.10
External debt0.380.090.250.09
Investment share−0.160.08
Banking crisis1.530.42
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, PIP = post-inclusion probability, WALS = weighted average least squares.

Looking at the upper panel of Table 4.7, we find that, according to the probit analysis, high levels of public debt, banking crises, better terms of trade (in level), and adverse terms-of-trade shocks all increase the probability of experiencing a slowdown. One interpretation of the significance of terms of trade, both in level and difference but with opposite signs, is that it captures two different scenarios in our sample. These are the growth collapses that affected poor economies importing fuel during the energy crisis in the late 1970s and the mechanical slowdowns that affected fuel exporters during the oil glut of the early 1980s. In the first case, the unprecedented increase in oil prices simply triggered a strong deterioration of local terms of trade and a decrease in growth rates. In the second case, countries with originally high terms of trade experienced a growth collapse in the wake of the drop in global demand for oil. The significance of banking crises, on the other hand, is not surprising and illustrates the high vulnerability of low-income countries to banking crises over the past 40 years.25 Finally, the positive effect of public debt on the probability of slowdown is reminiscent of the debt overhang problem (Krugman 1988) that affected many low-income countries in the 1990s and 2000s. In fact, our sample contains several heavily indebted poor countries—most of them in sub-Saharan Africa—that registered very poor economic performance after reaching unsustainable levels of public debt.26

Other Determinants

In the next module (Table 4.8), we consider variables that do not fit easily in any of the previous economic categories. ELF is an index of ethno-linguistic fractionalization, which has often been associated with poor social capital and negative growth outcomes (Easterly and Levine 1997; La Porta and others 1999). “Tropics” measures the fraction of a country’s land area that lies in the tropical zone. Various features of this climatic zone, such as poorer land productivity and conditions more favorable to vector-borne diseases could have an adverse impact on growth (Masters and McMillan 2001). Having a large Buddhist population was found by Sala-i-Martin, Doppelhofer, and Miller (2004) to be significantly associated with growth even after controlling for other institutional and cultural factors. Finally, in Table 4.8, the variables of wars and civil conflicts, and natural disasters can clearly depress growth. Looking closely at Table A4.1, we see that several slowdown episodes coincide with notable war and civil conflict episodes, such as the Russian army’s invasion of Afghanistan in the 1970s or the Hutu-Tutsi conflicts in Burundi in the 2000s. Floods in Guyana and Zimbabwe in 2005 and droughts in Zimbabwe in 1991 were also associated with growth collapses.

Table 4.8Other Determinants
I. Final Probit Specification
LevelsDifferences
VariableCoefficientP > zCoefficientP > z
War and civil conflicts0.520.05
Natural disasters0.410.09
Pseudo R20.03
Observations314
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, ELF = ethno-linguistic fractionalization index; PIP = post-inclusion probability, WALS = weighted average least squares.
II. Bayesian Averaging Robustness Tests
LevelsDifferences
WALSBMAWALSBMA
tPIPtPIP
Tropics0.970.09
Buddhist−1.260.12
ELF0.970.06
War and civil conflicts2.070.36
Natural disasters1.760.22
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, ELF = ethno-linguistic fractionalization index; PIP = post-inclusion probability, WALS = weighted average least squares.
Source: Authors’ calculation.Notes: BMA = Bayesian model averaging, ELF = ethno-linguistic fractionalization index; PIP = post-inclusion probability, WALS = weighted average least squares.

Note that because the variables in this module are either time invariant or plausibly exogenous, they enter the specifications contemporaneously rather than with a lag. Moreover, the nature of the variables considered means that they only enter in levels, not in differences. Using these variables, we find, unsurprisingly, that both wars and civil conflicts and natural disasters are significant. The robustness of war is clearly in line with the vast literature studying the economic causes and consequences of conflicts on growth. In the case of low-income countries, a recent contribution by Arbache and Page (2008) has found that conflict in Africa is one of the primary factors associated with growth collapses and stagnation. On the other hand, the importance of natural disasters in driving slowdown episodes reminds us that many low-income countries still have to overcome unfavorable geographic location and severe climatic conditions.

Summary of Findings

Before turning to policy implications, this section summarizes our main findings. Gathering the different pieces of evidence from the seven modules, we see that slowdowns in low-income countries broadly follow three types of scenarios. The first, which we term “noneconomic,” is provoked by events that are not economic in essence, such as exogenous natural disasters (droughts, floods, and so on) or political conflicts that do not have economic proximate causes. For instance, we argue that even though the Afghanistan invasion or the Hutu-Tutsi conflict had very disruptive effects on growth, both conflicts were motivated by strategic or ethnic factors rather than by economic interests. The second type of slowdown, which we denote “economic,” characterizes low-income countries that are particularly sensitive to external demand shocks, especially developing economies that are dependent on exports of primary commodities. As we have emphasized, evidence suggests that these external shocks are in general poorly managed and cause substantial contractions in output. Third and finally, we term “institutional” the slowdowns that are primarily generated by institutional or social deficiencies. In this case, growth accelerations and sudden rises in income generate adverse institutional outcomes, such as rent seeking, civil wars, or excessive public sector growth that make it impossible for countries to sustain this temporary growth or transform past income into productive investments.

Policy Implications

Having identified the different type of slowdowns as well as their robust determinants, we turn to policy implications. In the light of our findings, which are the most vulnerable countries? For instance, are dynamic Asian and African low-income countries likely to experience a slowdown? Eventually, is there room for policy intervention?

To assess the potential risk, Table 4.9 constructs an illustrative “vulnerability map,” which assesses the vulnerabilities of 31 low-income countries in each of the seven dimensions identified as significant in the previous sections. To do so, for each category and each country, we use the latest available data at the country level and look at the country rankings of the 31 countries in each dimension.27 By construction, a country ranked 1 displays the greatest risk of slowdown in that category, whereas a country ranked 31 is considered the least risky.28 To enhance the interpretation, we also use a color scale in which red indicates a riskier ranking and green denotes a lower ranking relative to other economies featured in the table.

Table 4.9Vulnerability Map
RegionCountryNatural

Resources

Rent
Political

Risk
InstitutionsPublic

Debt
Output

Diversification
Food

Exports

(% GDP)
Fuel Exports

(% GDP)
Terms of

Trade
Latin AmericaGuyana8113612618
AsiaBangladesh223261213231322
Cambodia2921262724
Mongolia421714254
Nepal2021172225
Philippines231118112120919
Papua New Guinea2169153
AfricaBenin261020510191
Burundi741012181811
Cameroon9131727101739
Central African Republic1315192272123
Cote d’Ivoire113244192113
Gambia, The242078242425
Ghana15111411488
Kenya2757991017
Lesotho281152471520
Liberia6714
Malawi16161318932012
Mali51212278721
Mauritania152522
Mozambique10212016171257
Niger2172613145
Rwanda191625621224
Senegal25621131615610
Sierra Leone1518141
Tanzania12136520141716
Togo1772533162315
Uganda142122422111614
Zambia31918232319126
Zimbabwe1872818611
Source: Authors’ calculation.
Source: Authors’ calculation.

Looking at Table 4.9, we first see that, with few exceptions (Guyana, for example), the highest risks in each dimension are concentrated in the African region. Second, some African countries, such as Ghana, Liberia, and Mauritania, happen to be “structurally” risky in almost all dimensions. For instance, Mauritania has a combination of poor economic diversification, high dependence on primary exports (fuel and food), poor institutions, and high public debt. This makes such countries vulnerable to both economic and institutional slowdowns. However, several African countries display more specific vulnerabilities in one or two dimensions only. For instance, Côte d’Ivoire combines a high dependence on primary exports with high political risk; Burundi combines high natural resources rent with poor institutions. This suggests that Côte d’Ivoire may be more susceptible to both economic and noneconomic slowdowns, whereas Burundi may be more sensitive to institutional slowdowns. And finally, we see that with the exception of Mongolia and Papua New Guinea, low-income countries in Asia (Bangladesh, Cambodia, Nepal, Philippines) tend to have very low levels of risk in most dimensions, implying that Asian low-income countries may be generally at a lower risk of a growth slowdown.

Having identified the relative strength and weaknesses in each dimension for each country, we are now in a position to identify the Asian and African low-income countries that are most likely to experience a slowdown in the coming years. To do so, Figure 4.2 plots the average rating over all dimensions for a sample of 11 countries in Africa and Asia against the average growth rate over 2008. We find that Ghana, Liberia, and Zimbabwe show very rapid growth over the past four years with a high level of risk. On the other hand, the dynamic low-income countries in east and south Asia, which have also experienced sustained growth periods, all have low levels of risk. Overall, this implies that the recent growth dynamics in Asian low-income countries may be more sustainable than in some African countries.

Figure 4.2Slowdown Risk Map

Source: Author’s calculation.

Conclusion

Although many empirical contributions have emphasized that growth is hard to generate in poor countries, we find that sustaining growth is in fact as challenging as achieving it. Using a conditional convergence framework, we found that that 32 growth slowdowns affected low-income countries between 1960 and 2005. Most occurred in sub-Saharan Africa, with a third of all slowdown periods concentrated in four countries. Moreover, probit analysis and Bayesian averaging tests highlighted that several factors can generate substantial and protracted deviations from the predicted growth path, ranging from institutional misalignments and poor economic diversification to exogenous natural disasters. Broadly speaking, the identified slowdowns fall into three categories. The first is noneconomic: the underlying drivers are natural disasters or political conflicts that do not have economic causes. The second is external shocks, especially in countries relying on primary commodity exports. The third type of slowdown is more related to institutional factors: sudden rises in income lead to adverse institutional outcomes, such as rent seeking, civil wars, or excessive public sector growth, making growth unsustainable.

After identifying the slowdowns and their underlying drivers, we then assessed the vulnerability of Asian and African low-income countries to each risk factor. Overall, we found that most Asian low-income countries seem less at risk than their African counterparts, albeit with a few exceptions (for example, Mongolia and Papua New Guinea rely heavily on natural resources, and political risks are high in Bangladesh). On the other hand, most African low-income countries show high levels of risk. While the vulnerability of some countries lies in specific dimensions, some African countries face significant slowdown risks along all the dimensions (Mauritania, for example, has a combination of poor economic diversification, high dependence on primary exports, poor institutions, and high public debt).

Appendix
Table A4.1Growth Slowdown Episodes (by income group)
High IncomeMiddle IncomeLow Income
Japan1970–1975Algeria1980–1985Haiti1980–1985Papua New Guinea1995–2000Afghanistan1985–1990Pakistan1965–1970
Japan1990–1995Algeria1985–1990Honduras1960–1965Paraguay1980–1985Benin1985–1990Sierra Leone1990–1995
Finland2000–2005Argentina1980–1985Honduras1980–1985Peru1975–1980Burundi1970–1975Sudan2000–2005
Ireland2000–2005Argentina1995–2000Indonesia1995–2000Peru1980–1985Burundi2000–2005Togo1990–1995
Malta2000–2005Belize1990–1995Iran1970–1975Poland1980–1985Cameroon1985–1990Uganda1970–1975
Portugal1990–1995Bolivia1975–1980Iran1975–1980Portugal1970–1975Congo, Republic of1970–1975Zambia1975–1980
Portugal2000–2005Botswana1975–1980Iraq1980–1985Romania1975–1980Cote d’Ivoire1970–1975Zimbabwe1975–1980
Spain1975–1980Botswana2000–2005Jamaica1970–1975Romania1980–1985Egypt1965–1970Zimbabwe1990–1995
Spain2000–2005Brazil1975–1980Jamaica1990–1995South Africa1980–1985Ghana1970–1975Zimbabwe2000–2005
Barbados1970–1975Brazil1980–1985Jordan1965–1970Spain1965–1970Indonesia1975–1980
Barbados1980–1985Bulgaria1980–1985Jordan1980–1985Swaziland1990–1995Kenya1990–1995
Barbados2000–2005Chile1995–2000Korea, Republic of1970–1975Syria1975–1980Lao P.D.R.1985–1990
Bahrain1980–1985Congo, Republic of1985–1990Malaysia1980–1985Syria1980–1985Liberia1980–1985
Cyprus1990–1995Cyprus1980–1985Malaysia1995–2000Syria1995–2000Liberia1985–1990
Israel1975–1980Dominican Republic1975–1980Maldives1985–1990Thailand1995–2000Liberia2000–2005
Kuwait1995–2000Ecuador1975–1980Malta1980–1985Tonga1985–1990Malawi1970–1975
Brunei1980–1985Ecuador1980–1985Mauritius1975–1980Trinidad and Tobago1960–1965Malawi1975–1980
Hong Kong SAR1980–1985Egypt1995–2000Mexico1980–1985Trinidad and Tobago1980–1985Malawi1980–1985
Hong Kong SAR1990–1995El Salvador1975–1980Namibia1970–1975Tunisia1975–1980Mauritania1975–1980
Korea, Republic of1990–1995El Salvador1995–2000Nicaragua1965–1970Uruguay1995–2000Mongolia1990–1995
Korea, Republic of1995–2000Gabon1975–1980Nicaragua1985–1990Venezuela1975–1980Morocco1965–1970
Singapore1995–2000Guatemala1980–1985Panama1980–1985Yemen2000–2005Mozambique1975–1980
Guyana2000–2005Papua New Guinea1980–1985Zambia1970–1975Niger1980–1985
Source: Authors’ compilation.
Source: Authors’ compilation.
Table A4.2Independent Variables: Unit and Sources
DescriptionsSourcesCategoryStartEndFrequency
Fertility rate, total (births per woman)WDIDemography19602009Annual
Dependency ratioUNDemography195020055-year
Sex ratioUNDemography195020055-year
Agriculture share of value added (% of GDP)WDIEconomic Structure19702011Annual
Services share of value added (% of GDP)WDIEconomic Structure19702011Annual
Industry share value added (% of GDP)WDIEconomic Structure19702011Annual
Output diversificationPapageorgiou and Spatafora (2012)Economic Structure20002010Annual
Telephone linesCalderon and Serven (2004)Infrastructure196019955-year
Power generating capacityCalderon and Serven (2004)Infrastructure196019955-year
RoadsCalderon and Serven (2004)Infrastructure196019955-year
Size of governmentEconomic Freedom DatasetInstitutions196020105-year
Rule of lawEconomic Freedom DatasetInstitutions196020105-year
Freedom to trade internationallyEconomic Freedom DatasetInstitutions196020105-year
RegulationEconomic Freedom DatasetInstitutions196020105-year
Financial opennessChinn and Ito (2006)Institutions19702009Annual
Gross capital inflows as % GDPWEOMACRO19702009Annual
Gross capital outflows as % GDPWEOMACRO19702009Annual
Banking crisis dummyLaeven and Valencia (2012)MACRO19752008Annual
Real exchange rateIMF staff calculationsMACRO19502009Annual
Trade openness at 2005 constant prices (%)PWTMACRO19502009Annual
CPI inflationWDIMACRO19702010Annual
Price level of investmentPWTMACRO19502009Annual
External debt (net)-to-GDP ratioLane and Milesi-Ferretti (2007)MACRO19702010Annual
Public debt-to-GDP ratioAbbas and others (2010)MACRO19502010Annual
Terms of tradeWEOMACRO19702009Annual
Reserves-to-GDP ratioWEOMACRO19702010Annual
Investment share of PPP GDP per capita at 2005 constant pricesPWTMACRO19602010Annual
Oil exporters’ price shockIMF staff calculationsMACRO19502010Annual
Food exporters’ price shockIMF staff calculationsMACRO19502010Annual
Oil importers’ price shockIMF staff calculationsMACRO19502010Annual
Food importers’ price shockIMF staff calculationsMACRO19502010Annual
Fraction of country in TropicsSala-i-Martin, Doppelhofer, and Miller (2004)Other19502010Annual
Spanish colonySala-i-Martin, Doppelhofer, and Miller (2004)Other19502010Annual
Fraction BuddhistSala-i-Martin, Doppelhofer, and Miller (2004)Other19502010Annual
Ethno-linguistic fractionalizationSala-i-Martin, Doppelhofer, and Miller (2004)Other19502010Annual
War and civil conflictsCorrelates of War ProjectOther19502010Annual
Natural disastersInternational Disaster DatabaseOther19502010Annual
Distance (GDP weighted)World BankTRADE19502010Annual
Regional integrationIMF staff calculationsTRADE19602010Annual
Trade diversification: Theil IndexPapageorgiou and Spatafora (2012)TRADE19602010Annual
Source: IMF staff.Note: CPI = consumer price index, PPP = purchasing power parity, PWT = Penn World Tables, UN = United Nations, WDI = World Bank, World Development Indicators database, WEO = IMF, World Economic Outlook database.
Source: IMF staff.Note: CPI = consumer price index, PPP = purchasing power parity, PWT = Penn World Tables, UN = United Nations, WDI = World Bank, World Development Indicators database, WEO = IMF, World Economic Outlook database.
References

    AbbasS.N.BelhocineA.ElGanainy and M.Horton.2010. “A Historical Public Debt Database.” IMF Working Paper 10/245International Monetary FundWashington.

    Abu-GhaidaD. and S.Klasen.2004. “The Costs of Missing the Millennium Development Goal on Gender Equity.” World Development32 (7): 10751107.

    AgosinM.2007. “Export Diversification and Growth in Emerging Economies.” Working Papers 233University of Chile, Department of EconomicsSantiago.

    AiyarS.R.DuvalD.PuyY.Wu and L.Zhang.2013. “Growth Slowdowns and the Middle-Income Trap.” Working Paper 13/71International Monetary FundWashington.

    AiyarS. and A.Mody.2011. “The Demographic Dividend: Evidence from the Indian States?Working Paper 11/38International Monetary FundWashington.

    ArbacheS. and J.Page.2008. “Hunting for Leopards: Long Run Country Income Dynamics in Africa.” Policy Research Working Paper Series 4715World BankWashington.

    BarroR. J.1991. “Economic Growth in a Cross-Section of Countries.” Quarterly Journal of Economics106 (2): 40744.

    Ben-DavidD.1993. “Equalizing Exchange: Trade Liberalization and Income Convergence.” Quarterly Journal of Economics108 (3): 65379.

    BergA.J. D.Ostry and J.Zettelmeyer.2012. “What Makes Growth Sustained?Journal of Development Economics98 (2): 14966.

    BloomD.D.Canning and P.Malaney.2000. “Demographic Change and Economic Growth in Asia.” Population and Development Review26 (Supplement): 25790.

    BloomD. and D.Canning2004. “Global Demographic Change: Dimensions and Economic Significance.” In Global Demographic Change: Economic Impacts and Policy Challengesproceedings of a symposium sponsored by the Federal Reserve Bank of Kansas CityJackson Hole, WyomingAugust26–28.

    BloomD. and J.Williamson.1998. “Demographic Transitions and Economic Miracles in Emerging Asia.” World Bank Economic Review12 (3): 41955.

    BoserupE.1975. “The Impact of Population Growth on Agricultural Output.” Quarterly Journal of Economics89 (2): 25770.

    BoulholH. and A.Serres.2010. “Have Developed Countries Escaped the Curse of Distance?Journal of Economic Geography10 (1): 11339.

    BrunoM. and W.Easterly.1998. “Inflation Crises and Long-Run Growth.” Journal of Monetary Economics41 (1): 326.

    BuchananJ. and G.Tullock.1962. The Calculus of Consent: Logical Foundations of Constitutional Democracy. Ann Arbor, Michigan: University of Michigan Press.

    CalderonC. and L.Serven.2004. “The Effects of Infrastructure Development on Growth and Income Distribution.” Policy Research Working Paper Series 3400World BankWashington.

    CaldwellJ. C.1976. “Toward A Restatement of Demographic Transition Theory.” Population and Development Review2 (3/4): 32166.

    CaselliF.G.Esquivel and F.Lefort.1996. “Reopening the Convergence Debate: A New Look at Cross-Country Growth Empirics.” Journal of Economic Growth1 (3): 36389.

    ChinnM. D. and H.Ito.2006. “What Matters for Financial Development? Capital Controls, Institutions, and Interactions.” Journal of Development Economics81 (1): 16392.

    CollierP. and J.Gunning.1999. “Why Has Africa Grown Slowly?Journal of Economic Perspectives13 (3): 322.

    CollierP. and A.Hoeffler.2001. “On the Duration of Civil War.” Policy Research Working Paper Series 2681World BankWashington.

    Commission on Growth and Development. 2008. The Growth Report: Strategies for Sustained Growth and Inclusive Development. Washington: World Bank.

    De LongJ. and A.Shleifer.1993. “Princes and Merchants: European City Growth before the Industrial Revolution.” Journal of Law and Economics36 (2): 671702.

    DurlaufS. N.A.Kourtellos and C. M.Tan.2008. “Are Any Growth Theories Robust?Economic Journal118 (527): 32946.

    EasterlyW.M.KremerL.Pritchett and L. H.Summers.1993. “Good Policy or Good Luck? Country Growth Performance and Temporary Shocks.” Journal of Monetary Economics32 (3): 45983.

    EasterlyW. and R.Levine.1997. “Africa’s Growth Tragedy: Policies and Ethnic Division.” Quarterly Journal of Economics112 (4): 120350.

    EdwardsS.2001. “Capital Mobility and Economic Performance: Are Emerging Economies Different?NBER Working Paper 8076National Bureau of Economic ResearchCambridge, Massachusetts.

    EgertB.T.Kozluk and D.Sutherland.2009. “Infrastructure and Growth: Empirical Evidence.” William Davidson Institute Working Paper 957University of Michigan Business School, William Davidson InstituteMichigan.

    EichengreenB.D.Park and K.Shin.2011. “When Fast Growing Economies Slow Down: International Evidence and Implications for China.” NBER Working Paper 16919National Bureau of Economic ResearchCambridge, Massachusetts.

    ElbadawiI. A.B. J.Ndulu and N.Ndungu.1997. “Debt Overhang and Economic Growth in sub-Saharan Africa.” In External Finance for Low-Income Countries edited by Z.Iqbal and R.Kanbur.Washington: International Monetary Fund.

    FischerS.1993. “The Role of Macroeconomic Factors in Growth.” Journal of Monetary Economics32 (3): 485512.

    GaertnerM. and C.Papageorgiou.2011. “Sustaining Tanzania’s High Growth Path through Exports.” Unpublished. International Monetary FundWashington.

    GrilliV. and G.Milesi-Ferretti.1995. “Economic Effects and Structural Determinants of Capital Controls.” Working Paper 95/31International Monetary FundWashington.

    GusehJ. S.1997. “Government Size and Economic Growth in Developing Countries: A Political-Economy Framework.” Journal of Macroeconomics19 (1): 17592.

    HagenE. E.1975. The Economics of Development. Homewood, Illinois: Irwin.

    HarrisJ. and M.Todaro.1970. “Migration, Unemployment and Development: A Two-Sector Analysis.” American Economic Review60 (1): 12642.

    HausmannR.F.Rodriquez and R.Wagner.2006. “Growth Collapses.” Working Paper Series 06-046Harvard University, John F. Kennedy School of GovernmentCambridge, Massachusetts.

    HigginsM.1998. “Demography, National Savings, and International Capital Flows.” International Economic Review39 (2): 34369.

    HoriM.2007. “Investment Recovery from Financial Crises: A View from Cross Country Experiences.” In Thailand: Selected Issues IMF Country Report 07/231. Washington: International Monetary Fund.

    ImbsJ. and R.Wacziarg.2003. “Stages of Diversification.” American Economic Review93 (1): 6386.

    IslamN.1995. “Growth Empirics: A Panel Data Approach.” Quarterly Journal of Economics110 (4): 112770.

    KnackS. and P.Keefer.1997. “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation.” Quarterly Journal of Economics112 (4): 125188.

    KelleyA. and S.Schmidt.1996. “Saving, Dependency and Development.” Journal of Population Economics9 (4): 36586.

    KrugmanP.1988. “Financing vs. Forgiving a Debt Overhang.” Journal of Development Economics29 (3): 25368.

    KorenM. and S.Tenreyro.2007. “Volatility and Development.” Quarterly Journal of Economics122 (1): 243287

    KuznetsS.1966. Modern Economic Growth, Rate, Structure and Spread. New Haven, Connecticut and London: Yale University Press.

    LaneP. R. and G. M.Milesi Ferretti.2007. “The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970-2004.” Journal of International Economics73 (November): 22350.

    LaevenL. and P.Valencia.2012. “Systemic Banking Crises Database: An Update.” Working Paper 12/163International Monetary FundWashington.

    La PortaR.F.Lopez-de-SilanesA.Shleifer and R.Vishny.1997. “Legal Determinants of External Finance.” Journal of Finance52 (3): 113150.

    La PortaR.F.Lopez-de-SilanesA.Shleifer and R.Vishny.1998. “Law and Finance.” Journal of Political Economy106 (6): 111355.

    La PortaR.F.Lopez-de-SilanesA.Shleifer and R.Vishny.1999. “The Quality of Government.” Journal of Law Economics and Organization15 (1): 22279.

    LeamerE. E.1979. “Specification Search: Ad Hoc Inference with Nonexperimental Data.” Journal of Economic Literature17 (2): 56668.

    LewisW. A.1980. “The Slowing Down of the Engine of Growth.” American Economic Review70 (4): 55564.

    MagnusJ. R.O.Powell and P.Prüfer.2010. “A Comparison of Two Model-Averaging Techniques with an Application to Growth Empirics.” Journal of Econometrics154 (2): 13953.

    MasanjalaW. and C.Papageorgiou.2008. “Rough and Lonely Road to Prosperity: A Reexamination of the Sources of Growth in Africa Using Bayesian Model Averaging.” Journal of Applied Econometrics23 (5): 67182.

    MasonA.2001. “Population, Capital, and Labor.” In Population Change and Economic Development in East Asia: Challenges Met Opportunities Seized edited by A.Mason.Stanford: Stanford University Press.

    MastersW. and M.McMillan.2001. “Climate and Scale in Economic Growth.” Journal of Economic Growth6 (3): 16786.

    MauroP.1995. “Corruption and Growth.” Quarterly Journal of Economics110 (3): 681712.

    MendozaE.1997. “Terms of Trade Uncertainty and Economic Growth.” Journal of Economic Development54 (2): 32356.

    MinskyH.1986. Stabilizing an Unstable Economy. New Haven, Connecticut: Yale University Press.

    MinskyH.1992. “The Financial Instability Hypothesis.” Jerome Levy Economics Institute Working Paper 74Jerome Levy Economics Institute, Annandale-on-HudsonNew York.

    MironovM.1995. “Corruption and Growth.” Quarterly Journal of Economics110 (3): 681712.

    NorthD.1981. Structure and Change in Economic History. New York: Norton & Co.

    NorthD.1990. Institutions Institutional Change and Economic Performance. Cambridge, United Kingdom: Cambridge University Press.

    PapageorgiouC. and NikolaSpatafora.2012. “Economic Diversification in Low-Income Countries: Stylized Facts and Macroeconomic Implications.” IMF Staff Discussion Note 12/13. International Monetary FundWashington.

    PradhanS.1996. “Evaluating Public Spending: A Framework for Public Expenditure Reviews.” World Bank Discussion PapersWorld Bank, Washington.

    PrasadE.R.Rajan and A.Subramanian.2007. “Foreign Capital and Economic Growth.” NBER Working Paper 13619National Bureau of Economic ResearchCambridge, Massachusetts.

    PritchettL.2000. “Understanding Patterns of Economic Growth: Searching for Hills among Plateaus, Mountains, and Plains.” World Bank Economic Review14 (2): 22150.

    QuinnD.1997. “The Correlates of Change in International Financial Regulation.” American Political Science Review91(September): 53151.

    ReddingS. and A.Venables.2004. “Geography and Export Performance: External Market Access and Internal Supply Capacity.” In Challenges to Globalization: Analyzing the Economics edited by R. E.Baldwin and L. A.Winters.Chicago: University of Chicago Press.

    ReddyS. and C.Minoiu.2009. “Real Income Stagnation of Countries 1960-2001.” Journal of Development Studies45 (1): 123.

    ReinhartC. and K.Rogoff.2008. “Banking Crises: An Equal Opportunity Menace.” NBER Working Paper 14587National Bureau of Economic ResearchCambridge, Massachusetts.

    RollerL. and L.Waverman.2001. “Telecommunications Infrastructure and Economic Development: A Simultaneous Approach.” American Economic Review91 (4): 90923.

    RompW. and J.de Haan.2007. “Public Capital and Economic Growth: A Critical Survey.” Perspektiven der Wirtschaftspolitik8 (1): 652.

    Sala-i-MartinX.G.Doppelhofer and R. I.Miller.2004. “Determinants of Long-Term Growth: A Bayesian Averaging of Classical Estimates (BACE) Approach.” American Economic Review94 (4): 81335.

    Sala-i-MartinX. and A.Subramanian.2003. “Addressing the Natural Resource Curse: An Illustration from Nigeria.” NBER Working Paper 9804National Bureau of Economic ResearchCambridge, Massachusetts.

    SchuknechtL.1999. “Tying Governments’ Hands in Commodity Taxation.” Journal of African Economies8 (2): 15281.

    SchumpeterJ.1912. The Theory of Economic Development. Cambridge, Massachusetts.

    ScullyG. W.1988. “The Institutional Framework and Economic Development.” Journal of Political Economy96 (3): 65262.

    SenA.1992. Inequality Reexamined. Oxford: Oxford University Press.

    SimonJ. L. and R.Gobin.1980. ‘The Relationship between Population and Economic Growth in LCD’s.” In Research in Population Economics 2 edited by J. L.Simon and J.DaVanzo.London: J A I Press.

    TiffenM.1995. “Population Density, Economic Growth and Societies in Transition: Boserup Reconsidered in a Kenyan Case-Study.” Development and Change26 (1): 3166.

After comparing slowdown episodes with those identified by Berg, Ostry, and Zettelmeyer (2012), we find that more than half of the slowdowns we capture are also captured by their study, and about two-thirds when using a looser criterion of 25 percent significance threshold. However, a substantial number of episodes identified in their study do not qualify in our analysis. Our conditional convergence framework, which is based on a more restrictive definition of slowdowns, therefore yields a substantially lower number of slowdowns compared with Berg, Ostry, and Zettelmeyer (2012).

See in this respect the role of infrastructure, government size, and regulation. Masanjala and Papageorgiou (2008) also emphasize the peculiarity of growth determinants in low-income countries, showing that determinants of growth in Africa are strikingly different from those in the rest of the world.

We use five-year rolling geometric averages.

The final period, 2006–09, covers only four years due to a lack of subsequent data in the Penn World Tables 7.0.

This represents the most parsimonious established framework for conditional convergence using panel data. The rate of investment in physical capital is taken from the Penn World Tables. The rate of investment in human capital across countries is unavailable, so we follow the standard practice of using the stock of human capital instead; for example, Islam (1995) and Caselli, Esquivel, and Lefort (1996). See Aiyar and others (2013) for detailed calculation methods.

Note that these conditions imply that we cannot identify slowdowns either in our sample’s initial period (1955–60), because there is no prior period for comparison, or in the final period (2005–09) because there is no subsequent period for comparison.

See Aiyar and others (2013) for an extensive discussion of the middle-income trap.

Although the total sample of low-income countries is made of 441 observations, the sample of “possible” slowdowns is smaller because, by construction, slowdowns cannot be identified in the initial period or in the last one.

More generally we support the analysis of Easterly and others (1993), who point out that long-term growth averages “mask important and distinct periods of success and failure” and call for more attention to growth dynamics.

Despite a relatively large body of empirical literature, theories behind growth slowdowns are still at an early stage of development. Here we allow the data to “speak for itself” using variables borrowed from the growth literature.

See Durlauf, Kourtellos, and Tan (2008) for a thorough review.

Note that we use the lagged levels and difference to minimize the possible endogeneity issues. As for the interpretation, the use of lagged values implies that for a slowdown episode over 1975–80, the 1975 level of variable X is used together with the change in that variable between 1970 and 1975.

Sierra Leone, which also experienced a slowdown in early 1990s, followed the same pattern. Other cases supporting the institutional interpretation of the resource-curse hypothesis have also been studied. For instance, the Nigerian experience–where waste and corruption accounted for a poor long-term economic performance–is explored in detail by Sala-i-Martin and Subramanian (2003).

At this time, the agricultural sector in Malawi accounted for 40 percent of its GDP and 94 percent of exports, most in the form of tobacco exports.

This work has engendered several subsequent contributions emphasizing the importance of legal institutions more broadly.

Guseh (1997) found that, once fixed effects are controlled for, growth in government size has negative effects on economic growth, but the negative effects are three times as great in nondemocratic socialist systems as in democratic market systems.

According to Collier and Gunning (1999), public employment was expanded, often as an end in itself. For example, in Ghana by the late 1970s the public sector accounted for three-quarters of formal wage employment), and even in a more market-oriented economy like Kenya, the figure was 50 percent as of 1990.

This poor service delivery was also said to handicap firms through unreliable transport and power, inadequate telecommunications networks, and unreliable courts. This in turn might have increased the likelihood of a slowdown. See Collier and Gunning (1999) for a review.

The size of government remains significant at 1 percent.

More recent contributions, and studies using more direct measures of infrastructure, have generally found a more positive impact of public capital on growth (Roller and Waverman 2001; Calderon and Serven 2004; Egert, Kozluk, and Sutherland 2009).

More generally, the Commission on Growth and Development (2008) found that urbanization is related to stronger market forces and a shifting and deepening of the knowledge base of the economy.

For example, Boulhol and de Serres (2010) demonstrate that the relationship is valid even within a panel of advanced economies. Furthermore, economies that take advantage of their geographic location by pursuing regional integration might be thought to improve their growth prospects. Ben-David (1993) showed that trade agreements in Europe have furthered convergence among member countries.

The example of Malawi, detailed earlier, exemplifies this channel.

Note that using a probit specification and only the export diversification index as a regressor, we find a result similar to our earlier result: poor export diversification is associated with a higher probability of slowdown.

For a thorough examination of banking crises in both advanced and developing economies, see Reinhart and Rogoff (2008), who state that “the incidence of banking crises proves to be remarkably similar in the high- and middle-to-low-income countries.”

“It is now widely understood that the crux of sub-Saharan Africa’s debt problem is excessive debt overhang, which has led many countries in the region to be classified as insolvent. The severity of the debt crises has impacted negatively on growth in incomes per capita and private investment rates in the region.” Elbadawi, Ndulu, and Ndungu (1997).

We omit infrastructure, which did not yield noteworthy results, as well as population density and natural disaster, which cover variables largely irrelevant from a policy perspective. On the other hand, we proxy for the risk of war and civil conflict using (external and internal) conflict risk ratings computed by the International Country Risk Guide.

In practice, some data points are missing for some countries in each category, so the maximum rank ranges between 30 and 27. Missing data points are reported in gray.

    Other Resources Citing This Publication