A. Definition of State Fragility: Literature and our Approach

While there are many ways to define fragile countries, reflecting its complexity, they seem to have common characteristics (IMF, 2008). These include: (i) significant institutional and policy implementation weakness; (ii) a fractious political context; (iii) severe domestic resource constraints; and (iv) high vulnerability to shocks (e.g., IMF, 2012).

However, there is no universally accepted operational definition of “fragility” to conduct empirical studies. The complex and multidimensional nature of state fragility has militated against a strict definition of FCS, with various definitions used by different institutions and scholars (Appendix I provides a summary of operational definitions used in different institutions).

• In most of the academic studies, states are considered as fragile when their weak institutional capacity, political instability and weak governance severely limit the state’s ability to guarantee security to its citizens and deliver basic public services. For example, Collier (2020) suggests the following characteristics of a fragile state: little or no overarching shared identity, lack of state legitimacy, lack of capacity, existential uncertainties, underdeveloped private sector, and high exposure to political and economic shocks. Acemoglu and Robinson (2021) define a fragile state by the strength of its capacity and its political regime. Limited political and institutional ability to implement appropriate policies to address structural challenges and exogenous shocks have resulted in weak economic performance, chronic humanitarian crises, persistent social tensions, and often violence or the legacy of armed conflicts in fragile states.

• On the other hand, in most of the empirical studies, state fragility is defined empirically as low Country Policy and Institutional Assessment (CPIA) scores or high values of a fragility index.⁵ In most of the IMF/World Bank (WB) papers, state fragility is defined as having either weak institutional capacity as measured by the World Bank’s CPIA score (average of 3.2 or lower) or experience of conflict (signaled by the presence of a peace-keeping or peace-building operation in the most recent three-year period). Recent work by the OECD breaks down the drivers of fragility for each country and reveals different patterns of vulnerability.⁶ While sensible, this approach requires very big datasets and does not provide a long time series. In most cases, fragile countries have limited data, making the compilation of data with sufficient coverage of countries and time periods very challenging.

Given the above, this study follows the definition used for operational and analytical work by the WB and the IMF. It is based on the CPIA ratings and the presence of UN and regional peace keeping and peace building operations (PKO/PBO).^7,8 There are a couple of advantages in this approach. First, using the WB and IMF definition makes this study’s results operationally relevant for both institutions. Second, it can be applied more or less consistently backwards to the late 1970s, which would enable us to have a wider data coverage. This is not the case for almost all of the other definitions. Third, the definition covers both “slow moving” aspects of state fragility (by the CPIA component) and “fast moving” aspects (by the PKO/PBO one)—see below for detailed discussion on the two aspects. In any event, the empirical strategy employed in the paper could be applied to datasets compiled by any definition, subject to data availability.

B. Determinants of State Fragility and Policies to Address Fragility

Understanding the nature and drivers of fragility is important to devise policies that will help countries. Many theoretical and empirical studies have examined the determinants of state fragility to identify the manifestations and a better understanding of state fragility to inform policy making. Meanwhile, IFIs have worked closely with the authorities in fragile countries to help them to strengthen capacity.

Theoretical studies found political instability and violence, insecurity, weak institutions, and corruption as key drivers of FCS. Vallings and Moreno-Torres (2005) noted that weak institutions are the central driver of state fragility, relative to economic factors. Kaplan (2008) pointed out that fragile states have two structural problems, political identity fragmentation and weak national institutions, which reinforce each other. Andrimihaja et al., (2011) noted that three features—political instability and violence, insecure property rights and unenforceable contracts, and corruption—conspire to create a slow-growth-poor-governance equilibrium.

On the other hand, empirical studies found macroeconomic variables, institutions, predetermined factors, and social indicators as major determinants of fragility. Feeny et al., (2015) found that macroeconomic variables such as income levels and economic growth are important determinants of state fragility. Similarly, poorer countries tend to be more fragile than richer countries on average and countries that are more open to trade tended to be less fragile (Carment et al., 2008, 2011). Also, some studies identified weak institutions as the main determinants of state fragility. For example, Bertocchi and Guerzoni (2012) used probit regressions to identify that restrictions of civil liberties and the number of revolutions raise the probability of fragility in Sub-Saharan African countries. Predetermined factors affect too, including: country size (Feeny et al., 2015); and ethnic risk and ethnic diversity (Carment et al., 2008). Also, state fragility depends on social indicators, including low values of the Human Development Indicator (HDI), higher infant mortality rates, and a lower level of education (Feeny et al., 2015; Carment et al., 2008, 2011).

Some studies found the importance of strengthening institutions and governance in crafting policies to tackle fragility. There is in general no one-size-fits-all approach to addressing fragility, given that FCS are hererogeneous in their economic structure, sources and degree of fragility, and macroeconomic performance (IMF, 2017a). Therefore, specific policies to address fragility need to take into account country-specific circumstances and reflect the country’s degree of fragility (Baer et al., 2021). Case studies have so far provided useful insights. For example, enhancing the capacity to mobilize revenues more efficiently and strengthening control in budget and financial management helped countries to exit from fragility (IMF, 2017b). Fiscal institutions and fiscal space are significantly and robustly associated with building resilience (Delechat et al., 2018). Absorptive capacity constraints hinder capacity building from delivering real and lasting institutional change (IMF, 2012; 2015). Benefits of capacity development support to fragile state could be enhanced by increasing the use of on-the-ground experts, employing realistic impact assessment tools, and making efforts to secure adequatre financial resources for capacity building (IMF IEO, 2018). It is also important to develop forward-looking, holistic country strategies and practical steps that integrate the roles of policy advice, financial support and capacity building. Moreover, greater flexibility should be built into program design and more attention to the political economy in Fund supported arrangements for fragile countries (IMF, 2011). An evaluation conducted by the World Bank Group in 2016, based on case studies, emphasized that strengthening legitimate institutions and governance to provide citizen security, justice, and jobs is crucial to break cycles of violence and help restore stable development path in fragile countries.

C. Recent Studies to Address a “Fragility Trap”: Any Asymmetry and Implications?

Recent studies on state fragility emphasize the persistence of state fragility (“fragility trap”) caused by the nexus of politics, institution building social or economic inclusion, and economic policies. Fragile countries may be further entrenched by the reinforcing interactions between fragility, economic and political instability, and daunting structural challenges, such as poor governance, weak institutional capacity, and lack of political and social inclusion (IGC, 2018). This emphasizes the importance of addressing both slow-moving and fast-moving factors to avoid being trapped by and facilitating exit from state fragility. The slow-moving factors include significant institutional and policy implementation weaknesses. In this context, departing from “extractive” political and economic systems that impede economic growth (Acemoglu and Robinson, 2012) and policy intervention to change institutional, political and economic structure should help countries avoid fragility as well as facilitate exit from it.⁹ On the other hand, the fast moving factors include political conflict, exogenous shocks, and economic/social stability, including economic policy errors, that could affect state fragility and push a country over the cliff toward fragility (Vallings and Moreno-Torres, 2005; Gelbard et al., 2015). Other studies found that building tax capacity could support state building, improve institutional capacity, and foster economic development (Besley and Persson, 2010; Gaspar et al., 2016a, 2016b; and Akanbi, 2019).

The nexus of the two factors also suggests considerable hurdles for successful exit from a fragility trap, with substantial benefits of avoiding fragility in the first place. Acemoglu and Robinson (2019) argue that exiting from fragility requires empowering both the state and the society to create a strong and inclusive state. Meanwhile, Besley and Persson (2011b) and Collier (2020) argue for the criticality of policy intervention and institution building, especially at a critical juncture (a “pivotal moment”). Relatedly, Hegre et al., (2017) pointed to the oversized influence of conflict in low-income countries: (i) a single onset of conflict could be enough to push the country into the trap; and (ii) even a couple of years refraining from a major conflict will reduce the risk of conflict over the next decades substantially.

Recent studies indicate the need of a new framework to secure development in FCS at the brink of entering or entrenched in “fragility trap.” Zoelick (2008) noted that a new framework is required which involves building security, legitimacy, governance, and economy. After identifying specific traps—the conflict trap; the natural resources trap; the trap of being landlocked with bad neighbors; and the trap of bad governance in a small country, Collier (2007) concludes that a new mix of policy instruments is required to escape from these traps, where (i) the policy mix for tackling these traps should include: aid; security; laws and charters; and trade policies but (ii) aid should not be viewed as the main solution because it is unable to stop conflict although it can help after a conflict has ended. Other studies support this: aid only works in good policy environments, and weak absorptive capacity as well as the failure of conditionality to buy reforms hinder the effectiveness of aid in FCS (Burnside and Dollar, 2000; Carment et al., 2008).¹⁰ Smarter demand-driven aid modalities and instruments should therefore be implemented (OECD, 2015).

D. A Gap: Factors Explaining the Turning Points

Past studies examined the root causes of fragility rather than the turning points. A review of the literature reveals that previous theoretical and empirical studies have focused on the common characteristics of countries in fragile situations and the determinants of fragility. Meanwhile, IFIs have focused on how to improve engagement to help build capacity in fragile countries. There is, however, a dearth of literature on what causes countries to enter into and exit fragility, or turning points, as well as how to stay out of fragility after exit. Moreover, currently, many low income countries close to fragility or those that exited fragility during 2017–2019 have faced the huge exogenous shocks from the COVID-19 pandemic and its economic spillovers. It is therefore important for policymakers to understand the factors that can prevent their countries from falling into fragility and the measures which can be implemented to exit the fragility trap. This study aims to analyze these issues.

A. Event Study Analysis

We conduct event studies on a wide range of indicators. These include: (i) institutional and governance indicators (e.g., Polity IV and change in executive powers); (ii) pivotal moments; (iii) macroeconomic indicators (e.g., GDP growth and inflation); (iv) fiscal indicators (e.g., tax revenue, government expenditure, education expenditure, military expenditure, health expenditure, and capital spending); (v) social indicators (e.g., HDI, infant mortality, and life expectancy); and (vi) shocks (e.g., terms of trade, natural disasters, conflicts, and financial crises). These indicators were used in a pairwise comparison of the identified entry/exit events to their counterfactuals to check for any meaningful difference between them.²² The Mood’s median test is used to test whether the difference between the two medians is statistically significant (if the number of observations allows it). While event studies cannot escape endogeneity concerns, to minimize the effects of endogeneity, we focus on developments of indicators during T-4 and T-2 (where Year T is the identified year of entry into or exit from fragility) in most of the analysis. In the paper, we report the economically important results with statistically significant differences; the rest are available upon request.

The results from the event studies show important differences between the identified entry/exit events and their counterfactuals for several indicators: ²³

• Political competition affects entry into fragility as well as sustained exit from fragility (Figure 1). By construction, the political competition variable could be regarded as a proxy for political inclusion.²⁴ The result of the Mood’s median test shows that the identified events and their counterfactuals have statistically-significant different medians.²⁵ This may reflect that more political competition could facilitate a departure from “extractive” political and economic systems that impede economic growth (Acemoglu and Robinson, 2012). On the other hand, executive constraints, a measure of the strength of the executive vs. other institutions, do not appear to differentiate identified entry and exit events from their counterfactuals before the events. However, sustained exit cases have stronger executive constraints than those returning to fragility after they exited from fragility (T+2-T+4)―the median test indicates the difference is significant at 10 percent level (Figure 1). While its interpretation should be careful (because it may be contaminated by endogeneity), this observation implies that more institutionalized constraints on the decision making powers of executives at the early stage of exit could help countries to sustain exit from fragility, perhaps through making recovery/economic activity more private-driven.

• The size of social spending (health/education) helps sustained exit from fragility (Figure 2). Both health and education spending are significant at one percent level by the median test. Also, countries with exit events had statistically stronger human development fundamentals as measured by Human Development Index (HDI) than their counterfactuals (Figure 2), based on the median test too. These findings may indicate that better social conditions could affect confidence in government, political support, and traction of reforms through promoting more inclusive growth, supporting a virtus cycle to exit from fragility and its sustainability. While these findings are based on simple event studies, they apper in line with the existing studies as well as other finding in the paper too. For example, IMF (2019) and Gelbard et al., (2015) emphasize the importance of social spending to keep macroeconomic, social and political stability in fragile states.²⁶ Moreover, an increase in GDP growth could help to prevent countries from entering into fragility and raise the probability of exiting from fragility, as noted in the findings from the logit model (see Section IV-C).

• Pivotal moments seem to be more associated with the cases of exit from fragility. The share of exit events with pivotal moments around the exit is higher than that in their counterfactuals (about 20 percent vs. 10 percent, statistically significant at 1 percent level).²⁷ This holds for sustained exit as well as early exit from fragility (about 25 percent vs. 5 percent, and about 25 percent vs. 10 percent, statistically significant at 1 percent level).²⁸ These observations are in line with the discussions in Collier (2020).

• Conflict may have lasting effect on the likelihood of exiting from fragility. The share of the cases with conflict in the identified exit events is statistically smaller than that of their counterfactuals (30 percent vs. 40 percent, statistically significant at 5 percent).²⁹ This seems to be in line with other studies that highlighted that the legacy of conflict has made it difficult for fragile states to escape from fragility (Gupta et al., 2005; Cillers and Sisk, 2013; Hegre et al., 2017).

• We did a couple of robustness check on some of the above findings. Due to limited data availability for some indicators, especially social spending in percent of overall government spending, the available sample is too small to apply the median test. However, many of the findings seem to survive after the robustness checks.

  • “Policy efforts” to preserve social spending may affect early exit from fragility. In the above, we use social spending in percent of GDP. However, in general, LICs face tight financing constraints and reprioritize their spending to create space for social spending. This is particularly relevant when they are at the brink of falling into fragility or in state fragility. This suggests policy efforts to protect social spending could be better measured by social spending in percent of overall fiscal spending instead of in percent of GDP, as in Besley and Persson (2010). While the small numbers of sample do not allow for the application of statistical tests, Figure 3 seems to indicate preserving health spending could help early exit from state fragility.

  • Dropping conflict episodes do not appear to affect many of the above results. The definition of state fragility includes both slow moving factors (e.g., institutions, proxied by CPIA) and fast moving factors (conflicts). Therefore, it would be interesting to check whether the above results hold after stripping out the identified entry/exit cases with conflicts.³⁰ Even after dropping the identified turning points with conflicts, (i) political competition affects falling into fragility (i.e., the medians of the entry events without conflicts and the counterfactuals are statistically different at one percent level), and (ii) HDI affects exit from fragility (statistically significant at one percent level). On the other hand, the differences in the medians of health and education spending (in percent of GDP) between exit events returning to fragility and sustained exit cases become smaller after dropping the conflict cases. Unfortunately, however, the small sample hampers applying the statistical methodology. Therefore, at this stage, it is difficult to get concrete results for social spending.

Event Studies: Political Competition and Executive Constraints ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ For political competition, entry events (sustained exit) have a statistically different median from their counterfactuals (return to fragility within 10 years), based on the median test. For executive constraints, the difference in medians between sustained exit and returned to fragility within 10 years cases is marginally significant (10 percent level).^2/ Higher value of Polity IV means more political inclusion and more executive constraints.Sources: Polity IV Project and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Event Studies: Political Competition and Executive Constraints ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ For political competition, entry events (sustained exit) have a statistically different median from their counterfactuals (return to fragility within 10 years), based on the median test. For executive constraints, the difference in medians between sustained exit and returned to fragility within 10 years cases is marginally significant (10 percent level).^2/ Higher value of Polity IV means more political inclusion and more executive constraints.Sources: Polity IV Project and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Event Studies: Political Competition and Executive Constraints ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ For political competition, entry events (sustained exit) have a statistically different median from their counterfactuals (return to fragility within 10 years), based on the median test. For executive constraints, the difference in medians between sustained exit and returned to fragility within 10 years cases is marginally significant (10 percent level).^2/ Higher value of Polity IV means more political inclusion and more executive constraints.Sources: Polity IV Project and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Event Studies: Health and Education Spending and HDI ^1/

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Health spending (education spending) has different medians across sustained exit and returned to fragility within 10 years cases (sustained exit and returned to fragility within 10 years cases). Human development index has different median across exit events and their counterfactuals.Sources: World Development Indicators (WDI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Event Studies: Health and Education Spending and HDI ^1/

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Health spending (education spending) has different medians across sustained exit and returned to fragility within 10 years cases (sustained exit and returned to fragility within 10 years cases). Human development index has different median across exit events and their counterfactuals.Sources: World Development Indicators (WDI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Event Studies: Health and Education Spending and HDI ^1/

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Health spending (education spending) has different medians across sustained exit and returned to fragility within 10 years cases (sustained exit and returned to fragility within 10 years cases). Human development index has different median across exit events and their counterfactuals.Sources: World Development Indicators (WDI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Event Studies: Health and Education Spending in percent of Overall Spending

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Event Studies: Health and Education Spending in percent of Overall Spending

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

Event Studies: Health and Education Spending in percent of Overall Spending

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

B. Synthetic Control Method (SCM)

Next, we will apply the SCM to look for systematic differences in the behavior of indicators across identified entry/exit events and their counterfactuals. The SCM is a data-driven procedure for comparing the performance of an indicator in one country against a control group with structured guidelines for the selection of the control group and weights to compute a synthetic projection (Abadie et al., 2010).³¹ The method demonstrates the counterfactual effect of a trigger event (or policy intervention) by estimating what the outcome variable would have been in the absence of a trigger event (or policy intervention). Therefore, relative to simple event studies that put the same weight across samples, the SCM provides a safeguard against extrapolation and uses weighted average of the selected control groups.

This paper focuses on investigating any meaningful difference in economic activity, measured by real GDP per capita, between the identified events and their counterfactuals. The SCM requires relatively long time series for its application, and this considerably limits the range of indicators to which the SCM can be applied in this study. Therefore, the paper focuses on the level of GDP per capita, as a proxy of overall economic activity.³² The estimated synthetic projection describes a counterfactual that could have materialized if a country had not entered or exited from fragility. A control group consists of the counterfactuals identified in Section IV and those derived from a probability model that follows Gelbard et al. (2015). Therefore, a control group for entry into fragility includes countries that were likely to enter but did not enter fragility. Likewise, for a control group for exiting fragility, we selected cases that were likely to exit but did not exit from fragility.

The synthetic projections suggest significant costs of entering into and staying in fragility. The left-hand chart in Figure 4 plots median and quartile band of the “gap” between actual real GDP per capita and that derived from the SCM (i.e., the control group) at 3 and 7 years after entering into fragility.³³ We further divide the group into two subgroups: staying in fragility and early exit groups. The left-hand chart shows that early exit cases tend to have better real GDP per capita gap than cases that stayed in fragility and indicates two interesting observations.

• Cost of staying in fragility. Our analysis shows that the median gap of per capita GDP between countries staying in fragility and the control group (non-fragile cases)—interpreted as cost of staying in fragility―is about -12 (-24) percentage points after 3 (7) years, with a quartile band [-4, -23] percentage points ([-11, -36] percentage points after 7 years).³⁴ This is in line with the finding that the economic costs of fragility have been capitalized to much lower average growth rates over time (World Bank 2018; Mueller and Tobias 2016). Also, the estimated magnitude is comparable to existing literature on macroeconomic cost of fragility and conflicts (e.g., Novta & Pugacheva, 2020; Mueller, 2013; Ncube et al., 2014).³⁵

• Benefit of early exit. The median deviation for countries that remained in fragility 3 years after entering stands at about 8 percentage points of lower GDP per capita (i.e., the difference between “Early exit” and “Stay FS” at T+3 in the left hand chart in Figure 4) compared to those that exited early. The difference increases to about 12 percentage points 7 years after entering fragility. The difference is statistically significant at 10 percent level for year T+3 and 5 percent for T+7 by the Mood median test, suggesting the substantial benefit of early exit.

GDP Per Capita Developments for Identified Events

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

GDP Per Capita Developments for Identified Events

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

GDP Per Capita Developments for Identified Events

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

• Download Figure
• Download figure as PowerPoint slide

For the cases of sustained exit, “pivotal moments” appear closely associated with stronger economic activity relative to the control group, supported by improved policy and institutions.³⁶

• In the right hand chart in Figure 4, real GDP per capita in countries with pivotal moments around the exit rises strongly for the sustained exit cases, more strongly than for those without pivotal moments. The average median deviation for sustained exit countries with pivotal moments is about 9 (11) percentage points on average during the first 4 (7) years relative to the countries without pivotal moments.³⁷ The differences are statistically significant (at 5 percent level).

• Interestingly, the sustained exit cases with pivotal moments have systematically higher CPIA scoreSi as a possible proxy for the quality of a country’s present policy and institutional framework, than those without the moments (at 10 percent significant level by the median test) after the exit, while the cases with pivotal moments have systematically lower CPIA scores than cases without the moments before the exit (at 5 percent level).³⁸ This observation seems consistent with the discussion in Collier (2019) and IGC (2018): some countries have grabbed the chance of “pivotal moments” to implement critical reforms that facilitated successful exit from fragility with sustainable growth. During “pivotal moments,” policy makers have an opportunity to change the expectations of their citizens for the better, which would strengthen the traction of government policies as well as societal support to the state’s institutions, which in turn strengthens their effectiveness.

In sum, the above analysis indicates a possible nexus of growth/resilience of an economy and quality of institution/effectiveness of government, especially the cases of sustained exit from fragility. This issue will be further investigated in the next subsection by employing a logit model.

C. Logit Model

The third empirical approach applies a logit model to ascertain the importance of the various factors influencing turning points of state fragility. In line with existing literature of empirical studies on the determinants of fragility, using a logit model is motivated by the fact that entry or exit from fragility is a binary variable. Also, the logit model is by its construction non-linear—the first derivative of a specific regressor (i.e., its coefficient in a linear regression) depends on other regressors, suggesting that the logit model already implicitly captures any interaction between regressors. However, different from typical panel regressions on the determinants of fragility, this paper, focusing on turning points of fragility, does not use panel data for a large number of countries. Instead, our sample includes identified events and their counterfactuals without a time dimension. Therefore, the analysis in this subsection aims to illuminate the influence of factors that affect turning points at the moment of entering/exiting from fragility.

We investigate both fast-moving factors (e.g. GDP per capita growth, government expenditure per capita, etc.) and slow moving factors (e.g., executive constraints, government effectiveness, etc.) as possible factors contributing to the turning points.³⁹ Let us assume that:

where y_i is equal to 1 if country i in a given year is entering or existing from fragility and 0 otherwise. The conditional probability of entering or exiting fragility given the influencing factors is Pr(y_i = 1 |x_i, β), with β representing the marginal effect of regressors on the conditional probability. Under the assumption that ξ_i is an i.i.d. logistic distribution, the conditional probability can be written as:

Applying the maximum likelihood estimator (MLE) gives the estimate of the marginal effect of regressors on the conditional probability. Equation (2) will be separately estimated for the two samples: the one including entry events and their counterfactuals; and the other including the exit events and their counterfactuals. In order to reduce endogeneity, we use lagged values for explanatory variables.

The very persistent nature of fragility poses big challenges to get a sufficiently large number of observations to estimate the logit model. Combining the identified entry events with their counterfactuals gives 85 observations in the “entry” logit model and 69 observations in the “exit” logit model. This relatively small number of observations considerably limits the number of explanatory variables that could be included in the right hand side of equation (1). In particular, van der Ploeg et al. (2014) discusses that at least 20 observations per predictor are required to achieve good stability and high expected performance, and the estimated marginal effects are robust to sample size as long as the size meets this rule of thumb (Bergtold et al., 2017). As a result, we limit the number of explanatory variables in equation (1) to 3 or 4.⁴⁰

The most robust determinants of entry/exit event are real GDP per capita growth and government effectiveness.^{41 42} Given the constraints on the number of regressors on the right hand side, we tried a number of combinations of indicators in equation (1) (see Appendix III). Since an intuitive interpretation of β is not straightforward in the logit model, the paper shows the effects of the two most robust explanatory variables on the implied probability, instead of discussing the value of the estimated parameters. Figure 5 displays the implied probability of entry/exit, depending on the two most significant variables―government effectiveness and real GDP per capita growth, evaluated at different values of the explanatory indicator specified by the horizontal axis and at the mean for other regressors. In the left chart in Figure 5, with the values of indicators other than real GDP per capita growth set at their means, higher growth would reduce the implied probability of entering into fragility: from nearly 100 percent probability when the growth is -5 percent to about 25 percent when the growth is 5 percent. The shadowed area shows the confidence interval at 95 percent level.

The Determinants of Entry/Exit from Fragility ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Shadow areas indicate 95 percent confidence interval.^2/ Government effectiveness is an index that has mean zero and one standard deviation (for details, see https://info.worldbank.org/governance/wgi/Home/Documents).Source: Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

The Determinants of Entry/Exit from Fragility ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Shadow areas indicate 95 percent confidence interval.^2/ Government effectiveness is an index that has mean zero and one standard deviation (for details, see https://info.worldbank.org/governance/wgi/Home/Documents).Source: Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

The Determinants of Entry/Exit from Fragility ^{1/ 2/}

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

^1/ Shadow areas indicate 95 percent confidence interval.^2/ Government effectiveness is an index that has mean zero and one standard deviation (for details, see https://info.worldbank.org/governance/wgi/Home/Documents).Source: Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

• Growth of real GDP per capita and government effectiveness affect the probability of entry or exit points. Higher (lower) real GDP per capita growth, a proxy for “fast moving” fragility-inducing factors, is likely to reduce (raise) the likelihood of entries and raise (reduce) the likelihood of exits from fragility. Likewise, stronger (weaker) government effectiveness, a proxy for “slow moving” factors, is likely to reduce (raise) the likelihood of entry and raise (reduce) the likelihood of exit from fragility.

• There is an asymmetry in the implied impact of the two indicators on the conditional probability of entry or exit. Quantitatively, real GDP per capita growth seems to matter more for entry than for exit as reflected in the steeper slope of the implied probability of “entry” than that of “exit.” In other words, the implied probability of entering fragility would change more than that of exiting from fragility for the same change in real GDP per capita growth (e.g., the implied probability of entry declines by 75 percentage points by raising growth from -5 percent to 5 percent, while that would increase the probability of exiting from fragility by only 20 percentage points). For government effectiveness, the effect on the probabilities of entry or exit is more symmetric, but likely to be non-linear, or depending on initial conditions, as implied by a “hot” zone where the slope of the schedule of the implied probability is pretty steep relative to that in other areas (e.g., somewhere like domain [-1.5, 1]). This suggests more benefits of improving government effectiveness when countries have already obtained some level of government effectiveness.

There is another interesting asymmetry in the impact of growth and government effectiveness on the conditional probability of entry or exit. To investigate this, let’s plot a change in the conditional probability schedule when growth declines or government effectiveness improves, while keeping other explanatory variables unchanged at their means. The span of the variation (2 percentage points for growth and 0.25 points for government effectiveness) is selected because it is close to the change from the median to the 25^th percentile for growth and the median to the 75^th percentile for government effectiveness in the samples used for the logit model. The vertical axis shows a change in the implied probability (0.25 indicates 25 percentage points) when government effectiveness (per capital growth) changes, given a specific value of per capita growth (government effectiveness).

• A decline in growth has asymmetric size of impact on the implied probability of entry or exit from fragility. Lowering growth raises (reduces) the probability of entering (exiting from) fragility, given the level of government effectiveness (the left hand chart in Figure 6). But the impact of growth is much larger for “entry” cases than “exit case.” Moreover, there is a “hot zone” for a change in the conditional probability of entering fragility, within some range of government effectiveness [-1, 0.5].⁴³ This may suggest the importance of maintaining growth (or maintaining short-term stability and avoiding an output collapse) to avoid falling into fragility. This benefit seems particularly large for countries in the mid-range of government effectiveness, or development of institutions.

• On the other hand, improving government effectiveness has more uniform effect on the implied probability than growth, with more powerful effect on the probability of exit. Improving government effectiveness will shift the probability conditional on growth in a favorable direction—a lower probability of entry and a higher probability of exit (the right chart in Figure 6). However, different from the impact of growth, its effect seems more uniform (i.e., with gentler slope of the “bell” curve, or no “hot zone”) and matters more for “exit” cases.

Nexus of Growth and Government Effectiveness

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

Source: Worldwide Governance Indicators (WGI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Nexus of Growth and Government Effectiveness

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

Source: Worldwide Governance Indicators (WGI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Nexus of Growth and Government Effectiveness

Citation: IMF Working Papers 2021, 133; 10.5089/9781513573687.001.A001

Source: Worldwide Governance Indicators (WGI) and Fund staff estimates.

• Download Figure
• Download figure as PowerPoint slide

The above observations would suggest differential effects of growth (a proxy for fast-moving factors) and government effectiveness (a proxy for slow-moving factors) on the likelihood of entering or exiting fragility. Avoiding a growth collapse should help a country both avoid turning into and exiting from fragility. But it helps more a country at the brink of falling into fragility, especially a country with some level/quality of institution. On the other hand, improving government effectiveness would raise the probability of exit and make it more robust to shocks to growth, but not so much help avoid falling into fragility, especially when adverse growth shocks cause an economic contraction. In sum, there are two implications. The first is that preventing a growth collapse by timely and appropriate counter-cyclical policies is critical to prevent a country from falling into fragility, while strengthening government effectiveness/institutions seems the main factor for a successful exit from fragility. The second is, however, that the effectiveness of maintaining growth to avoid entering into fragility depends on the initial level/quality/effectiveness of government institutions. For a country with very weak institutions (less than -1), the benefits of avoiding a collapse of growth diminish substantially— one could even imagine a situation where aggressive macroeconomic policies to avoid a collapse of economic activity amidst weak government effectiveness result in macroeconomic and financial instability.