IMF-Supported Programs in Low Income Countries: Economic Impact over the Short and Longer Term1

This paper studies the short and longer-term impact of IMF engagement in Low-Income Countries (LICs) over nearly three decades. In contrast to earlier studies, we focus on a sample composed exclusively of LICs and disentangle the different effects of IMF longer-term engagement and short-term financing using a propensity score matching approach to control for selection bias. Our results indicate that longer-term IMF support (at least five years of program engagement per decade) helped LICs sustain economic growth and boost resilience by building fiscal buffers. Interestingly, the size of IMF financing has no significant impact on economic growth, possibly pointing to the prominent role of IMF policy advice and institutional capacity building in the context of longer-term engagement. We also present evidence that the short-term IMF engagement through augmentations of existing programs or short-term and emergency facilities is positively associated with a wide range of macroeconomic outcomes. Notably, the IMF financial support has the greatest impact on short-term growth when LICs are faced with substantial macroeconomic imbalances or exogenous shocks.

Abstract

This paper studies the short and longer-term impact of IMF engagement in Low-Income Countries (LICs) over nearly three decades. In contrast to earlier studies, we focus on a sample composed exclusively of LICs and disentangle the different effects of IMF longer-term engagement and short-term financing using a propensity score matching approach to control for selection bias. Our results indicate that longer-term IMF support (at least five years of program engagement per decade) helped LICs sustain economic growth and boost resilience by building fiscal buffers. Interestingly, the size of IMF financing has no significant impact on economic growth, possibly pointing to the prominent role of IMF policy advice and institutional capacity building in the context of longer-term engagement. We also present evidence that the short-term IMF engagement through augmentations of existing programs or short-term and emergency facilities is positively associated with a wide range of macroeconomic outcomes. Notably, the IMF financial support has the greatest impact on short-term growth when LICs are faced with substantial macroeconomic imbalances or exogenous shocks.

I. Overview

The last 25 years have witnessed a profound transformation in the economic fortunes of low-income countries (LICs). A marked improvement in macroeconomic policies has resulted in improved fiscal performance, stronger external positions, and, most importantly, significant long-term increases in real GDP per capita growth together with reductions in poverty.

This paper assesses how the involvement of LICs in IMF-supported programs may have affected these economic developments in LICs over the past quarter century. 2 During this period, the IMF has engaged in financial or non-financial arrangements with more than half of all LICs, and more than three-quarters of all IMF-supported programs have been with LIC members.

Disentangling the specific impact of an IMF-supported program from the broader economic and development trends in LICs is no easy task. The vast academic literature on this subject—which has typically focused on a mixed sample of LICs and middle-income economies—has found both positive and negative effects of IMF-supported programs on economic performance, depending on the econometric methodology and sample used.

The fundamental methodological challenge in assessing the impact of IMF-supportedprograms is selection bias: countries that approach the IMF often do so because they are already facing economic difficulties or expect to experience problems in the near future. Thus a simple comparison of performance of IMF-supported program countries with non-program countries can be misleading.

Another difficulty in evaluating the impact of IMF support is the vast differences in country characteristics and circumstances. Our study starts with the premise that mixing LICs and middle-income economies, as most studies do (see Section II), would tend to overlook the unique characteristics of LICs as well as the distinct nature and objectives of IMF engagement in these countries, such as:

  • (i) Nature of shocks: While emerging market countries may experience “sudden stop”types of capital account crises, LICs are more vulnerable to other domestic and external shocks (e.g., terms of trade shocks, demand shocks, natural disasters, domestic or regional instability, etc.) that tend to occur frequently and reflect these countries’ lack of economic development and diversification.

  • (ii) Access to financing: LICs have much less access to domestic or external financing than emerging market countries, making them more dependent on donor assistance and periodically on IMF-supported programs that can help catalyze such assistance.

  • (iii) Longer-term challenges: IMF-supported programs with LICs tend to focus heavily on medium- and longer-term objectives that are important for poverty reduction and growth, and which tend to extend well beyond the duration of an individual program. In this context, these programs emphasize more capacity and institution building rather than just provision of financing and short-term policy advice.

  • (iv) As a result of these factors, IMF-supported program engagement with most LICs has been less episodic than with other countries, and more continuous in nature. Consequently, analyzing the impact of IMF support by looking at snapshots of performance right before and after an individual program, as most studies do, tends to ignore the repeated nature of IMF engagement with most LICs and does not measure progress toward the longer-term objectives that are pursued under these programs (Figure 1).

Figure 1.
Figure 1.

Years Spent by LICs Under IMF-Supported Programs, 1986–2010

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF staff calculations.Notes: Program years are defined as years when a country had a SAF, ESAF, PRGF, ECF or PSI for at least six months. The sample is composed of 75 out of the 78 LICs as of January 2010. Somalia, Timor-Leste, and Tonga had no available data.

A related limitation of the existing literature is that it has usually not differentiated IMF-supported programs by types of instruments. To tailor its support to member countries, the IMF offers a diverse range of instruments—medium-term support, episodic short term and emergency financing, precautionary financing, and non-financial policy support. Economic objectives tend to differ under these instruments. In particular, medium-term instruments place greater emphasis on addressing entrenched imbalances and institutional weaknesses, while short-term instruments are more focused on financing and adjustment to shocks. These differences can have important implications for the examination of the impact of IMF engagement on macroeconomic outcomes.

This study breaks down IMF-supported programs with LICs into two subsets: those aiming to provide more prolonged support, and those aiming to provide short-term financing in response to shocks. We also introduce a number of other methodological refinements to the existing literature, including taking into account the implementation of IMF programs and examining a wider range of macroeconomic and social outcomes using Propensity Score Matching (PSM) approach to correct for selection bias.

Using these techniques, our evidence suggests that longer-term IMF program support may indeed have helped LICs sustain economic growth and boost resilience by building fiscal buffers (Section III). Our findings indicate that while the majority of LICs improved their longer-term macroeconomic performance, this tendency was more pronounced for those with longer-term IMF support (at least five years of program engagement per decade). Specifically, controlling for selection bias, the study shows that between 1986 and 2010, LICs with IMF-supported programs experienced, on average, significantly higher real per capita GDP growth, fiscal balances, foreign investment, and social spending compared to LICs without such programs. At the same time, countries with longer-term IMF engagement tended to attain significant reductions in poverty, income inequality, inflation, and growth volatility relative to their control group.

A further noteworthy finding is that, controlling for the presence of longer-term IMF engagement, the scale of IMF financing does not appear to be significant in determining economic growth over long time frames. This finding may point to the role of the IMF as policy advisor and an implicit institutional capacity building aspect of IMF-supported programs, which may dominate its lending role when the focus is on long-term growth.

This is not to suggest that concessional financial support from the IMF necessarily takes a back seat to policy support or other forms of assistance. In Section IV, we present evidence that suggest that IMF financial support has the greatest impact when LICs are faced with substantial short-term macro-economic imbalances or exogenous shocks. Notably, after controlling for selection bias, stepped-up IMF financing through augmentations of existing programs or short-term and emergency facilities is positively associated with short-term growth and indicators of macroeconomic stability.

II. Overview of the Empirical Literature and Contribution of the Current Study

Existing literature and methodological challenges

While there is a large literature on the macroeconomic consequences of IMF-supported programs, country samples have varied significantly across studies, and very few papers have focused exclusively on LICs. However, regardless of the country sample, the main challenge of these studies has always been the treatment of the endogeneity/self-selection bias related to the participation in IMF-supported programs, and the identification of relevant macroeconomic outcomes of such programs. Selection bias arises from systematically different initial macroeconomic and structural conditions for program versus a non-program countries. Countries that approach the IMF often do so because they are already facing economic difficulties. Structural vulnerabilities such as commodity dependence or poor governance may also lead to longer-term use of Fund facilities, and may result in increased exposure to shocks and a decreased ability to implement appropriate macroeconomic policies in the face of these shocks.

If econometric estimations of the impact of IMF programs ignore these systematic differences between program and non-program countries, the estimated effect of IMF engagement on growth and other macroeconomic indicators would likely be biased. Because most of the variables that determine the participation into an IMF program are also likely to have an independent impact on macroeconomic outcomes, it is particularly difficult to find exogenous sources of variation for IMF programs that can serve as valid instruments to address the selection bias.

The self-selection bias (explained by the fact that countries actually “request” an IMF arrangement when they need it the most) remains a concern when studying the impact of longer-term IMF engagement on various outcomes. Amongst the various factors that may explain the status of a country as a “longer term user of IMF facilities,” one critical role is played by structural vulnerabilities (which increase the exposure to and the severity of shocks when those happen and prevent sustainable macro-economic policies from being put in place). These factors include structural exposure to global shocks, weak external buffers, governance and lack of democracy, and lack of domestic resources. In addition to these factors, one can also consider the role of the political proximity vis-à-vis main donors.

In order to disentangle the different factors—and, hence, isolate the specific contribution of IMF engagement—many studies begin by attempting to assess the determinants of countries’ participation in IMF-supported programs.3 While early research emphasized the economic determinants of participation in IMF programs, the low predictive power of these models led researchers to include in participation equations political variables that would affect the “supply” side of programs.4 Evidence on the significance of these factors is again mixed. Although some individual political factors appeared to be significant, they did not significantly improve the models’ predictive power (Bird and Rowlands, 2001). Moser and Sturm (2011) distinguished nonconcessional and concessional loans and found that although several economic and political variables robustly explain the approval of nonconcessional loans only three variables (international reserves to imports, past IMF programs and an election in the past year) pass the robustness test for concessional loans.5

Steinwand and Stone (2008, p.129) conclude that “the variety of models used to explain participation in IMF programs and the plethora of contradictory results they produce indicates that existing models are far from definitive. This unfinished business is the strongest reason to urge caution in rushing to judgment about the effects of IMF lending.” In light of the little consensus in the literature on which variables really matter for participation in IMF-supported programs Bird (2007) concluded that ‘the empirical evidence so far may imply that important determining variables may still have been omitted… or that there is no one overall explanation of IMF arrangements. Rather certain things are important in some cases but not in others, such that their significance effectively cancels out in large sample studies’. Therefore, he suggests that future research could look into subsets of country cases, distinguishing the traditional current account crisis, capital account crisis, and LICs.

Recently just a handful of studies have disaggregated the analysis of participation in IMF arrangements by country income groups. Ghosh and others (2005) and Cerutti (2007) examine IMF engagement with Middle Income Countries (MICs) and emerging market economies and find several economic variables significant. Bird and Rowlands (2009) report significant differences between their regression specifications for LICs and MICs; however, the results for LIC specification are weaker than for MICs.6 Bal Gündüz (2009) focuses on a specific subset of IMF arrangements with LICs addressing policy and exogenous shocks and reports various economic variables as being statistically significant. The latter study is particularly relevant for our research as the selection equation in assessing the short-term impact of concessional programs draws on this empirical model.

While the literature has primarily focused on explaining the annual participation in IMF-supported programs in any given year, a few studies look into the factors behind the prolonged use of Fund resources.7 This line of research is closely related to the selection model we estimate to assess the impact of longer-term IMF engagement. Overall, this limited literature suggests that repeated use is peculiar to LICs and is explained by both economic and structural variables.

Turning to the impact of IMF engagement, the large empirical literature has reached some consensus that IMF-supported programs are associated with significant improvement in the balance of payments and have some effect on inflation; however, results are mixed regarding the impact on growth (Table 1).8 A few observations are noteworthy about this literature:

  1. Most of the previous research examines only non-concessional programs (Stand-By Arrangements (SBAs) and Extended Fund Facilities (EFFs)) on a mixed sample of countries. Few studies have focused on just LICs or concessional programs, and they have identified some positive effects of IMF engagement on macroeconomic performance, but not in all areas.9 A few studies have examined the social impact of IMF programs with mixed results.10

  2. Only a few studies explore the impact of prolonged engagement on longer-term growth.11 Independent Evaluation Office (IEO, 2002) concluded that IMF lending appears to have negative effects on growth for some prolonged users, though not for those under concessional arrangements. Drawing on a mixed sample of LICs and MICs, IMF (2006) noted that macroeconomic problems were reduced in many countries with longer-term program engagement, while structural problems often persisted.

  3. Although the literature widely acknowledges that whether IMF-supported programs are fully implemented or not is a key issue in properly assessing their impact, most studies do not take into account compliance with programs.12

  4. Correcting for selection bias has become a standard component of the analysis only more recently, while most studies having applied either the Heckman two-stage methodology or instrumental variable (IV) regressions (notable exceptions implementing the PSM are Garuda (2000), Hardoy (2003), Hutchison (2004), Atoyan and Conway (2006), and IMF (2012d)).13 However, the key challenge associated with these approaches was to identify exclusion restrictions, i.e., finding variables strongly correlated with the likelihood of having an IMF-supported program but not correlated with the macro-economic outcome of interest.

  5. The more recent literature builds on techniques borrowed from the microeconometric impact evaluation literature to correct for selection bias. Under this approach, each IMF-supported program country observation is matched to a counterfactual nonprogram-country observation with a similar predicted probability of having a program, and their macroeconomic outcomes are then compared. Using this technique, Atoyan and Conway (2006) found little statistical support that IMF programs contemporaneously improve real economic growth in participating countries (though this was the case for the fiscal and current account balances), but found stronger evidence of an improvement in economic growth in years following a program.

Table 1.

Summary of Literature on the Impact of IMF Programs, 2000–12

article image
Source: Draws on Steinwand and Stone (2008), expanded by authors to include selected key aspects of previous studies as well as recent literature.Note: Heckman = Heckman two-step estimator for correcting selection bias; IV = Instrumental variable estimator; PSM = Propensity Score Matching; DID= Difference-in-difference; EMs = Emerging Markets; EFF = Extended Fund Facility; ESAF = Enhanced Structural Adjustment Facility; LICs = Low-Income Countries; MICs = Middle Income Countries; SAF = Structural Adjustment Facillty; SBA = Stand-By Arrangement. +* Significantly positive -* Significcntly negative; + Positive but insignificant; - Negative but insignificant; 0 Very close to zero.

Countries with low propensity scores show improvement, while for those with high propensity scores inequality deteriorates.

Significant only for SAF/ESAF, positive but insignificant in mixed sample.

This study applies Heckman correction to growth equation, however, the inverse mills ratio (IMR) turns insignificant. The author notes that his participation equation is not strong. Therefore, it is difficult to know whether the insignificance of the IMR is because a stable participation equation is not identified or participation is random.

Finds significant negative effect for prolonged users only, while the impact on growth is insignificant for temporary users.

Results from IV regressions are very close to zero.

Easterly (2005) notes that his instruments are weak.

They report a significantly positive impact from the interaction of IMF and World Bank programs.

Based on descriptive comparison vis-à-vis the control group constructed by the PSM, therefore, significance level is not reported.

Positive effect is reported for years following the initiation of programs.

The findings are reversed for the period 2000–09 with IMF programs leading to lower poverty and lower inequality.

Contribution of the current paper

This paper contributes to the existing literature on the impact of IMF-supported programs in at least four ways:

  • Focus on LICs. By focusing entirely on LICs, the paper contributes to a nascent literature that distinguishes the impact of IMF–supported programs by country-income groups, explicitly recognizing that LICs’ unique characteristics (discussed in Section I) set them apart from other countries and warrant a distinct approach to assessing program impact.

  • Longer and short-term support. The paper studies two homogenous and complementary subsets of IMF-supported programs with LICs (longer-term prolonged support versus short-term episodic support) not examined by the earlier research. This level of disaggregation significantly improves the identification of economic and structural factors in participation models, which is the key step to correct for selection bias, while also making it possible to distinguish between short-run effects and effects of prolonged use of IMF facilities for LICs;

  • Matching techniques. Based on a sample covering nearly three decades and ending in 2010, the current paper examines a wide range of macroeconomic and social outcomes using the PSM, previously implemented only by a handful of studies for a few outcome indicators and for a mixed sample of countries, to correct for selection bias.

  • Transmission channels. The paper investigates a number of potential channels of transmission through which longer-term IMF engagement can affect long-term growth and distinguishes between effects due to IMF financing versus the role of its policy advice and capacity development.

III. Impact of Longer-Term IMF Engagement in LICs

As noted in Section I, macroeconomic conditions have improved substantially over the last two decades for most LICs, regardless of whether they were engaged with the IMF.14 On average, LICs experienced long-term increases in real GDP per capita growth, government balances, reserves, current account balances, foreign direct investment (FDI), exports, institutional quality, and social spending while also achieving noticeable reductions in economic volatility, inflation, external debt, as well as poverty (see Figures 2 and 3). This finding holds across country sizes (small versus non-small economies), geographical groupings (coastal versus landlocked), institutional capacity (as measured by the World Bank’s CPIA), and per-capita income (see Figure 4).

Figure 2.
Figure 2.

Changes in Average Decadal GDP Per Capita Growth and Poverty Gaps, 1986–2010

(In percent)

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF staff calculations using World Bank data.Note: The sample is composed of 75 low-income countries and four overlapping decadal period averages:1986–95; 1991–2000; 1996–2005; and 2001–10. A country is considered to have longer-term (LT) engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years. The figure shows the distribution of decadal changes across countries by quartiles. Poverty gap is defined as the mean shortfall from the poverty line (counting the nonpoor as having zero shortfall), expressed as a percentage of the poverty line. A more negative change in the figure implies a bigger reduction in the poverty gap.
Figure 3.
Figure 3.
Figure 3.

Macroeconomic Conditions in LICs Across Decades

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF stafff calculations.Notes: The sample is composed of 75 low-income countries (LICs.) Each value represents an unweighted average (except inflation which shows the median) over each decade. Longer-term engagement is defined as 10 or more years of having an IMF financial arrangement or Policy Support Instrument in place during 1991–2010, for at least six months in each of these years. CPIA = Country Policy and Institutional Assessment; FDI = foreign direct investment.
Figure 4.
Figure 4.

Macroeconomic Conditions in LICs Across Decades and County Groupings

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF stafff calculations.Note: The sample is composed of 75 low-income countries (LICs). Each value represents an unweighted average over each decade. Longer-term engagement is defined as 10 or more years of having an IMF financial arrangement or Policy Support Instrument in place during 1991–2010, for at least six months in each of these years.

LICs with longer-term IMF program engagement have experienced, on average, a comparatively stronger improvement in longer-term economic performance. Looking at the past three decades, countries with extensive program engagement faced comparatively weaker initial economic conditions in the 1980s, and experienced on average larger increases in real GDP per capita growth, government balance, exports, FDI and social spending than countries without such extensive engagement.15 LICs with longer-term program engagement also achieved a more marked reduction in economic volatility, inflation, and external debt. This stylized fact was first reported in IMF (2009) and continues to hold after updating the data to include the most recent years covering the global financial crisis. This strong economic improvement of extensive program users has largely eliminated the performance gap that existed relative to other LICs around the time the Enhanced Structural Adjustment Facility (ESAF) was created in 1987. Figure 5 shows a similar result when looking at the change in decadal averages of economic indicators and splitting the country sample into LICs with longer-term engagement (at least five years within the second decade) and those without such engagement.

Figure 5.
Figure 5.
Figure 5.
Figure 5.

Changes in Macroeconomic Performance of LICs

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF staff calculations.Notes: The sample is composed of 75 low-income countries and four overlapping decadal period averages:1986–95; 1991–00; 1996–05; and 2001–10. A country is considered to have longer-term (LT) engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years. The figure shows the distribution of decadal changes across countries by quartiles. CPIA = Country Policy and Institutional Assessment; FDI = Foreign Direct Investment.

Econometric analysis

The analysis that follows investigates to what extent the positive association of longer-term IMF engagement and economic performance discussed above and presented in Figures 2-5 holds up when controlling for other factors and addressing the sample selection bias. The questions addressed in this section are the following:

  1. How does longer-term IMF program engagement affect macroeconomic performance, including growth and institutional variables? The approach used is the PSM, an econometric matching technique, which is a two-stage process where (i) a first-stage regression estimates the propensity score (probability) of a country becoming a longer-term user of IMF-supported programs and (ii) then the average economic performance of countries over a 10-year period is compared between longer-term program users and others with similar propensity scores.

  2. What is the longer-term impact of IMF engagement on economic growth, and what are the associated transmission channels? We run panel regressions based on 10-year period averages that control for the traditional determinants of long-run growth commonly studied in the economic literature as well as a dummy identifying longer-term IMF engagement. The goal of the panel growth regressions is to identify the channels through which the IMF support impacts longer-term growth performance—namely macroeconomic stabilization, institutional development, and provision of development financing.

The analysis uses a panel dataset of 75 LICs and decadal averages spanning the period 1986–2010. Given the focus on longer-term engagement we work with decadal averages where periods share a 50 percent overlap with each other.16 We also worked with yearly rolling decadal averages but considered them suboptimal given the stronger serial correlation generated by the repetition of the bulk of the observations. For any given 10-year period, longer-term IMF engagement is captured by a dummy variable that takes the value of 1 if a country has had five or more years of IMF-supported programs in the 10-year period and zero otherwise. The qualifying programs are all IMF financial arrangements available to LICs, primarily the Extended Credit Facility (ECF) and its predecessors (Poverty Reduction and Growth Facility (PRGF), ESAF, and Structural Adjustment Facility (SAF)) but also the SBA, Exogenous Shocks Facility-High Access Component (ESF-HAC), and Standby Credit Facility (SCF), as well as the Policy Support Instrument (PSI). Program years have been purged of episodes when there were prolonged program interruptions to address a shortcoming of the existing literature in failing to take into account program implementation when investigating the impact of IMF engagement.17

The Propensity Score Matching (PSM) Approach To control for selection bias, a PSM selection equation is specified to estimate the determinants of longer-term IMF engagement. We specify a pooled panel probit model to avoid the well-known “incidental parameters problem” when estimating probit models with country fixed-effects.18 The independent variables are chosen broadly in line with the literature’s approach of including both demand and supply factors determining IMF support, with the aim of identifying a parsimonious set of variables that achieves a relatively good fit based on the historical data series. Longer-term IMF engagement is assumed to be determined by a country’s initial macroeconomic buffers, its structural and institutional characteristics, as well as external demand conditions during the period, but also by the role of IMF quotas in determining countries’ available financing from the Fund. Initial macroeconomic buffers are proxied by the reserve coverage ratio and the foreign aid to GDP ratio at the beginning of each decade. Structural characteristics are proxied by a dummy variable identifying landlocked countries, countries’ political connectedness, natural resource rents, and institutional characteristics, with the latter stemming from the more recent empirical focus on political and institutional influences on IMF agreements. Trading partners’ real GDP growth captures exogenous external demand. Finally, countries’ access to IMF resources is proxied by their IMF quota. See Annex I for additional discussion and estimation results.

The results of the PSM suggest that longer-term IMF engagement has been associated with improved macroeconomic and socio-institutional outcomes. Table 2 presents the PSM results for dependent variables measured in changes in order to capture relative differences between countries with and without longer-term IMF engagement in their macroeconomic outcomes. Measuring the outcomes variables as decadal changes within countries also allows for controllingcountry specific time-invariant characteristics that may be correlated with macroeconomic performance between countries. The PSM estimations are run using four different matching approaches (nearest-neighbor matching, five-nearest neighbor matching, radius matching, and Kernel matching) to test the robustness of the estimates.19

Table 2.

Impact of Longer-Term IMF Engagement on Economic Performance

article image
Source: IMF staff calculations.Note: Bootstrapped standard errors in parentheses. Each coefficient represents a separate estimation. All coefficient estimates share the same first-stage regression on the determinants of longer-term IMF engagement. Analysis is based on four 10-year period averages between 1986 and 2010 where periods overlap by 50 percent. A country is considered to have longer-term engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years. Changes in each varaible refer to first decadal differences. *10 percent significance; **5 percent significance; ***1 percent significance. CPIA = Country Policy and Institutional Assessment; FDI = Foreign Direct Investment.

The main findings are as follows:

  • Longer-term IMF engagement leads to significantly higher long-term real per capita GDP growth compared to the initial conditions. The panel growth regressions that follow attempt to identify the possible channels through which IMF engagement may affect growth performance.

  • Longer-term IMF users have significantly higher reductions in growth volatility and inflation, corroborating the role of continued Fund engagement in restoring or fostering macroeconomic stability. Improvements in the government balance are larger for longer-term users.

  • Longer-term IMF engagement is associated with significantly larger increases in FDI.

  • Changes in social spending, in particular education spending, are larger for countries with longer-term IMF engagement. They are also positive but not statistically significant for health spending.

  • The poverty gap decreased more for countries with longer-term IMF engagement. Declines in poverty rates are also larger for countries with longer-term IMF engagement, but they are not statistically signifi cant. Data availability for poverty gaps and poverty rates is limited, especially for earlier years, leading to a signifi cantly smaller regression sample and possibly less variation in the data.

  • Longer-term IMF engagement is associated with significantly greater reductions in income inequality. Like the poverty data, data on income inequality are limited for the earlier years. Yet, the regression coefficients are consistently significant across matching techniques.

  • Changes in reserve coverage, tax revenue, and CPIA are larger for countries with longer-term IMF engagement but are not significantly different from the control group (with the exception of the CPIA under one estimation). The relationship between longer-term IMF engagement and changes in aid (a test of the so-called “catalytic effect”), external debt, and the current account is not conclusive under the four estimation techniques

At first sight some of these last findings of insignificant or inconclusive results may seem surprising. For reserves, one possible explanation could be that the presence of the IMF implies that countries are often able to adjust less, since Fund presence serves to some extent as insurance when a balance of payments need arises. Therefore, countries may have less need to accumulate reserves, especially considering the high opportunity cost of doing so in countries where development needs are vast. Furthermore, oftentimes LICs requesting IMF financial support face protracted balance of payments needs and the necessity to undertake major structural reforms, so the role of Fund programs may not necessarily be to boost reserves. This is also suggested by the first-stage regression, which links long-term Fund engagement with initial levels of reserves. As for aid and debt, it is quite plausible that the effect of IMF engagement is weakened when measured as an average over the decade but may be nevertheless significant at the start of an IMF program, or at the point when debt relief is granted in the context of the program.20

What are the channels through which longer-term IMF engagement spurs these outcomes? Some of the macroeconomic variables captured above are often directly linked to IMF program conditionality. For instance, fiscal targets often aim to create fiscal space, changes in the composition of spending to favor health and education outlays or to better target spending on the poorest and most vulnerable, or capital projects with growth benefits. In addition, the IMF assists country authorities in designing consistent macroeconomic frameworks, provides regular and independent policy advice, has been a key player in debt relief initiatives, as well as an important source of capacity building through its policy support and technical assistance.

Panel growth regressions

Panel regressions are used as a complementary approach to estimate the impact of longer-term IMF engagement on growth and to identify the associated transmission channels of longer-term IMF engagement. The starting point for the growth specification follows a large strand of empirical growth literature that seeks to link economic growth performance to economic as well as institutional variables in a panel dataset context. All regressions also control for the endogeneity of longer-term Fund engagement through the inverse Mills Ratio estimated in the first stage selection equation model.21 The analysis uses the dynamic panel generalized method of moments (GMM) procedure that addresses endogeneity of the other explanatory variables and controls for unobserved country-specific factors through the presence of country fixed-effects.22 Under an initial regression specification we include as explanatory variables certain growth determinants that have received attention in the literature but exclude variables that are likely to be under the direct influence of IMF-supported programs. In subsequent regressions, we augment our specification by including explanatory variables that are likely to be influenced directly by IMF engagement as identified in the PSM analysis above, and we study the change in magnitude of the coefficients associated with the longer-term IMF engagement dummy along with the changes in their statistical significance.23 A variable will be considered as a likely transmission channel if it is significant and the coefficient associated with the IMF dummy decreases in size and/or significance relative to the benchmark model.

The panel growth regressions corroborate the PSM findings that longer-term IMF engagement appears to support higher real per capita GDP growth in LICs. They also help to identify some of the transmission channels through which this impact is achieved (Table 3).

Table 3.

Determinants of Long-Term Real Per Capita GDP Growth

article image
Source: IMF staff calculations.Note: Robust standard errors in parentheses. Analysis is based on four 10-year period averages between 1986 and 2010 where periods overlap by 50 percent. A country is considered to have longer-term engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years.

10 percent significance;

5 percent significance;

1 percent significance.

  • The regression results confirm the PSM finding above that longer-term IMF program engagement appears to have a positive impact on long-term real per capita GDP growth.

  • Based on the different specifications of the panel regressions, it appears that the decline in real per capita GDP growth volatility and inflation are likely transmission channels of the IMF longer-term impact on per capita GDP growth given the observance of a lower coefficient on the IMF engagement dummy when these factors are controlled for.

  • When controlling for both the longer-term IMF engagement and the size of net IMF disbursements in the decade, only the longer-term IMF engagement dummy is significant. This suggests that for longer-term growth performance, it is the IMF’s policy support that matters rather than the overall level of financing provided in this context.24

Robustness

We performed a series of robustness tests to confirm the validity of the findings discussed above. These tests pertained to: (i) choice of the period of analysis; (ii) alternative specification of the [paritcipationequation; (iii) role of IMF program implementation; (iv) impact of IMF program engagement in the context of donor aid. What follows is a brief discussion of each of these areas.25

We investigated whether our results would be different if alternative periods of analysis were chosen. We divided our sample into two sub-samples broadly consisting of a pre-2000s period and a post-2000s period. The results remained qualitatively similar (see Tables 7 and 8). The main differences were that the effect of longer-term IMF engagement on reserve coverage turned out to be positive and significant in the pre-2000s period, while the effect on government balance was no longer statistically significant. In addition, there were insufficient poverty data for the pre-2000s sample, and the results of the post-2000s sample were statistically weaker. We also reran our regressions with a sample of non-overlapping observations, and, although the sample size was significantly reduced, we found that the results remained qualitatively similar to those reported above.

Table 4.

Results: Impact of Short-Term IMF Engagement by Propensity Score Matching

article image
Source: IMF staff calculations.Note: PS stands for the propensity score indicating the likelihood of IMF programs addressing immediate balance of payments needs. Changes in macroeconomic outcomes refer to first differences of the variables in the top panel. The sample is composed of 58 low-income countries (LICs) and covers 1980–2010. Significant at 10 percent:*; 5 percent:**; and 1 percent:***. Standard errors in parentheses. FDI = Foreign Direct Investment; ODA = Official Development Assisstance.

All variables except for health and education spending and change in Real Effective Exchange Rate (REER) for which data is more limited.

Table 5.

Determinants of Longer-Term IMF Engagement

article image
Source: IMF staff calculations.Note: Robust standard errors in parentheses. A country is considered to have longer-term engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years.

10 percent significance;

5 percent significance;

1 percent significance.

Table 6.

Demand for IMF Financing in Response to Policy and/or External Shocks

article image
Source: Bal Gündüz (2009).Note: Demand for IMF financing in response to policy and/or exogenous shocks excluding natural disasters is estimated by a correlated random effects probit model. Significant at 10 percent:*; 5 percent:**; and 1 percent:***, t-statistics in paranthesis. Country-specific averages are calculated as the sample average of variables for each country. FDI = foreign direct investment; LR = likelihood ratio test.

The CFA franc zone consists of 14 countries in sub-Saharan Africa, each affiliated with one of two monetary unions maintaining the same currency, the CFA Franc.

Table 7.

Impact of Longer-Term IMF Engagement on Economic Performance Using Pre-2000s and Post-2000s Periods

article image
Source: IMF staff calculations.Notes: Bootstrapped standard errors in parentheses. Each coefficient represents a separate estimation. All coefficient estimates share the same first-stage regression on the determinants of longer-term Fund engagement. Analysis is based on two 10-year period averages between 1986 and 2005 where periods overlap by 50 percent. A country is considered to have longer-term (LT) engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years. Changes in each variable refer to first decadal differences.

10 percent significance;

5 percent significance;

1 percent significance.

Implies that there are no data to run the estimation.

Table 8.

Impact of Longer-Term IMF Engagement on Economic Performance Using Non-Overlapping Periods

article image

10 percent significance;

5 percent significance;

1 percent significance.

Notes: Bootstrapped standard errors in parentheses. Each coefficient represents a separate estimation. All coefficient estimates share the same first-stage regression on the determinants of longer-term Fund engagement. Analysis is based on two 10-year period averages between 1996 and 2010 where periods overlap by 50 percent. A country is considered to have longer-term (LT) engagement in a given decade if in five or more years it had a financial arrangement or a PSI in place, for at least six months in each of these years. Changes in each variable refer to first decadal differences.

We ran an alternative participation equation that controls for a lagged dummy of longer-term IMF program engagement. The dummy turns out to be significant implying that there may be other factors driving countries’ decisions to participate in IMF programs beyond the structural and macroeconomic factors that we have explicitly controlled for. Nevertheless, the second-stage regressions remain qualitatively similar, and we choose to present the results without the lagged longer-term engagement dummy given that at least for our type of analysis prior involvement does not necessarily imply continuity of engagement since the program dummy is defined as at least five years of program engagement in a decade. Although this lagged variable may capture unidentified deepseated structural problems, more work would be needed to identify potential additional variables to be included in the participation equation.

In the analysis presented above we took the view that it is not just the presence of an IMF program that matters but also the degree of implementation. For this reason, and unlike most of the existing literature, we purged our dataset of program years when the IMF program was significantly interrupted, defined as a delay of more than six months in completing a review owing to noncompliance with macroeconomic performance criteria. Table 9 presents the PSM results when there is no adjustment for program implementation. We do, indeed, find that while the results are qualitatively similar, the coefficients tend to be larger when program implementation is taken into account. In particular, this holds for the change in GDP per capita growth variable. Additional research could try to investigate this question even further by devising ways of measuring program performance (i.e., number of conditions met, timing of reviews, etc.). We interpret this finding to suggest that the size of our coefficients may be a lower bound of the true impact of longer-term IMF engagement and that a more refined measure of longer-term IMF engagement could lead to more significant results for some of the variables analyzed which were insignificant in the current analysis.

Table 9.

Impact of Longer-Term IMF Engagement on Economic Performance With No Adjustment for Program Implementation

article image

10 percent significance;

5 percent significance;

percent significance.

Notes: Bootstrapped standard errors in parentheses. Each coefficient represents a separate estimation. All coefficient estimates share the same first-stage regression on the determinants of longer-term Fund engagement. Analysis is based on four 10-year period averages between 1986 and 2010 where periods overlap by 50 percent. A country is considered to have longer-term (LT) engagement in a given decade if in five or more years it had a financial arrangement or a Policy Support Instrument in place, for at least six months in each of these years. Changes in each varaible refer to first decadal differences.

The panel growth regressions presented in Table 3 were re-estimated by including foreign aid as an additional explanatory variable. The rationale was to rule out that the positive impact attributed to IMF engagement was not in fact capturing that of other donors. The alternative specification did not change the results substantially, and aid was not statistically significant in any of the five regressions of Table 3.

Sensitivity analysis of matching estimates to unobserved heterogeneity indicate that the estimated impacts of longer-term IMF engagement are not sensitive to hidden bias. Standard matching procedures are based on the conditional independence assumption implying that selection in the treatment group (in our context, the group of countries with longer-term engagement with the IMF) is only based on observable characteristics. In our context, this means that the selection into longer- term engagement with the IMF is “entirely” explained or modeled with observable factors. Recent empirical work employing PSM techniques in micro-econometric studies checked the sensitivity of the PSM estimated results with respect to deviations from this identifying assumption. Following Clément (2011), we carry out a sensitivity analysis using the Rosenbaum bounds method (Rosenbaum, 2002). Our paper is among the first studies having explicitly tested the robustness and the relevance of the matching framework with respect to hidden bias. The purpose of this procedure is to determine if the average effect of longer-term IMF engagement may be modified or altered by unobserved variables, thus creating a hidden bias. One wants to determine how strongly unobservable factors must influence the selection process to undermine the implications of the matching analysis. Indeed, when considering Hodges-Lehmann point estimates, the results indicate that the unobserved country characteristics would have to increase the odds ratio of being in a longer-term engagement with the IMF by more than 300 percent (well above the rule-of-thumb of 100 percent suggested by Aakvik, 2001) before they would bias the estimated baseline results. These results hold regardless of the matching algorithm used in the propensity score methodology.26

IV. Impact of Short-Term IMF Engagement in LICs

This section explores the short-term macroeconomic effects of IMF financial support to LICs experiencing immediate balance of payment needs as a result of policy slippages or external shocks.27 The nature of IMF support evaluated in this section differs from the more extensive program support via successive medium-term arrangements, as discussed above. Here, we focus on short-term IMF financial support, either through augmentations of access under existing medium-term financial arrangements or through short term or emergency financing instruments. Such support would often be called for when a country faces a pressing balance of payments problem, which would require a combination of macroeconomic adjustment and external financing. The IMF engagement in these cases would typically involve understandings on short-term macroeconomic adjustment accompanied by IMF financing, which could potentially have catalytic effects inducing additional bilateral and multilateral financing.

Sample selection bias is an even greater methodological challenge when the short-term impact of IMF-supported programs is studied. If countries that are experiencing balance of payments crises owing to policy slippages or exogenous shocks are more likely to participate in IMF-supported programs, failing to correct for selection bias could lead to a flawed conclusion that programs “cause” these crises along with adverse effects on macroeconomic outcomes. As in the previous section, this study implements the PSM methodology. In the first stage, the annual probability of participating in IMF-supported programs is estimated conditional on observable economic conditions and country characteristics. The second stage uses these probabilities, or propensity scores, to match program countries to non-program countries, and thereby, construct a statistical comparison, or control, group (see Annex I for details).

Empirical analysis

The probability of participation in IMF-supported programs that address policy and/or exogenous shocks increases with the deterioration in the pre-shock macroeconomic conditions and the magnitude of the adverse external shocks. The selection model for LIC participation in IMF-supported programs draws on Bal Gündüz (2009). This study finds that a lower reserve coverage, a deterioration in the current account balance, a weaker real GDP growth, increased macroeconomic instability (evident in higher fiscal deficits, inflation and exchange market pressures), and adverse terms of trade shocks would increase the likelihood of IMF financing. Moreover, global conditions, including changes in real oil and non-oil commodity prices and world trade, are also significant determinants of participation in IMF-supported programs which could potentially create cycles in demand for IMF financing as a result of adverse global shocks. Finally, persistent differences in debt service burden and resource inflows among LICs seem to be significantly associated with unobserved country heterogeneity.

The results suggest that IMF-supported programs lead to significantly better outcomes particularly for LICs experiencing substantial prior macroeconomic imbalances and/or severe adverse external shocks. Table 4, Figure 6 and Figure 7 present the differences in various macroeconomic outcomes between program countries and the control groups.

Figure 6.
Figure 6.

The Impact of Short-Term IMF Engagement on Macroeconomic Outcomes

(By propensity scores: Fund programs minus the control group)

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF staff estimates.Notes: Estimated impact of short-term Fund engagement relative to the control group having similar propensity scores. PS stands for the propensity score indicating the likelihood of Fund programs addressing immediate balance of payments needs. The sample is composed of 58 LICs and covers 1980–2010.
Figure 7.
Figure 7.

The Impact of Short-Term IMF Engagement on Changes in Macroeconomic Outcomes

(By propensity scores: Fund programs minus the control group)

Citation: IMF Working Papers 2013, 273; 10.5089/9781484356203.001.A001

Source: IMF staff estimates.Notes: Estimated impact of short-term Fund engagement on changes in macroeconomic outcomes relative to the control group having similar propensity scores. PS stands for the propensity score indicating the likelihood of Fund programs addressing immediate balance of payments needs. Changes in macroeconomic outcomes refer to first differences of the outcome variables. The sample is composed of 58 LICs and covers 1980–2010.

While growth is estimated to be 0.9 percent higher than the control group for the full sample, the impact rises to 1¼–1¾ percent and becomes significant only for countries with high propensity scores, which indicate immediate balance of payments problems brought about by existing macroeconomic imbalances and/or external shocks. Furthermore, change in growth is positive but turns out significant only for those with high propensity scores.

Overall, program countries attain significantly higher current account balances, and reserve coverage, as well as lower inflation and fiscal deficits compared to their control groups. Moreover, reflecting the stabilization achieved under IMF-supported programs, these variables tend to post significant improvements during the program, with the impact especially pronounced for the high propensity group. Although program countries tend to have more depreciated real exchange rates, differences with the control groups are not significant. Changes in real health and education spending per capita are not statistically different from those of the control group.

The estimated positive impact on growth could be attributed to IMF financing (along with its potential catalytic effects) easing the burden of the short-term adjustment as well as restrotation of macroeconomic stability, especially for countries experiencing significant levels of instability prior to the program. Both commitments and disbursements of official developments assistance (ODA) are significantly higher for the program group. Lower differences in disbursements than those of commitments compared to the control group may suggest room for improving the utilization and predictability of ODA for program countries. However, contrary to the presumed catalytic role of IMF-supported programs, no significant change in ODA is detected. One explanation could be that some countries with high propensity scores could avoid or delay requesting IMF assistance thanks to an ad hoc increase in ODA flows, weakening the estimated catalytic impact for this group. Another explanation could be that ODA provided as budget support may be more responsive to IMF programs than the project support. Exploration of such heterogeneity in the catalytic effect of IMF-supported programs is left for further research.

Robustness checks

This section explores the robustness of results to four sensitivity analyses: (i) relaxing the adjustment made for the implementation record of programs; (ii) setting the sample to 1980–1999 to improve the comparability of results to earlier research; (iii) conditioning matching on propensity score and ODA disbursements to explore the IMF impact at similar levels of assistance; and (iv) conditioning matching on propensity score and lagged GDP growth to look into whether the positive growth impact is driven by a cyclical recovery in program countries having a very weak growth prior to the program. Results turn out to be robust to these adjustments with some noteworthy changes highlighted below.

When the IMF program dummy is not adjusted for the implementation record, a dummy variable marking interruptions of six months or longer in completing a review, the estimated IMF impact on both the level and the change in growth and other macroeconomic variables remains qualitatively similar, however, the size of the impact tends to get lower (Table 10). Adjustment for implementation is one of the key departures of this work from the previous literature as only a few studies took it explicitly into account.28 As our measure of implementation accounts for only serious interruptions in programs and does not assess the strength of programs across a more differentiated and continuous scale, the adjusted estimates may reflect the lower bound for the impact of a strongly-implemented program.

Table 10.

Impact of Short-Term IMF Engagement With No Adjustment for Implementation

article image
Source: IMF staff calculations.Note: PS stands for the propensity score indicating the likelihood of Fund programs addressing immediate balance of payments needs. Changes in macroeconomic outcomes refer to first differences of the variables in the top panel. needs. The sample is composed of 58 LICs and covers 1980–2010. Significant at 10 percent:*; 5 percent:**; and 1 percent:***. Standard errors in parentheses. FDI = Foreign Direct Investment; REER = Real Effective Exchange Rate.

All variables except for health and education spending and change in REER for which data is more limited.

In order to facilitate the comparability of our results to earlier research, the impact is estimated for 1980–1999 sample studied by most of earlier research. The results turn out to be qualitatively similar and estimated quantitative impact on growth and other macroeconomic indicators is even stronger (Table 11). Therefore, the benchmark findings are not driven by differences in the sample.

Table 11.

Impact of Short-Term IMF Engagement, 1980–99

article image
Source: IMF staff calculations.Note: PS stands for the propensity score indicating the likelihood of Fund programs addressing immediate balance of payments needs. Changes in macroeconomic outcomes refer to first differences of the variables in the top panel. The sample is composed of 58 LICs and covers 1980–1999. Significant at 10 percent:*; 5 percent:**; and 1 percent:***. Standard errors in parentheses. FDI = Foreign Direct Investment; REER = Real Effective Exchange Rate. ODA = Official Development Assistance.

All variables except for health and education spending and change in REER for which data is more limited.

Some countries with high propensity scores might avoid resquesting IMF assistance owing to high donor assistance that would ease or delay the necessary adjustment. A comparison of performances of program versus non-program countries having both similar propensity scores and ODA disbursements as percentage of GDP would be insightful to explore this issue. Results for levels of macroeconomic outcomes are qualitatively similar while the estimated impact, including on growth, gets stronger for program countries compared to the control group (Table 12). This finding may suggest that some non-program countries with high propensity scores seemed to have avoided a sharp adjustment thanks to high ODA disbursements. Differences in improvement in growth and government balances between program and non-program countries, although positive, become insignificant, indicating broadly similar outcomes for program versus non-program countries after controlling for ODA disbursements, suggesting that the catalytic effect of the short-term IMF-supported programs may be an important channel of transmission explaining their macroeconomic impact.

Table 12.

Impact of Short-Term IMF Engagement, Matching on Propensity Score and ODA Disbursements

(Percent of GDP)

article image
Source: IMF staff calculations.Note: PS stands for the propensity score indicating the likelihood of Fund programs addressing immediate balance of payments needs. Changes in macroeconomic outcomes refer to first differences of the variables in the top panel. The sample is composed of 58 LICs and covers 1980–2010. Significant at 10 percent:*; 5 percent:**; and 1 percent:***. Standard errors in parentheses. FDI = Foreign Direct Investment; REER = Real Effective Exchange Rate.

All variables except for health and education spending and change in REER for which data is more limited.