Do IMF Programs Stimulate Private Sector Investment?
Author:
Pietro Bomprezzi
Search for other papers by Pietro Bomprezzi in
Current site
Google Scholar
PubMedClose
,
Silvia Marchesi
Search for other papers by Silvia Marchesi in
Current site
Google Scholar
PubMedClose
, and
Ms. Rima A Turk
Search for other papers by Ms. Rima A Turk in
Current site
Google Scholar
PubMedClose
https://orcid.org/0000-0001-5321-6524
Publication Date:
29 Jul 2022
eISBN:
9798400216800
ISBN:
9798400216800
Language:
English
Keywords:
IMF; Firm investment; Local Projection; Financial Frictions; Difference-in-Differences; IMF arrangement; IMF working papers; firm investment response; AIPW estimate; program approval; Private investment; Financial statements; Balance of payments need; Sub-Saharan Africa; firm friction; private sector investment; financing constraint; sectoral uncertainty; investment dynamics
Download PDF (1.7 MB)

This paper provides new evidence on the role of IMF programs in stimulating private sector investments. Using detailed firm-level data on tangible fixed assets and a local projection methodology, we first estimate the dynamic response of firm investments to the approval of an IMF arrangement. We find that distinguishing between GRA and PRGT financing matters for the path of firm investment and its growth, and we also document the presence of two financial channels; the degree of firms’ external financial dependence and firms’ sectoral uncertainty. Exploiting these firm-level characteristics, we employ a difference-in-differences approach to understand the mechanisms through which the approval of an IMF arrangement propagates in the private sector. We find that the more firms rely on external finance and the more they are subject to uncertainty, the less binding these financial frictions become, and hence the more firms invest following a program approval. Finally, using ownership data, we find that private investments are stimulated more for domestic firms. The presence of a private investment transmission channel could help improve our understanding of what factors could affect the success and effectiveness of IMF programs.

Abstract

This paper provides new evidence on the role of IMF programs in stimulating private sector investments. Using detailed firm-level data on tangible fixed assets and a local projection methodology, we first estimate the dynamic response of firm investments to the approval of an IMF arrangement. We find that distinguishing between GRA and PRGT financing matters for the path of firm investment and its growth, and we also document the presence of two financial channels; the degree of firms’ external financial dependence and firms’ sectoral uncertainty. Exploiting these firm-level characteristics, we employ a difference-in-differences approach to understand the mechanisms through which the approval of an IMF arrangement propagates in the private sector. We find that the more firms rely on external finance and the more they are subject to uncertainty, the less binding these financial frictions become, and hence the more firms invest following a program approval. Finally, using ownership data, we find that private investments are stimulated more for domestic firms. The presence of a private investment transmission channel could help improve our understanding of what factors could affect the success and effectiveness of IMF programs.

Introduction

The economic headwinds over the past several years, from pandemics, to supply crises, to geopolitical tensions faced by countries has reinvigorated the role of multilateral lenders such as the International Monetary Fund (IMF) in global finance. IMF resources have been tapped over the past decade to deal with systemic debt crises in advanced economies such as in the Euro area, as well as reviving its role among developing and fragile economies.2 Independent of these examples, as Horn et al. (2020) document, official lending has historically been larger than commonly thought, and has been growing as a share of total cross-border capital flows.

Traditionally, financial support by the IMF aims to create breathing room for countries hit by crises as they implement adjustment policies to restore macroeconomic stability and growth. While policies depend on country circumstances, the set of corrective actions provide a seal of approval that appropriate policies are adopted, helping mitigate crises and boosting market confidence during periods of heightened risks. In the absence of economic stability, inefficiency or misallocation of resources lead to lower aggregate productivity and investments, which are leading explanations for economic disparities across countries (Hsieh and Klenow, 2009). In this paper we aim to show how the IMF influences private sector dynamics, and through which channels private investments are influenced by the approval of a Fund program. The goal of the IMF in spurring private sector growth as the engine of economic growth is well understood, and while there is some narrative evidence of the interplay between the IMF and private firm dynamics in the affected country, and this paper tries to quantify these effects.3

At the macro level, the effects of IMF programs on the economy have been extensively investigated focusing on two main channels. The “seal of good faith”, or signaling, argument is typically used to explain a catalytic finance effect, whereby foreign capital or creditors return to the country (Corsetti et al. 2006; Marchesi and Thomas 1999; Marchesi 2003; Morris and Shin 2006; Zwart 2007). Another strand of the literature has also looked at the liquidity effects of IMF credit injections, which can reduce the probability of self-fulfilling runs arising from illiquidity problems (Boockmann and Dreher 2003; Dreher 2006; Dreher and Vaubel 2004; Zettelmeyer 2000).

In this paper, we think of whether the signaling effect of an IMF program operates within the confines of the country itself. Consistent with acting as a “seal of good faith” to outside investors who would unlock portfolio flows to the country, an IMF program can be viewed positively by domestic investors (in our case, firms in the private sector). By strengthening policy frameworks and institutions, and adopting concrete and credible policy actions, IMF programs can therefore be viewed as triggering a policy reduction mechanism domestically for the private sector to undertake positive net present value capital investments.

An increasing amount of research focuses on the impact of uncertainty on firms (among others see Abel 1983; Bernanke 1983; Bloom et al. 2007; Gilchrist et al. 2014; Gulen and Ion 2015; Handley and Limao 2015; Huihua et al. 2020; Julio and Yook 2014; Kang et al. 2014). While few early studies contend that uncertainty may in some cases stimulate companies’ investment (Bar-Ilan and Strange, 1996; Abel, 1983), more recent research predicts that uncertainty would inhibit corporate investment (Gulen and Ion 2016 and Rao et al. 2017). This paper leverages balance sheet data to quantify how this uncertainty interacts with IMF programs to determine firm investments. Since tangible fixed assets investments tend to be non-reversible, firms favor precautionary delays in long-term decisions until policy uncertainty subsides, as may happen following the approval of an IMF adjustment program. By focusing on firm investment dynamics following the approval of IMF programs, we aim to assess the reaction of the private sector which is the engine for achieving better growth outcomes. The presence of a private investment uncertainty channel could help improve our understanding of what factors affect the success and effectiveness of IMF programs.

Using detailed firm-level data on tangible fixed investments and a local projection methodology, in this paper we estimate the dynamic response of firm investments to the approval of an IMF arrangement. We find that distinguishing between GRA and PRGT financing matters for the path of firm investment. Then we document the presence of two financial channels: the degree of firms’ external financial dependence and firms’ sectoral uncertainty. To exploit these firm-level characteristics, we employ a difference-in-differences approach to understand the mechanisms through which the approval of an IMF arrangement propagates in the private sector. We find that following an IMF program approval, for firms that rely more on external finance or are more subject to uncertainty, the less binding these financial frictions become, and hence the more firms invest following a program approval. Finally, using ownership data, we find that private investments are stimulated more for domestic firms. The presence of a private investment transmission channel could help improve our understanding of what factors could affect the success and effectiveness of IMF programs. To the best of our knowledge, this is the first paper that investigates whether IMF programs, as well as improving a country’s creditworthiness for external investors, may also make “internal” investors more willing to invest in their own country, by reducing the impact of uncertainty.

The remainder of the paper is organized as follows. Section 2 provides a review of the main literature on both the effects of IMF financing and firm investment under policy uncertainty. Section 3 presents our data, while Section 4 illustrates the methodology and identification strategy. Section 5 presents our main empirical results and Section 6 summarizes robustness tests. Section 7 concludes.

Literature Review

Effects of IMF Programs

Traditionally the literature on IMF effectiveness focuses on broad country-level outcomes. Formulating a successful macroeconomic adjustment strategy requires a thorough understanding of a country’s idiosyncratic characteristics and the environment in which it operates. The literature on the effectiveness of IMF programs is vast (Przeworkski and Vreeland 2000; Barro and Lee 2005; Easterly 2005; Dreher 2006; Marchesi and Sirtori 2011; Bas and Stone 2014). Among these studies, more recent ones have focused on the specific objectives of IMF policy conditions in pursuing macroeconomic stability. For example, some argue monetary stability, debt management, and the containment of external arrears as key goals of IMF programs (Kentikelenis, Stubbs, and King 2016). Also, IMF programs have been associated with reduced inflation and monetary growth, lower risk of currency crises and banking crises, and improved market performance of banks (Dreher and Walter 2010; Papi et al. 2015; Steinwand and Stone 2008). In sum, the evidence suggests some positive adjustment effects regarding financial, fiscal, and monetary positions, though the benefits have generally fallen short of expectations, especially in terms of GDP growth and debt reduction (IMF 2018; IEO 2021).

The success of IMF programs, however, largely hinges on its catalytic effect, namely the propensity of private capital to flow into the country following the approval of an IMF program. The signaling role of an IMF-supported adjustment program and its catalytic effects have both been extensively analyzed in the literature with mixed findings (e.g., Chapman et al. 2015; Corsetti et al. 2006; Gehring and Lang 2020; Krahnke 2020; Marchesi and Thomas 1999; Marchesi 2003; Mody and Saravia 2006; Morris and Shin 2006). While policy conditionality can reassure international investors that adequate policies are being implemented to resolve the balance of payments needs (Tirole 2002), the preferred creditor status of the IMF could make foreign investors fear penalization in case of a debt restructuring (Mody and Saravia 2003). In addition, moral hazard incentives by borrowing countries expecting a bail-out could also be a concern (Dreher 2006).

This paper belongs more generally to the growing body of literature that focuses on the effects of official intervention at the subnational level. Recent studies have re-investigated economic outcomes following official capital flows at a more disaggregated levels than using broad macroeconomic aggregates (Bluhm et al. 2020; Bomprezzi and Marchesi 2021; Chauvet and Ehrhart 2018; Dreher and Lohman 2015; Dreher et al. 2021; Marchesi et al. 2021).4 The paper which most closely relates to ours is by Bomprezzi and Marchesi (2021), who evaluate the effects of IMF programs on firm-level outcomes by considering two dimensions: participation in a Fund-supported program and scope of conditionality. They find a positive impact of IMF programs on firms’ sales growth, with the effect being persistent over time, and that corporate performance improves through the alleviation of the firm financing constraint. However, more severe conditionality seems to worsen firm performance in the short run before benefits materialize in the longer term.

More broadly, a growing literature is investigating the effects of official intervention on domestic firms. Broner et al. (2021) find that sovereign debt inflows reduce the domestic interest rate by raising the price of government debt, thereby encouraging banks to expand domestic credit, and benefitting domestic firms, especially those that are financially constrained. To the extent that the signaling role and the catalytic effect of an IMF program work to reassure financial markets, sovereign risk of the recipient country can also be reduced. A large theoretical literature explains how an increase in sovereign risk may propagate to firms (Mendoza and Yue, 2012; Hébert and Schreger 2017; Andrade and Chhaochharia 2018), including how the sovereign default fiscal channel affects firms (Corsetti et al. 2006; Gourinchas et al. 2017). In general, the international macroeconomics literature on the effects of capital flows on the real economy is vast, especially for foreign direct investment, banks, and equity portfolio flows (Broner et al. 2020; Schnabl 2012; Baskaya et al. 2017).

This paper takes a different approach. It contributes to the extensive literature that focuses on firm-level decisions by exploring the effect of IMF programs on private sector investment. It proposes a domestic signaling effect, under which economic agents (firms) responsible for undertaking non-reversible long-term investment decisions are sensitive to the current and expected macroeconomic environment. Under this hypothesis, the reduction of domestic policy uncertainty that accompanies IMF programs induces firms to increase private sector investment.

Firm Investment Under Uncertainty

The literature on determinants of investment dynamics emphasizes the role of firm and sector-specific factors such as size, profitability, asset tangibility, and industry median leverage (Myers, 1984; Myers and Majluf, 1984; Titman and Wessels, 1988; Harris and Raviv, 1991; Booth and et al. 2001; Baker and Wurgler, 2002; Lemmon et al. 2008; Frank and Goyal, 2009; Gungoraydinoglu and Öztekin, 2011; Graham et al. 2015; De Angelo and Roll, 2015; Öztekin, 2015). Another strand of literature underscores the critical role of country-specific macroeconomic and institutional factors in determining firm outcomes (Borio, 1990; Rajan and Zingales, 1998; Kayo and Kimura, 2011; Cevik and Miryugin, 2018), as well the role of political instability (Herrala and Turk-Ariss, 2016).

More closely related to our paper is the growing strand of literature that focuses on the adverse impact of uncertainty on firm investment. A common strategy is to proxy exposure to uncertainty through the volatility of returns of stock prices (Leahy and Whited 1996; Bloom et al. 2007; Baum et al. 2010; Bloom 2009; Panousi and Papanikolaou 2012). Bloom et al. (2007) present a model in which uncertainty reduces firms’ irreversible long-term investments in response to shocks to sales, arguing that firms become more cautious during times of heightened stock price volatility (interpreted as demand shocks). Using data on U.S. firms over 1970–2005, Panousi and Papanikolaou (2012) show that firm-level idiosyncratic risk (or the volatility of stock price returns that is not explained by market or industry returns) associates negatively with corporate investment. Sector-specific sensitivities of firms to the macroeconomic environment are also documented in the literature. Durante et al. (2020) find that the transmission of monetary policy depends on the degree of durability of the goods produced, as well as financial frictions faced by the firm or the sector to which it belongs, with sectors that are active in the production of durable goods reacting more than others.

Other studies distinguish between the effects of macro and micro sources of uncertainty on firm investment, as uncertainty can express itself at different levels. Temple et al. (2001) use survey responses regarding expectations of uncertainty on demand by firms in the U.K., finding that both sources of uncertainty have a negative impact on firm investment. Baum et al. (2010) distinguish between own uncertainty (based on a firm’s stock returns), market uncertainty (derived from the returns on a stock index), and a measure of covariance between the two types of uncertainty. They confirm that an increase in market uncertainty inhibits firm-level investment, also finding that the sign of the effect of the other measures of uncertainty on firm-level investment depends on interaction with the firm’s cash flow. Kang et al. (2014) also report that firm-level investment is influenced by the interplay between firm-level uncertainty (i.e., micro uncertainty) and aggregate economic policy uncertainty (i.e., macro uncertainty). They find that economic policy uncertainty depresses firms’ investment decisions, and the effect of economic policy uncertainty on firm-level investment is greater for firms with higher firm-level uncertainty (proxied by stock price volatility).

Recent methodological advances allow for improved measurements of uncertainty, though mostly at the aggregate level. Measuring overall, systemic, uncertainty is a complicated task because of the multitude of factors to consider, and proxies can suffer from a host of theoretical and empirical drawbacks. For example, Jurado et al. (2015) provide rigorous econometric estimates of aggregate uncertainty, showing that popular uncertainty proxies overestimate the number of quantitatively important uncertainty episodes. Baker et al. (2013) deviate from traditional proxies by constructing a three-part index containing news-based, future tax provisions, and economic forecast components. Gulen and Ion (2016) use this index to show that the news-based component is the most relevant in explaining the negative relationship between aggregate uncertainty and capital investments and highlighting how the magnitude of the effect varies by the degree of investment irreversibility. There is a rapidly growing literature using text-based measures of aggregate uncertainty, such as Alexopoulos and Cohen (2015), who find a negative relationship between their New York Times based uncertainty index and real and financial economic outcomes. Finally, and directly relevant for our paper, Alfaro, Bloom, and Lin (2021) construct a firm-level dataset of uncertainty measures as well as firm-level instruments to address endogeneity concerns. They provide two different proxies of uncertainty at the micro level: (i) realized stock return volatility of daily returns from the Center for Research in Security Prices (CRSP) and (ii) implied volatility, as constructed from a mix of put and call-at-the-money options. As will be shown in later sections, we employ the first of these indicators as our preferred measure of firm-level uncertainty.

This paper contributes to the empirical literature on uncertainty by incorporating these micro-level indicators in the context of international capital flows, such as IMF lending. In the next section we introduce our two main data sources, IMF program data and firm-level data from Orbis.

Data

Identifying IMF Programs

We focus on the pre COVID-19 period, drawing data on IMF programs from the IMF’s Monitoring of Fund Arrangements (MONA) database between 2002 and 2019. We consider the main lending instruments in the IMF’s toolkit, which are tailored to different types of balance of payments needs as well as the specific circumstances of the diverse membership. However, unlike previous work that we are familiar with, we focus separately on financing through the GRA and the PRGT.5 Whereas GRA financial support is available to all member countries on non-concessional terms, the IMF also provides concessional financing through the PRGT to cater to the diversity and needs of low-income countries.

We make the explicit distinction between GRA and PRGT lending facilities because the policy ramifications of financing differ between the two. Financing under the GRA requires that policy measures be taken within the program period and the macroeconomic adjustment be completed by the time repurchases (or repayment) to the IMF begin.6 Under GRA financing, the member’s balance of payment needs should be resolved by the end of the program period and no follow-up arrangement would in principle be expected. In contrast, financing under the PRGT is tied to achieving or making progress towards a stable and sustainable macroeconomic position consistent with strong and durable poverty reduction and growth, without necessarily expecting a resolution of the balance of payments problem during the program period.7 The distinction between GRA and PRGT arrangements is important because it signifies that, unlike for the GRA, repeated programs under the PRGT can be expected for sustained engagement to deliver progress towards macroeconomic stability consistent with poverty reduction and growth.

Our treatment variable of interest is an indicator that takes the value 1 if a country approved an IMF-supported program during the year as reported in MONA, but no later than October.8 In this latter case, the subsequent year is coded as the program approval year. The treatment dummies identifying program start for each financing type, GRA or PRGT are then equal to one for that country-year pair.9 This coding allows us to explicitly account for program effects that would not manifest in a certain year if the program was effectively adopted later in the year.

Our sample contains countries under an IMF arrangement. This setup helps mitigate problems of endogeneity, whereby estimates of the effects of an IMF arrangement approval on investment dynamics could be biased by selection into the sample, or they could potentially be subject to reverse causality under the assumption that arrangements have a “crowding out” effect on investments. By focusing only on a set of countries under an IMF arrangement, the research question shifts to the dynamics of investment following an arrangement approval as opposed to before entering a program. Secondly, with a sample of treated countries (i.e., countries having an IMF program), the focus can shift to the heterogeneity among arrangements.

Figure 1 plots the number of unique programs recorded per year in the MONA database for the 2 types of arrangements considered (under the GRA and the PRGT). GRA arrangements make up the bulk of programs over the full sample, while PRGT arrangements represent a smaller share, generally not surpassing 5 per year. On average, the overall number of programs per year increased in the latter half of our sample.

Figure 1.
Figure 1.

Distribution of IMF Programs per Year

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Number of unique IMF programs signed per year, by program type. Blue bars are for the PRGT category, red bars are for the GRA category.

Aside from our main variables of interest discussed in the next subsection, we consider a series of country-level controls in our analysis (see Table A1 in Appendix A for a list, description, and sources of all variables). These variables fall into three broad categories of economic, financial, and political factors. We consider both the growth rate of real GDP and the log of real GDP per capita, which capture growth opportunities for the firm. We proxy for the size of the banking sector and financial development using the log of claims by depository institutions on the private sector. The real interest rate captures both the representative lending rate offered in the economy as well as inflation risk to investments. Finally, we use the International Country Risk Guide (2021) index of political risk to control for the broad perception of investment risk within the country. We also control for the actual program years, excluding the first year which is the approval year and therefore the baseline treatment.

In the following sub-section, we introduce our measure of firm investment, as well as other firm-level data and controls. In Section 4 we put all variables together to explain our identification strategy and baseline model.

Firm Tangible Fixed Asset Investment and Balance Sheet Data

We retrieve detailed firm-level balance sheet data from the Orbis database provided by Bureau Van Dijk. To assess the influence of uncertainty on firm investment decisions, it is important to focus on tangible investments because of their non-reversible and long-term nature. Generically, tangible investment refers to investment in physical assets (e.g., property, plants, and equipment) acquired by a firm for long-term use and which have tangible value. Considering these investment types, as opposed to other more generic investment categories (financial or intangible), allows us to capture how firms react to changes or potential changes in the macroeconomic environment. We therefore take the ratio of tangible fixed assets to total assets to compute investments.

The Orbis database allows us to also use other balance sheet data as firm-level controls in our main specification. We follow the specification of Kalemli-Özcan et al. (2019) in identifying our main Orbis firm-level controls. These include a set of balance sheet variables and ratios that are standard in the corporate finance literature focusing on firm investment. First, we use the log of total assets to proxy for firm size. Leverage is measured as the ratio of total debt to total assets, where total debt is in turn the sum of all long-term debt, loans, credits, and other current liabilities. Debt maturity is proxied by the ratio of long-term debt to total debt to capture the rollover risk of firms. Companies with a longer debt maturity structure (i.e., greater share of long-term debt in total debt) are more “locked-in” in their investment paths and have lower rollover risk, meaning that they are less likely to rollover their debt in the short-term to finance new investments. To capture the drag that past debt has on current finances, we include the ratio of interest expense to earnings before taxes (EBIT). Our full specification also includes macroeconomic variables that proxy for growth opportunities, which we also capture at the firm level using sales growth. Finally, we control for cash flows scaled by total assets as is standard in the literature. Table 1 provides the full summary statistics for these and our country level variables. The Orbis database also provides information such on firm ownership, incorporation dates, geographic areas, and sectors of operation. From these we construct sector-year fixed effects to account for time-varying, sector specific heterogeneity. Table A3 in Appendix A reports the average tangible investment (computed as the change in the ratio tangible assets/ total assets) by NACE main sector for each year across all firms.

Table 1.

Summary Statistics

article image
Notes: Summary statistics run on winsorized sample.

There are several data issues with the Orbis data base that we need to handle. First and foremost, firm coverage of Orbis varies by region and by country (see Table A4 in Appendix A). Orbis collects data from a range of sources, including publicly available national company registries. The principal reason for the variation in firm coverage by country is the difference in legal standards for companies when it comes to financial documents. For countries where the filing of financial information is mandatory for all, the Orbis sample is more comprehensive (Kalemli-Özcan et al. 2015).10 By nature of funding needs, countries in the sample that have had an IMF program are for the most part middle and lower income, and highly concentrated in Africa, Latin America, the Caribbean, Eastern Europe, and Southeast or Central Asia. Orbis, just like other data service providers with this type of information, has typically more limited firm financial data in these countries compared with firms from other parts of the world, particularly with respect to Western Europe and the Americas. The data is particularly scant for Africa, where more than 90 percent of companies have no financial information in the Orbis database. Figure 2 gives a graphical representation of countries having had an IMF program (from the MONA database), showing a clear concentration in the African continent. The size of the bubbles indicates the number of unique firms for which we have detailed balance sheet data in each country that has had an IMF program. The Orbis coverage of firm-level data for African countries is provided for half of the MONA sample from Africa, whereas most of the firm representation is from MONA countries in Eastern Europe, Latin and South America, and Central Asia.

Figure 2.
Figure 2.

IMF Programs and Firms

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: The figure plots average SDR access (MONA) over sample years and number of unique firms in Orbis sample for a given country. Light blue indicates no programs between 2002 and 2020, darker color indicates greater average access, larger bubbles indicate larger panel of firms.

As with any data provider of balance sheet data, there are issues of missing or unreliable data that need to be acknowledged. As Kalemli-Özcan et al. (2015) show, utilizing a single vintage, i.e., a single download with time series data, leads to inconsistent coverage by Orbis of certain variables because different vintages may have varying coverage. Furthermore, industry classifications may be misleading because over time firms move in and out of industries through expanding operations or statistical offices change definitions. Instead, we follow the procedure outlined by the Kalemli-Özcan et al. (2015) and rely on historical Orbis data, downloading year-specific vintages and then matching firms over time with Orbis’ unique firm identifiers. This produces firm samples which are more nationally representative and mitigates the need to re-weigh the data.

We adopt some simple data cleaning to our sample and our main variables. First, we drop financial firms, government sector firms, and other firms which operate primarily in service activities.11 We also avoid double counting by considering only consolidated financial statements when available, otherwise unconsolidated. We then clean the data to remove cases of erroneously reported balance sheet items, such as negative reported costs. As per Kalemli-Ozcan et al. (2019), all balance sheet variables are winsorized or trimmed so that their kurtosis falls to a value around 10. Our final firm sample is an unbalanced panel of 43,949 firms for 69 countries from 2000 to 2019. We plot some descriptive statistics in Figure 3. We categorize firms by age according to their year of establishment as either young (less than 15 years old), mature (established between 15 and 34 years), or well-established (more than 35 years old). The plot shows the average investment growth by firm age for each year in the sample. As would be expected investment at younger firms generally grows faster than for other firms. There is also yearly variation across groups, with drops in the middle of the sample likely due to a series of global shocks like the Great Financial Crises, adverse commodity price shocks, and “taper tantrums”, all which threw developing countries into turmoil.

Figure 3.
Figure 3.

Average Investment by Firm Age

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Average firm investment growth by firm age. Young firms are between 0 and 14 years old, mature between 15 and 34, well-established are 35 and above. Investment growth is measured as the average per firm-age-category across countries and sectors each year.

Methodology and Identification Strategy

We hypothesize that entry into a program signals a reduction in future macroeconomic uncertainty, inducing firms to undertake non-reversible investments even if no real macroeconomic effects have had time to materialize yet. We are therefore interested in the dynamic response of firm investment to the approval of an IMF program. We estimate impulse response functions using a local projections (LP) methodology, which have become a popular alternative to VAR models in the macro-empirical literature because of their flexibility and simplicity. As described by Jordà (2005), local projections can be estimated by simple regression models and are in general more robust to misspecification errors than other related methods.

We not only aim to track the evolution of firm investment dynamics over time following the approval of an IMF program, but also estimate the average treatment effect (ATE) of such programs on firm investment. Calculating the unbiased effect of a Fund program on investments would require comparing two contrasting scenarios: one where we can measure the change in investment for a firm following a program approval, and another where we measure the change in investment when no such event has occurred, ceteris paribus. If the decision to enter an IMF program were fully exogenous, we would simply compare average investment across the two groups.

However, the decision to enter a program is endogenous to several observable and non-observable factors that contemporaneously determine firm investment. Dealing with the endogeneity of IMF programs is an issue that is tackled in several different ways in the literature. Initial papers focused on instrumental (IV) strategies that relied on political determinants (e.g., Barro and Lee 2005). As the exclusion restrictions for political variable IVs have been challenged, another type of IVs popularized by Lang (2021) and Gehring and Lang (2020) has become a preferred strategy. This shift-share IV exploits the differential effects of a country’s past participation in an IMF program given the IMF’s budget constraint to predict current participation in another program. However, crucially for our empirical strategy, this IV is suitable for the identification into an IMF program but not into program type, as we wish to explore the differential effects of IMF programs looking at GRA and PRGT programs separately.

To account for the endogeneity of an IMF program approval, we exploit a methodology developed by Jordà and Taylor (2016) that uses a propensity-score based method, combined with local projections (Jordà 2005) to find the ATE of an IMF program on the firm tangible fixed assets investment rate. With this method, we accept the endogeneity of entering an IMF program and first attempt to explicitly model for it. If the probability of entering a program is modeled correctly, we can re-balance the sample as if the decisions were taken at random. Then, in a second stage we use as the potential outcome variable the cumulative change in the ratio of tangible fixed assets to total assets. The final estimator gives the average treatment effect which is referred to as the Adjusted Inverse Propensity Weighted (AIPW) estimator, as developed and explained by Jordà and Taylor (2016). The AIPW estimator incorporates the flexibility of local projections with a method for reducing endogeneity bias. The two-stage method described above is doubly robust, in that the estimator will be unbiased if either of the two stages is correctly specified. The underlying idea is that the predictor set in the first stage, and then the control set in the second stage should be expansive enough to capture as much of the variation in program approval as possible.12

In our first stage we model the probability of being under a specific program type by estimating a propensity score for each element in our sample. Our dependent variable is the dummy variable identifying IMF program years as indicated in the MONA dataset. The propensity score for being under a program as predicted by the multinomial logit model:

P(i,t,p)=λ(β,Z(t−1,i))(1)

Where λ is the multinomial logistic distribution function and Z is a vector of country-specific controls including macro and political variables as well as region fixed effects.13 We estimate then the probabilities of either a) having no program, b) having a GRA program, c) having a PRGT program. In the model, the base values are the non-program years, and we estimate the propensity scores for each outcome, GRA or PRGT. This allows us to capture the heterogeneity of program type, and simultaneously the types of countries under programs, and so account for these factors implicitly through the weighting procedure in the second stage.

This first stage specification follows Dreher et al. (2009) and includes a dummy if a country was under a program in the past, a measure of autocracy, the country’s investment to GDP ratio, the log of real GDP per capita measured in PPP, total debt service, the budget balance, ratio of reserves to imports to capture import sustainability, real GDP growth, changes in reserves for external sustainability, the current account balance to GDP, and two measures of political quality including a dummy for election years and the log of checks-and-balances. Table A1 in the Appendix describes the predictor variables in detail.

The estimated P(i,t,p) is then the predicted probability of being under program p, for country i at time t given our set of predictor variables. From this, the second stage re-balances to create a synthetic sample where the decision to be under an IMF arrangement is as good as random. Using our logit estimates, we can estimate the extent of the non-randomness in our sample. Specifically, a highly endogenous event would be predictable based on observables and have a high P(i,t,p), while a highly endogenous control country would have a low P(i,t,p). We assign the weights 1P(i,t,p) to the treatment group and 1(1−P(i,t,p)) to the control group. The average treatment effect, given the re-balanced sample, will then be the difference of the average weighted potential outcomes of the two groups across our sample.

Table B1 in Appendix B reports the estimated coefficients for the first stage. The results are in line with the literature. There is strong evidence of path dependency, where countries that have participated in programs in the past are more likely to enter a new program. GDP per capita and GDP growth are both negatively associated with the likelihood of entering a PRGT arrangement, as more well-off countries typically have less of a need for these programs and countries which are growing are likely not to be in the sort of economic crises that typically warrant an adjustment program. The positive coefficient on GDP per capita when treatment is GRA is justified by the fact that among our sample of always-taker IMF countries, the richer ones are eligible for GRA arrangements only. An increase in reserves is also negatively correlated with IMF arrangements, indicating the importance of reserves in staving off balance of payment crises which can lead to an IMF program. It may be surprising that variables such as current account to GDP are not significant in some cases, given the Fund’s mandate to help countries in a balance of payment crises, but this result is in line with previous work (Conway, 1994). Finally, we find some evidence of the role of political variables in our sample. The literature speaks to different reasons as to why these variables might influence entering into a program.14 For example, combative elections might make the stigma of a program unappealing for incumbent politicians, which reflects the negative sign on our legislative election dummy.

The outcome variable, which is modeled in the second stage, is the cumulative change in the firm tangible fixed assets scaled by total assets, which captures investment throughout the years. Our baseline model models the outcome variable as measured with a local projection (Jordà 2005) according to the following baseline specification:

Δyi,j,k,t+h=α+βZi,j,k,t−1+δXj,t−1+γDj,t+μi+τk,t+εi,j,k,t h=1,2,....5(2)

Where Δyi,j,k,t+h is thus the conditional forecast of the change in tangible fixed assets from time t to t+h, which denotes firm investment over the forecast time horizon h of up to five years. This measure of the outcomes is also referred to as the cumulative impulse response function and is standard practice with this methodology (Jordà and Taylor 2016). The outcome is measured for firm i, in country), and sector k. Zi,j,k,t-1is a vector of firm control variables as described in Section 3 which also controls for the lagged difference in investment Δyi,j,k,t-1 to account for serial correlation, while Xj,t-1is a vector of country-level controls and lagged treatment variables, also described in Section 3. Dj,t is our country-level treatment variable, which is equal to one for the year when the country enters into an IMF program as described in Section 3.1. We also control for the remaining program years. Finally, we include firm fixed effects μi and sector-year time-varying heterogeneity τk,t. This way, we can account for both global factors determining investment dynamics as well as industry-specific unobservable characteristics tied to investment choices. Standard errors are clustered at the country level. Ei,j,k,t is the error term.

Regression equation (2) is run for each point in horizon h on the rebalanced sample to obtain the desired average treatment effect, ATE:

ATEh1nΣilΣtT{[(Δyi,j,k,t+h)(Di,t)Pi,t,p−(Δyi,j,k,t+h)(1−Di,t)1−Pi,t,p]−Di,t−Pi,t,pPi,t,p(1−Pi,t,p)[(1−Pi,t,p)m1h(Zi,t−1,Xi,j,k,t−1)+(Pi,t,p)m0h(Zi,t−1,Xi,j,k,t−1)]}(3)

Where Δyi,j,k,t+h are the estimated conditional forecasts for the local projections (Equation 2), and Di,t is the dummy variable to indicate treatment, in our case being program approval. Pi,t,p are the estimated propensity scores from Equation 1. The first part of Equation 3 for computing the estimator is a standard inverse propensity-score weighted ATE. Intuitively, this is like a group-means comparison between countries that have signed a program and those that have not, with the additional step that we correct for allocation bias of the treatment by modeling it in Equation 1, reducing it to a unidimensional element which is the estimated propensity score, and inverting to achieve a random distribution. The second part is an adjustment term consisting of the weighted average of the two independent regression estimates. The purpose of the adjustment term is to stabilize the estimator as the propensity scores get close to the extremes (0 or 1) and therefore alleviates the need to truncate weights.15

In conclusion, the use of local projections for our estimation strategy is motivated by several factors. First, local projections are free of structural constraints that would otherwise be imposed on a parallel VAR model, thereby allowing for the response of investment to an IMF program approval to vary non-linearly over the forecast horizon, making them useful for computing dynamic effects. Local projections are also easier to compute and can be estimated using ordinary least squares (OLS). In evaluating the properties of local projections, Plagborg-Møller and Wolf (2021) and Montiel Olea and Plagborg-Møller (2021) argue for the use of lag-augmented local projections as a requirement for robustness. However, local projections are not without drawbacks. Since the estimation does not impose any direct link between impulse responses at times h and h + 1, estimates can sometimes display erratic behavior (Ramey and Zubairy 2014). Furthermore, as the horizon increases, observations are lost on both sides especially when including lags, which can lead to loss of efficiency. Therefore, local projections are optimal for short to medium term projections, and the efficiency of the estimator is a function of forecast horizon over the total size of the time dimension T. Because we forecast the impulse response of investment up to a max of 5 years over a 20-year period, our choice of method remains safe. As a robustness, we test the sensitivity of results by restricting estimates to groups of firms with data over a full forecast and lag horizon.

In the next section we present our main results. We consider the effects of entering any of the programs in the standard IMF toolkit, whether using the GRA or the PRGT.

Empirical Findings

Local Projections Baseline Results

Table 2 presents our main results for the AIPW estimator, also plotted in Figure 4. Each panel in Figure 4 shows the ATE at time t+h for programs approved at time t separately for GRA and PRGT financing. We find that the effect of GRA programs is increasing over time, peaking at four years after program approval. On average, tangible assets scaled by total assets grow over four years by a cumulative amount of almost four percentage points with respect to the approval year. This effect takes time to manifest, as we find no significant change in the first year after the program approval. For PRGT programs on the other hand, we find only a very weak temporary effect. In the first year after program approval, tangibles accumulate marginally, with a value around one percentage point above the reference level, but afterwards we find no evidence of continued investment.

Table 2.

Program Signing and Firm Investment Response, AIPW Estimates

article image
Notes: Average treatment effect of a Fund program approval estimators for each h-step ahead forecast on the cumulative change in firm tangibles/TA, with h=1,2,3,4,5. Standard errors clustered at the country level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Figure 4.
Figure 4.

Program Approval and Firm Investment Response, AIPW Estimates

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Panel A shows AIPW average treatment effects for each h-step ahead cumulative change in tangible fixed asset investment rate with respect to base year (yt+h – yt) following the signing of the respective IMF program (GRA or PRGT). Shaded areas show 90 and 95% confidence intervals, standard errors clustered at the country level.

We interpret our baseline results as follows. The delayed effect of GRA arrangements on private sector investment suggests that investors in GRA-only countries are cautious and they generally prefer to wait before making long-term non-reversible capital investments. Also, it could be that firms operating in GRA-only countries may be relatively more mature firms than firms in PRGT-eligible countries, with less initial room to grow and potentially more skeptical about the materialization of the pay-offs of reforms. However, one this initial skepticism is cleared, the private sector seems to embrace the opportunity to increase investment. On the flip side, there are likely many factors other than a confidence effect at play for firms operating in PRGT-eligible countries. The positive signaling effects of a Fund program are not enough to offset the drag on private sector investment due to poorer access to credit, lower quality of institutions, and fewer cash generating opportunities that are associated with the markets in which these firms operate.

The differential effects of the type of IMF financing on investment can also be explained by the nature of these arrangements and the underlying balance of payments problem faced by countries. As indicated, the length of GRA arrangements is typically up to four years16, whereas it is common for PRGT arrangements to last for five years, which could explain the more muted impact in PRGT contexts. Furthermore, as mentioned in section 3.1, in case of a protracted balance of payment problem under the PRGT, repeated arrangements with the country can be expected, to achieve progress toward a stable and sustainable macroeconomic position. Thus, it is likely that repeated IMF engagements under the PRGT would provide firms with the kind of confidence boost needed to match GRA effects, since macroeconomic stability would be safeguarded over an extended period.

Firm Financial Frictions

One of the advantages of local projections is its flexibility. A large literature on fiscal multipliers uses these methods to investigate the state-contingent effects of treatments. Since we wish to see how a program adoption affects firms differentially through their degree of perceived uncertainty, we can speculate that a relevant channel of transmission of this effect is the degree of financial frictions that the firm is subject to. Depending on such frictions, firms could be more sensitive to the positive, uncertainty-reducing effects of a program adoption, as opposed to a negative signal that is sometimes hypothesized in the literature. In the latter case, a program adoption can be seen as the acknowledgement of policy failures or of incoming adverse macroeconomic shocks, and therefore become a self-fulfilling prophecy that drives down investments. However, it could also be that, when distinguishing by the degree of financial constraint that the firm is subject to, the effects will vary since more constrained firms will benefit disproportionately. To this end, we adapt our baseline specification in Equation 2 following the flexible local projection methods of Jordà and Taylor (2016) as follows:

Δyi,j,k,t+h=Ii,t−1j[α+βZi,j,k,t−1+δXj,t−1+γDj,t+μi+τk,t]+(1−Ii,t−1j)[α+βZi,j,k,t−1+δXj,t−1+γDj,t+μi+τk,t]+εi,j,k,th=1,2.....5(5)

Where we introduce an indicator variable I(i,t1)j that takes the value 1 when an observation falls under state j in the year t- 1. In other words, the indicator variable sorts the data into two groups based on the observed state of the firm. We consider two different dimensions of firm uncertainty and implied financial frictions, and subsequently estimate Equation 5 by binning the data along these dimensions. In the remainder of the section, the firm characteristics along which the binning is performed are explained in detail, along with the binning strategy. With respect to Equation 5, the rest of the specification follows the baseline specification in Equation 2.

The methodology described above allows us to consider if the heterogenous response of firm investments to Fund program approval is attributable not just to the program type, but also to the exposed firm type. We hypothesize that one way which the IMF can influence private investment dynamics is through a financial frictions channel, such that the presence of a Fund program invigorates private investments for those firms which are under greater financial uncertainty.17 Then we expect that firms react differentially to an IMF arrangement based on their degree of financial frictions. To capture financial frictions at the firm level, we follow a large literature which points to the use of firm age as a useful proxy (Gertler (1988), Hadlock and Pierce (2010), Cloyne et al. (2018), Bahaj et al. (2019)). There is an obvious disadvantage to using direct measures of financial frictions such as size, leverage, or liquidity because they respond endogenously to shocks such as the approval of IMF arrangements, making it difficult to interpret ex-post effects as driven by ex-ante heterogeneity.

For these reasons, firm age has been shown to be an appropriate proxy. Especially in countries with less developed financial markets, younger firms are more leveraged, less liquid, and smaller in size. Highly leveraged firms will be sensitive to financing conditions in the country, as new debt will be more expensive, ceteris paribus. In a similar manner, less liquid firms are more sensitive to liquidity shocks or changes in borrowing conditions since they rely on external finance. All these elements highlight the importance of shocks to the financial system which reduce macroeconomic uncertainty and systemic risk. In this sense, if IMF programs are associated with these positive effects, then we expect that younger (more financially constrained) firms will benefit more.

For our empirical framework, we therefore estimate a state-contingent local projection as described in Equation 5. The indicator variable I(i,t1)j takes the value 1 for those firms which are above the median age of firms in the sample, where age is computed as the number of years from incorporation. Figure 5 plots the AIPW average treatment effect for the two groups.

Figure 5.
Figure 5.

AIPW and Firm Age

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: AIPW average treatment effects of program signing on firm tangible fixed assets investment rate for groups of firms based on age. Firms are divided into two groups: mature firms are those with above-median age, young firms below-median age. Areas indicate 90 and 95% confidence intervals, standard errors are clustered at the country level.

We find effects which mirror our baseline. For both PRGT and GRA programs, younger, more financially constrained firms benefit more from an IMF arrangement. Compared with our baseline estimates, we find a significant albeit short-lived effect on investment for PRGT programs. For GRA programs, the effects are notable; five years after program approval, there is on average a positive 10 percentage point cumulative change in tangibles above the reference year value. There is also a positive but much smaller effect for mature firms. Under PRGT programs, we find no effects for mature firms similar to our baseline estimates, but a positive effect of around 2 percentage points for the first three years following program approval. The positive response of younger firms in PRGT countries reflects a willingness of younger, more dynamic firms to expand in an environment with greater institutional and financial frictions. However, for the same reasons we commented in our baseline results, these effects are short lived.

Financial Frictions and Uncertainty: A Stacked Difference-in-Differences Approach

Our state-contingent local projection results on firm age provided a good first proxy for the role of financial constraints at the firm level as channels through which uncertainty and financial risk reduction following IMF arrangements can influence investments. However, it can also be argued that such a measure is not fully exogeneous because of a survivorship bias or changes in ownership–younger firms tend to be more likely to go bust because of those same characteristics just defined or, when they do survive, they are more likely to be absorbed by older, larger firms in M&A operations. Furthermore we are unable to find more granular effects of firm heterogeneity by simply looking at group-wise splitting. Therefore, in this section we first rely on a common methodology in the corporate finance literature, the Rajan and Zingales (1998) (henceforth RZ) external financial dependence indicator, to capture the differential effects of financial constraints on investment using a novel Difference-in-Differences (DiD) approach. We then apply this approach to our firm-level uncertainty measures.

To capture the differential effects of an IMF arrangement on private sector investment for firms with varying degrees of external financial dependence, we turn to a DiD approach. While for a long time the generalized (two-way fixed effects) DiD was the workhorse method in panel setups such as ours, the recent explosion in DiD methodologies provides a new set of tools to deal with the many limitations of the original models. New estimators from de Chaisemartin and d’Haultfoeuille (2021), Callaway et al. (2021), and Goodman-Bacon (2021) have recently been developed to account for issues such as multiple time periods of treatment, staggered adoption, or treatment switching. In our case, all these issues are relevant. Our sample consists of countries that can have an IMF program at different points in time and crucially switch in and out of treatment continuously. To understand the extent of this issue, Figure A2 in Appendix A plots the treatment status by country and program type for each year in the sample. Each row represents a country, while each column represents a year. The coloring of each point in this matrix is then a representation of the treatment status. Grey blocks are years with no treatment (i.e., no IMF program), red blocks correspond to years where a country was under a GRA program, and blue blocks are for PRGT program-years. White blocks are years for which we have no data on the country. As such, the dynamic nature of treatment is evident in our sample, motivating the use of a more robust methodology.

Given the structure of our data, we rely on a recent causal inference methodology popularized by Cengiz et al. (2019) and Deshpande and Li (2019), the stacked DiD method. The procedure is an event-study based approach to analyze the effects of multiple events (such as multiple IMF programs) in a way that eliminates the typically problematic comparisons between treated and control groups that occur in a generalized DiD. In brief, the method consists in splitting the data into n sub-experiments, where each sub-experiment represents a unique calendar year where treatment (program approval) occurred for any cross-sectional group (country). A treatment window is then defined, such that only observations with a subsequent treatment at least k years ahead are considered as controls. As a result, all observations within one sub-experiment will have a program adoption year which is the same, and the control group consists of observations sufficiently far into the future so as not to be confounded by other program adoptions. These sub-experiments are then stacked on top of each other to create a dataset which consists of n independent panel event studies identified between each other by a sub-experiment id.

For evaluating the role of financial frictions, the seminal paper by RZ was a crucial step forward in determining the causal direction of financial development on economic outcomes. The underlying idea is that the role of financial markets is to reduce problems of moral hazard and adverse selection, thereby reducing the costs of the firm in raising funds. Financial development, or any structural shock to the financial system of a country, should disproportionately help firms which are more dependent on external financing. In our context, we consider our main treatment (program approval) as such a shock, because of the signaling effects discussed in the previous sections. The advent of a program approval reduces macroeconomic uncertainty with positive spillover effects on the financial system. Then the response of private sector investment will be contingent on a firm’s reliance on external finance.

The RZ index is a sector-specific, time-invariant measure of the share of investment that is not financed by internal cash flow in the median listed U.S firm over the 1980s. The guiding assumption to this approach is that the U.S capital market is a good proxy for a frictionless market, and credit demand is driven by industry-specific technological fundamentals. In a cross-country framework such as ours, the second assumption is that these industry fundamentals are constant across countries. While a small literature compares the original index with a few country-specific measures (Eppinger and Neugabauer, 2022), the RZ index has been widely used and shown to be consistent across countries (Cetorelli and Gambera, 2001; Beck and Levine, 2002; Fisman and Love, 2003,, 2007; Kroszner et al., 2007; Pagano and Pica, 2012). Furthermore, the use of country-specific indices computed on our data would have an additional drawback specific to our research setting. The index is constructed using firm cash flows. Previous work by Bomprezzi and Marchesi (2022) have shown that IMF programs have a strong quantitative effect on sales and therefore on cash flows of firms, making such an index endogenous.

For these reasons, we use the indices computed by Eppinger and Neugabauer (2022) following the RZ methodology. From Compustat, the authors define the index of external financial dependence for U.S firms over the years 1990–2005. Being closer in time to our sample, it is a better proxy of technological demands of an industry. External financial dependence is then defined as capital expenditures minus cash flow from operations for each firm, then divided by the sum of capital expenditure over the period, and finally using the median value by industry as a measure.18 We then merge these industry values reported as NACE sectors with our Orbis data. Table A6 in Appendix A reports the values of the EFD indices. As in RZ, the indices are only computed for a set of firms in manufacturing-oriented industries.

A final assumption is made here regarding the structure of the financial system in the countries in our sample. Since we deal mostly with middle to lower income countries, the links between banks or credit institutions and firms are strong and these institutions remain the predominant source of financing in those countries, due to the underdeveloped equity markets or availability of other funding sources. To the extent that this is true, the effects we find reflect how traditional credit institutions perceive the level of risk within a country, after the approval of an IMF arrangement.

Given the setup, we fit a DiD to the stacked data. It allows us to evaluate the differential effect on tangibles scaled by total assets among firms with a high degree of external financial dependence (or volatility) versus those with low degrees of financial dependence, in the post-program period as compared with the period before. The model contains the same country and firm controls as in our baseline specifications, fixed effects, and sector-year fixed effects to account for time-varying heterogeneity. A further advantage of a stacked DiD setup is the ability to compute dynamic effects. As in our baseline local projection specifications, we are interested in the time-varying effects of an increment in uncertainty or dependence on external finance. We specify a model as shown in Equation 7, where we identify the two years before and the five years after program approval (with year 0 as the reference year) with a set of indicator variables YSE (years since event):

Tan/TAi,j,k,t=α+βZi,j,k,t−1+δXj,t−1+Σj=−kakbγj(FIRMINDEXi/i,t*1(YESt=j))+σFIRMINDEXi/i,t+Σj=−kakbρj*1(YESt=j)+μj/i+τk,t+εi,j,k,t(5)

Where our parameter of interest is yj, representing the interaction between the indicator for the year before/after the program approval and our firm index of choice, either external financial dependence or uncertainty.

Table 3 presents our results. First, we find no evidence of an anticipation effect or of pre-treatment trends. When considering external financial dependence (Panel A in Table 3), in the case of GRA programs, we find that tangible investments grow disproportionately more relative to the base year for firms operating in sectors that are characterized by a high degree of external financial dependence, at every point in time after the program approval up to 5 years. For example, for the industry which is at the bottom 5th percentile of external financial dependence (publishing and printing), the expected effect after 1 year is small and negative, at around -0.17 percentage points. For the firms in the industry at the top 95th percentile (communication equipment) the effect is 0.5 percentage points. For PRGT programs the size of the effect is smaller, but we do find that in the long run there is a strong and significant effect, indicating that more financially constrained firms in PRGT-eligible countries do benefit from IMF arrangements.

Table 3.

Firm Frictions and Dynamic Stacked DiD Estimates

article image
Notes: Year-specific DiD effect of a treatment d on tangibles/TA in a stacked event study setup. Panel A considers the interaction between the degree of external finance dependence and a dummy equal to 1 for the year t before/after the program approval. Panel B considers as the interacting term the measure of realized volatility. All specifications include full controls and sector-year fixed effects. Panel A uses country fixed effects, Panel B uses firm fixed effects. Standard errors clustered at the country level. * p < 0.1, ** p < 0.05, *** p < 0.01.

To capture uncertainty at the firm-level, we adopt the measures constructed by Alfaro, Bloom, and Lin (2021), which are the annualized stock return volatility. The first measure is constructed as the annualized 12-month standard deviation of daily CRSP returns of a sample of U.S firms. Furthermore, the authors provide firm level measures of 12-month compounded stock returns and Tobin’s Q as additional controls to tease out first-moment effects. Their data spans from 1992 to 2019 and provides the year-by-year 2 digit SIC industry codes. We therefore aggregate these measures by taking the median sector-year value and match them with our firm data also at the sector-year level. By matching U.S data with our sample at the sector level we are also constructing a measure of firm uncertainty that is exogenous. These measurements should be interpreted as an industry-specific characteristic which are comparable across countries, à la Rajan and Zingales (1998). We argue for this approach because uncertainty, after controlling for country-level macroeconomic and institutional factors as well as firm characteristics, is an industry-specific process that is driven by elements such as supply chain networks and product-specific demand elasticities.

For our measure of volatility (Panels B in Table 3), we find overall stronger effects when considering the realized volatility (annualized 12-month standard deviation of daily CRSP stock returns). As before, for GRA programs, there is evidence that firm investments increase as sector-wide volatility, and therefore uncertainty, increases. Greater volatility of yearly returns under a GRA program is associated with a 3-percentage point increase in tangibles after three years relative to the base year. Under PRGT programs, there is no consistent long-term effect.19 We take this as evidence of our initial hypothesis that the institutional and macro-financial environment of PRGT eligible countries is a strong deterrent to firm investments, and a single program is insufficient to reverse investor tendencies.

Domestic Ownership of Firms

Our main hypothesis considers that firms will regard an IMF arrangement as a guarantee of lower future economic and political uncertainty. In the previous sections, we considered measures of financial frictions and quantitative estimates of uncertainty as channels through which this effect can influence private sector investments, by lowering financial risk and therefore benefitting firms which are more sensitive to macro-financial conditions. In this section, we try to capture the perception of uncertainty by the firm at the organizational level by considering changes in its ownership structure.

An advantage of Orbis is the availability of historical ownership data. From historical vintages, we can download ownership data for the firms beginning from 2007. We retrieve information on the global ultimate owner (GUO) and the global ultimate consolidated owner where it exists. These are the ultimate owners, net of all intermediate ownership connections, with at least 50% of direct or indirect ownership in the firm. We can extract information on the owner country, and therefore classify a firm as having a foreign or domestic owner in a given year, identifying foreign ownership as the case where the country of operation of the firm is different from the country of the GUO.

Simply comparing domestically owned versus foreign-owned firms however is misleading. Foreign-owned firms are likely to be by nature larger and more successful because they are part of a multinational corporate group. Furthermore, their ownership changes occur quite frequently, and often they are cross-border and likely driven by economic expansions or recessions. We want to identify a set of firms which, using reasonably exogenous characteristics, are tied to the country more so than a counterpart firm, such that they would be disproportionately affected by the uncertainty-reducing effects of an IMF program. To that end, we consider firms which do not switch ownership in the immediate years before and after program approval. This behavior indicates that these firms remain “committed” to the country and are therefore more sensitive to the changes in uncertainty brought on by an IMF program. In contrast, firms that do switch to foreign ownership become less sensitive to what is happening in the country, because they gain a sort of natural hedge by being part of a foreign group. We call those groups of firms that are tied to the country and don’t switch ownership “never-leavers.”

We report the dynamic stacked DiD estimates for “never-leavers” in Table 4. The change in investments for these firms that remain exposed to the Fund programs throughout the treatment period is positive with respect to firms which changed in ownership. This effect is here again stronger for GRA countries. The magnitude of the effect in PRGT countries is around 0.3 percentage points increase in tangibles after five years, while for GRA countries it is around one percentage point. We attribute the differential effect to a combination of differences in ownership structures across countries and the environment in which these firms operate. GRA countries are more likely to host multinationals which could retreat from the market when adverse shocks or negative signals occur. This reaction of multinationals leaves more room for those firms that do stay in the country to expand once the positive signal from an IMF arrangement kicks in. In PRGT-eligible countries, our indicator of firms that do not leave the country may simply be picking up those firms which are unable to leave the country because of capital, institutional, or financial constraints. Regardless, the fact that we find a positive effect indicates that these firms still benefit from remaining on the ground throughout the IMF program years.

Table 4.

Ownership Switches and Dynamic Stacked DiD Estimates

article image
Notes: Year-specific DiD effect of a treatment d on tangibles/TA in a stacked event study setup. Results show the interaction between a dummy identifying “never-leavers” and a dummy equal to 1 for the year t before/after the program approval. Specifications include full controls and firm and sector-year fixed effects. Standard errors clustered at the country level. * p < 0.1, ** p < 0.05, *** p < 0.01.

Robustness and Alternative Specifications

In our appendix, we provide a series of robustness tests to our baseline analysis. First, one could wonder if the dynamics we estimate in the baseline results are driven by pre-treatment trends. It could be that investments were already growing before program approval. Our identification strategy captures primarily the systemic differences between countries that select into a program type, GRA or PRGT. As sensitivity check, we test for the presence of anticipation effects to see if, in the lead up to a program approval, there are detectable investment dynamics. To do this, we estimate a simple fixed effects model, regressing investment at time t, on dummies for a program approval that will occur at t+h. Table C1 in Appendix C shows the results. Each estimate represents the response of investment at time t to a program signing which will occur h years ahead. For both GRA and PRGT programs, we find no evidence of systematic anticipation effects. In general, there is weak evidence that two or three years before program approval, investment is falling, but there are no significant effects as we get closer to the program signing date.

We also want to test the persistence of program effects after the program ends. Since we find positive effects of a program approval, which induces firms to undertake investments following a reduction in perceived uncertainty, then the effects we observe after a program end could inform us on how the IMF arrangement affects firm perceived uncertainty in the long run. Table C2 in Appendix C addresses this question. Using our AIPW estimator, we use the final program year as treatment in the second stage estimation. This dimension allows us to make a further classification according to the program review status. Based on the number of reviews, a program can be either classified as completed if all reviews were completed, or it can be off-track if at most the first review is completed. To capture only the effects of a program ending and not a subsequent program that may follow, we restrict the sample to exclude instances of a successor program in the five years after a program ends. For GRA programs, we find effects which are positive and significant regardless of whether the program went off track or not. In some cases, we find that the magnitude of the effect is slightly larger when the program goes off-track. Compared with programs that were completed, where tangibles over total assets grew by 2.57 percentage points after 5 years with respect to the base year, the effect is 3.16 percentage points for off-track programs at the same horizon. Overall, the difference in the effects is not significantly different. For PRGT arrangements, we find that investment growth drops in the first years after the end of a program and, when the program goes off track, these negative effects persist, differently from the baseline estimates. Three years after program approval, tangibles are 4.21 percentage points lower than the reference year. This suggests both that firms were dependent on the effects of the program, as well as the fact that repeated programs are expected under PRGT schemes. Therefore, in instances where this doesn’t occur, firms internalize the lost opportunity from future programs.

We also run a series of tests to check whether the results are driven by the sample or by peculiarities in the data. We report all these results in table C3 in Appendix C. First, since we use a forecast horizon of up to five years, we reduce our sample to only those firms for which we have balance sheet data on investment for the full 5-year horizon. With differences in cross-country coverage in Orbis, it could be that our local projection estimates are driven by firms with more rigorous reporting standards, or which cover more years. By considering only spells of observations equal to or greater than our maximum forecast horizon, we can test for this. The findings are consistent with our baseline estimates.

It could also be argued that the program signaling effect should be strong enough to induce firms to undertake capital investments. As we described in the previous results of Table C2, a share of programs approved never make it to completion, and if countries go off track sufficiently early, firms could react adversely. Therefore, we run our baseline estimates dropping programs that are classified as off-track. Unlike the previous results, our treatment is again the program approval year. We find no significant differences in the investment dynamics relative to the baseline model, except that some of the estimates for PRGT programs are slightly less significant.

Finally, we consider the case where results might be affected by different country groups. As we describe in the data presentation and show in Table A4, Orbis coverage varies by country. In Table C4 we systematically drop, one at a time, countries belonging to the different IMF regional departments, thereby showing how results vary to the exclusion of Sub-Saharan Africa, for example. Overall, we find that our GRA results are robust to these sensitivity checks. For example, when Sub-Saharan Africa is removed, we still find a positive and significant effect, such that tangibles over total assets increase up to about four percentage points four years after program approval. This finding is consistent across the exercise of dropping regions. When we drop Europe, results for GRA are not significant. We explain this result in two ways. First, because our sample includes the European Debt Crises, it is unsurprising that removing this event attenuates the effect that the Fund has on private enterprise, given the large role it had in Europe during this period. Secondly, our sample shrinks considerably, making the results harder to interpret. Our AIPW estimates for PRGT programs are also weaker when dropping regions like Sub-Saharan Africa or Middle East and Central Asia, since PRGT-programs are extensive in those regions especially in Sub-Saharan Africa.

Conclusion

This paper provides new evidence on the role of IMF programs in stimulating private sector investments. Using detailed firm-level data on tangible fixed investments and a local projection methodology, we estimate the dynamic response of firm investments to the approval of an IMF arrangement. We find that distinguishing between GRA and PRGT financing matters for the path of firm investment, and that GRA programs seem to induce a stronger reaction of investors in the country. Leveraging a novel DiD methodology, we document the presence of two financial channels: the degree of firms’ external financial dependence and firms’ sectoral uncertainty. We find evidence that investments are higher for firms relying more on external finance or those which are exposed to greater uncertainty under a GRA arrangement, while we do not find similar effects for PRGT arrangements. The results speak to the differential role these programs play in stimulating private sector investments. Finally, using ownership data, we find that private investments are stimulated more for domestic firms. The presence of a private investment transmission channel could help improve our understanding of what factors could affect the success and effectiveness of IMF programs. To the best of our knowledge, this is the first paper that investigates whether IMF programs, as well as improving a country’s creditworthiness for external investors, may also make “internal” investors more willing to invest in their own country, by reducing the impact of uncertainty.

Bibliography

  • Abel, Andrew B. 1983. Optimal investment under uncertainty. The American Economic Review, 73 (1), 228–233.

  • Alexopoulos, Michelle and Cohen, Jon. 2015. The power of print: Uncertainty shocks, markets, and the economy. International Review of Economics & Finance, 40(C), 8–28.

    • Search Google Scholar
    • Export Citation

  • Alfaro, Iván, Bloom, Nicholas, and Lin, Xiaoji. 2021. The Finance Uncertainty Multiplier. Available at SSRN: https://ssrn.com/abstract=3093871 or http://dx.doi.org/10.2139/ssrn.3093871

    • Search Google Scholar
    • Export Citation

  • Andrade, Sandro C., and Chhaochharia, Vidhi. 2018. The costs of sovereign default: evidence from the stock market. Review of Financial Studies, 31: 1707–1751

    • Search Google Scholar
    • Export Citation

  • Bahaj, Saleem, Pinter, Gabor, Foulis, Angus, and Surico, Paolo. 2019. Employment and the Collateral Channel of Monetary Policy. Staff Working Paper No. 827, Bank of England.

    • Search Google Scholar
    • Export Citation

  • Baker, Malcom, and Wurgler, Jeffrey. 2002. Market Timing and Capital Structure. The Journal of Finance, 57: 1–32.

  • Baker, Scott R., Bloom, Nicholas, and Davis, Steven J. 2016. Measuring Economic Policy Uncertainty. The Quarterly Journal of Economics, 131 (4), 1593–1636.

    • Search Google Scholar
    • Export Citation

  • Bar-Ilan, Avner, and Strange, William. C. 1996. Investment Lags. The American Economic Review, 86 (3), 610–622.

  • Barro, Robert J. and Lee, Jong-Wha. 2005. IMF-programs: Who is chosen and what are the effects? Journal of Monetary Economics, 52, 1245–1269.

    • Search Google Scholar
    • Export Citation

  • Bas, Muhammet A., and Stone, Randall W. 2014. Adverse selection and growth under IMF programs. The Review of International Organization, 9:1—28.

    • Search Google Scholar
    • Export Citation

  • Baskaya, Yusuf, di Giovanni, Julian, Kalemli-Özcan, §ebnem, Peydro, Jose-Luis, and Ulu, Mehmet. 2017. Capital flows and the international credit channel. Journal of International Economics, 108: 15—22.

    • Search Google Scholar
    • Export Citation

  • Baum, Christopher F., Mustafa O., and Talavera, Oleksandr. 2010. On the Sensitivity of Firms’ Investment to Cash Flow and Uncertainty. Oxford Economic Papers, 62, 286–306.

    • Search Google Scholar
    • Export Citation

  • Beck, Thorsten, and Ross, Levine. Industry Growth and Capital Allocation: Does Having A Market- Or Bank-Based System Matter? Journal of Financial Economics, 64 (2), 147–180.

    • Search Google Scholar
    • Export Citation

  • Bernanke, Ben. 1983. Irreversibility, uncertainty, and cyclical investment. The Quarterly Journal of Economics, 98(1), 85–106

  • Bloom, Nicholas, Bond, Stephen, and van Reenen, John. 2007. Uncertainty and Investment Dynamics. Review of Economic Studies, 74, 391–415.

    • Search Google Scholar
    • Export Citation

  • Bloom, Nicholas, 2009. The Impact of Uncertainty Shocks. Econometrica, 77 (3), 623–85.

  • Bluhm, Richard, Dreher, Axel, Fuchs, Andreas, Parks, Bradley C., Strange, Austin M., and Tierney, Michael J. 2020. Connective Financing: Chinese Infrastructure Projects and the Diffusion of Economic Activity in Developing Countries. AidData Working Paper #103. Williamsburg, VA: AidData at William & Mary.

    • Search Google Scholar
    • Export Citation

  • Bomprezzi, Pietro, and Marchesi, Silvia. 2021. A firm level approach on the effects of IMF intervention. CefES Working Paper 476.

  • Boockmann, Bernhard, and Dreher, Axel. 2003. The Contribution of the IMF and the World Bank to Economic Freedom. European Journal of Political Economy 19 (3), 633–649.

    • Search Google Scholar
    • Export Citation

  • Booth, L., Aivazian, V., Demirguc-Kunt, A. and Maksimovic, V. 2001. Capital Structures in Developing Countries. The Journal of Finance, 56: 87–130.

    • Search Google Scholar
    • Export Citation

  • Borio, Claudio E. V. 1990. Leverage and Financing of Non-Financial Companies: An international Perspective. Economic Papers 27, Bank for International Settlements.

    • Search Google Scholar
    • Export Citation

  • Broner, Fernando, Martin, Alberto, Pandolfi, Lorenzo, and Williams, Tomas. 2021. Winners and losers from sovereign debt inflows. Journal of International Economics, 130: 103–446.

    • Search Google Scholar
    • Export Citation

  • Callaway, Brantly, Goodman-Bacon, Andrew, and Sant’Anna, Pedro H.C. 2021. Difference-in-Differences with a Continuous Treatment. Papers 2107.02637, arXiv.org, revised Jul 2021.

    • Search Google Scholar
    • Export Citation

  • Cengiz, Doruk, Dube, Arindrajit, Lindner, Attila, and Zipperer, Ben. The effect of minimum wages on low-wage jobs. The Quarterly Journal of Economics, 134, no. 3 (2019): 1405–1454.

    • Search Google Scholar
    • Export Citation

  • Cetorelli, Nicola. and Gambera, Michele. 2001. Banking Market Structure, Financial Dependence and Growth: International Evidence from Industry Data. The Journal of Finance, 56: 617–648.

    • Search Google Scholar
    • Export Citation

  • Cevik, Serhan, and Miryugin, Fedor. 2018. Does Taxation Stifle Corporate Investment? Firm-Level Evidence from ASEAN Countries. IMF Working Papers, International Monetary Fund, No 2018/034.

    • Search Google Scholar
    • Export Citation

  • Chauvet, Lisa and Ehrhart, Héleène. 2018. Aid and growth: evidence from firm-level data. Journal of Development Economics 135: 461–477.

    • Search Google Scholar
    • Export Citation

  • Chapman, Terrence, Fang, Songying, Li, Xin, and Stone, Randall W. 2015. Mixed Signals: IMF Lending and Capital Markets. British Journal of Political Science, 47.

    • Search Google Scholar
    • Export Citation

  • Cloyne, James, Ferreira, Clodomiro, Froemel, Maren, and Surico, Paolo. 2018. Monetary Policy, Corporate Finance and Investment. NBER Working Papers No. 25366, National Bureau of Economic Research

    • Search Google Scholar
    • Export Citation

  • Conway, Patrick. 1994. IMF lending programs: participation and impact. Journal of Development Economics, 45, 365–391.

  • Corsetti Giancarlo, Guimaraes, Bernardo, and Roubini, Nouriel. 2006. International lending of last resort and moral hazard: A model of IMF’s catalytic finance. Journal of International Economics, 53: 441–471

    • Search Google Scholar
    • Export Citation

  • de Chaisemartin, Clément, and d’Haultfoeuille, Xavier. Difference-in-Differences Estimators of Intertemporal Treatment Effects. NBER Working Paper Series No. 29873.

    • Search Google Scholar
    • Export Citation

  • Deshpande, Manasi, and Li, Yue. Who is screened out? Application costs and the targeting of disability programs. American Economic Journal: Economic Policy 11, no. 4 (2019): 213–48.

    • Search Google Scholar
    • Export Citation

  • Dreher, Axel. 2006. IMF and Economic Growth: The Effects of Programs, Loans, and Compliance with Conditionality. World Development 34 (5), 769–788.

    • Search Google Scholar
    • Export Citation

  • Dreher, Axel, and Vaubel, Roland. 2004. Do IMF and IBRD Cause Moral Hazard and Political Business Cycles? Evidence from Panel Data. Open Economies Review 15 (1), 5–22.

    • Search Google Scholar
    • Export Citation

  • Dreher, Axel, Sturm, Jan Egbert, Vreeland, James Raymond. 2009. Global horse trading: IMF loans for votes in the United Nations Security Council. European Economic Review, 53 (7), 742–757.

    • Search Google Scholar
    • Export Citation

  • Dreher, Axel, and Lohmann, Steffen. 2015. Aid and growth at the regional level. Oxford Review of Economic Policy, 31(3–4): 420–446.

  • Dreher, Axel, Fuchs, Andreas, Hodler, Roland, Parks, Bradley C., Raschky, Paul A., and Tierney, Michael J. 2021. Is Favoritism a Threat to Chinese Aid Effectiveness? A Subnational Analysis of Chinese Development Projects. World Development, 139: 105291.

    • Search Google Scholar
    • Export Citation

  • Easterly, W. 2005. What Did Structural Adjustment Adjust? The Association of Policies and Growth with Repeated IMF and World Bank Adjustment Loans. Journal of Development Economics, 76(1): 1–22.

    • Search Google Scholar
    • Export Citation

  • Eppinger, P. S., and Neugebauer, K. 2022. External financial dependence and firms’ crisis performance across Europe. Empirical Economics, 62, 887–904.

    • Search Google Scholar
    • Export Citation

  • Fisman, Raymond, and Love, Inessa. 2003. Trade Credit, Financial Intermediary Development, and Industry Growth. The Journal of Finance, 58: 353–374.

    • Search Google Scholar
    • Export Citation

  • Fisman, Raymond, and Love, Inessa. 2007. Financial Dependence and Growth Revisited. Journal of the European Economic Association, 5(2/3), 470–479.

    • Search Google Scholar
    • Export Citation

  • Frank, Murray Z., and Goyal, Vidhan K. 2009. Capital Structure Decisions: Which Factors Are Reliably Important? Financial Management, 38: 1–37.

    • Search Google Scholar
    • Export Citation

  • Gehring, Kai, and Lang, Valentin. 2020. Stigma or Cushion? IMF Programs and Sovereign Creditworthiness. Journal of Development Economics, 146: 1–14

    • Search Google Scholar
    • Export Citation

  • Gertler, Mark. 1988. Financial Structure and Aggregate Economic Activity: An Overview. Journal of Money, Credit and Banking, 20 (3), 559–588.

    • Search Google Scholar
    • Export Citation

  • Gilchrist, Simon, Sim, Jae W., and Zakrajsek, Egon. 2014. Uncertainty, financial frictions, and investment dynamics. NBER Working paper series No. 20038.

    • Search Google Scholar
    • Export Citation

  • Goodman-Bacon, Andrew. 2021. Difference-in-Differences with Variation in Treatment Timing. NBER Working Paper Series No. 25018.

  • Gourinchas, Pierre-Olivier, Philippon, Thomas, and Vayanos, Dimitri. 2017. The Analytics of the Greek Crisis. NBER Macroeconomics Annual, 31(1): 1–81.

    • Search Google Scholar
    • Export Citation

  • Graham, John R., Leary, Mark T., Roberts, Michael R. 2015. A century of capital structure: The leveraging of corporate America. Journal of Financial Economics, 118 (3), 658–683.

    • Search Google Scholar
    • Export Citation

  • Gulen, Huseyin, and Ion, Mihai. 2016. Policy uncertainty and corporate investment. Review of Financial Studies, 29 (3), 523–64.

  • Gungoraydinoglu, Ali, and Öztekin, Özde. 2011. Firm- and country-level determinants of corporate leverage: Some new international evidence. Journal of Corporate Finance, 17 (5), 1457–1474.

    • Search Google Scholar
    • Export Citation

  • Hadlock, Charles J., and Pierce, Joshua R. 2010. New Evidence on Measuring Financial Constraints: Moving Beyond the KZ Index. Review of Financial Studies 23 (5), 1909–1940.

    • Search Google Scholar
    • Export Citation

  • Handley, Kyle, and Limão, Nuno. 2015. Trade and investment under policy uncertainty: theory and firm evidence. American Economic Journal: Economic Policy, 7, 189–222.

    • Search Google Scholar
    • Export Citation

  • Harris, Milton, and Raviv, Artur. 1991. The Theory of Capital Structure. The Journal of Finance, 46: 297–355.

  • Hébert, Benjamin, and Schreger, Jesse. 2017. The Costs of Sovereign Default: Evidence from Argentina. American Economic Review, 107 (10): 3119–45

    • Search Google Scholar
    • Export Citation

  • Herrala, Risto, and Turk-Ariss, Rima. 2016. Capital accumulation in a politically unstable region, Journal of International Money and Finance, 64, 1–15.

    • Search Google Scholar
    • Export Citation

  • Horn, Sebastian, Reinhart, Carmen M., and Trebesch, Christoph. 2020. Coping with Disaster: Two Centuries of International Official Lending. CEPR Discussion Paper 14902.

    • Search Google Scholar
    • Export Citation

  • Hsieh, Chang-Tai, and Klenow, Peter J. 2009. Misallocation and Manufacturing TFP in China and India. The Quarterly Journal of Economics, 124 (4), 1403–1448.

    • Search Google Scholar
    • Export Citation

  • Independent Evaluation Office of the International Monetary Fund. 2021. Growth and Adjustment in IMF-supported Programs: Evaluation Report 2021. International Monetary Fund, Washington D.C.

    • Search Google Scholar
    • Export Citation

  • Jordà, Òscar. 2005. Estimation and Inference of Impulse Responses by Local Projections. American Economic Review, 95 (1): 161–182.

  • Jordà, Òscar, and Taylor, Alan M. 2016. The Time for Austerity: Estimating the Average Treatment Effect of Fiscal Policy. The Economic Journal, 126: 219–255.

    • Search Google Scholar
    • Export Citation

  • Julio, Brandon, and Yook, Youngsuk. 2014. Political uncertainty and corporate investment cycles. The Journal of Finance, 67 (1), 45–83.

    • Search Google Scholar
    • Export Citation

  • Kang, Wensheng, Lee, Kiseok, and Ratti, Ronald. 2014. Economic policy uncertainty and firm-level investment. Journal of Macroeconomics, 39, 42–53.

    • Search Google Scholar
    • Export Citation

  • Kalemli-Özcan, §ebnem, Sørensen, Bent, Villegas-Sanchez, Carolina, Volosovych, Vadym, and Yesiltas, Sevcan. 2015. How to Construct Nationally Representative Firm Level data from the ORBIS Global Database. NBER Working paper series, No. 21558.

    • Search Google Scholar
    • Export Citation

  • Kalemli-Özcan, §ebnem, Laeven, Luc, and Moreno, David. 2019. Debt overhang, rollover risk, and corporate investment: evidence from the European crisis. Working Paper Series 2241, European Central Bank.

    • Search Google Scholar
    • Export Citation

  • Kayo, Eduardo K., and Kimura, Herbert. 2011. Hierarchical determinants of capital structure. Journal of Banking & Finance, 35 (2), 358–371.

    • Search Google Scholar
    • Export Citation

  • Kentikelenis, Alexander E., Stubbs, Thomas H., and King, Lawrence P. 2016. IMF Conditionality and Development Policy Space, 1985–2014. Review of International Political Economy, 23(4): 543–82.

    • Search Google Scholar
    • Export Citation

  • Krahnke, Tobias. 2020. Doing more with less: the catalytic function of IMF lending and the role of program size. Bundesbank Discussion Paper No 18/2020.

    • Search Google Scholar
    • Export Citation

  • Kroszner, Randall S., Laeven, Luc, and Klingebiel, Daniela. 2007. Banking crises, financial dependence, and growth. Journal of Financial Economics, 84(1), 187–228.

    • Search Google Scholar
    • Export Citation

  • Lang, Valentin. 2021. The Economics of the Democratic Deficit: IMF Programs and Inequality. The Review of International Organizations, 16: 599:623.

    • Search Google Scholar
    • Export Citation

  • Leahy, John V., and Whited, Toni M. 1996. The Effect of Uncertainty on Investment: Some Stylized Facts. Journal of Money, Credit and Banking, 28 (1), 64–83.

    • Search Google Scholar
    • Export Citation

  • Lemmon, Michael L., Roberts, Michael R. and Zender, Jaime F. 2008. Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure. The Journal of Finance, 63: 1575–1608.

    • Search Google Scholar
    • Export Citation

  • Marchesi, Silvia and Thomas, Jonathan. 1999. IMF conditionality as a screening device. The Economic Journal, 109: 111–125.

  • Marchesi, Silvia. 2003. Adoption of an IMF Programme and Debt Rescheduling. An empirical analysis. Journal of Development Economics, 70: 403–423.

    • Search Google Scholar
    • Export Citation

  • Marchesi Silvia, and Sirtori, Emanuela. 2011. Is two better than one? Effects on growth of Bank-Fund interaction. Review of International Organization, 6, 287–306.

    • Search Google Scholar
    • Export Citation

  • Marchesi, Silvia, Masi, Tania, and Paul, Saumik. 2021. Aid projects and firm performance. IZA Discussion Paper 14705.

  • Mendoza, Enrique G., Yue, Vivian Z., 2012. A General Equilibrium Model of Sovereign Default and Business Cycles. Quarterly Journal of Economics, 127 (2): 889–946

    • Search Google Scholar
    • Export Citation

  • Mody, Ashoka, and Saravia, Diego. 2003. Catalyzing capital flows: do IMF-supported programs work as commitment devices? IMF Working paper No. 100.

    • Search Google Scholar
    • Export Citation

  • Montiel Olea, José Luis, and Plagborg-Møller, Mikkel. 2021. Local Projection Inference Is Simpler and More Robust Than You Think. Econometrica, 89: 1789–1823.

    • Search Google Scholar
    • Export Citation

  • Morris, Stephen, and Shin, Hyun Shong. 2006. Catalytic Finance: When Does It Work? Journal of International Economics, 70(1): 161—77.

  • Myers, Stewart C. 1984. The capital structure puzzle. Journal of Finance, 39, 575–592.

  • Myers, Stewart C., and Majluf, Nicholas S. 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics, 13, 187–221.

    • Search Google Scholar
    • Export Citation

  • Öztekin, Özde. 2015. Capital Structure Decisions around the World: Which Factors Are Reliably Important? Journal of Financial and Quantitative Analysis, 50 (3), 301–323.

    • Search Google Scholar
    • Export Citation

  • Pagano, Marco, Pica, Giovanni, Beck, Thorsten, and Kramarz, Francis. 2012. Finance and employment. Economic Policy, 27 (69), 5–55.

  • Panousi, Vasia, and Papanikolaou, Dimitris. 2012. Investment, Idiosyncratic Risk, and Ownership. The Journal of Finance, 67: 1113–1148.

    • Search Google Scholar
    • Export Citation

  • Papi, Luca, Presbitero, Andrea, and Zazzaro, Alberto. 2015. IMF Lending and Banking Crises. IMF Working Papers 63(3): 644—91.

  • Plagborg-Møller, Mikkel, and Wolf, Christian K. 2021. Local Projections and VARs Estimate the Same Impulse Responses. Econometrica, 89: 955–980

    • Search Google Scholar
    • Export Citation

  • Przeworkski, Adam, and Vreeland, James Raymond. 2000. The Effects of IMF Programs on Economic Growth. Journal of Development Economics, 62: 385–421

    • Search Google Scholar
    • Export Citation

  • Ramey, Valerie A., and Zubairy, Sarah. 2014. Government Spending Multipliers in Good Times and in Bad: Evidence from US Historical Data. Journal of Political Economy, 126 (2), 850–901.

    • Search Google Scholar
    • Export Citation

  • Rajan, Raghuram G., and Zingales, Luigi. 1998. Financial Dependence and Growth. The American Economic Review, 88 (3), 559–586.

  • Schnabl, Philipp. 2012. The international transmission of Bank liquidity shocks: evidence from an emerging market. Journal of Finance, 67: 897—932.

    • Search Google Scholar
    • Export Citation

  • Steinwand, Martin, and Stone, Randall W. 2008. The International Monetary Fund: A Review of the Recent Evidence. The Review of International Organizations, 3(2): 123—49.

    • Search Google Scholar
    • Export Citation

  • Temple, Paul, Urga, Giovanni, and Driver, Ciaran. 2001. The Influence of Uncertainty on Investment in the UK: A Macro or Micro Phenomenon? Scottish Journal of Political Economy, 48 (4): 361–3

    • Search Google Scholar
    • Export Citation

  • Tirole, Jean. 2002. Financial Crises, Liquidity, and the International Monetary System. Princeton University Press.

  • Titman, Sheridan., and Wessels, Roberto. 1988. The Determinants of Capital Structure Choice. The Journal of Finance, 43: 1–19.

  • Zettelmeyer, Jeromin. 2000. Can Official Crisis Lending be Counterproductive in the Short Run? Economic Notes, 29 (1), 13–29.

  • Zwart, Sanne. 2007. The mixed blessing of IMF intervention: Signalling versus liquidity support. Journal of Financial Stability, 3 (2), 149–174.

    • Search Google Scholar
    • Export Citation

Appendix

A. Sample and Descriptive Statistics

Table A1.

Variable Descriptions and Sources

article image
article image
Table A2.

Program Completion Status

article image
Notes: Tabulation of programs and their final review status as of 2020. For each program type, indicates the number of programs that were completed, offtrack, partially completed, or ongoing, as well as total number of unique programs. Offtrack is defined as programs that failed to complete more than two reviews, partially entails the completion of more than two but less than the total number of expected final reviews (IMF 2018 Review of Conditionality, 2019). Others refers to precautionary and non-disbursing programs which are not considered in the sample.
Table A3.

Yearly Average Firm Investment by Primary NACE Sector

article image
Notes: Table shows the year-sector firm average for investment for the full set of countries. Sectors are the NACE Rev. 2 main sections, excluding Financial, Public administration and defense, Real estate activities, Administrative and support services, Human health and social work, Other service activities, Activities of the household, and Extraterritorial sections.
Table A4.

Panel Summary

article image
Notes: Number of observations and unique firms available per country of Orbis tangible fixed asset investment data.
Table A5.

EFD by Sector

article image
Eppinger and Neugabauer (2022) EFD indices computed from Compustat according to RZ (1998) methodology.
Figure A1.
Figure A1.

Investment Growth by Country

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Average firm tangible fixed asset investment by country across sample years, considering all firms and sectors. Ranked by investment.
Figure A2.
Figure A2.

Treatment Status, by Program Type

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Treatment status by year for countries in sample. Shaded bars indicate a country is under a given program for a specific year; red for GRA, blue for PRGT. Grey bars indicate no program, while white bars indicate missing years for the dependent variable (tangible fixed assets investments) due to Orbis missing data. Effective treatment status of observations therefore defined by years for which there exists Orbis data for at least one firm for a given country.

B. Additional Tables and Augmented Inverse Propensity Score Weighted

Table B1.

AIPW First Stages

article image
Notes: The model uses predictors listed in Table A1 in the first stage and region dummies as fixed effect. T- statistics in parenthesis, standard errors clustered at the country level. * p < 0.1, ** p < 0.05, *** p < 0.01.
Table B2.

AIPW Estimates, Second Stage

article image
article image
Notes: Control coefficient estimates for second stage regression in AIPW estimates, baseline model. Standard errors clustered at the country-sector level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Table B3.

AIPW Estimates for State-Contingent Local Projections

article image
article image
Notes: AIPW estimators under flexible local projections for each time horizon h=1,2,3,4,5 by firm age categories, where younger firms are those below the sample median value of firm age and mature firms are those above the sample median. Standard errors clustered at the country level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Figure B1.
Figure B1.

Balance Tests for Propensity Scores

Citation: IMF Working Papers 2022, 157; 10.5089/9798400216800.001.A001

Notes: Plots show the estimated propensity scores for different outcome levels in the first stage multinomial logit model, where untreated is the base value “no program”, and treated is either GRA or PRGT.

C. Alternative Specification

Table C1.

Anticipation Effects

article image
Notes: Change in firm tangibles/TA investment rate in the h years leading up to program approval, with h=1,2,3,4,5. Model is a fixed effects regression with baseline controls, firm and sector-year fixed effects. Standard errors clustered at the country level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Table C2.

End-of-Program Effects by Completion Status

article image
Notes: AIPW average treatment effect of a program end, by completion status, for each time horizon h=1,2,3,4,5. Standard errors clustered at the country level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Table C3.

AIPW Robustness Tests

article image
Notes: AIPW estimators for each time horizon h=1,2,3,4,5 under different conditions. Spell length restricts the sample to firms with a series of yearly observations spanning at least 5 years to cover the full projection horizon. No offtrack drops programs from the treatment dummy that were classified as off track. No advanced drops countries from the 2010 European Union sovereign debt crisis that required IMF intervention. Standard errors clustered at the country-sector level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
Table C4.

AIPW Dropping Regions

article image
article image
Notes: AIPW average treatment effects for each time horizon h=1,2,3,4,5 when region m is dropped. Regions correspond to IMF Regional Department groups. Standard errors clustered at the country-sector level, T-statistics in parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01.
2

For example, the IMF has introduced a set of measures aimed to help developing economies tackling both liquidity (e.g., the Short-term Liquidity Line, or SLL) and solvency problems caused by the pandemic (e.g., the Catastrophe Containment and Relief Trust, or CCRT). Most importantly, the new issuance of $650 billions of new IMF special drawing rights (SDRs) should boost emerging economies’ balance sheets. At the same time, the IMF, together with the World Bank, urged G20 countries to establish the DSSI, a form of debt relief that eases financing constraints by deferring debt service repayments.

3

In 2010 The Economist Schumpeter columnist wrote on the private sector investment response to an IMF SBA arrangement in Jamaica that had recently been approved, reporting how Jamaican business owners were optimistic and expanding in light of the new program (The Economist, May 14th 2010. “Island Stories”)

4

Dreher and Lohman (2015) were among the first to apply a macro-micro approach to evaluate the effectiveness of official capital flows. Using night-time light intensity, evaluate the effects of World Bank aid on development. Similarly, Marchesi et al. (2021) use survey data confront Chinese and World Bank project aid effects on firm sales. Bluhm et al. (2020) explore the equality inducing effects of Chinse infrastructure investments. Chauvet and Ehrhart (2018) use survey data to evaluate the effects of multilateral and bilateral aid flows on firm sales, finding a positive effect which manifest through the alleviation of an infrastructural constraint as well as a financing constraint.

5

Lending instruments under the GRA include the Extended Fund Facility (EFF) for medium-term support to address protracted balance of payments problems, the Stand-By Arrangement (SBA) for short-term or potential balance of payments problems and, for members with already strong policies, and the Flexible Credit Line (FCL) and the Precautionary and Liquidity Line (PLL) to help prevent or mitigate crises and boost market confidence during periods of heightened risks. As for the concessional PRGT, two lending facilities are considered; (i) the Extended Credit Facility (ECF) for sustained medium- to long-term engagement in case of protracted balance of payments problems and (ii) the Standby Credit Facility (SCF) to address short-term balance of payments and adjustment needs caused by domestic or external shocks, or policy slippages.

6

Amounts drawn under an EFF are to be repaid over 4½–10 years in 12 equal semiannual installments, whereas credit provided under an SBA are repaid over 3¼–5 years.

7

Repayments under the ECF carries a grace period of 5½ years and a final maturity of 10 years, whereas the SCF has a grace period of 4 years and a final maturity of 8 years.

8

We follow the IMF Independent Evaluation Office (IEO, 2021) strategy for coding program start years.

9

In the case of blended resources, we code them as PRGT programs.

10

For example, most private U.S companies are not required to disclose financial information to the public after incorporation.

11

We drop firms with a main NACE Rev. 2 category of Financial, Public administration and defense, Real estate activities, Administrative and support services, Human health and social work, Other service activities, Activities of the household, and Extraterritorial.

12

With this approach, we do not need to rely on exclusion restrictions. Even if all our variables were endogenous, if there is no unexplained deviation from the conditional forecasted change in ratings, the ATE will be unbiased (Jordà and Taylor 2016).

13

A notable difference with the reference paper is that our dependent varies based on the program type, as opposed to considering all programs together. The result is that country fixed effects would be collinear with the outcome in certain groups, when a country for example never has a PRGT arrangement. The solution is to proxy with region fixed effects.

14

See for example Przeworski and Vreeland (2000), Dreher and Vaubel (2004), and Sturm (2005).

15

Jordà and Taylor (2016) show that their AIPW estimator has properties such that extreme values of the propensity scores are offset by the adjustment term, in contrast to a standard IPW estimator.

16

EFFs may be approved for periods of 4 years to implement deep and sustained structural reforms.

17

Work by Gilchrist et al. (2014) shows the role of firm financial frictions in investment dynamics in the context of uncertainty shocks, pointing to the relevance of distortions in financial markets.

18

See Eppinger and Neugabauer (2021) Appendix A for a detailed methodology on the construction of the index.

19

We also consider an alternative uncertainty indicator from Alfaro, Bloom, and Lin (2021), which is the 365-day implied volatility build from a mix of put and call-at-the-money option. We find results which mirror these results but are statistically weaker.

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Working Papers
Do IMF Programs Stimulate Private Sector Investment?
Author:
Pietro Bomprezzi
,
Silvia Marchesi
, and
Ms. Rima A Turk
Volume/Issue:
Volume 2022: Issue 157
Publisher:
International Monetary Fund
ISBN:
9798400216800
ISSN:
1018-5941
Pages:
50
DOI:
https://doi.org/10.5089/9798400216800.001
Figure 1.

Distribution of IMF Programs per Year
Figure 2.

IMF Programs and Firms
Figure 3.

Average Investment by Firm Age
Figure 4.

Program Approval and Firm Investment Response, AIPW Estimates
Figure 5.

AIPW and Firm Age
Figure A1.

Investment Growth by Country
Figure A2.

Treatment Status, by Program Type
Figure B1.

Balance Tests for Propensity Scores