## wpiea2022157-print-pdf

## Source details

**Canonical URL:** [wpiea2022157-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022157-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022157-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022157-print-pdf.pdf.json)

---

### Introduction and research question
- Context and motivation:
  - Recent headwinds—pandemics, supply crises, geopolitical tensions—have increased demand for multilateral lenders such as the IMF.
  - IMF resources have been tapped for systemic debt crises in advanced economies and to support developing and fragile economies.
  - IMF programs provide corrective actions and a "seal of approval" that can mitigate crises and boost market confidence.
- Research question and contribution:
  - Does the approval of an IMF program influence domestic private firm tangible fixed asset investment by reducing policy uncertainty (a private investment transmission channel distinct from external catalytic finance)?
  - Paper investigates whether IMF program approval makes "internal" investors more willing to invest by reducing uncertainty.
- Theoretical framing:
  - Two macro channels in IMF literature: signaling/catalytic finance and liquidity effects of IMF credit injections.
  - This paper studies whether signaling reduces domestic policy uncertainty and thereby induces non-reversible private investment.

### Data sources and sample
- Firm-level data:
  - Orbis (Bureau Van Dijk) balance-sheet data on tangible fixed investments and firm characteristics.
  - Sample: unbalanced panel of 43,949 firms for 69 countries from 2000 to 2019.
- Program identification:
  - MONA database between 2002 and 2019 (pre COVID-19).
  - Distinguishes GRA (non-concessional) and PRGT (concessional) financing; in blended cases coded as PRGT.
  - Treatment indicator = 1 if a country approved an IMF-supported program during the year as reported in MONA, but no later than October; otherwise coded in subsequent year.
- Key descriptive statistics (winsorized sample):
  - Dependent variable (Tangibles over total assets): Observations 277,572; Mean 31.08; Sd 27.25; Max 100; Min 0.
  - Real PC GDP growth: Observations 277,818; Mean 3.25; Sd 4.28; Max 81.79; Min -29.
  - Log real PC GDP: Observations 277,780; Mean 9.91; Sd 0.51; Max 11.37; Min 6.63.
  - Log Total Assets (firm control): Observations 277,816; Mean 15.89; Sd 1.94; Max 35.73; Min .693.
  - Leverage: Observations 277,816; Mean 19.53; Sd 21.91; Max 100.00; Min 0.
  - Cash flow to TA: Observations 192,994; Mean 8.09; Sd 11.15; Max 60.96; Min -28.2.
- Coverage and cleaning notes:
  - Orbis coverage varies; limited in Africa ("more than 90 percent of companies have no financial information in the Orbis database").
  - Exclusions: financial firms, government sector firms, selected service activities; use consolidated statements when available; winsorize variables to target kurtosis ~10.

### Methodology and identification
- Estimation strategy:
  - Local projections (Jordà 2005) to estimate impulse responses of firm investment up to five years (h = 1..5) after program approval.
  - Two-stage Augmented Inverse Propensity Weighted (AIPW) estimator (Jordà and Taylor 2016) to address endogeneity: multinomial logit first stage predicting probabilities for outcomes (no program, GRA, PRGT), second-stage weighted local projections; estimator is doubly robust.
- First-stage predictors Z(t−1,i) include past program dummy, autocracy measure, investment to GDP ratio, log real GDP per capita (PPP), total debt service, budget balance, reserves to imports, real GDP growth, changes in reserves, current account to GDP, election year dummy, log checks-and-balances, and region fixed effects.
- Outcome variable:
  - Cumulative change in firm tangible fixed assets scaled by total assets (Δy_{i,j,k,t+h}) for horizons h = 1,2,...,5.
- Fixed effects and clustering:
  - Firm fixed effects μ_i and sector-year fixed effects τ_{k,t}; standard errors clustered at country level (or country-sector as noted).

### Main empirical findings — baseline AIPW local projections
- Baseline cumulative AIPW estimates for tangibles/total assets (N per horizon shown):
  - GRA (N per horizon: 21,643; 19,002; 16,516; 14,337; 12,608)
    - h=1: AIPW -0.123 (t-stat -1.48)
    - h=2: AIPW 1.096* (t-stat 2.78)
    - h=3: AIPW 2.036** (t-stat 3.21)
    - h=4: AIPW 3.986** (t-stat 3.72)
    - h=5: AIPW 3.260* (t-stat 2.34)
  - PRGT (N per horizon: 21,643; 19,002; 16,516; 14,337; 12,608)
    - h=1: AIPW 1.019* (t-stat 2.09)
    - h=2: AIPW 0.989 (t-stat 1.00)
    - h=3: AIPW -0.0792 (t-stat -0.06)
    - h=4: AIPW -1.254 (t-stat -0.79)
    - h=5: AIPW 0.359 (t-stat 0.21)
- Interpretation:
  - GRA programs: delayed and growing positive effect, peaking around four years after approval with a cumulative increase of almost 4 percentage points in tangibles/total assets relative to the approval year.
  - PRGT programs: weak and temporary effects; marginal increase in year 1 (~1 percentage point) but no sustained investment growth thereafter.

### Heterogeneity: firm age, external financial dependence, and uncertainty
- Firm age (AIPW average treatment effects; Figure 5 described):
  - 퐼 = 1 for firms above median age; "young" defined as below median.
  - Younger, more financially constrained firms benefit more from IMF arrangements for both PRGT and GRA.
  - PRGT: significant but short-lived effect for younger firms; positive effect ~2 percentage points for first three years after approval.
  - GRA: five years after approval, younger firms show a positive 10 percentage point cumulative change in tangibles above the reference year; mature firms show smaller positive effects.
- External Financial Dependence (Rajan and Zingales (RZ) index; stacked DiD results, Panel A Table 3):
  - No pre-treatment trends detected.
  - GRA programs: firms in sectors with higher external financial dependence experience larger investment growth up to 5 years after approval.
    - GRA coefficients by years since treatment (Panel A): -2: 0.186 (t-stat 0.87); -1: 0.275 (1.41); 1: 0.427* (2.08); 2: 0.501** (2.73); 3: 0.543** (2.98); 4: 0.591** (3.19); 5: 0.630** (3.29). N = 34,416 each year.
  - PRGT programs: smaller initial effects but a long-run significant effect; PRGT coefficients (Panel A): -2: 0.0869 (0.84); -1: -0.104 (-1.14); 1: 0.0809 (1.48); 2: 0.168* (2.26); 3: 0.164 (1.50); 4: 0.189* (1.81); 5: 0.310** (2.43). N = 75,835 each year.
- Realized volatility / firm-level uncertainty (Panel B Table 3):
  - Uncertainty proxy: annualized 12-month standard deviation of daily CRSP returns, aggregated to median sector-year.
  - GRA programs: greater sector-wide volatility under a GRA program associated with ~3 percentage point increase in tangibles after three years relative to base year.
    - GRA coefficients (Panel B): -2: 1.756 (t-stat 1.37); -1: 1.548 (1.53); 1: 1.571 (1.55); 2: 1.744* (1.71); 3: 1.835 (1.63); 4: 1.961* (1.75); 5: 2.334* (1.89). N = 77,554 each year.
  - PRGT programs: no consistent long-term effect using realized volatility; PRGT coefficients (Panel B): -2: 0.153 (1.38); -1: 0.189* (2.10); 1: -0.121 (-0.95); 2: 0.260 (0.97); 3: 0.194* (1.84); 4: 0.150 (1.29); 5: 0.0863 (0.35). N = 181,005 each year.
  - Alternative implied volatility indicator yields similar but weaker statistical results.

### Ownership and "never-leavers"
- Ownership indicator constructed from Orbis historical vintages (GUO and global ultimate consolidated owner ≥50%).
- "Never-leavers": firms that do not switch ownership immediately before and after program approval; interpreted as committed domestic firms.
- Stacked DiD dynamic estimates (Table 4):
  - GRA interaction with never-leavers (years since event): -2: 0.341 (t-stat 0.41); -1: 0.509 (0.68); 1: 1.341** (2.41); 2: 1.332** (2.38); 3: 1.091* (1.83); 4: 1.075* (1.73); 5: 1.110* (1.76). N = 66,845 each year.
  - PRGT never-leaver coefficients mirror Panel A PRGT row: five-year effect ~0.310** (t-stat 2.43). N = 75,835 each year.
- Magnitudes:
  - PRGT: ~0.3 percentage point increase in tangibles after five years for never-leavers.
  - GRA: ~1 percentage point increase after five years for never-leavers.
- Interpretation:
  - GRA countries may host multinationals that can retreat in adverse shocks, allowing committed domestic firms to expand post-signaling.
  - PRGT never-leavers may be firms unable to leave due to constraints; small positive effects suggest benefit from remaining through program years.

### Robustness and additional tests
- Anticipation effects (Table C1):
  - No systematic anticipation effects for GRA and PRGT; some weak evidence investment falls two or three years before approval but not closer to signing.
- Persistence after program end (Table C2; AIPW with final program year as treatment; exclude successor programs within five years):
  - GRA completed programs: tangibles/total assets grew by 2.57 percentage points after 5 years vs base year.
  - GRA off-track programs: effect 3.16 percentage points at same horizon; differences not significantly different.
  - PRGT completed programs: mixed—investment growth drops in first years after program end; when program goes off track, negative effects persist (three years after approval, tangibles 4.21 percentage points lower than reference year).
- Sample and coverage checks:
  - Restricting to firms with full 5-year balance-sheet spells yields consistent results.
  - Dropping off-track programs produces results similar to baseline, with some PRGT estimates slightly less significant.
  - Region-dropping exercises:
    - GRA results robust to dropping regions; e.g., removing Sub-Saharan Africa still shows up to ~4 percentage points increase four years after approval.
    - Dropping Europe weakens GRA significance (sample shrinkage and European Debt Crises effects).
    - PRGT AIPW estimates weaker when dropping regions like Sub-Saharan Africa or Middle East and Central Asia because PRGT programs concentrated there.

### Additional empirical details (selected tables and findings)
- First-stage predictors (AIPW first stages, Table B1; observations = 806):
  - Past program: 2.195*** (GRA), 2.053*** (PRGT).
  - GFCF/GDP: -0.097*** (GRA).
  - Total debt service to GNI: -0.145** (PRGT).
  - Budget surplus: 0.138*** (PRGT).
  - Real GDP growth: -0.097*** (PRGT).
  - Total reserves/imports: -0.112* (GRA), -0.175* (PRGT).
- Heterogeneous second-stage patterns (Table B3 examples):
  - GRA, younger firms (horizons h=1..5): -0.227, 1.251, 3.513*, 9.155***, 10.22***.
  - GRA, mature firms: -0.249, 1.136**, 1.817**, 2.203*, 1.442.
  - High realized volatility under GRA: 0.349, 1.755, 1.586**, 3.236***, 3.206***.
  - Low realized volatility under GRA: -0.947**, -0.644, 1.305***, -0.437, -2.293.
- Program completion counts (Table A2):
  - PRGT: Completed 23; Off track 6; Ongoing 8; Total 37.
  - GRA: Completed 23; Off track 10; Ongoing 4; Partially completed 42; Total 79.
  - Total across program types: 75 Completed; 18 Off track; 8 Ongoing; 54 Partially; 155 Total.

### Conclusion and policy implications
- Distinguishing GRA and PRGT financing matters for the path of firm investment following program approval.
  - GRA programs associated with stronger, delayed, and more persistent increases in firm-level tangible fixed asset investment.
  - PRGT programs show weaker, more heterogeneous, and often short-lived effects.
- Two financial channels documented:
  - Degree of firms’ external financial dependence (sectors more dependent on external finance benefit more, especially under GRA).
  - Firms’ sectoral uncertainty (higher realized volatility sectors see larger GRA-driven investment increases).
- Domestic ownership matters: private investments stimulated more for domestic ("never-leaver") firms, especially under GRA.
- Interpretation:
  - IMF programs may reduce domestic policy uncertainty and thereby encourage internal investors to undertake non-reversible long-term investments.
  - Program design (GRA vs PRGT), firm characteristics (age, external finance dependence, uncertainty exposure), program completion status, and regional context influence the private investment transmission channel.
- Policy relevance:
  - Monitoring program implementation and addressing liquidity/credit constraints, along with consideration of firm heterogeneity and program design, could improve private-sector investment outcomes following IMF-supported programs.

*Source: IMF Working Paper — Introduction, Data, Methods, Empirical Results, and Conclusion sections from wpiea2022157-print-pdf.*

### Introduction ...........................................................................................................

### Introduction

### Context and motivation
- Recent economic headwinds—pandemics, supply crises, geopolitical tensions—have reinvigorated the role of multilateral lenders such as the International Monetary Fund (IMF).
- IMF resources have been tapped to deal with systemic debt crises in advanced economies and to revive its role among developing and fragile economies.
- Official lending has historically been larger than commonly thought, and has been growing as a share of total cross-border capital flows (Horn et al. (2020) referenced).
- The IMF’s traditional financial support aims to create breathing room for countries hit by crises as they implement adjustment policies to restore macroeconomic stability and growth.
- Corrective actions associated with programs provide a "seal of approval" that appropriate policies are adopted, helping mitigate crises and boosting market confidence during periods of heightened risks.

### Research question and contribution
- The paper investigates how the approval of an IMF program influences private sector dynamics, specifically private firm tangible fixed asset investment.
- Key question: Does the signaling or "seal of good faith" effect of an IMF program operate within the confines of the country to stimulate domestic private investment by reducing policy uncertainty?
- The paper aims to quantify whether IMF programs make internal investors more willing to invest in their own country by reducing the impact of uncertainty — a private investment transmission channel distinct from better-known external catalytic finance effects.

### Theoretical framing
- Two macro channels in the IMF literature:
  - Signaling/catalytic finance ("seal of good faith") that attracts foreign capital or creditors back to the country (Corsetti et al. 2006; Marchesi and Thomas 1999; Marchesi 2003; Morris and Shin 2006; Zwart 2007).
  - Liquidity effects of IMF credit injections that can reduce the probability of self-fulfilling runs from illiquidity problems (Boockmann and Dreher 2003; Dreher 2006; Dreher and Vaubel 2004; Zettelmeyer 2000).
- This paper considers whether signaling also affects domestic private firms by strengthening policy frameworks and institutions, thereby triggering a policy reduction mechanism that encourages positive net present value capital investments.

### Uncertainty and firm investment literature
- Growing literature on the impact of uncertainty on firms: 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.
- Mixed theoretical predictions: some early studies argue uncertainty may stimulate investment (Bar-Ilan and Strange, 1996; Abel, 1983), but more recent research predicts uncertainty inhibits corporate investment (Gulen and Ion 2016 and Rao et al. 2017).
- Tangible fixed assets investments tend to be non-reversible, so firms may delay long-term decisions under policy uncertainty until uncertainty subsides (which may occur following IMF program approval).

### Data and methodology overview

### Data sources and focus
- The paper uses detailed firm-level balance sheet data on tangible fixed investments and other firm characteristics.
- Identification of IMF programs and matching to firm-level observations is a central data task.
- The paper distinguishes between types of IMF financing, specifically noting differences between GRA and PRGT financing.

### Empirical approach
- Uses a local projection methodology to estimate the dynamic response of firm investments to the approval of an IMF arrangement.
- Employs a difference-in-differences approach (including a stacked DiD variant) to exploit firm-level heterogeneity and identify mechanisms.
- Leverages firm-level characteristics (external financial dependence, sectoral uncertainty, ownership) to test channels.

### Key empirical findings (summary)

- Distinguishing between GRA and PRGT financing matters for the path of firm investment following program approval.
- Two firm-level financial channels are documented:
  - Degree of firms’ external financial dependence affects investment responses.
  - Firms’ sectoral uncertainty affects investment responses.
- Following an IMF program approval, firms that rely more on external finance or are more subject to uncertainty experience less binding financial frictions and hence invest more.
- Using ownership data, private investments are stimulated more for domestic firms than for others.
- The presence of a private investment transmission channel helps explain factors that affect the success and effectiveness of IMF programs.
- To the authors’ knowledge, this is the first paper that investigates whether IMF programs, beyond improving a country's creditworthiness for external investors, also make "internal" investors more willing to invest by reducing uncertainty.

### Methods and identification highlights
- Local projections estimate dynamic firm investment responses to program approval.
- Stacked difference-in-differences approach used to exploit firm heterogeneity and to identify the roles of financial frictions and uncertainty.
- Augmented inverse propensity score weighting and robustness checks are employed (detailed tables and appendices referenced).

### Notable contextual details and examples
- IMF measures mentioned that target developing economies include the Short-term Liquidity Line (SLL) and the Catastrophe Containment and Relief Trust (CCRT).
- The IMF issued $650 billions of new IMF special drawing rights (SDRs) (as referenced in the source text).
- Example narrative: The Economist reported positive private sector expectations following an IMF SBA arrangement in Jamaica (May 14th 2010, "Island Stories").

### Organization of the paper
- Section 2: literature review on IMF financing effects and firm investment under policy uncertainty.
- Section 3: data description.
- Subsequent sections present methodology, empirical findings, robustness checks, and conclusions (detailed table of contents lists additional sections, figures, tables, and appendices).

*Source: IMF Working Paper — Introduction section (as provided).*

### Section 4 illustrates the methodology and identification strategy. Section 5 presents our main empirical results

### wpiea2022157-print-pdf

### Literature Review — Effects of IMF Programs
- Prior literature focuses on broad country-level outcomes; recent work emphasizes specific IMF policy objectives: monetary stability, debt management, and containment of external arrears (Kentikelenis, Stubbs, and King 2016).
- Empirical associations reported:
  - Reduced inflation and monetary growth.
  - Lower risk of currency crises and banking crises.
  - Improved market performance of banks (Dreher and Walter 2010; Papi et al. 2015; Steinwand and Stone 2008).
- Overall evidence: some positive adjustment effects on financial, fiscal, and monetary positions, but benefits generally fall short of expectations for GDP growth and debt reduction (IMF 2018; IEO 2021).
- Catalytic/signaling role of IMF programs debated:
  - Can reassure international investors that adequate policies are being implemented (Tirole 2002).
  - Concerns: preferred creditor status could deter foreign investors fearing penalization in debt restructuring (Mody and Saravia 2003); moral hazard if borrowing countries expect bailouts (Dreher 2006).
  - Mixed findings in literature on catalytic effects (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).
- Related micro-level literature:
  - Studies of official intervention at subnational or firm-level show heterogeneous effects (Bluhm et al. 2020; Bomprezzi and Marchesi 2021; Chauvet and Ehrhart 2018; Dreher and Lohman 2015; Dreher et al. 2021; Marchesi et al. 2021).
  - Bomprezzi and Marchesi (2021): IMF programs positively affect firms' sales growth and alleviate financing constraints; however, more severe conditionality can worsen firm performance short-term before longer-term benefits.
  - Broner et al. (2021): sovereign debt inflows reduce domestic interest rates, encourage banks to expand domestic credit, and benefit financially constrained firms.
  - Theoretical channels: sovereign risk propagation to firms; sovereign default fiscal channel (Mendoza and Yue 2012; Hébert and Schreger 2017; Andrade and Chhaochharia 2018; Corsetti et al. 2006; Gourinchas et al. 2017).
- Contribution of this paper:
  - Focuses on firm-level private sector investment responses to IMF programs.
  - Proposes a domestic signaling effect: reduction of domestic policy uncertainty accompanying IMF programs induces firms to increase private sector (non-reversible long-term) investment.

### Literature Review — Firm Investment Under Uncertainty
- Firm- and sector-specific determinants emphasized: size, profitability, asset tangibility, industry median leverage (Myers 1984; Myers and Majluf 1984; Titman and Wessels 1988; Harris and Raviv 1991; Booth 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).
- Country-specific macro and institutional factors also important (Borio 1990; Rajan and Zingales 1998; Kayo and Kimura 2011; Cevik and Miryugin 2018); political instability matters (Herrala and Turk-Ariss 2016).
- Uncertainty typically reduces irreversible long-term investment:
  - Proxies: volatility of stock returns (Leahy and Whited 1996; Bloom et al. 2007; Baum et al. 2010; Bloom 2009; Panousi and Papanikolaou 2012).
  - Bloom et al. (2007): uncertainty reduces firms’ irreversible long-term investments in response to sales shocks.
  - Panousi and Papanikolaou (2012): firm-level idiosyncratic risk associates negatively with corporate investment.
- Interaction of micro and macro uncertainty:
  - Temple et al. (2001): both firm-level and aggregate uncertainty negatively impact firm investment.
  - Baum et al. (2010): market uncertainty inhibits firm-level investment; interaction with firm cash flow matters.
  - Kang et al. (2014): economic policy uncertainty depresses firms’ investment, especially for firms with higher firm-level uncertainty.
- Measurement advances and micro-level uncertainty proxies:
  - Jurado et al. (2015): critique of popular uncertainty proxies.
  - Baker et al. (2013): three-part index (news-based, future tax provisions, economic forecast); Gulen and Ion (2016) find news-based component most relevant.
  - Text-based measures are growing (Alexopoulos and Cohen 2015).
  - Alfaro, Bloom, and Lin (2021): firm-level uncertainty proxies — (i) realized stock return volatility (CRSP daily returns) and (ii) implied volatility from options; this paper employs the first as preferred firm-level uncertainty indicator.

### Data — Identifying IMF Programs
- Sample period: pre COVID-19, MONA database between 2002 and 2019.
- Focused lending instruments: distinction between GRA and PRGT financing.
  - GRA: non-concessional; available to all member countries.
  - PRGT: concessional; targeted to low-income countries.
- Rationale for distinguishing GRA vs PRGT:
  - GRA financing requires policy measures to be taken within the program period and balance of payments needs resolved by the time repurchases begin.
  - PRGT financing is tied to achieving or making progress toward a stable and sustainable macroeconomic position consistent with poverty reduction and growth, without necessarily resolving the balance of payments problem during the program period.
  - Repeated programs under PRGT can be expected for sustained engagement; GRA typically expects resolution by program end and no follow-up arrangement in principle.
- Examples of instruments under GRA and PRGT (as described in the source):
  - GRA instruments include the Extended Fund Facility (EFF), Stand-By Arrangement (SBA), Flexible Credit Line (FCL), and Precautionary and Liquidity Line (PLL).
  - PRGT instruments include the Extended Credit Facility (ECF) and the Standby Credit Facility (SCF).
- Repayment and maturity details (as stated):
  - EFF repayments: repaid over 4½–10 years in 12 equal semiannual installments.
  - SBA repayments: repaid over 3¼–5 years.
  - ECF: grace period of 5½ years and final maturity of 10 years.
  - SCF: grace period of 4 years and final maturity of 8 years.
- Treatment variable coding:
  - Indicator = 1 if a country approved an IMF-supported program during the year as reported in MONA, but no later than October.
  - If program approved after October, the subsequent year is coded as program approval year.
  - Treatment dummies for program start are created separately for GRA and PRGT; in blended resources cases, coded as PRGT programs.
  - This coding accounts for program effects that would not manifest if the program were adopted later in the year.
- Sample design and identification strategy rationale:
  - Sample contains countries under an IMF arrangement to mitigate selection endogeneity and reverse causality concerns.
  - Focus shifts to dynamics of investment following arrangement approval rather than selection into programs, and to heterogeneity among arrangements.
- Empirical observations:
  - Figure 1 (described): distribution of unique IMF programs per year by type — GRA arrangements constitute the bulk of programs; 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 the sample.

### Data — Firm Tangible Fixed Asset Investment and Balance Sheet Data
- Firm-level source: Orbis database (Bureau Van Dijk).
- Focus on tangible investments because of non-reversible, long-term nature (property, plant, equipment).
- Investment measure: ratio of tangible fixed assets to total assets to compute investments.
- Firm-level controls (following Kalemli-Özcan et al. 2019 specification):
  - Log of total assets (proxy for firm size).
  - Leverage = ratio of total debt to total assets (total debt = sum of long-term debt, loans, credits, other current liabilities).
  - Debt maturity = ratio of long-term debt to total debt (captures rollover risk).
  - Interest burden = ratio of interest expense to earnings before taxes (EBIT).
  - Sales growth (firm-level proxy for growth opportunities).
  - Cash flows scaled by total assets.
- Additional Orbis data used to construct controls:
  - Firm ownership, incorporation dates, geographic areas, and sectors of operation.
  - Sector-year fixed effects constructed to account for time-varying sector-specific heterogeneity.
- Appendix/Additional tables referenced:
  - Table A1 lists country-level controls and sources.
  - Table 1 provides full summary statistics for firm and country-level variables.
  - Table A3 referenced for further details.

*Source: IMF Working Paper (excerpt from wpiea2022157-print-pdf).*

### Appendix A reports the average tangible investment (computed as the change in the ratio tangible assets/ total

### wpiea2022157-print-pdf - 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.

### Summary statistics
- Sample: unbalanced panel of 43,949 firms for 69 countries from 2000 to 2019.
- Summary statistics run on winsorized sample.
- Dependent variable
  - Tangibles over total assets: Observations 277,572; Mean 31.08; Sd 27.25; Max 100; Min 0.
- Country controls
  - Real PC GDP growth: Observations 277,818; Mean 3.25; Sd 4.28; Max 81.79; Min -29.
  - Log real PC GDP: Observations 277,780; Mean 9.91; Sd 0.51; Max 11.37; Min 6.63.
  - Log claims by depository institutions: Observations 263,772; Mean 12.90; Sd 2.18; Max 20.12; Min 6.11.
  - Real lending rate: Observations 147,993; Mean 5.44; Sd 9.05; Max 93.92; Min -25.7.
  - Political Risk Rating: Observations 264,879; Mean 67.22; Sd 10.18; Max 92.50; Min 31.
- Firm controls
  - Log Total Assets: Observations 277,816; Mean 15.89; Sd 1.94; Max 35.73; Min .693.
  - Long-term to total debt: Observations 231,150; Mean 39.26; Sd 40.02; Max 100.00; Min 0.
  - Leverage: Observations 277,816; Mean 19.53; Sd 21.91; Max 100.00; Min 0.
  - Interest expense to EBIT: Observations 169,973; Mean 27.47; Sd 437.62; Max 10000.00; Min 0.
  - Cash flow to TA: Observations 192,994; Mean 8.09; Sd 11.15; Max 60.96; Min -28.2.
  - Sales growth: Observations 169,952; Mean 14.29; Sd 54.29; Max 582.72; Min -92.

### Data coverage and cleaning
- Orbis firm coverage varies by region and country; variation driven by differences in legal standards for filing financial documents.
- Orbis coverage particularly limited in Africa: "more than 90 percent of companies have no financial information in the Orbis database."
- Historical vintage approach adopted: download year-specific vintages and match firms over time using Orbis unique firm identifiers to improve national representativeness and mitigate re-weighting needs.
- Sample exclusions and cleaning:
  - Drop financial firms, government sector firms, and firms primarily in selected service activities (specific NACE Rev. 2 categories listed in the source).
  - Use consolidated financial statements when available, otherwise unconsolidated, to avoid double counting.
  - Remove erroneous balance sheet items (e.g., negative reported costs).
  - Winsorize or trim balance sheet variables so that their kurtosis falls to a value around 10 (as per Kalemli-Ozcan et al. (2019)).

### Methodology and identification strategy
- Hypothesis: entry into an IMF program signals a reduction in future macroeconomic uncertainty, inducing firms to undertake non-reversible investments; interest in dynamic response over up to five years.
- Estimation approach:
  - Use impulse response functions via local projections (Jordà 2005) to estimate dynamic responses.
  - Address endogeneity of IMF program approval via a two-stage Adjusted Inverse Propensity Weighted (AIPW) estimator (Jordà and Taylor 2016) combining propensity-score based rebalancing and local projections. The estimator is doubly robust.
- First stage (propensity score):
  - Multinomial logit predicting probabilities P(i,t,p) for outcomes: no program, GRA program, PRGT program.
  - Predictor set Z(t−1,i) includes past program dummy, autocracy measure, investment to GDP ratio, log real GDP per capita (PPP), total debt service, budget balance, reserves to imports, real GDP growth, changes in reserves, current account to GDP, election year dummy, log checks-and-balances, and region fixed effects.
  - Weights assigned: treatment group 1/P(i,t,p); control group 1/(1−P(i,t,p)).
- Second stage (outcome and local projections):
  - Outcome: cumulative change in firm tangible fixed assets scaled by total assets (Δy_{i,j,k,t+h}) for horizons h = 1,2,...,5.
  - Baseline local projection specification includes lagged firm controls Z_{i,j,k,t−1}, country-level controls X_{j,t−1}, treatment dummy D_{j,t}, firm fixed effects μ_i, and sector-year fixed effects τ_{k,t}. Standard errors clustered at country level.
  - AIPW estimator formula incorporates IPW term and adjustment term to stabilize extreme propensity scores and avoid truncation.

### Empirical findings — Local Projections baseline results
- Main AIPW results for cumulative change in tangibles/total assets at horizons h = 1..5 (Table 2):
  - GRA (N per horizon: 21643, 19002, 16516, 14337, 12608)
    - h=1: AIPW -0.123 (t-stat -1.48)
    - h=2: AIPW 1.096* (t-stat 2.78)
    - h=3: AIPW 2.036** (t-stat 3.21)
    - h=4: AIPW 3.986** (t-stat 3.72)
    - h=5: AIPW 3.260* (t-stat 2.34)
  - PRGT (N per horizon: 21643, 19002, 16516, 14337, 12608)
    - h=1: AIPW 1.019* (t-stat 2.09)
    - h=2: AIPW 0.989 (t-stat 1.00)
    - h=3: AIPW -0.0792 (t-stat -0.06)
    - h=4: AIPW -1.254 (t-stat -0.79)
    - h=5: AIPW 0.359 (t-stat 0.21)
- Interpretation:
  - GRA programs: effect increases over time, peaking at four years after program approval with a cumulative increase in tangible assets scaled by total assets of almost four percentage points relative to the approval year. No significant change in year 1.
  - PRGT programs: only a weak and temporary effect; marginal increase around one percentage point in year 1, but no evidence of sustained investment thereafter.
- Possible explanations:
  - Delayed GRA effect consistent with investor caution and time needed for confidence to materialize; GRA countries may host more mature firms with less initial room to grow.
  - PRGT contexts may feature poorer access to credit, lower institutional quality, and fewer cash-generating opportunities that dampen private investment despite signaling effects.
  - Program length differences: GRA arrangements typically up to four years; PRGT arrangements commonly last five years and may involve repeated engagements—repeated PRGT arrangements could provide longer-term confidence similar to GRA effects.

### Firm financial frictions and state-contingent responses
- Motivation: investigate heterogenous firm responses to IMF program approval via degree of financial frictions and perceived uncertainty.
- Strategy: state-contingent local projections with binning indicator I_{i,t−1}^j to estimate separate local projections for firms in different states (Equation 5).
- Proxy for financial frictions: firm age (young < 15 years; mature 15–34 years; well-established 35+ years) used as an ex-ante indicator of likely financial constraints, since direct measures (size, leverage, liquidity) respond endogenously to shocks.
- Hypothesis: younger (more financially constrained) firms are more sensitive to improvements in macroeconomic uncertainty; IMF programs that reduce systemic risk should disproportionately boost investment for these firms.

*Source: IMF Working Paper content unit wpiea2022157-print-pdf (Appendix A and associated methodological and empirical sections).*

### 5. The indicator variable 퐼

### 5. The indicator variable 퐼

### Firm age: AIPW average treatment effects
- 퐼 takes the value 1 for firms above the median age (age = years from incorporation); firms below median are "young".
- Figure 5 (AIPW average treatment effects of program signing on firm tangible fixed assets investment rate) finds:
  - For both PRGT and GRA programs, younger, more financially constrained firms benefit more from an IMF arrangement.
  - PRGT programs: a significant albeit short-lived effect on investment for younger firms; no effects for mature firms similar to baseline, but a positive effect of around 2 percentage points for the first three years following program approval.
  - GRA programs: five years after program approval, on average a positive 10 percentage point cumulative change in tangibles above the reference year value for younger firms; smaller positive effect for mature firms.
- Interpretation: younger firms in PRGT countries expand in an environment with greater institutional and financial frictions, but effects are short lived.

### Stacked Difference-in-Differences approach and RZ external financial dependence
- Motivation:
  - Firm age proxy for financial constraints may be endogenous (survivorship bias, M&A absorption).
  - Need for more granular firm heterogeneity analysis.
- Methodology:
  - Use Rajan and Zingales (1998) (RZ) external financial dependence (EFD) indicator with a stacked DiD (event-study) approach popularized by Cengiz et al. (2019) and Deshpande and Li (2019).
  - Stacked DiD: split data into n sub-experiments by calendar year of treatment; define treatment window and controls (observations with subsequent treatment at least k years ahead); stack sub-experiments to form independent panel event studies identified by sub-experiment id.
- RZ index:
  - Sector-specific, time-invariant measure: share of investment not financed by internal cash flow in the median listed U.S firm over the 1980s.
  - Assumptions: U.S capital market as proxy for frictionless market; industry technological fundamentals constant across countries.
  - Authors use indices computed by Eppinger and Neugabauer (2022) from Compustat for 1990-2005; external financial dependence defined as (capital expenditures minus cash flow from operations) divided by sum of capital expenditure over period, median by industry; merged with Orbis via NACE sectors.
  - Indices computed for a set of firms in manufacturing-oriented industries.
- Identification:
  - Treat program approval as shock reducing macroeconomic uncertainty and financial risk; response of private sector investment is contingent on firm reliance on external finance.
  - Stacked DiD model (Equation 5) includes country and firm controls, fixed effects, sector-year fixed effects; parameter of interest is 훾_j representing interaction between year-since-event indicators and firm index (EFD or uncertainty).

### Results — External Financial Dependence (Panel A, Table 3)
- No evidence of anticipation effect or pre-treatment trends.
- GRA programs:
  - Tangible investments grow disproportionately more for firms in sectors with high external financial dependence at every point up to 5 years after approval.
  - Examples across percentiles:
    - Industry at bottom 5th percentile (publishing and printing): expected effect after 1 year ≈ -0.17 percentage points.
    - Industry at top 95th percentile (communication equipment): effect after 1 year = 0.5 percentage points.
- PRGT programs:
  - Size of effect smaller than GRA, but in the long run there is a strong and significant effect, indicating more financially constrained firms in PRGT-eligible countries benefit from IMF arrangements.
- Table 3 (Panel A estimates for EFD interaction, years since treatment):
  - GRA program coefficients: -2: 0.186; -1: 0.275; 1: 0.427*; 2: 0.501**; 3: 0.543**; 4: 0.591**; 5: 0.630**.
    - Standard errors (t-statistics reported in parentheses): (0.87) (1.41) (2.08) (2.73) (2.98) (3.19) (3.29).
    - N = 34,416 for each year.
  - PRGT program coefficients: -2: 0.0869; -1: -0.104; 1: 0.0809; 2: 0.168*; 3: 0.164; 4: 0.189*; 5: 0.310**.
    - Standard errors (t-statistics): (0.84) (-1.14) (1.48) (2.26) (1.50) (1.81) (2.43).
    - N = 75,835 for each year.
- Notes: Panel A uses country fixed effects; standard errors clustered at the country level. * p < 0.1, ** p < 0.05, *** p < 0.01.

### Results — Realized Volatility / Firm-level uncertainty (Panel B, Table 3)
- Uncertainty measure: annualized 12-month standard deviation of daily CRSP returns (Alfaro, Bloom, and Lin (2021)), aggregated to median sector-year value and matched to firms at sector-year level; interpreted as exogenous industry-specific characteristic.
- Findings:
  - Stronger effects for realized volatility measure.
  - GRA programs:
    - Greater sector-wide volatility under a GRA program associated with a 3-percentage point increase in tangibles after three years relative to the base year.
  - PRGT programs:
    - No consistent long-term effect.
    - Interpreted as PRGT countries' institutional and macro-financial environment deterring firm investments; a single program insufficient to reverse investor tendencies.
  - Alternative uncertainty indicator (365-day implied volatility from options) yields mirror results but statistically weaker.
- Table 3 (Panel B estimates for realized volatility interaction, years since treatment):
  - GRA programs coefficients: -2: 1.756; -1: 1.548; 1: 1.571; 2: 1.744*; 3: 1.835; 4: 1.961*; 5: 2.334*.
    - Standard errors (t-statistics): (1.37) (1.53) (1.55) (1.71) (1.63) (1.75) (1.89).
    - N = 77,554 for each year.
  - PRGT programs coefficients: -2: 0.153; -1: 0.189*; 1: -0.121; 2: 0.260; 3: 0.194*; 4: 0.150; 5: 0.0863.
    - Standard errors (t-statistics): (1.38) (2.10) (-0.95) (0.97) (1.84) (1.29) (0.35).
    - N = 181,005 for each year.
- Notes: Panel B uses firm fixed effects; standard errors clustered at the country level. * p < 0.1, ** p < 0.05, *** p < 0.01.

### Domestic ownership and “never-leavers” (Table 4)
- Construct ownership indicator using Orbis historical vintages (since 2007): global ultimate owner (GUO) and global ultimate consolidated owner (≥50% direct or indirect ownership); classify firm as foreign-owned if GUO country differs from firm country of operation.
- Define “never-leavers”: firms that do not switch ownership in the immediate years before and after program approval — interpreted as tied/committed to the country and more sensitive to domestic uncertainty changes; contrasted with firms that change to foreign ownership.
- Findings (dynamic stacked DiD estimates, interaction between "never-leavers" dummy and years-since-event):
  - GRA program coefficients: -2: 0.341; -1: 0.509; 1: 1.341**; 2: 1.332**; 3: 1.091*; 4: 1.075*; 5: 1.110*.
    - Standard errors (t-statistics): (0.41) (0.68) (2.41) (2.38) (1.83) (1.73) (1.76).
    - N = 66,845 each year.
  - PRGT program coefficients match Panel A PRGT row: -2: 0.0869; -1: -0.104; 1: 0.0809; 2: 0.168*; 3: 0.164; 4: 0.189*; 5: 0.310**.
    - N = 75,835 each year.
- Magnitudes:
  - PRGT countries: around 0.3 percentage points increase in tangibles after five years for never-leavers.
  - GRA countries: around 1 percentage point increase after five years for never-leavers.
- Interpretation:
  - GRA countries more likely to host multinationals that may retreat during adverse shocks, leaving room for committed domestic firms to expand after positive IMF signaling.
  - In PRGT countries, never-leavers might be firms unable to leave due to capital/institutional/financial constraints; positive effect indicates benefit from remaining through program years.

### Robustness and alternative specifications
- Tests for anticipation effects:
  - Simple fixed effects regressions of investment at time t on dummies for program approval occurring at t+h (Table C1) find no systematic anticipation effects for both GRA and PRGT programs; weak evidence investment falls two or three years before approval but no significant effects closer to signing.
- Persistence after program end (Table C2, using AIPW with final program year as treatment):
  - Restrict sample to exclude successor programs within five years after program end.
  - GRA programs: positive and significant effects regardless of program going off-track; completed programs: tangibles/total assets grew by 2.57 percentage points after 5 years vs base year; off-track programs: effect 3.16 percentage points at same horizon; differences not significantly different.
  - PRGT programs: investment growth drops in first years after program end; when program goes off track, negative effects persist; three years after program approval, tangibles are 4.21 percentage points lower than reference year.
  - Interpretation: PRGT firms may be dependent on program effects and on repeated programs; absence of subsequent programs leads to negative outcomes.
- Sample and data-coverage checks (Table C3, C4):
  - Restrict to firms with full 5-year balance-sheet data spells: findings consistent with baseline.
  - Drop off-track programs from baseline: no significant differences relative to baseline, except some PRGT estimates slightly less significant.
  - Exclude regions one at a time to test sensitivity (Table C4):
    - GRA results robust to dropping regions; e.g., removing Sub-Saharan Africa still shows tangibles/total assets increase up to about four percentage points four years after approval.
    - Dropping Europe makes GRA results not significant (sample shrinkage and inclusion of European Debt Crises affect interpretation).
    - AIPW estimates for PRGT programs weaker when dropping regions like Sub-Saharan Africa or Middle East and Central Asia because PRGT programs are extensive in those regions.

*IMF Working Paper — Section 5: The indicator variable 퐼 (content as presented in the chapter)*

### Conclusion

### Conclusion

### Objectives and methodology
- Uses detailed firm-level data on tangible fixed investments.
- Employs a local projection methodology to estimate the dynamic response of firm investments to the approval of an IMF arrangement.
- Leverages a novel DiD methodology to identify financial channels.

### Key findings
- Distinguishing between GRA and PRGT financing matters for the path of firm investment.
- GRA programs seem to induce a stronger reaction of investors in the country than PRGT programs.
- Two financial channels are documented:
  - The degree of firms’ external financial dependence.
  - Firms’ sectoral uncertainty.
- Under a GRA arrangement:
  - Investments are higher for firms relying more on external finance.
  - Investments are higher for firms exposed to greater uncertainty.
- No similar investment effects are found for PRGT arrangements.
- Using ownership data, private investments are stimulated more for domestic firms.

### Interpretation and implications
- Results point to a differential role played by GRA versus PRGT programs in stimulating private sector investments.
- The presence of a private investment transmission channel suggests IMF programs may not only improve a country's creditworthiness for external investors but may also make "internal" investors more willing to invest by reducing the impact of uncertainty.
- Findings help improve understanding of factors that could affect the success and effectiveness of IMF programs.

*Source: Conclusion from wpiea2022157-print-pdf*

### 2021.  Is  Favoritism  a  Threat  to  Chinese  Aid  Effectiveness?  A  Subnational  Analysis  of  Chinese  Development

### Do IMF Programs Stimulate Private Sector Investment? Working Paper No. WP/2022/157

### Overview
- Research question: Whether IMF programs (GRA and PRGT) stimulate private sector investment measured as annual percentage change of tangible fixed assets (firm-level).
- Outcome variable: Investment — Annual percentage change of tangible fixed assets investment (BvD Orbis (2021)).
- Treatment definitions:
  - GRA: Dummy = 1 if country signs GRA program within the first 9 months of the year (MONA).
  - PRGT: Dummy = 1 if country signs PRGT program within the first 9 months of the year (MONA).

### Data and Methodology
- Firm-level data source: BvD Orbis (2021) for tangible fixed asset investment.
- Country-level macro controls: World Economic Outlook (October 2021); World Development Indicators (2021); International Financial Statistics (2021); International Country Risk guide (2021); Database of Political Institutions (2020).
- Key predictors/controls include: Real GDP growth, Log real GDPPC (2017 PPP dollars), GFCF to GDP, Total debt service to GNI, Budget surplus, Total reserves/imports (months), Inflation, Change in reserves, Current account/GDP, Legislative election dummy, Log legislative checks, Log claims by depository institutions, Real interest rate, Political risk rating, Program years dummy.
- Empirical approach: Two-stage Augmented Inverse Propensity Weighted (AIPW) estimator with first-stage logit (or multinomial logit distinguishing GRA, PRGT, no program) and second-stage firm-level local projections; firm and sector-year fixed effects; standard errors clustered at country or country-sector level as noted.
- Sample size variation across horizons: N reported in second-stage tables (examples): 21,817; 18,560; 15,900; 13,685; 11,899 (see Table B2).

### Main Findings (AIPW estimates and local projections)
- First-stage predictors (AIPW first stages, Table B1):
  - Past program: 2.195*** (GRA), 2.053*** (PRGT) — strong predictor of signing a program. (T-statistics in parenthesis shown in table.)
  - GFCF/GDP: -0.097*** (GRA).
  - Total debt service to GNI: -0.145** (PRGT).
  - Budget surplus: 0.138*** (PRGT).
  - Real GDP growth: -0.097*** (PRGT).
  - Total reserves/imports: -0.112* (GRA), -0.175* (PRGT).
  - Change in reserves: -0.006** (GRA).
  - Current account/GDP: 0.088** (GRA).
  - Observations in first-stage models: 806 (for both GRA and PRGT).
- Second-stage AIPW baseline estimates (Table B2):
  - GRA dummy coefficients across models (columns 1–5): 0.122*, 0.119, 0.147*, 0.115*, 0.102** (T-statistics reported).
  - GRA years (program-years excluding year of signing) coefficients: 0.122**, 0.121*, 0.0930**, -0.0216, -0.00637.
  - PRGT dummy coefficients across models (columns 1–5): 0.123, 0.112, 0.0259, -0.0614, 0.0205 (not consistently significant).
  - Lagged investments often negative and significant in some specifications (examples: -0.0239*; -0.0280*).
  - Firm characteristics with significant coefficients in various specifications include Log(total assets), Leverage, Cash flows/TA (often negative), Sales growth (negative and often significant).
  - R-squared values (examples across GRA models): 0.425, 0.454, 0.468, 0.562, 0.598.
  - Sample sizes for second-stage reported: N = 21,817; 18,560; 15,900; 13,685; 11,899.
- Heterogeneous and state-contingent effects (Table B3):
  - GRA effects vary by firm age and volatility:
    - Younger firms: coefficients by horizon (h=1..5): -0.227, 1.251, 3.513*, 9.155***, 10.22*** (N reported by horizon).
    - Mature firms: -0.249, 1.136**, 1.817**, 2.203*, 1.442.
    - High realized volatility: 0.349, 1.755, 1.586**, 3.236***, 3.206***.
    - Low realized volatility: -0.947**, -0.644, 1.305***, -0.437, -2.293.
    - High implied volatility: 0.0472, 0.695**, 0.878**, 2.987***, 3.262***.
    - Low implied volatility: -0.595**, -1.021, 0.517*, -0.509, -2.805.
  - PRGT shows different patterns with some large positive effects for younger or low-volatility firms at short horizons and negative or insignificant effects for high volatility or later horizons (details in Table B3).
- Anticipation effects (Table C1):
  - Years to program impacts on investment growth (h = -5 to -1):
    - GRA: 0.03, 0.01*, -0.03***, 0.01, 0.01 (N = 27,585 each).
    - PRGT: 0.01, 0.02, -0.02, -0.03***, 0.04 (N = 27,585 each).
- End-of-program effects by completion status (Table C2):
  - GRA (Completed programs): horizons 1–5: -0.23***, 0.56, 1.77**, 3.46***, 2.57* (N values provided per horizon).
  - GRA (Offtrack programs): 0.27, 1.15**, 2.01***, 3.91***, 3.16**.
  - PRGT (Completed programs): -1.05*, -0.99, -1.65, -0.27, 2.72.
  - PRGT (Offtrack programs): -2.58***, -1.17, -4.21**, -2.04, -0.76.
- Robustness and sample-split analyses (Table C3 and C4):
  - Results robust to restricting to firms with spell length >= 5 years and dropping offtrack programs in many specifications.
  - Region-dropping exercises show regional heterogeneity; examples for GRA dropping regions:
    - Asia Pacific: -0.09, 1.24**, 2.20***, 4.44***, 3.63**.
    - Europe: -0.02, 0.29, -1.88, -5.13**, 0.05.
    - Sub-Saharan Africa (SSA): -0.13, 0.99***, 1.81***, 4.10***, 3.25**.
  - PRGT regional patterns show substantial variation and some negative coefficients in Europe across horizons.

### Sample and Descriptive Statistics (selected)
- Program completion status by type (Table A2):
  - PRGT: Completed 23; Off track 6; Ongoing 8; Total 37.
  - GRA: Completed 23; Off track 10; Ongoing 4; Partially completed 42; Total 79. (Note: table shows columns “Completed Off track Ongoing Partially Total” with numbers aligned per program type.)
  - Others: Completed 29; Off track 2; Ongoing 4; Partially 4; Total 39.
  - Total across program types: 75 Completed; 18 Off track; 8 Ongoing; 54 Partially; 155 Total (as tabulated).
- Panel coverage (Table A4): Number of observations and unique firms by country ISO examples:
  - PL (Poland): 100,859 observations; 9,919 unique firms.
  - PT (Portugal): 40,587 obs; 3,198 firms.
  - RO (Romania): 45,738 obs; 3,614 firms.
  - TR (Turkey): 44,411 obs; 7,944 firms.
  - Examples of small coverage: AF 6 obs; 1 firm. DM 1 obs; 1 firm.
- Sectoral and industry-level metrics:
  - Table A3 provides yearly average firm investment by primary NACE sector (yearly sector averages 2000–2019).
  - Table A5 provides EFD by sector (NACE Rev 1.1) with values (examples): Tobacco -3.4462; Pharmaceuticals 8.6029; Radio/ TV/ communication equipment 1.1559; Medical/ precision/ optical instruments 1.0336.

### Robustness, Inference, and Tests
- Balance tests for propensity scores and sensitivity checks performed (Figure B1 referenced).
- Standard inference: T-statistics shown in tables; significance markers used: * p < 0.1, ** p < 0.05, *** p < 0.01.
- Clustered standard errors: country-level and country-sector level clustering used depending on table.

### Policy Implications and Interpretation
- GRA programs are associated with positive and sometimes significant increases in firm-level tangible fixed asset investment in several specifications and horizons, particularly for certain firm types (younger firms, high volatility firms) and regions.
- PRGT effects are more heterogeneous and less consistently positive; some negative effects appear for PRGT in certain samples or completion-status splits.
- Heterogeneity suggests program design, firm characteristics (age, volatility), program completion status, and regional context matter for the investment response to IMF programs.
- Monitoring program implementation and addressing factors that influence heterogeneous firm responses (liquidity, credit conditions, firm size and age) could improve private-sector investment outcomes following IMF-supported programs.

*Do IMF Programs Stimulate Private Sector Investment? Working Paper No. WP/2022/157*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022157-print-pdf.pdf_