## 1. Current Account and Saving-Investment Balances with Internal Imbalances

## Source details

**Canonical URL:** [1. Current Account and Saving-Investment Balances with Internal Imbalances](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021198-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021198-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021198-print-pdf.pdf.json)

---

### Key objectives and scope
- Study dynamics of external accounts during economic recessions over the past 60 years using Jordà’s (2005) local projection method.
- Horizon: up to 10 years.
- Sample: 278 recession episodes covering advanced economies (AEs) and emerging markets (EMs), including natural disasters and epidemics.
- Baseline yearly panel: 1960–2019 for 49 AEs and EMs (~90 percent of global GDP).

### Aggregate empirical findings on external-sector responses to recessions
- Global historical pattern:
  - Large post-WWII downturns: global imbalances narrowed by around 1½ percent of global GDP and remained at this level for several years.
  - COVID-19 shock: global imbalances increased by around ½ percent of global GDP (deficits and surpluses widened).
- Typical CA and component dynamics following recessions:
  - Current account (CA) increases by around 1½ percent of GDP within a year after a recession and this increase is maintained up to five years, then gradually unwinds.
  - CA improvement is driven primarily by a sharp and persistent decline in investment.
- Saving and investment decomposition (CA = Saving − Investment):
  - Investment: sharp and persistent decline of around 2½ percent of GDP in the year following the recession.
  - Aggregate saving: falls by 1 percent (post-recession year), with private saving increasing and public dissaving almost 2 percent of GDP in the post-recession year.
  - Private saving increase mainly driven by household deleveraging (subsample of 20 countries).
  - Saving response is short-lived (~3 years); public saving returns to and beyond pre-crisis levels relatively quickly.
- Trade and exchange rates:
  - Imports: fall around 13 percent year on year (volumes) following recessions; suppressed on average for more than 10 years.
  - Exports: decline by around 3 percent and recover over the medium term.
  - Exchange rates: NEER falls roughly 40 percent and REER 10 percent; exchange rate developments mainly driven by emerging markets in the sample.
- Persistence and heterogeneity:
  - Effects more persistent in AEs than in EMs.
  - 5-year rolling CA response one year after recession: oscillates around 2 percent historically, spiked around 4 percent in late 1980s–mid-1990s, trended down to around 1 percent of GDP in the past decade.
  - Distributional analysis: CA response evenly distributed across countries (10th, 50th, 90th percentiles); private saving shows large dispersion across quantiles.

### Role of internal imbalances (pre-recession borrowing, credit booms, public debt)
- General:
  - Presence of internal imbalances amplifies CA strengthening after recessions, particularly in the medium to longer term.
- Credit booms:
  - Recessions associated with credit booms trigger a contemporaneous CA increase of around 2 percent of GDP which rises to 3 percent over the estimation horizon.
  - Investment falls by around 4 percent of GDP in credit boom economies versus ~2 percent without such booms.
  - Identification: 48 recessions associated with credit booms in the baseline sample (credit boom recessions up to 3 years after the end of a credit boom).
  - Short- to medium-term saving responses in credit boom recessions are dominated by public dissaving rather than private deleveraging.
- Banking crises:
  - Similar pattern to credit booms but with a more pronounced role of private deleveraging.
- High public debt:
  - Higher public debt prior to recessions (3-year average above the 75th percentile by income group) is associated with stronger CA increases in the medium term.
  - High public debt economies show a significantly stronger aggregate saving response throughout the horizon.

### Role of external imbalances (pre-recession CA deficits/surpluses, external debt)
- CA deficit vs surplus behavior:
  - Pre-recession CA deficit economies typically experience CA increases of around 2 percent of GDP.
  - Pre-recession CA surplus economies remain relatively stable as buffers are used to dampen investment falls.
  - Private saving increases strongly in CA deficit economies and mirrors a steeper short-term fall in investment.
- Financing-constraint amplifiers:
  - Sudden stops amplify CA impact to around 3 percent of GDP with similar response paths.
  - Higher stocks of external debt liabilities produce comparable results, with a somewhat more pronounced role of public saving.
  - High foreign currency exposure is not associated with a larger increase in the CA balance; investment may return somewhat faster in high FX liability economies.
- Definitions and sample averages:
  - High external debt dummy: 3-year average of external gross debt liabilities to GDP prior to the recession above the median by income group. Sample average over time: 62 percent for AEs and 35 percent for EMs.
  - High foreign currency exposure dummy: 3-year average of foreign currency-denominated debt liabilities to GDP above the median by income group. Sample average over time: 46 percent for AEs and 31 percent for EMs.

### Interaction with synchronization and global conditions
- Greater synchronization of recessions across countries reduces CA adjustment because exports fall in tandem with imports.
- Global crises (global 25: >25 percent of global GDP in recession):
  - Average CA response during global crises: around 1 percent of GDP short term (lower than ~2 percent in less synchronized recessions).
  - Exports decline more strongly; investment to GDP falls somewhat less; saving declines more strongly.
    - Private deleveraging hits 2 percent of GDP in global crises.
    - Public dissaving reaches 3 percent of GDP in global crises.
  - Synchronization implies a less severe decline in investment by around ½ percent of GDP (statistically significant).
- Historical 1870–1960 era (more synchronized): smaller CA improvements (~½ percent initially, rising to ~1 percent longer term).

### Recessions from disasters and epidemics
- Natural disasters (severe, 75th percentile mortality):
  - Sample: ~178 countries, 576 events (75th percentile cutoff).
  - Short-term impact: CA deteriorates by ¾ percent of GDP shortly after a large natural disaster, driven by falling exports.
  - Effect fades about three years after the disaster.
- Severe epidemics (75th percentile mortality; global sample 1900–2018):
  - Historical sample: 176 countries and 252 epidemics (75th percentile cutoff).
  - Medium-term effect: CA increases by 1 to 1½ percent of GDP; short-term CA fluctuations not significant.
  - Exports drop initially and recover later; imports increase immediately during the epidemic.
- Severe epidemics in baseline 49 AEs and EMs (reduced coverage):
  - 66 epidemics with CA data; 37 events for saving regressions; 35 events for investment (Spanish Flu excluded in some panels).
  - Estimations show a significant short-term dip in CA during the first year, followed by a strong CA increase to 1½ percent of GDP.
  - Public dissaving reaches 1 percent of GDP and returns to pre-recession levels after three years.
  - Investment declines with wide error bands.
- Note on mortality cutoffs: 75th percentile cutoff ≈ 5 deaths per 100 000 inhabitants; COVID-19 was about 20–40 times as lethal in most countries.

### How the COVID-19 shock differed
- Synchronization and pre-crisis imbalances:
  - COVID-19 was highly synchronized: 80 percent of global GDP in recession (47 of the 49 countries in baseline sample had negative real GDP growth in 2020).
  - Fewer pre-crisis internal imbalances: only two countries signaled credit booms in 2017–2018 and none in 2019 (vs 12 in the three years preceding the GFC).
  - Sum of surpluses and deficits: 5½ percent of global GDP in 2006 vs 3 percent in 2019.
- Model predictions vs outturn for 2020:
  - Regression models predict a muted, not statistically significant CA response short term.
    - Without epidemics variable: predicted CA increase of around 1 percent of GDP.
    - Including epidemics: predicted short-term CA weakening of around ¾ percent.
    - Both predict a statistically significant CA increase in subsequent years; effects fade after three years due to smaller and less persistent investment decline.
  - 2020 outturn: CA balances of recession economies declined by 0.04 percent of GDP on average (non-weighted).
- Saving deviations due to policy in 2020:
  - Realized decline in saving was around 1½ percent of GDP in 2020 (model without epidemics predicted similar decline).
  - Decomposition of deviations:
    - Public saving declined by 7 percent of GDP on average in 2020 (vs predicted fall of 3 percent of GDP).
    - Private saving increased by 5½ percent of GDP (more than 3 times higher than past-recession evidence would predict).
  - Drivers: unprecedented fiscal support → massive fiscal deficits; government transfers; ample central bank liquidity avoided corporate deleveraging; lockdowns, border closures, and precautionary motives induced forced and precautionary private saving.

### Robustness and relation to prior literature
- Findings robust to:
  - Inclusion of additional controls (exchange rate movements, openness, relative growth).
  - Alternative recession and vulnerability definitions.
  - Cook’s distance correction for historic pre-1960 recessions.
- Consistency with literature:
  - Aligns with studies showing large and persistent declines in output and scarring after recessions (Cerra and Saxena 2008; Blanchard et al. 2015).
  - Consistent with Abiad et al. (2011) on persistent import declines after financial crises.
  - Complements CA adjustment and early-warning literature (Milesi-Ferretti and Razin 1998; Lane and Milesi-Ferretti 2012; Freund and Warnock 2007).

### Policy-relevant implications and near- to medium-term considerations
- Historical pattern:
  - Past recessions: CA strengthened significantly and persistently, driven by falling investment and private deleveraging; excessive pre-existing borrowing magnified CA responses.
  - Buffer use: countries with larger buffers attenuated investment falls by drawing on buffers.
- COVID-19 divergence:
  - CAs of economies in recession were broadly unchanged in 2020, reflecting fewer pre-crisis imbalances and the shock’s global nature.
  - Policy interventions produced unprecedented shifts: public dissaving more than offset record private saving.
- Near- to medium-term considerations:
  - The translation of private saving into consumption, and the paths of fiscal and monetary normalization, will materially affect external balances—especially if adjustments occur asymmetrically across countries.
  - Rapid household dissaving combined with gradual public consolidation could temporarily lower CA balances in some countries.
  - Faster government rebuilding of buffers would raise CA balances in line with historical medium-term improvements, especially for economies with larger vulnerabilities.
- Research gaps:
  - Sample focuses on AEs and non-commodity exporting EMs; more research needed on commodity exporters and lower-income economies.

*Source: wpiea2021198-print-pdf.*

### 1. Current Account and Saving-Investment Balances with Internal Imbalances .....................16

### 1. Current Account and Saving-Investment Balances with Internal Imbalances

### Key objectives and scope
- Study dynamics of external accounts during economic recessions over the past 60 years.
- Focus on responses of current accounts (CA) and components for a horizon of up to 10 years using Jordà’s (2005) local projection method.
- Sample covers a broad set of countries, including advanced economies (AEs) and emerging markets (EMs), and various downturn types, including natural disasters and epidemics.
- Analysis is based on 278 recession episodes.

### Aggregate empirical findings on external-sector responses to recessions
- Global historical pattern:
  - Large post-WWII downturns were typically followed by narrowing global current account balances; example episodes include the 1980s debt and oil crises and the Global Financial Crisis (GFC).
  - Global imbalances narrowed by around 1½ percent of global GDP after such crises and remained at this level for several years.
  - The COVID-19 shock exhibited a contrasting pattern: global imbalances increased by around ½ percent of global GDP (deficits and surpluses widened across economies).
- Typical CA and component dynamics following recessions:
  - Current account increases by around 1½ percent of GDP within a year after a recession.
  - This CA increase is maintained up to five years and then gradually unwinds.
  - The CA improvement is driven primarily by a sharp and persistent decline in investment.
  - Aggregate saving:
    - Drops temporarily, mainly driven by countercyclical fiscal policy.
    - Turns positive over the medium to long term as private deleveraging fuels saving.
  - Persistence of effects:
    - Impacts on the external sector are more persistent in advanced economies (AEs) than in emerging markets (EMs).

### Role of internal imbalances (pre-recession borrowing, credit booms, public debt)
- Presence of internal imbalances amplifies CA strengthening after recessions:
  - Strengthening of the CA is significantly more pronounced when pre-recession private or public borrowing is elevated.
  - Example magnitudes:
    - Investment falls by 4 percent during recessions associated with credit booms, versus 2 percent in recessions without such booms.
  - Countries with larger buffers (e.g., CA surplus economies or countries with lower public debt) used buffers to dissave during recessions, attenuating the fall in investment.
- Timing:
  - Internal imbalances push up the CA balance particularly in the medium to longer term.

### Role of external imbalances (pre-recession CA deficits/surpluses, external debt)
- External imbalances affect the CA path more immediately:
  - Countries with excessive pre-crisis CA balances experienced the largest contractions in their external balance in past episodes (consistent with Lane and Milesi-Feretti (2012) findings).
  - Adjustment in deficit countries historically was achieved primarily through demand compression rather than expenditure switching.
- Other external vulnerability indicators noted in related literature include the stock of foreign liabilities and its currency denomination.

### Interaction with synchronization and global conditions
- Greater synchronization of recessions across countries reduces CA adjustment:
  - CA increases by less when recessions are more synchronized because exports fall in tandem with imports due to faltering global demand.
- COVID-19 specifics:
  - The COVID shock featured comparatively moderate pre-existing imbalances but high synchronization.
  - This combination led to a muted global CA effect relative to the GFC.
  - Compositional changes in 2020 were unusual: exceptional private saving offset historical public dissaving while investment fell only mildly.

### Recessions from disasters and epidemics
- Severe natural disasters or epidemics:
  - CAs tend to weaken in the short term during severe natural disasters or epidemics.
  - The paper documents distinct CA and trade responses for severe natural disasters and epidemics across the historical record.

### Robustness and relation to prior literature
- Findings are robust to:
  - Inclusion of additional controls.
  - Alternative definitions of recessions and vulnerabilities.
- Relation to prior studies:
  - Results align with literature showing large and persistent declines in output and scarring after recessions (e.g., Cerra and Saxena 2008, Blanchard et al. 2015).
  - Results are consistent with Abiad et al. (2011) on persistent import declines after financial crises.
  - The paper complements early warning and CA adjustment literature (e.g., Milesi-Feretti and Razin 1998; Lane and Milesi-Feretti 2012; Freund and Warnock 2007).

*Source: IMF working paper chapter "1. Current Account and Saving-Investment Balances with Internal Imbalances" (excerpt).*

### section IV analyzes vulnerabilities and conditions affecting the path of external accounts

### wpiea2021198-print-pdf - section IV analyzes vulnerabilities and conditions affecting the path of external accounts

### Data and methodology
- Sample: yearly, unbalanced panel data over 1960 to 2019 for a baseline sample of 49 advanced and emerging economies representing around 90 percent of global GDP.
- Extended samples: 35 advanced economies (AEs), 43 emerging markets (EMs) with further differentiation by main commodity exporter status; extended subsample counts cited include 35 AEs, 22 non-commodity exporting EMs, and 21 commodity-exporting EMs for some analyses.
- Recession definition: a year with negative real GDP growth; baseline sample includes 278 recession events up to 2019.
- Historical augmentation: Jordà et al. 2017 dataset (1870–2016 for 17 countries) plus EM-DAT natural disaster data and Li and Coppo 2020 epidemics data (1900–2018) to study epidemics and natural disasters' impact on current accounts.
- Estimation approach:
  - Local projection methodology (Jordà 2005) with Driscoll Kraay standard errors.
  - Equation (1) estimates cumulative response of external sector variables 풀′
௜,௧ to recession 푅
௜,௧ across horizons h≥0.
  - Controls: country fixed effects 훼
௜
௛, time fixed effects 훾
௧
௛, two lags of dependent variable and recession dummy, and the residual from horizon h−1 included to address efficiency and bias concerns (Teulings and Zubanov 2014).
  - Robustness: Cook’s distance correction for historic pre-1960 recessions, natural disasters, epidemics, and regressions with EM subsample; robustness checks using only single-year recessions.
- Treatment of vulnerabilities: pre-crisis conditions 푉푢푙푛
௜,௧ interact with recession dummy (푅
௜,௧∗푉푢푙푛
௜,௧) to compare responses when vulnerability is present versus absent. Time horizon of pre-crisis conditioning variable varies by vulnerability type (e.g., credit boom peak up to 3 years prior; sustained CA deficits averaged over 5 years prior).

### How do current account balances behave during recessions?
- Aggregate CA response:
  - Recessions trigger an average improvement in the current account (CA) balance of around 1½ percent of GDP for up to five years after the recession.
  - The CA path is highly persistent and returns to pre-crisis levels only after around 10 years.
- Saving and investment decomposition (퐶퐴 = 훴Saving − Investment):
  - Investment: sharp and persistent decline of around 2½ percent of GDP in the year following the recession.
  - Aggregate saving: falls by 1 percent (private saving increases while public dissaving is almost 2 percent of GDP in the post-recession year).
  - Private saving increase is mainly driven by household sector deleveraging (subsampIe of 20 countries).
  - Saving response duration: short-lived, around 3 years, with public saving quick to return to and beyond pre-crisis levels.
- Trade and exchange rates:
  - Imports: fall around 13 percent year on year (in volumes) following recessions; import volumes suppressed on average for more than 10 years.
  - Exports: decline by around 3 percent (mild effect) and recover over the medium term.
  - Exchange rates: NEER falls roughly 40 percent and REER 10 percent; exchange rate developments mainly driven by emerging markets in the sample.
- Comparison with prior literature:
  - Results broadly consistent with Abiad et al. 2011 (declines of around 19 percent for imports and 4 percent for exports in earlier sample), though magnitudes differ given sample differences.
- Temporal evolution of CA sensitivity:
  - 5-year rolling estimations of CA response one year after recession oscillate around 2 percent except for late 1980s–mid-1990s when the CA response spiked to around 4 percent then dropped to insignificant levels; trended downward to around 1 percent of GDP in the past decade (including during the GFC).
  - Persistence appears to have increased in the aftermath of the GFC (medium-term impact t+5).
- Distributional analysis (quantile regressions):
  - CA response is evenly distributed across countries (10th, 50th, 90th percentiles).
  - Investment responses show very small dispersion across quantiles.
  - Private saving divergence: lowest decile experiences little accumulation of saving (if not dissaving) while highest decile leads deleveraging.
- Income groups and commodity exporters:
  - Non-commodity EMs: CA balances return to pre-crisis levels in around 5 years.
  - AEs: CA increase remains above pre-recession level even at a 10 year horizon.
  - Commodity exporters: no significant CA improvement detected.
  - Differences driven by saving response: in AEs private saving increases by around one percent and remains elevated; in EMs private saving returns to pre-crisis levels within two years.
  - Imports and exports by group:
    - EMs: imports fall by around 17 percent in the first year after the recession; exports fall around 10 percent (high dispersion).
    - AEs: imports fall by around 6 percent in the first year after the recession; exports only mildly affected.

### What factors shape the CA response? — Internal imbalances
- Internal imbalances proxied by credit booms, banking crises, and public debt levels are linked to stronger and more persistent CA adjustments after recessions.
- Credit booms:
  - Recessions associated with credit booms trigger a contemporaneous increase in the CA balance of around 2 percent of GDP which rises to 3 percent over the estimation horizon.
  - Recessions without prior excessive borrowing see CA balances gradually return to pre-crisis levels.
  - Mechanisms:
    - Investment falls by around 4 percent of GDP in credit boom economies.
    - Stronger private deleveraging in credit boom recessions reins in past excesses.
    - Short-term countercyclical fiscal policy is more expansive among credit boom economies; public dissaving more than offsets private saving in the short and medium term, causing aggregate saving to decline more in credit boom economies before gradually returning and amplifying CA response divergence over time.
  - Identification: 48 recessions associated with credit booms identified in the baseline sample; credit boom recessions defined as recessions occurring up to 3 years after the end of a credit boom (credit boom episodes from Dell’Ariccia et al. 2020).
- Banking crises:
  - Produce a similar pattern to credit booms but with a more pronounced role of private deleveraging, consistent with rapid tightening of financial conditions in banking crises.
  - Data differences: banking crises in the 1990s were less often accompanied by credit booms; more countries faced a credit boom than a banking crisis during the GFC.
- High public debt (introductory finding):
  - Higher public debt levels prior to recessions are associated with stronger CA increases in the medium term.
  - High public debt dummy: equals one if the 3-year average of a country’s public debt to GDP ratio prior to the recession year is above the 75th percentile of the same horizon by income group.

*Italic: Source — wpiea2021198-print-pdf (section IV analyzes vulnerabilities and conditions affecting the path of external accounts).*

### Annex Figure 5 for the sample’s density of public debt by income group, noting the unbalanced time horizon

### Annex Figure 5 for the sample’s density of public debt by income group, noting the unbalanced time horizon

### Recessions, High Public Debt, and Credit Booms
- Short-run responses:
  - Recessions are associated with a significant countercyclical fiscal policy response in the short run.
  - Recessions in low-public debt economies trigger somewhat more private deleveraging in the short term, but aggregate saving response is significantly stronger in high-public debt economies throughout the considered horizon.
- Medium-term dynamics:
  - High public debt pushes up saving, notably in the medium term.
  - Credit booms paralyze investment persistently; falling investment is a dominant transmission channel for recessions associated with credit booms.
- Empirical magnitudes and significance:
  - Higher public debt is defined as debt above the 75th percentile, lower public debt as debt below the 25th percentile within income group (AEs and EMs).
  - Effects are robust when both high credit growth and high public debt are included in estimations; additional controls (notably relative growth) lower short-term effects but do not change qualitative conclusions.

### External Imbalances: CA Deficits, Sudden Stops, and External Debt
- CA deficit vs surplus behavior:
  - Economies with pre-recession CA deficits typically experience CA increases of around 2 percent of GDP.
  - Pre-recession CA surpluses remain relatively stable as surplus economies draw on buffers and dampen recession-induced falls in investment.
  - Private saving increases strongly in CA deficit economies (mirrored by a steeper fall in investment in the short term); private saving remains constant during recessions in surplus economies.
- Financing-constraint amplifiers:
  - Using sudden stops instead of CA deficits amplifies CA impact to around 3 percent of GDP and yields similar response paths.
  - Higher/lower stocks of external debt liabilities produce comparable results, with a somewhat more pronounced role of public saving.
  - High foreign currency exposure is not associated with a larger increase in the CA balance; estimations suggest investment returns somewhat faster in high FX liability economies.
- Definitions and sample averages:
  - High external debt dummy: 3-year average of external gross debt liabilities to GDP prior to the recession above the median, by income group. Sample average over time: 62 percent for AEs and 35 percent for EMs.
  - High foreign currency exposure dummy: 3-year average of foreign currency-denominated debt liabilities to GDP above the median, by income group. Sample average over time: 46 percent for AEs and 31 percent for EMs.

### Combined Internal and External Imbalances (Summary from Table 2)
- Timing and channels:
  - External imbalances affect CA balances already in the short term while internal imbalances push up the CA in the medium term.
  - Muted short-term CA effect explained by offsetting forces in saving and investment responses.
- Credit boom distinctiveness:
  - Falling investment plays an exceptional role in recessions associated with credit booms: coefficient substantially larger and highly persistent.
  - Recessions associated with credit booms show short- to medium-term saving responses dominated by public dissaving rather than private deleveraging.
- Robustness:
  - Results hold when including additional controls such as exchange rate movements, openness, and relative growth.

### Globally Synchronized Recessions
- Definition and historic episodes:
  - Global crisis dummy equals one when more than 25 percent of countries worldwide (GDP-weighted) are in recession. Episodes include 1974-75 first oil crisis, 1980 and 1982 crises, 1991 and 1993 crises, and 2008-09 Global Financial Crisis (COVID-19 2020 excluded from sample).
- Impact on CA and components:
  - Average CA response during global crises is around 1 percent of GDP in the short term, lower than the 2 percent observed in less synchronized recessions.
  - Exports decline more strongly during global downturns.
  - Investment to GDP falls somewhat less in synchronized recessions (fewer investment opportunities abroad), while saving declines more strongly.
    - Private deleveraging hits 2 percent of GDP in global crises.
    - Public dissaving reaches 3 percent of GDP in global crises.
  - Synchronization implies a less severe decline in investment by around ½ percent of GDP (statistically significant in multivariate tables).
- Historical support:
  - CA responses in 1870–1960 (more synchronized era) show smaller improvements (~½ percent initially, rising to ~1 percent longer term), lower than post-1960s estimates.

### Natural Disasters and Severe Epidemics (1900–2018)
- Natural disasters (severe, 75th percentile mortality):
  - Sample: about 178 countries, 576 events with CA data available at the 75th percentile cutoff.
  - Short-term impact: CA balances deteriorate by ¾ percent of GDP shortly after a large natural disaster, driven by falling exports.
  - Effect fades about three years after the disaster; imports rise early but effect not statistically significant.
- Severe epidemics (75th percentile mortality; global sample):
  - Historical sample (1900 onwards): 176 countries and 252 epidemics (75th percentile cutoff).
  - Medium-term effect: CA increases by 1 to 1½ percent of GDP; short-term CA fluctuations not significant.
  - Exports drop initially and recover later; imports increase immediately during the epidemic.
- Severe epidemics in baseline 49 AEs and EMs:
  - With reduced coverage (66 epidemics with CA data coverage; 37 events for saving regressions; 35 events for investment, excluding Spanish Flu in some panels), estimations show:
    - Significant short-term dip in CA during the first year of the epidemic, followed by a strong increase in CA balance to 1½ percent of GDP.
    - Public dissaving reaches 1 percent of GDP and returns to pre-recession levels after three years.
    - Investment declines but with wide error bands.
- Notes on cutoffs and data:
  - 75th percentile cutoff equivalent to 5 deaths per 100 000 inhabitants; COVID-19 was about 20–40 times as lethal in most countries.
  - Using the 90th or 95th percentile reduces observations; 90th yields broadly similar results for exports and imports to GDP.

### How the COVID-19 Shock Was Different
- Synchronization and pre-crisis imbalances:
  - COVID-19 was one of the most globally synchronized downturns: 80 percent of global GDP in recession (47 of the 49 countries in baseline sample experienced negative real GDP growth in 2020).
  - Economies entered 2020 with fewer prior internal and external imbalances: only two countries in sample show credit boom signs in 2017–2018 and none in 2019 versus 12 in the three years preceding the GFC.
  - Sum of surpluses and deficits: 5½ percent of global GDP in 2006 vs 3 percent in 2019.
- Model predictions vs outturn:
  - Regression models (Table 3 and Annex specifications) predict muted, not statistically significant CA response for economies in recession in the short term.
    - Without epidemics variable: model predicts CA increase of around 1 percent of GDP.
    - Including epidemics: CA shows some weakening of around ¾ percent in the short term.
  - Both specifications predict a statistically significant CA increase in subsequent years; effects no longer significant after three years due to smaller and less persistent decline in investment.
  - 2020 outturn: CA balances of recession economies declined by 0.04 percent of GDP on average (non-weighted average).
- Saving deviations due to policy:
  - Model (specification without epidemics) comes close to realized decline in saving of around 1½ percent of GDP in 2020.
  - Decomposition shows stark deviations:
    - Public saving declined by 7 percent of GDP on average in 2020 (vs predicted fall of 3 percent of GDP).
    - Private saving increased by 5½ percent of GDP, more than 3 times higher than evidence from past recessions would have predicted.
  - Drivers:
    - Unprecedented fiscal support → massive expansion of fiscal deficits → public dissaving.
    - Government transfers offset income losses for some households.
    - Ample central bank liquidity helped avoid corporate deleveraging and dampened falls in investment.
    - Lockdowns, border closures, and precautionary motives produced forced and precautionary private saving.

### Conclusion and Policy-Relevant Implications
- Historical pattern:
  - During past recessions, CA balances strengthened significantly and persistently, driven by falling investment and supported by private deleveraging.
  - Pre-existing excessive private, public, and/or external borrowing magnified CA responses; countries with larger buffers attenuated investment falls.
- COVID-19 divergence:
  - On average, CAs of economies in recession were broadly unchanged in 2020, consistent with fewer pre-crisis imbalances and the shock’s global nature.
  - Policy interventions produced unprecedented shifts in saving: public dissaving more than offset record private saving.
- Near- to medium-term considerations:
  - Translation of private saving into consumption, and fiscal and monetary normalization paths, will materially affect external balances—especially if adjustments occur asymmetrically across countries.
  - Rapid household dissaving combined with gradual public consolidation could temporarily lower CA balances in some countries.
  - Faster government rebuilding of buffers would raise CA balances in line with historical medium-term improvements, especially for economies with larger vulnerabilities.
- Research gaps:
  - The sample focuses on advanced and non-commodity exporting EMs; more research is needed on commodity exporters and lower-income economies to understand external account responses in those groups.

*Italic: Source — wpiea2021198-print-pdf - Annex Figure 5 for the sample’s density of public debt by income group, noting the unbalanced time horizon*

### REFERENCES

### REFERENCES

### Bibliographic references
- Abiad, Abdul, Petia Topalova, and Prachi Mishra, 2011, “How Does Trade Evolve in the Aftermath of Financial Crises?” IMF Working Paper No. 11/3 (Washington, International Monetary Fund).
- Adrian Tobias, Federico Grinberg, Nellie Liang, Sheheryar Malik, Jie Yu 2018, “The Term Structure of Growth-At-Risk,” IMF Working Paper No. 18/181 (Washington, International Monetary Fund).
- Adrian, Tobias, Nina Boyarchenko, and Domenico Giannone, 2019, “Vulnerable Growth,” American Economic Review, 109(4): 1263–89.
- Allen, Cian, 2019, “Revisiting External Imbalances: Insights from Sectoral Accounts,” Journal of International Money and Finance, Vol. 96(C), pp. 67–101.
- Barro, Robert J., Jose F. Ursua, and Joanna Weng, 2020, “The Coronavirus and the Great Influenza Pandemic: Lessons from the “Spanish Flu” for the Coronavirus’s Potential Effects on Mortality and Economic Activity,” Technical Report 26866, NBER Working Paper (Cambridge, Massachusetts: National Bureau for Economic Research).
- Bakker, Bas, Marta Korczak, and Krzysztof Krogulski, 2019, “Unemployment Surges in the EU: The Role of Risk Premium Shocks,” IMF Working Paper No. 19/56 (Washington, International Monetary Fund).
- Bénétrix, Agustin, Deepali Gautam, Luciana Juvenal, and Martin Schmitz, 2019, “Cross-Border Currency Exposures,” IMF Working Paper No. 19/299 (Washington: International Monetary Fund).
- Blanchard, Olivier, Eugenio Cerutti, and Lawrence Summers, 2015, “Inflation and Activity—Two Explorations and their Monetary Policy Implications,” IMF Working Paper No. 15/230 (Washington: International Monetary Fund).
- Cabezon, Ezequiel, Leni Hunter, Patrizia Tumbarello, Kazuaki Washimi, and Yiqun Wu, 2015, “Enhancing Macroeconomic Resilience to Natural Disasters and Climate Change in the Small States of the Pacific,” IMF Working Paper No. 15/125 (Washington: International Monetary Fund).
- Catão, Luis, 2006, “Sudden Stops and Currency Drops: A Historical Look,” IMF Working Paper No. 06/133 (Washington: International Monetary Fund).
- Cerdeiro, Diego, Andras Komaromi, 2020, “Supply Spillovers During the Pandemic: Evidence from High-Frequency Shipping Data,” IMF Working Paper No. 20/284 (Washington: International Monetary Fund).
- Cerra, Valerie and Sweta Chaman Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery,” American Economic Review, Vol. 98(1), pp. 439–57.
- Cerra, Valerie, Antonio Fatas, and Sweta Chaman Saxena, 2020, “Hysteresis and Business Cycles,” IMF Working Paper No. 20/73 (Washington: International Monetary Fund).
- Cubeddu, Luis, Swarnali Ahmed Hannan, and Pau Rabanal, 2021, “External Financing Risks: How Important is the Composition of the International Investment Position?” IMF Working Paper No. 2021/020 (Washington: International Monetary Fund), https://ssrn.com/abstract=3799612.
- Cuesta Aguirre, Juan Pablo, and Swarnali Ahmed Hannan, 2021, “Recoveries After Pandemics: The Role of Policies and Structural Features,” IMF Working Paper No. 21/181 (Washington: International Monetary Fund).
- Dell’Ariccia, Giovanni, Ehsan Ebrahimy, Deniz Igan, and Damien Puy, 2020, “Discerning Good from Bad Credit Booms: The Role of Construction,” IMF Staff Discussion Note No. SDN/20/02.
- Engler, Philipp, Nathalie Pouokam, Diego Rodriguez Guzman, and Irina Yakadina, 2020, “The Great Lockdown: International Risk Sharing Through Trade and Policy Coordination,” IMF Working Paper No. 20/242 (Washington: International Monetary Fund).
- Freund, Caroline, and Frank Warnock, 2007, “Current Account Deficits in Industrial Countries: The Bigger They Are, the Harder They Fall?” in G7 Current Account Imbalances: Sustainability and Adjustment ed. by Richard Clarida (University of Chicago Press).
- International Monetary Fund, 2016, IMF Policy Paper, “Small States’ Resilience to Natural Disasters and Climate Change—Role for the IMF” (Washington).
- International Monetary Fund, 2021a, “World Economic Outlook: Managing Divergent Recoveries,” April 2021 (Washington).
- International Monetary Fund, 2021b. “External Sector Report: Divergent Recoveries and Global Imbalances (Washington).
- Jordà, Òscar, 2005, “Estimation and Inference of Impulse Responses by Local Projections,” American Economic Review, 95(1): 161–82.
- Jordà, Òscar, Moritz Schularick, and Alan M. Taylor, 2017, “Macrofinancial History and the New Business Cycle Facts,” in NBER Macroeconomics Annual 2016, Vol. 31, ed. by Martin Eichenbaum and Jonathan A. Parker (Chicago: University of Chicago Press).
- Jordà, Òscar, Martin Kornejew, Moritz Schularick, and Alan M. Taylor, 2020a, “Zombies at Large? Corporate Debt Overhang and the Macroeconomy,” Federal Reserve Bank of San Francisco Working Paper 2020-36 https://doi.org/10.24148/wp2020-36.
- Jordà, Òscar, Sanjay Singh, and Alan M. Taylor, 2020b, “Longer-Run Economic Consequences of Pandemics,” Federal Reserve Bank of San Francisco, Working Paper 2020-09 https://doi.org/10.24148/wp2020-09.
- Laeven, Luc, and Fabian Valencia, 2020, “Systemic Banking Crises Database II,” IMF Economic Review, Vol. 68, pp. 307–361 https://doi.org/10.1057/s41308-020-00107-3.
- Laframboise, Nicole, and Boileau Loko, 2012, “Natural Disasters: Mitigating Impact, Managing Risks,” IMF Working Paper No. 12/245 (Washington: International Monetary Fund).
- Lane, Philip R., and Gian Maria Milesi-Ferretti, 2012, External Adjustment and the Global Crisis, Journal of International Economics, Vol. 88(2): pp. 252–65.
- Li, Nan, and Mattia Coppo, 2020, “Severe Epidemics in Modern History: Growth, Debt and Civil Unrest,” IMF Research Department Special Series on COVID-19.
- Ma, Chang, John H. Rogers, and Sili Zhou, 2020, “Modern Pandemics: Recession and Recovery,” BOFIT Discussion Paper No. 16 https://ssrn.com/abstract=3565646.
- Milesi-Ferretti, Gian Maria, and Assaf Razin, 1998, “Current Account Reversals and Currency Crises: Empirical Regularities,” NBER Working Paper No. 6620 (Cambridge, Massachusetts: National Bureau of Economic Research).
- Teulings, Coen N., and Nikolay Zubanov, 2014, “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses,” Journal of Applied Econometrics, Vol. 29, No. 3, pp. 497–514.

### Annex — Data sources (key data coverage and definitions)
- Macroeconomic data, including the current account balance and its components, public debt, exchange rates, and relative growth, are from IMF IFS over the period 1960 to 2020. Data on household and non-financial corporate saving are taken from the OECD national accounts statistics database.
- Credit booms are based on Dell’Ariccia et al. 2020 and defined as episode during which the ratio of credit to GDP grows faster than implied by its past trend, with the deviation from trend greater than 1.5 times its standard deviation and the annual growth rate of the credit to-GDP ratio exceeding 10 percent; or the annual growth rate of the credit-to-GDP ratio exceeding 20 percent covering 1970-2016.
- Banking crises are from Laeven and Valencia 2020, covering 1970-2017.
- Sudden stops are defined, in line with the literature (Catão, 2006), as episodes when net private capital inflows are 1½ standard deviation below their mean and the annual decline is ¾ standard deviations from a year earlier (defined in real U.S. dollars); or have declined by at least 3 percentage points of GDP relative to the previous year and 2 percentage points from two years earlier, covering 1991-2018.
- External liabilities are from the Lane and Milesi-Ferretti database on the External Wealth of Nations and foreign currency composition is based on Benetix et al. 2019.
- Historic data on current account, exports and imports to GDP are based on Jordà et al. 2017 (covering 17 countries, 1870-2016, unbalanced). Where current account data are not available but data points on exports and imports to GDP exist, we proxy the current account by net exports.
- Natural disasters are taken from the EM-DAT database by the Centre for Research on the Epidemiology of Disasters (CRED) and cover the years 1900-2019.
- Epidemics are taken from Li and Coppo 2020, based on WHO, ECDC/CDC, and Journal of Infectious Diseases and cover the years 1900-2019.

### Annex — Figures (notes and coverage)
- Annex Figure 1: Sectoral Response to Recessions
  - Notes: Cumulative response in percent of GDP over a 5-year horizon. Dashed lines represent 90 percent confidence interval.
- Annex Figure 2: CA sensitivity over time, 5 year response to recessions, 5 year window
- Annex Figure 3: Trade Response By Income Group (Non-commodity EMs only)
  - Notes: Cumulative response of trade and exchange rates in percent over a 10-year horizon by country group. Dashed lines represent 90 percent confidence interval. Responses of AEs and non-commodity EMs are estimated in split samples.
- Annex Figure 4: Recessions Associated with Banking crises
  - Notes: Cumulative response of CA balances and components in percent of GDP over a 10-year horizon. Dashed lines represent 90 percent confidence interval. Banking crises are taken from Laeven and Valencia (2020), covering the period 1970-2017.
- Annex Figure 5: Density of Public Debt to GDP By Income Group
  - Notes: Lines depict normal density curve (rose) and kernal density estimate (green). AE and EM densities seem surprisingly similar around the mean of 50 percent, which is explained by the unbalanced time horizon. Most AE debt series start in the 1960s when public debt was relatively low while EM debt series are populated from the 1990s onwards.
- Annex Figure 6: Recessions Associated with Sudden Stops
- Annex Figure 7: Recessions and High Foreign Debt Liabilities
- Annex Figure 8: Historical Response to Recessions, 1870-1960
  - Notes: Cumulative response of CA balances and components in percent of GDP over a 10-year horizon. Dashed lines represent 90 percent confidence interval, estimated with Cook’s distance correction. Estimations cover 375 recession country-years, defined as negative real GDP growth, based on historical data from Jordà et al. 2017.

### Annex — Tables (content summary and key estimation details)
- Annex Table 1: Sample Economies
  - Lists Baseline Sample, Advanced Economies, Non-commodity Emerging Markets, Commodity Emerging Markets, and Historical Sample (1870-2016) with country names as presented.
- Annex Table 2: Current Account, Saving, and Investment in AEs and Non-commodity EMs
- Annex Table 3: The Role of Internal Imbalances, Additional Controls
  - Reports panel estimations with variables including Recession, Recession*EM, Δ Exchange rate, Openness, Relative growth, constants, Observations (e.g., 2,597; 2,141; 2,567; 2,111; 2,631; 2,175; 2,035; 1,579; 2,067; 1,611), Number of groups (57 repeated), Within R-squared (e.g., 0.528, 0.545, 0.507, 0.520, 0.616, 0.582, 0.525, 0.539, 0.536, 0.537, 0.626, 0.568). Notes: Standard errors in parentheses. Time and country fixed effects included. *** p<0.01, ** p<0.05, * p<0.1.
  - Additional panels for Investment, CA, Saving with interactions Recession*credit boom, Recession*high debt; Observations examples: 1,324; 1,227; 1,037; 1,327; 1,230; 1,036; 1,330; 1,233; 1,039. Number of groups: 49 repeated. Within R-squared examples: 0.549, 0.553, 0.585, 0.575, 0.598, 0.592, 0.683, 0.614, 0.604.
- Annex Table 4: The Role of External Imbalances, Additional Controls
- Annex Table 5: Combining Internal and External Imbalances, Additional Controls
  - Reports estimations with interactions: Recession*CA deficit, Recession*high external debt, Recession*credit boom, Recession*high debt, Δ Exchange rate, Openness, Relative Growth. Observations examples: 1,821; 1,625; 1,433; 1,815; 1,619; 1,425; 1,809; 1,613; 1,419. Number of groups: 49 repeated. Within R-squared examples: 0.553, 0.606, 0.617, 0.556, 0.580, 0.565, 0.615, 0.599, 0.575.
  - Additional subpanels for CA, Saving, Investment with detailed coefficients and significance levels; Observations examples: 1,324; 1,227; 1,037; 1,319; 1,222; 1,030; 1,322; 1,225; 1,033. Number of groups: 49 repeated. Within R-squared examples: 0.569, 0.606, 0.643, 0.579, 0.609, 0.605, 0.690, 0.631, 0.621.
- Annex Table 6: Synchronization, and Internal and External Imbalances
  - Reports estimations including Recession, Recession*credit boom, Recession*high debt, Recession*CA deficit, Recession*high external debt, Recession*global 25, with Observations examples: 1,480; 1,383; 1,193; 1,441; 1,344; 1,152; 1,440; 1,343; 1,151. Number of groups: 49 repeated. Within R-squared examples: 0.556, 0.593, 0.640, 0.576, 0.598, 0.587, 0.678, 0.619, 0.631.
  - Notes: Standard errors in parentheses. Variables included in the estimation but not shown are first and second lags of Recession and the dependent variable, the vulnerabilities, and time and country fixed effects. *** p<0.01, ** p<0.05, * p<0.1.

*Source: wpiea2021198-print-pdf - REFERENCES (IMF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021198-print-pdf.pdf_