## 1. International Investment Position and Income Flows

## Source details

**Canonical URL:** [1. International Investment Position and Income Flows](https://www.imf.org/-/media/files/publications/cr/2021/english/1colea2021002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2021/english/1colea2021002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2021/english/1colea2021002.pdf.json)

---

### Overview
- Colombia’s income balance improved in 2020, reaching a surplus of 1¼ percent of GDP in 2020.
- Resilient remittances boosted secondary income above 3 percent of GDP in 2020.
- The largest contribution to the 2020 improvement came from primary/investment income, which improved by more than 1 percent of GDP to -2 percent of GDP.

### Key findings on income balance and trade balance
- Correlations:
  - Trade balance vs income balance: almost -0.8.
  - Trade balance vs primary income balance: -0.9.
- Recent dynamics:
  - In 2019 and 2020, an improving income balance partially offset a deteriorating trade balance.
  - Historical pattern: during global economic and commodity price downturns (including the 2014-15 oil price decline), investment income receipts and expenditures both declined; as trade and activity recovered, investment income expenditures rebounded and the income balance deteriorated.
- Forecast:
  - Colombia’s income balance is forecast to worsen in 2021 and thereafter.

### NIIP, gross investment income flows, and asymmetry
- Gross investment income (credit + debit) and asymmetry (credit − debit) relative position:
  - Gross flows averaged 7 percent of GDP in 2017-19.
  - Asymmetry averaged 45 percent of gross flows in 2017-19.
- Investment income balance:
  - Averaged -3 percent of GDP for 2017-19.
  - The -3 percent of GDP balance is on par with what a bivariate cross-country regression with the NIIP would predict, but the balance is less negative during downturns.

### Returns, return differentials, and implications
- Implied annual returns:
  - Return on Colombia’s assets held abroad: broadly constant since before the GFC at 3 percent on average; fell slightly to 2¼ percent in 2020.
  - Return on liabilities held by non-residents: declined from 10 percent in 2006-8 to 5 percent in 2017-9 and to about 2½ percent in 2020.
- Interpretation:
  - Declining return differentials reflect lower risk premia and/or lower profitability of FDI (partly driven by lower commodity prices).
  - As a result, a steadily more negative NIIP has not yet translated into a more negative primary income balance.

### Composition of liabilities and risk-sharing properties
- Liability composition and dynamics:
  - Foreign Direct Investment (predominantly equity) accounts for more than half of the increase in foreign liabilities.
  - Portfolio (predominantly debt) and other (predominantly loans) liabilities also grew.
- Income payments variation:
  - Most variation occurred within direct investment (e.g., oil profit payments to non-residents were large in commodity boom years, negligible after the 2014-16 oil price crash, and moderate since).
  - The rise in external debt induced by the oil price crash translated into higher interest payments to foreigners who lent when national income was hurt by the shock.

### Foreign-currency positions and valuation
- Foreign-currency denomination shares:
  - Share of foreign liabilities denominated in foreign currency: 40 percent in Q3 of 2020 (including 37 percent in U.S. dollars).
  - Focusing on foreign debt liabilities: FX share was 83 percent in 2020 Q3 compared to almost 100 percent in 2000.
- Net foreign-currency position:
  - Combining FX exposure shares yields a net long foreign currency position of 16 percent of GDP in 2019.
  - In 2020, the net long foreign currency position is estimated to have increased to 21 percent of GDP despite a deterioration in the nominal NIIP.

### Exchange rate channels and effects on NIIP and income balance
- Three channels determining depreciation effects on income:
  i. Debtor countries’ negative income balance tends to become more negative after a depreciation because expenditures rise more than receipts in local currency terms.
  ii. A higher proportion of external assets denominated in foreign currency (as in Colombia) implies a larger local-currency increase in receipts following a depreciation, working in the opposite direction to (i).
  iii. Economic and behavioral effects (e.g., depreciation boosts domestic activity and profits, leading to larger income outflows; currency volatility can influence profit management).
- Quantitative illustration:
  - In Colombia, a depreciation of 10 percent would improve the NIIP by about 2 percent of GDP through valuation gains.
  - Annual income balance effects are smaller than stock valuation effects and vary year-to-year.

### Income balance semi-elasticities (mechanical and with behavioral effects)
- Definition (as in source): semi-elasticity of the investment income balance in pesos (IB/GDP) with respect to the nominal exchange rate E (pesos per unit of foreign currency): (∂IB/GDP)/(∂E/E) = η_R S_R − η_P S_P, where S_R and S_P are income receipts and payments as shares of GDP and η_R and η_P are shares of income receipts and payments denominated in foreign currency (mechanical/valuation exposure); incorporating behavioral effects relaxes mechanical assumptions.
- Colombia (Table 1 values):
  - 2019:
    - Mechanical only: 0.013
    - With possible behavioral effects: 0.010
    - Illustration: the 2019 depreciation improved the income balance by 0.14 percent of GDP (using the mechanical semi-elasticity).
  - 2020:
    - Mechanical only: 0.006
    - With possible behavioral effects: 0.004
- Comparative scale:
  - Semi-elasticities are small compared to trade balance semi-elasticity (0.12-0.13).
  - Incorporating economic/behavioral effects could make a depreciation less favorable to the income balance; estimated behavioral adjustments are included but not robustly significant across specifications.

### Secondary income and remittances
- Emigrant stock:
  - Grew from 1.4 million in 2000 to 2.6 million in 2010.
  - Grew by 0.1 million to 2015 and by 0.3 million to reach an estimated 3 million people in 2020 (United Nations).
- Remittances dynamics:
  - Rising steadily as a share of GDP since 2014; bottomed near 1 percent in 2014.
  - In 2020, remittances rose by ½ percent of GDP to 2½ percent of GDP and rose even in nominal U.S. dollar terms (by 3 percent).
  - Push factors in 2020: growth in U.S. remittances outflows of 2¼ percent in the first nine months of 2020 globally (linked to U.S. policies supporting household incomes).
  - Pull factors: Colombia’s relatively sharp economic contraction in 2020 and modest growth since the 2014 oil price decline may have prompted countercyclical remittances; evidence suggests remittances to Colombia are altruistic and countercyclical.

### Policy implications and considerations
- With persistent current deficits and the risk of higher global interest rates, fiscal and structural reforms would help preserve favorable external dynamics.
- Colombia’s positive foreign-currency position suggests no adverse income or balance sheet effects from depreciations so far, but ongoing monitoring is warranted given NIIP deterioration and potential valuation and income channels.
- Exchange rate policy:
  - Benefits to the current account through the investment income channel are at best negligible and fluctuate.
  - Valuation gains on the stock (NIIP) from depreciation can be meaningful (e.g., 10 percent depreciation → ~2 percent of GDP NIIP improvement), but annual income flow benefits are much smaller.

---

### Depreciations can complement the stabilizing role of remittances inflows. In peso

### Remittances, exchange rates, and stabilization
- In peso terms, inflows of remittances grew by 16 percent in 2020.
- Evidence and empirical references:
  - Yang (2008) estimate: a one percent appreciation in the home country increases remittances by 0.6 percent in recipient-country currency (implying a reduction in home-country currency terms).
  - Garavito-Acosta and others (2019): find no significant effect of the exchange rate on remittances measured in U.S. dollars for Colombia, which implies a depreciation would raise remittances in local currency or as a share of GDP.
  - Hakura and others (2009): provide evidence on the stabilizing effects of remittances on output.
- Policy interpretation:
  - Exchange rate flexibility in response to external shocks can help stabilize the current account through the secondary income channel and provide additional income to support domestic economic activity, reducing its volatility.

### NIIP and income flow risks; policy-relevant mitigations
- Short-to-medium term supportive factors:
  - Resilient remittances.
  - Success in securing funding in local currency.
  - Risk sharing properties of equity liabilities.
  - Lower global interest rates helping contain income outflows.
- Recent impact:
  - These properties contributed to the decline in Colombia’s current account deficit in 2020.
- Risks as the economy recovers:
  - The current account deficit is expected to rise towards 4 percent of GDP over the medium term, which would reinforce the decline in the NIIP and could increase income outflows.
  - If global interest rates normalize and if Colombia’s risk premium were to rise, income outflows would increase further.
- Policy-relevant mitigation:
  - Resuming the downward trend in FX liabilities would keep adverse balance sheet and income balance effects of depreciations at bay.

### Current account, fiscal and structural policy implications
- Colombia’s current account deficit underscores the importance of fiscal and structural reforms to boost competitiveness.
- Role of the flexible exchange rate:
  - Can help stabilize the current account modestly through the trade balance channel without adverse income balance effects.
- Fiscal reforms would support narrowing the current account deficit by reducing internal dissaving and preserving investor appetite for local-currency debt and low risk premia.
- Structural reforms to reduce the current account deficit could include:
  - Aid export diversification and competitiveness.
  - Close infrastructure gaps.
  - Lower transport costs.
  - Streamline non-tariff barriers.
  - Improve customs procedures.

---

### Labor market impact from COVID-19 and the uneven recovery

### A. COVID-19 impact: overview and macro labor outcomes
- Employment disruption:
  - At the height of pandemic-induced lockdowns, about a quarter of total employment was disrupted.
- Unemployment and participation:
  - Pre-pandemic unemployment: 12 percent.
  - At peak pandemic: above 20 percent (a historical record).
  - Labor force participation rate collapsed from 63 percent to a low of 52 percent in April 2020.
- Comparative losses:
  - Net employment losses in 2020Q1 (as a share of total employment) were higher than Brazil, Mexico and Chile but lower than Peru.
- Hours and transitions:
  - Hours worked fell significantly; majority of those who lost jobs transitioned directly out of the labor force.

### B. Distributional effects: inequality, demographics, informality
- Inequality effects:
  - Employment and income losses worsened inequality via decreases in mean and median household earnings and a leftward shift in household income distribution.
- Groups most affected:
  - Poorer households.
  - Those with low levels of education.
  - Informal workers or self-employed.
  - Women, the young, and low-educated in informal employment were most affected.
- Gender specifics:
  - Women comprised about 45 percent of employed but accounted for 52 percent of total job losses and 58 percent of those in the informal sector.
  - Only 46 percent of informal sector jobs regained during the rebound were regained by women.

### C. The transfer policy response
- Pre-pandemic coverage:
  - Over half of poorer households covered by some form of transfer before COVID-19.
  - Familias en Accion program: most important transfer mechanism in coverage and magnitude.
- Pandemic-era measures:
  - Emergency transfers and employment protection measures worth around 1.3 percent of GDP in 2020.
  - Wage subsidies equivalent to 40 percent of the minimum wage for formal workers at firms experiencing over 20 percent in revenue losses.
  - New unconditional transfer program (Ingreso Solidario) aimed at 3 million households at risk not covered under other programs.
- Targeting and coverage outcomes:
  - VAT-return transfer program focused on poorest among Familias en Accion participants and achieved better targeting.
  - Household surveys indicate a significant portion of transfers were allocated to households not among the poorest income deciles even before the pandemic, indicating room to improve focalization.
  - Expansion of coverage concentrated among poorer households but also reached above-median households due to program design (e.g., Ingreso Solidario).
  - Majority of those reporting job loss due to COVID-19 were not enrolled in any government transfer program.

### D. A swift but uneven rebound
- Labor market rebound in H2 2020:
  - Labor force participation snapped back 10 percentage points to 62 percent.
  - Unemployment decreased by near 7 percentage points to 13 percent by end-2020.
  - Over 75 percent of total employment recovery was from workers returning directly to employment—likely pre-pandemic jobs.
- Remaining gaps and setbacks:
  - Total number of unemployed remained 2-3 percent above pre-COVID level.
  - As reopening slowed, employment gains slowed; new lockdowns in January 2021 caused employment gains to stop, labor force participation to decline, and unemployment to rise to 17 percent.
- Informality role:
  - Largest job losses and fastest rebound occurred in the informal sector (self-employment).
  - Informal sector adjusted labor prices and hours faster; acted as shock absorber but recovery had gender imbalance with women slower to return to labor force.
- Sectoral status:
  - Most of the economy remained below pre-lockdown employment except real estate, agriculture, information and communication, and the public sector.
- Outstanding gap:
  - At end-2020, there was an outstanding 1.4 million job gap relative to pre-COVID employment level.
  - Education sector employment remained depressed due to school closures.

### E. Recovery at risk and employment-at-risk analysis
- Method:
  - Employment at risk uses sector-specific demand shocks, sector-specific supply shocks from lockdowns, occupation and telework capacity, upstream/downstream impacts, and firm-size differentials.
  - Computation uses individuals employed between August and October 2020 weighted by job loss probability.
- Key results:
  - Significant jobs remain at risk from prolonged lockdowns.
  - Risks concentrated among small firms or self-employed in sectors: construction, tourism, transportation and other selected services.
  - Employment at risk by firm type and scenario (percent of total employment): self-employed up to about 3.5 percent under prolonged full restrictions (visual scale); small, medium, large firms show lower shares (visual scale).
  - Employment at risk by sector under prolonged full and partial restrictions shows highest exposure in Hospitality, Trade, Transport, Construction, and selected services (visual scale).
- Policy implication:
  - Prolonged economic shutdowns can give rise to renewed employment losses and small-firm led insolvency problems of significant scale.

### F. Discussion and policy guidance
- Constraints to recovery:
  - Informal jobs absorbed much of the rebound, but these sectors remain vulnerable.
  - Formal job gains are slower; unemployment remains somewhat sticky.
  - A sizable share of women have not returned to the labor force due to school closures and unequal child/household care burdens.
  - Potential for more permanent losses in potential output.
- Policy recommendations and priorities:
  - Prompt control of the pandemic through a successful vaccination campaign to underpin a stronger cyclical recovery.
  - Continued fiscal, monetary, micro- and macro-prudential policy support and vigilant monitoring of risks.
  - Support measures targeted to vulnerable firms and workers, especially small firms and self-employed in pandemic-sensitive sectors, to avoid permanent job losses and insolvent firms.
  - Attention to labor market adjustment and structural rigidities to facilitate formal job recovery and mitigate a shift toward greater informality.

---

### High unemployment and sluggish recovery of the formal sector highlight need to address regulatory barriers

### Labor market and formal sector challenges
- Structural unemployment:
  - Colombia had a structural unemployment problem with rates above nine percent over the last decade.
- Identified barriers:
  - Comparatively high non-wage labor costs.
  - High minimum-to-median wage ratios (chart referenced).
  - Formal business start-up costs.
  - Regulatory barriers for full migrant-integration into formal markets (currently being tackled by recently announced migrant regularization measures).
- Policy objective:
  - Speed up formal labor market recovery, facilitate short-run adjustment, and lower structural levels of informality and unemployment in the medium term.
  - Transition of the workforce towards formal stable jobs is highlighted as the main challenge for achieving a more inclusive growth path.

### Pandemic impact channels on potential output
- Main channels:
  - Labor market channels:
    - Declines in labor inputs from higher unemployment and lower participation.
    - Loss of worker skills during unemployment and mismatches between workers’ skills and jobs.
    - A quarter of employed persons were temporarily displaced into unemployment or out of the labor force (source: labor market SIP).
    - Lower migration net inflows from Venezuela in 2020 reduced working age population; Venezuelan migrants tend to have higher human capital than local workers (2020 SIP).
  - Investment channels:
    - Sharp fall in investment and capital underutilization due to firm destruction or halted production amid mobility restrictions.
  - Sectoral channels:
    - Sectoral reallocation and network effects causing temporary underutilization of capital and labor, affecting TFP.
    - Cuts to R&D could reduce future productivity.

### Historical and comparative context
- 2020 recession:
  - The recession in 2020 was Colombia’s largest on record.
  - Comparable investment falls:
    - -40 percent year-on-year in 2020Q2 (pandemic peak).
    - -41 percent year-on-year in 1999Q2 (1999-2000 recession).
- Unemployment history:
  - Unemployment saw gradual increases starting in the mid-1990s, peaking in the 1999-2000 recession and rising moderately after the global financial crisis of 2008-09 and the oil price drop.

---

### Potential output: reassessment and pandemic impact

### Methodologies to estimate potential output
- Three approaches compared:
  - Production function (PF): Cobb-Douglas Y = TFP(K^(1−α) L^α). HP filter smooths labor input, capital services, and TFP; potential growth computed from aggregate production function.
  - Multivariate filter (MVF): Reduced-form relationships (Phillips curve, Okun’s law, CAPU) to identify slack and potential output from output, inflation, unemployment rate and CAPU. Estimated on annual data.
  - Semi-structural new Keynesian model (SSNK): Similar to MVF, includes Phillips curve and a role for oil prices; utilizes quarterly data and is augmented to include a “COVID-19” shock that lowers potential output and opens the output gap.
- Structural break corrections:
  - Potential output estimates prior to the pandemic are estimated using pre-pandemic data for MVF and SSNK and including January 2020 WEO projections for PF.
  - For estimating 2020 impact, MVF and SSNK use latest information on output, inflation, exchange rate, interest rate; MVF also uses unemployment and CAPU, SSNK uses oil price and risk premia. For PF, WEO projections for output, capital and labor are used with filters on capital, labor and TFP.

### Quantitative estimates of pandemic impact and recovery (percent)
- 2020 estimated contraction in potential output (estimates range by method):
  - Production function (PF): -3½ percent.
  - Multivariate filter (MVF): around -1 percent.
  - SSNK: around -½ percent.
- Drivers:
  - In PF and MVF: fall in labor input and, to a lesser extent, lower capital.
- Output gaps (2020):
  - MVF: opening output gap of around 7 percent.
  - SSNK: opening output gap of around 8 percent.
  - PF: less open output gap of around -4½ percent.
- 2020 component estimates (Percent):
  - PF: Y* -3.7; K* (cont.) -0.6; L* (cont.) -4.6; TFP 1.5.
  - MVF: Y* -1.1; K* (cont.) -0.9; L* (cont.) -1.6; TFP 1.3.
  - SSNK: Y* -0.3; K* na; L* na; TFP na.
- 2021 projected partial recovery in potential output:
  - PF: ½ percent (listed as 0.7 in component table).
  - MVF: 1.5 percent (component table shows 1.5).
  - SSNK: 2½ percent (component table shows 2.7).
- 2021 component estimates (Percent):
  - PF: Y* 0.7; K* (cont.) -0.1; L* (cont.) 0.9; TFP 0.0.
  - MVF: Y* 1.5; K* (cont.) -0.1; L* (cont.) 1.0; TFP 0.6.
  - SSNK: Y* 2.7; K* na; L* na; TFP na.
- Medium-term outlook:
  - Potential output growth is expected to gradually return to its pre-pandemic growth rate of between 3 and 3½ percent.

### Projections and model inputs
- Capital projections:
  - Benchmarked using an estimated investment equation including the price of oil, GDP and the user cost of capital.
- Labor projections:
  - Okun’s law and a vector autoregression (VAR) were estimated. Regressions estimated to 2019Q4; forecasts for 2021-26 use observed data for 2020, exogenous forecasts for the price of oil, GDP and the user cost of capital.
- Migration assumptions:
  - In 2020 the number of migrants from Venezuela in Colombia declined slightly by around 40 thousand as some migrants returned to Venezuela during the national lockdown.
  - The reduction represents around 2 percent of all migrants.
  - Staff assume the border with Venezuela will open in the second half of 2021 with modest arrivals thereafter relative to pre-pandemic trends.
- Projections point to:
  - A gradual increase in the investment to GDP ratio.
  - A gradual increase in the participation rate.
  - The unemployment rate projected to decline slowly towards pre-pandemic levels.

### Modeling specifics (annex highlights)
- VAR/VECM for capital and labor inputs:
  - Estimated variables: capital stock, gross fixed capital formation (investment), GDP and the price of oil (assumed exogenous).
  - Cointegration tests indicate two cointegrating relationships, identified using restrictions inspired by Ellis and Price (2002):
    - k = y − σ uc + θ
    - k = i − ln(δ + g)
  - Sample (adjusted): 2005Q3–2019Q4. Included observations: 58 after adjustments.
- Key VECM reported statistics (selected):
  - CointEq1 error correction coefficients: D(LOG(INV)) 0.344568; D(LOG(GDP)) −0.053629; D(LOG(K)) −0.021560; D(UC) 0.051414.
  - DLOG(POIL(-1)) coefficients: −0.021542 for D(LOG(INV)); 0.010224 for D(LOG(GDP)); −0.004273 for D(LOG(K)); 0.014421 for D(UC).
  - R-squared: D(LOG(INV)) 0.271359; D(LOG(GDP)) 0.323323; D(LOG(K)) 0.666137; D(UC) 0.164513.
  - Adj. R-squared: D(LOG(INV)) 0.185636; D(LOG(GDP)) 0.243714; D(LOG(K)) 0.626859; D(UC) 0.066221.

---

### VAR comprising GDP, unemployment and participation; MVF and SSNK model notes

### VAR identification and impulse responses
- Identification:
  - Cholesky and long-run identification (Blanchard-Quah) schemes are considered.
- Dynamics:
  - Under both identification techniques, unemployment and participation both take a long time to return to pre-shock levels (over twenty quarters).
- Interpretation of shocks:
  - First column of Cholesky and second column of long-run restrictions (Shock 2): consistent with demand-type shocks.
  - First column of long-run restriction (Shock 1): consistent with a supply-type shock.
  - Other shocks could be unemployment or labor force participation shocks.
- All columns point to sluggish responses in the labor market.
- Responses reported as: Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E.; Response to Structural VAR Innovations ± 2 S.E. (Long-Run Restrictions).

### MVF model (key equations and features)
- Based on Alichi and others (2017).
- Uses GDP, inflation, unemployment and capacity utilization.
- Key supply-side equations (presented in source):
  - y*t = y*t−1 − ε1(u*t − u*t−1) + ε2(capu*t − capu*t−1) + g*t + εty
  - g*t = (1 − γ)g*t−1 + γgss* + εtg
- Definitions:
  - y* potential output; u* NAIRU; capu* equilibrium capacity utilization; g* potential growth; ε shocks.

### SSNK model (key equations and features)
- Based on Gonzalez and others (2015).
- Uses GDP, inflation, and oil prices; estimated on quarterly data.
- Key supply-side equations (presented in source):
  - y*t = y*t−1 + g*t
  - g*t = (1 − γ)g*t−1 + γgss* + ω(∆poil*t − ∆poilss*) + εtg
- Features:
  - ∆poil is growth of the equilibrium oil price (autoregressive).
  - Incorporates a Covid-19 specific shock affecting demand, potential output and inflation.

### Model implementation and data used for potential output identification
- Inputs used:
  - Both models use data and projections for GDP, inflation, the policy rate and the exchange rate.
  - MVF additionally uses unemployment and capacity utilization.
  - SSNK additionally uses the oil price.
- Purpose:
  - Models use shocks and projections to infer impacts on potential output and related variables.

*Source: COLOMBIA — INTERNATIONAL MONETARY FUND, March 9, 2021.*

### 1. International Investment Position and Income Flows ____________________________________ 5

### 1. International Investment Position and Income Flows

### Overview
- Colombia’s income balance improved in 2020, reaching a surplus of 1¼ percent of GDP in 2020.
- Resilient remittances boosted secondary income above 3 percent of GDP in 2020.
- The largest contribution to the 2020 improvement came from primary/investment income, which improved by more than 1 percent of GDP to -2 percent of GDP.

### Key findings on income balance and trade balance
- The correlation between Colombia’s trade balance and income balance is almost -0.8; the correlation between its trade and primary income balance is -0.9.
- In 2019 and 2020, an improving income balance partially offset a deteriorating trade balance.
- Historical pattern: during global economic and commodity price downturns (including the 2014-15 oil price decline), investment income receipts and expenditures both declined, and as trade and activity recovered, investment income expenditures rebounded and the income balance deteriorated.
- Forecast: Colombia’s income balance is forecast to worsen in 2021 and thereafter.

### NIIP, gross investment income flows, and asymmetry
- Gross investment income flows (credit + debit) and their asymmetry (credit - debit) are roughly in the middle of the range in a sample of countries.
- Gross flows averaged 7 percent of GDP and the asymmetry averaged 45 percent of gross flows in 2017-19.
- Colombia’s investment income balance averaged -3 percent of GDP for 2017-19.
- The investment income balance of -3 percent of GDP is on par with what a bivariate cross-country regression with the NIIP would predict, but the balance is less negative during downturns.

### Returns, return differentials, and implications
- Implied annual return on Colombia’s assets held abroad:
  - Broadly constant since before the GFC at 3 percent on average
  - Fell slightly to 2¼ percent in 2020
- Return on liabilities held by non-residents:
  - Declined from 10 percent in 2006-8 to 5 percent in 2017-9 and to about 2½ percent in 2020
- Declining return differentials reflect lower risk premia and/or lower profitability of FDI (partly driven by lower commodity prices).
- As a result, a steadily more negative NIIP has not yet translated into a more negative primary income balance.

### Composition of liabilities and risk-sharing properties
- Foreign Direct Investment (predominantly equity) accounts for more than half of the increase in foreign liabilities, though portfolio (predominantly debt) and other (predominantly loans) liabilities also grew.
- Most variation in income payments occurred within direct investment (e.g., oil profit payments to non-residents were large in commodity boom years, negligible after the 2014-16 oil price crash, and moderate since).
- The rise in external debt induced by the oil price crash translated into higher interest payments to foreigners who lent when national income was hurt by the shock.

### Foreign-currency positions and valuation
- Colombia’s share of foreign liabilities denominated in foreign currency was 40 percent in Q3 of 2020 (including 37 percent in U.S. dollars).
- Focusing on foreign debt liabilities, the FX share was 83 percent in 2020 Q3 compared to almost 100 percent in 2000.
- Combining FX exposure shares yields a net long foreign currency position of 16 percent of GDP in 2019.
- In 2020, the net long foreign currency position is estimated to have increased to 21 percent of GDP despite a deterioration in the nominal NIIP.

### Exchange rate channels and effects on NIIP and income balance
- Three channels determine net effect of depreciation on income:
  i. Debtor countries’ negative income balance tends to become more negative after a depreciation because expenditures rise more than receipts in local currency terms.
  ii. A higher proportion of external assets denominated in foreign currency (as in Colombia) implies a larger local-currency increase in receipts following a depreciation, working in the opposite direction to (i).
  iii. Economic and behavioral effects (e.g., depreciation boosts domestic activity and profits, leading to larger income outflows; currency volatility can influence profit management).
- In Colombia, a depreciation of 10 percent would improve the NIIP by about 2 percent of GDP through valuation gains.
- Annual income balance effects are smaller than stock valuation effects and vary year-to-year.

### Income balance semi-elasticities (mechanical and with behavioral effects)
- Definition: semi-elasticity of the investment income balance in pesos (IB/GDP) with respect to the nominal exchange rate E (pesos per unit of foreign currency): (∂IB/GDP)/(∂E/E) = η_R S_R − η_P S_P, where S_R and S_P are income receipts and payments as shares of GDP and η_R and η_P are shares of income receipts and payments denominated in foreign currency (mechanical/valuation exposure); incorporating behavioral effects relaxes mechanical assumptions.
- For Colombia (Table 1 values):
  - 2019:
    - Mechanical only: 0.013
    - With possible behavioral effects: 0.010
    - Illustration: the 2019 depreciation improved the income balance by 0.14 percent of GDP (using the mechanical semi-elasticity).
  - 2020:
    - Mechanical only: 0.006
    - With possible behavioral effects: 0.004
- Semi-elasticities are small compared to trade balance semi-elasticity (0.12-0.13).
- Incorporating economic/behavioral effects could make a depreciation less favorable to the income balance; estimated behavioral adjustments are included but not robustly significant across specifications.

### Secondary income and remittances
- Stock of emigrants:
  - Grew from 1.4 million in 2000 to 2.6 million in 2010.
  - Grew by 0.1 million to 2015 and by 0.3 million to reach an estimated 3 million people in 2020 (United Nations).
- Remittances:
  - Rising steadily as a share of GDP since 2014; bottomed near 1 percent in 2014.
  - In 2020, remittances rose by ½ percent of GDP to 2½ percent of GDP and rose even in nominal U.S. dollar terms (by 3 percent).
  - Push factors in 2020: growth in U.S. remittances outflows of 2¼ percent in the first nine months of 2020 globally (linked to U.S. policies supporting household incomes).
  - Pull factors: Colombia’s relatively sharp economic contraction in 2020 and modest growth since the 2014 oil price decline may have prompted countercyclical remittances; evidence suggests remittances to Colombia are altruistic and countercyclical.

### Policy implications and considerations
- With persistent current deficits and the risk of higher global interest rates, fiscal and structural reforms would help preserve favorable external dynamics.
- Colombia’s positive foreign-currency position suggests no adverse income or balance sheet effects from depreciations so far, but ongoing monitoring is warranted given NIIP deterioration and potential valuation and income channels.
- Exchange rate policy: benefits to the current account through the investment income channel are at best negligible and fluctuate; valuation gains on the stock (NIIP) from depreciation can be meaningful (e.g., 10 percent depreciation → ~2 percent of GDP NIIP improvement), but annual income flow benefits are much smaller.

*Source: COLOMBIA — INTERNATIONAL MONETARY FUND, March 9, 2021.*

### 12.      Depreciations can complement  the stabilizing role of remittances inflows. In peso

### Depreciations can complement the stabilizing role of remittances inflows. In peso

### Remittances, exchange rates, and stabilization
- In peso terms, inflows of remittances grew by 16 percent in 2020.
- Evidence indicates depreciations in the home relative to host country can increase remittances inflows in local currency or as a share of GDP, though the magnitude depends on whether remittances are more sticky in host or home currency terms.
- Exchange rate flexibility in response to external shocks (for example, commodity prices) can help stabilize the current account through the secondary income channel and provide additional income to support domestic economic activity, thus reducing its volatility.
- Empirical references from the source:
  - Yang (2008) estimate: a one percent appreciation in the home country increases remittances by 0.6 percent in recipient-country currency (implying a reduction in home-country currency terms).
  - Garavito-Acosta and others (2019) find no significant effect of the exchange rate on remittances measured in U.S. dollars for Colombia, which implies a depreciation would raise remittances in local currency or as a share of GDP.
  - Hakura and others (2009) provide evidence on the stabilizing effects of remittances on output.

### Net international investment position (NIIP) and income flow risks
- Short-to-medium term factors that have supported Colombia:
  - Resilient remittances.
  - Success in securing funding in local currency.
  - Risk sharing properties of equity liabilities.
  - Lower global interest rates helping contain income outflows.
- Recent impact:
  - These properties contributed to the decline in Colombia’s current account deficit in 2020.
- Risks as the economy recovers:
  - The current account deficit is expected to rise towards 4 percent of GDP over the medium term, which would reinforce the decline in the NIIP and could increase income outflows.
  - If global interest rates normalize and if Colombia’s risk premium were to rise, income outflows would increase further.
  - Policy-relevant mitigation: resuming the downward trend in FX liabilities would keep adverse balance sheet and income balance effects of depreciations at bay.

### Current account, fiscal and structural policy implications
- Colombia’s current account deficit underscores the importance of fiscal and structural reforms to boost competitiveness.
- Role of the flexible exchange rate:
  - Can, if needed, help stabilize the current account modestly through the trade balance channel without adverse income balance effects.
- Fiscal reforms would support narrowing the current account deficit by reducing internal dissaving and preserving investor appetite for local-currency debt and low risk premia.
- Structural reforms to reduce the current account deficit could include:
  - Aid export diversification and competitiveness.
  - Close infrastructure gaps.
  - Lower transport costs.
  - Streamline non-tariff barriers.
  - Improve customs procedures.

---

### Labor market impact from COVID-19 and the uneven recovery

### A. COVID-19 impact: overview and macro labor outcomes
- At the height of pandemic-induced lockdowns, about a quarter of total employment was disrupted.
- Unemployment rate dynamics:
  - Pre-pandemic unemployment: 12 percent.
  - At peak pandemic: above 20 percent (a historical record).
- Labor force participation rate collapsed from 63 percent to a low of 52 percent in April 2020.
- Net employment losses in 2020Q1 (as a share of total employment) were higher than Brazil, Mexico and Chile but lower than Peru.
- Hours worked fell significantly; majority of those who lost jobs transitioned directly out of the labor force.

### B. Distributional effects: inequality, demographics, informality
- Employment and income losses worsened inequality via decreases in mean and median household earnings and a leftward shift in household income distribution.
- Job and labor income losses concentrated among:
  - Poorer households.
  - Those with low levels of education.
  - Informal workers or self-employed.
- Disproportionate impacts:
  - Women, the young, and low-educated in informal employment were most affected.
  - Women comprised about 45 percent of employed but accounted for 52 percent of total job losses and 58 percent of those in the informal sector.
  - Only 46 percent of informal sector jobs regained during the rebound were regained by women.
- Sectoral outcomes:
  - Construction and hospitality led both employment losses in H1 2020 and recovery in H2 2020.
  - At end-2020, there was an outstanding 1.4 million job gap relative to pre-COVID employment level.
  - Education sector employment remained depressed due to school closures.

### C. The transfer policy response
- Pre-pandemic coverage:
  - Over half of poorer households covered by some form of transfer before COVID-19.
  - Familias en Accion program: most important transfer mechanism in coverage and magnitude.
- Pandemic-era measures and scale:
  - Emergency transfers and employment protection measures worth around 1.3 percent of GDP in 2020.
  - Wage subsidies equivalent to 40 percent of the minimum wage for formal workers at firms experiencing over 20 percent in revenue losses.
  - New unconditional transfer program (Ingreso Solidario) aimed at 3 million households at risk not covered under other programs.
- Targeting and coverage outcomes:
  - VAT-return transfer program focused on poorest among Familias en Accion participants and achieved better targeting.
  - Household surveys indicate a significant portion of transfers were allocated to households not among the poorest income deciles even before the pandemic, indicating room to improve focalization.
  - Expansion of coverage concentrated among poorer households but also reached above-median households due to program design (e.g., Ingreso Solidario).
  - Majority of those reporting job loss due to COVID-19 were not enrolled in any government transfer program.

### D. A swift but uneven rebound
- Labor force participation and unemployment rebound in H2 2020:
  - Labor force participation snapped back 10 percentage points to 62 percent.
  - Unemployment decreased by near 7 percentage points to 13 percent by end-2020.
  - Over 75 percent of total employment recovery was from workers returning directly to employment—likely pre-pandemic jobs.
- Stickiness and risks:
  - Total number of unemployed remained 2-3 percent above pre-COVID level.
  - As reopening slowed, employment gains slowed; new lockdowns in January 2021 caused employment gains to stop, labor force participation to decline, and unemployment to rise to 17 percent.
- Informality as an adjustment margin:
  - Largest job losses and fastest rebound occurred in the informal sector (self-employment).
  - Informal sector adjusted labor prices and hours faster; acted as shock absorber but recovery had gender imbalance with women slower to return to labor force.
- Sectoral employment status:
  - Most of the economy remained below pre-lockdown employment except real estate, agriculture, information and communication, and the public sector.

### E. Recovery at risk and employment-at-risk analysis
- Method and drivers:
  - Employment at risk uses sector-specific demand shocks, sector-specific supply shocks from lockdowns, occupation and telework capacity, upstream/downstream impacts, and firm-size differentials.
  - Employment at risk computed using individuals employed between August and October 2020 weighted by job loss probability.
- Key results:
  - Significant jobs remain at risk (in terms of income, hours, or total job loss) from prolonged lockdowns.
  - Risks concentrated among small firms or self-employed in sectors: construction, tourism, transportation and other selected services.
  - Employment at risk by firm type and scenario (percent of total employment):
    - Self-employed: up to about 3.5 percent under prolonged full restrictions (visual scale).
    - Small firms, medium firms, large firms: varying lower percentages (visual scale).
  - Employment at risk by sector under prolonged full and partial restrictions (visual scale) shows highest exposure in Hospitality, Trade, Transport, Construction, and selected services.
- Policy implication:
  - Prolonged economic shutdowns can give rise to renewed employment losses and small-firm led insolvency problems of significant scale.

### F. Discussion and policy guidance
- Recovery faces resistance due to cyclical and structural factors:
  - Informal jobs absorbed much of the rebound, but these sectors remain vulnerable.
  - Formal job gains are slower; unemployment remains somewhat sticky.
  - A sizable share of women have not returned to the labor force due to school closures and unequal child/household care burdens.
  - Potential for more permanent losses in potential output.
- Policy recommendations and priorities:
  - Prompt control of the pandemic through a successful vaccination campaign to underpin a stronger cyclical recovery.
  - Continued fiscal, monetary, micro- and macro-prudential policy support and vigilant monitoring of risks.
  - Support measures targeted to vulnerable firms and workers, especially small firms and self-employed in pandemic-sensitive sectors, to avoid permanent job losses and insolvent firms.
  - Attention to labor market adjustment and structural rigidities to facilitate formal job recovery and mitigate a shift toward greater informality.

*International Monetary Fund — COLOMBIA country report excerpts*

### 14.      High unemployment and sluggish

### High unemployment and sluggish recovery of the formal sector highlight need to address regulatory barriers

### Labor market and formal sector challenges
- Colombia had a structural unemployment problem with rates above nine percent over the last decade.
- Key regulatory and cost barriers identified:
  - Comparatively high non-wage labor costs.
  - High minimum-to-median wage ratios (chart referenced).
  - Formal business start-up costs.
  - Regulatory barriers for full migrant-integration into formal markets (currently being tackled by recently announced migrant regularization measures).
- Policy objective: speed up formal labor market recovery, facilitate short-run adjustment, and lower structural levels of informality and unemployment in the medium term.
- Transition of the workforce towards formal stable jobs is highlighted as the main challenge for achieving a more inclusive growth path.

### Pandemic impact channels on potential output
- The pandemic likely affected potential output primarily through:
  - Labor market channels:
    - Declines in labor inputs from higher unemployment and lower participation.
    - Loss of worker skills during unemployment and mismatches between workers’ skills and jobs.
    - A quarter of employed persons were temporarily displaced into unemployment or out of the labor force (source: labor market SIP).
    - Lower migration net inflows from Venezuela in 2020 reduced working age population; Venezuelan migrants tend to have higher human capital than local workers (2020 SIP).
  - Investment channels:
    - Sharp fall in investment and capital underutilization due to firm destruction or halted production amid mobility restrictions.
  - Sectoral channels:
    - Sectoral reallocation and network effects causing temporary underutilization of capital and labor, affecting TFP.
    - Cuts to R&D could reduce future productivity.

### Historical and comparative context
- The recession in 2020 was Colombia’s largest on record.
- Comparable investment falls:
  - -40 percent year-on-year in 2020Q2 (pandemic peak).
  - -41 percent year-on-year in 1999Q2 (1999-2000 recession).
- Unemployment saw gradual increases starting in the mid-1990s, peaking in the 1999-2000 recession and rising moderately after the global financial crisis of 2008-09 and the oil price drop.

### Methodologies to estimate potential output
- Three approaches compared:
  - Production function (PF): Cobb-Douglas Y = TFP(K^(1−α) L^α). HP filter smooths labor input, capital services, and TFP; potential growth computed from aggregate production function.
  - Multivariate filter (MVF): Reduced-form relationships (Phillips curve, Okun’s law, CAPU) to identify slack and potential output from output, inflation, unemployment rate and CAPU. Estimated on annual data.
  - Semi-structural new Keynesian model (SSNK): Similar to MVF, includes Phillips curve and a role for oil prices; utilizes quarterly data and is augmented to include a “COVID-19” shock that lowers potential output and opens the output gap.
- Structural break corrections:
  - Potential output estimates prior to the pandemic are estimated using pre-pandemic data for MVF and SSNK and including January 2020 WEO projections for PF.
  - For estimating 2020 impact, MVF and SSNK use latest information on output, inflation, exchange rate, interest rate; MVF also uses unemployment and CAPU, SSNK uses oil price and risk premia. For PF, WEO projections for output, capital and labor are used with filters on capital, labor and TFP.

### Quantitative estimates of pandemic impact and recovery
- 2020 estimated contraction in potential output (estimates range by method):
  - Production function (PF): -3½ percent.
  - Multivariate filter (MVF): around -1 percent.
  - SSNK: around -½ percent.
- Drivers in PF and MVF: fall in labor input and, to a lesser extent, lower capital.
- Demand-side factors explain about half of the fall in GDP in 2020, implying large output gaps:
  - MVF: opening output gap of around 7 percent.
  - SSNK: opening output gap of around 8 percent.
  - PF: less open output gap of around -4½ percent.
- 2020 component estimates (Percent):
  - PF: Y* -3.7; K* (cont.) -0.6; L* (cont.) -4.6; TFP 1.5.
  - MVF: Y* -1.1; K* (cont.) -0.9; L* (cont.) -1.6; TFP 1.3.
  - SSNK: Y* -0.3; K* na; L* na; TFP na.
- 2021 projected partial recovery in potential output:
  - PF: ½ percent (listed as 0.7 in component table).
  - MVF: 1.5 percent (component table shows 1.5).
  - SSNK: 2½ percent (component table shows 2.7).
- 2021 component estimates (Percent):
  - PF: Y* 0.7; K* (cont.) -0.1; L* (cont.) 0.9; TFP 0.0.
  - MVF: Y* 1.5; K* (cont.) -0.1; L* (cont.) 1.0; TFP 0.6.
  - SSNK: Y* 2.7; K* na; L* na; TFP na.
- Medium-term outlook:
  - Potential output growth is expected to gradually return to its pre-pandemic growth rate of between 3 and 3½ percent.

### Projections and model inputs
- Projections for capital:
  - Benchmarked using an estimated investment equation including the price of oil, GDP and the user cost of capital.
- Projections for labor:
  - Okun’s law and a vector autoregression (VAR) were estimated. Regressions estimated to 2019Q4; forecasts for 2021-26 use observed data for 2020, exogenous forecasts for the price of oil, GDP and the user cost of capital.
- Labor projections incorporate updated Venezuelan migration assumptions:
  - In 2020 the number of migrants from Venezuela in Colombia declined slightly by around 40 thousand as some migrants returned to Venezuela during the national lockdown.
  - The reduction represents around 2 percent of all migrants.
  - Staff assume the border with Venezuela will open in the second half of 2021 with modest arrivals thereafter relative to pre-pandemic trends.
- Projections point to:
  - A gradual increase in the investment to GDP ratio.
  - A gradual increase in the participation rate.
  - The unemployment rate projected to decline slowly towards pre-pandemic levels.

### Modeling specifics (Annex highlights)
- VAR/VECM for capital and labor inputs:
  - Estimated variables: capital stock, gross fixed capital formation (investment), GDP and the price of oil (assumed exogenous).
  - Cointegration tests indicate two cointegrating relationships, identified using restrictions inspired by Ellis and Price (2002):
    - k = y − σ uc + θ
    - k = i − ln(δ + g)
  - Sample (adjusted): 2005Q3–2019Q4. Included observations: 58 after adjustments.
  - Key reported statistics from the VECM (selected):
    - CointEq1 error correction coefficients: D(LOG(INV)) 0.344568; D(LOG(GDP)) −0.053629; D(LOG(K)) −0.021560; D(UC) 0.051414.
    - DLOG(POIL(-1)) coefficients: −0.021542 for D(LOG(INV)); 0.010224 for D(LOG(GDP)); −0.004273 for D(LOG(K)); 0.014421 for D(UC).
    - R-squared: D(LOG(INV)) 0.271359; D(LOG(GDP)) 0.323323; D(LOG(K)) 0.666137; D(UC) 0.164513.
    - Adj. R-squared: D(LOG(INV)) 0.185636; D(LOG(GDP)) 0.243714; D(LOG(K)) 0.626859; D(UC) 0.066221.

*Source: IMF staff analysis as presented in the chapter "Potential Output: Reassessment and Pandemic Impact" (excerpt).*

### 2.      A VAR comprising GDP, the unemployment and participation rates was estimated. A

### 2.      A VAR comprising GDP, the unemployment and participation rates was estimated. A

### VAR identification and impulse responses
- Cholesky and long-run identification (Blanchard-Quah) schemes are considered to evaluate the impact on the labor variables to different identified shocks.
- Under both identification techniques, unemployment and participation both take a long time to return to its pre-shock levels (over twenty quarters).
- Interpretation of columns:
  - The first column of the Cholesky identification and the second column of the long-run restrictions (Shock 2) are consistent with demand-type shocks.
  - The first column of the long-run restriction (Shock 1) is consistent with a supply-type shock.
  - Other shocks have a less forward interpretation and could be considered as unemployment or labor force participation shocks.
- All columns point to sluggish responses in the labor market.
- Responses reported as: Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E.
- Responses reported as: Response to Structural VAR Innovations ± 2 S.E. (Long-Run Restrictions)

### MVF model (key equations and features)
- The MVF model is based on Alichi and others (2017) and exploits information from GDP, inflation, unemployment and capacity utilization.
- Key supply-side equations (as presented):
  - 푦푡∗ = 푦푡−1∗ − 휀1(푢푡∗ − 푢푡−1∗) + 휀2(캡푢푡∗ − 캡푢푡−1∗) + 푔푡∗ + 휖푡푦    (3)
  - 푔푡∗ = (1 − 훾)푔푡−1∗ + 훾푔푠푠∗ + 휖푡푔                            (4)
- Definitions and modeling assumptions:
  - y* denotes potential output.
  - u* is the NAIRU (structural unemployment).
  - capu* is the equilibrium capacity utilization rate.
  - g* is potential growth.
  - 휖 are shocks to the level and growth rate of potential.
  - The model identifies the output gap using a Phillips Curve for inflation.
  - Both the unemployment and capacity utilization gaps are assumed to be proportional to the output gap.
  - The structural unemployment (NAIRU) and capacity utilization equations are assumed to have an autoregressive structural form with their own shocks.
  - The model is estimated using annual data.

### SSNK model (key equations and features)
- The SSNK model is based on Gonzalez and others (2015) and exploits information from GDP, inflation, and oil prices.
- Key supply-side equations (as presented):
  - 푦푡∗ = 푦푡−1∗ + 푔푡∗                                            (5)
  - 푔푡∗ = (1 − 훾)푔푡−1∗ + 훾푔푠푠∗ + ω(∆poil푡∗ − ∆poil푠푠∗) + 휖푡푔    (6)
- Definitions and modeling assumptions:
  - ∆poil is the growth of the equilibrium oil price, which is assumed to have an autoregressive process.
  - A Phillips Curve for inflation helps to identify the output gap.
  - The model is estimated using quarterly data.
  - The model incorporates a Covid-19 specific shock that can affect both demand, potential output and inflation in line with recent literature (e.g. Guerreri and others, 2020 or Baqaee and Farhi, 2021).

### Model implementation and data used for potential output identification
- A subset of model shocks is used to identify the impact on potential using data and projections for key data.
- Both models use data and projections for:
  - GDP
  - inflation
  - the policy rate
  - the exchange rate
- Additional inputs:
  - MVF uses unemployment and capacity utilization.
  - SSNK uses the oil price.
- The models use shocks and projections to infer impacts on potential output and related variables.

*IMF staff compilation from chapter text.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2021/english/1colea2021002.pdf_
