## UNPACKING LOW PRODUCTIVITY IN COLOMBIA: EVIDENCE FROM FIRM-LEVEL DATA

## Source details

**Canonical URL:** [UNPACKING LOW PRODUCTIVITY IN COLOMBIA: EVIDENCE FROM FIRM-LEVEL DATA](https://www.imf.org/-/media/files/publications/cr/2024/english/1colea2024002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1colea2024002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1colea2024002.pdf.json)

---

### A. Stylized Facts about TFP Growth
- Colombia’s Total Factor Productivity (TFP) declined by more than 10 percent in the 1990s and has stagnated at that low level until 2019.
- Recent estimates suggest that TFP has not grown since 2019, as productivity growth in the post-pandemic years was just enough to offset the pandemic decline (DANE, 2023).
- Regional comparisons:
  - High-productivity-growth emerging markets (China, India, Korea) recorded productivity growth of some 20-40 percent.
  - Peers: Brazil and Mexico show similar declines; Peru and Chile fared better.
- Firm-level dataset and methodology:
  - Orbis database: 600,000 firms in Colombia; 150,000 firms have sufficient data to estimate TFP (2005-2019).
  - Firm-level productivity estimated using De Loecker and Warzynski (2012) and Ackerberg, Caves, and Frazer (2015) methodologies.
- Sectoral outcomes:
  - Mining saw the largest TFP decline, reaching a trough of -30 percent following the oil price collapse in late 2014, with some subsequent recovery.
  - Manufacturing TFP declined by around 7 percent between 2005-2019, with largest declines in manufacturing of concrete, electrical equipment, basic iron and steel, and special-purpose machinery.
  - TFP improved by 5-10 percent in construction, trade, and transport during 2014-2019, coinciding with implementation of the 4th Generation (4G) infrastructure agenda after 2014.

### B. Understanding the Nature of TFP Decline: A Decomposition
- Decomposition framework:
  - Dynamic Olley-Pakes decomposition (Melitz and Polanec, 2015) into four channels: Technology, Reallocation, Entry, Exit.
  - Formal expression: aggregate TFP for sector s is 푧_{s,t} = ∑_{f∈i} ω_{f,t} z_{f,t}; change decomposed into Technology + Reallocation + Entry − Exit.
- Empirical decomposition results (sectoral heterogeneity):
  - Mining: TFP decline mostly reflects the technology channel (firm-level productivity declines) rather than composition effects.
  - Manufacturing: between 2005-2015, positive reallocation (high TFP firms grew) improved TFP; after 2015, negative technology channel dominated causing overall TFP decline.
  - Agriculture: firm-level productivity decline dominates but partially mitigated by reallocation and exit of inefficient firms.
  - Construction: firm-level productivity improvements drove sector TFP increases.
- Business dynamism:
  - Less than 10 percent of firms in the sample are “new” firms; they are smaller and tend to have lower TFP than average incumbents.
  - Firm exits contribute significantly to aggregate TFP dynamics, while lack of productive entry is a notable constraint.

### C. Quantifying Misallocation and Its Growth Impact
- Misallocation measurement:
  - Hsieh and Klenow (2009) methodology using dispersion in revenue productivity (TFPR) to infer distortions.
  - Loss of TFP relative to first-best approximated by −(σ^2/2) var(log(TFPR_i)) under specified assumptions.
- Key numeric findings on misallocation:
  - Standard deviation of log revenue productivity in Colombia: around 0.7 in 2005.
  - Comparable benchmarks: China 0.63 and India 0.67 in 2005 (Hsieh and Klenow, 2009); United States around 0.45.
  - Using elasticity of substitution between goods σ = 3, reducing misallocation to U.S. level would increase TFP by 3/2 * (0.7^2 − 0.45^2) ≈43 percent.
  - Back-of-envelope scenario: reducing misallocation from 0.7 to 0.6 over two decades could result in 20 percent higher TFP, or almost 1 percent higher potential growth annually.
- Sectoral patterns:
  - Resource allocation has improved in most sectors since 2005, especially in agriculture, utilities, and mining.
  - Misallocation remains high in services, notably postal, accommodation, and telecommunication.
- Distortions affecting capital-labor input choice:
  - Two distortion types: τ^K (affecting capital-labor choice) and τ^Y (general firm-level distortions).
  - The ratio of the variance of τ^K to τ^Y has generally been larger than 1 and rising since 2005, indicating an increasing role of capital-labor input distortions.
  - Potential drivers to investigate: collateral constraints, financial market underdevelopment, and heterogeneous impacts of labor-cost reforms or measurement changes (discussion and caveats provided).

### D. Conclusion and Policy Implications
- Core conclusions:
  - Low productivity in Colombia is broad-based across many sectors and has weighed on growth for decades.
  - Resource misallocation is sizable and a meaningful contributor to low TFP.
  - Lack of entry by productive new firms is an important factor behind missing TFP growth.
- Policy directions highlighted:
  - Improve business dynamism: encourage firm entry, reduce barriers to entry, and facilitate firm growth and innovation.
  - Alleviate distortions affecting capital-labor choices and access to external financing for small firms (e.g., address financial market frictions and collateral constraints).
  - Complement TFP improvements with policies to increase labor force participation (e.g., integrate Venezuelan immigrants) and capital accumulation (encourage investment).
- Expected gains:
  - Reducing misallocation could substantially raise TFP and sustainable growth; specific numerical scenarios outlined above illustrate large potential payoffs.

### Misallocation: findings and scope (chapter synthesis)
- The paper finds that misallocation of factors of production appears to pose a sizable growth cost for Colombia, though progress in reducing misallocation has been made in recent years.
- The paper does not identify the causes of misallocation; possible causes could include industry- and firm-specific bottlenecks such as taxes, adverse incentives, access to financing, labor market rigidities, red tape, or policy uncertainty.
- More work is needed to better understand how relative changes impact firms of different sizes and potential implications for misallocation, including potential relative changes from the current labor reform proposal that could raise wage and non-wage costs.

### Policy principles and recommendations to address misallocation
- Policies should be derived from studies that explore the causes of misallocation and industry-specific problems.
- General principles: leveling the playing field, encouraging investment, and removing distortions.
- Improving resource allocation would include:
  - Streamlining regulations;
  - Lowering labor market rigidity;
  - Providing more policy certainty (e.g., on regulated prices);
  - Removing bottlenecks that keep firms smaller and outside the recorded economy.

### Export concentration, energy transition risks, and fiscal exposure
- Export composition and exposure:
  - Since 2000, oil exports accounted for about 20-30 percent of annual total goods and services exports, except for the oil boom after the 2008 Global Financial Crisis when the oil share reached a peak of almost 50 percent in 2013.
  - Coal export shares increased from 5.6 percent in 2000 to a peak of 16.8 percent in 2022.
  - In 2023, oil and coal together account for about 47 percent of goods exports, the equivalent of around 5 percent of GDP.
  - Oil-related revenues account for an average of 9 percent of the non-financial public sector’s revenue.
- Energy transition and production projections:
  - IEA (2023) projections: Oil and natural gas consumption would peak in 2030 and decline slowly and only by 5 percent by 2050; Coal consumption would peak over the next few years and decline by 40 percent by 2050.
  - Boer et. al. (2023) suggests oil prices could be between 25 USD and 130 USD per barrel by 2030, depending on policy pathways.
  - Proven reserves are only enough to cover 7.5 years of production at the current rate (National Hydrocarbons Agency, 2023).
  - Annual oil production currently stands at around 275 million barrels, is expected to halve by 2030 and approaches zero by 2040.
  - Oil reserves are expected to drop by 75 percent in this decade, though enhanced recovery and contingent reserves could moderate the pace of decline.

### Export diversification: historical trends and lessons
- Rationale:
  - Export diversification needs to be integral to Colombia’s development strategy given high concentration in energy; services account for less than one-third of total exports (IMF, 2023).
- Historical diversification (Theil index, IMF dataset covering 1962-2014):
  - Latin America significantly diversified exports between 1964-2000 but progress stalled since 2000.
  - Colombia reduced its Theil index by more than 2 points between 1962 and 2002, catching up to Chile, Peru, and Mexico by 2000.
  - The post-2008 oil boom reversed two decades of diversification progress; reversal larger in Colombia compared to LA5 peers (except Brazil).
  - The reversal was driven both by a booming mining sector and by dented growth in non-mining exports: farming and industrial export values doubled between 2000-2008, then fell and stagnated after 2008.
- Margin of diversification:
  - More than ¾ of Colombia’s diversification gain during 1962-2000 was due to the intensive margin rather than exporting new products.

### Comparative advantage dynamics and implications for diversification strategy
- Persistence and speed:
  - Industries with initial RCA > 1 tend to grow more and faster than industries with initial RCA < 1.
  - More than 60 percent of industries with initial RCA < 1 have speed lower than 0.05, implying these industries would require more than 2 decades (1/0.05 = 20 years) to develop comparative advantage from RCA = 0 to RCA = 1.
  - The distribution for industries with initial RCA > 1 has a mean speed of 0.44.
- Policy implication:
  - Diversification is more likely to succeed and be faster if built on existing comparative advantage.
  - Identified opportunities where RCA > 1 but export share < 1 percent include:
    - Agriculture products (coffee extracts, live animals, coconut, sugar cane, soya-bean oil, etc.);
    - Plastic and polymer products;
    - Chemical manufacturing (inorganic chemicals/carbon, organic chemicals/acid, hygiene products);
    - Metal products (aluminum building materials, precious-metal ores, copper, steel pipe).

### Services, quality upgrading, and future avenues
- Services export trends:
  - Services export share rose from an average of 13 percent in the 2000s to around 23 percent in 2023.
  - Composition shifts (2000 to 2023):
    - Travel exports account for half of services exports; tourism’s share of services exports has not changed between 2000-2023.
    - “Other Business Services” grew from almost 4 percent in 2000 to 18 percent in 2023, led by a sharp growth in call center services.
    - The 14pp increase in export share of business services is roughly matched by a 10pp decline of transportation services and 3pp decline of financial services.
- Opportunities:
  - Remote work and teleworkability present promising avenues for increasing services exports given Colombia’s labor cost advantage relative to advanced economies (Hausmann and Bustos, 2021).
  - Quality upgrading could raise exports of higher value-added goods and improve diversification.

### Quality upgrading: measurement, trends, and implications
- Measurement:
  - Export quality measured by the unit value of exports purged of contaminating factors (Papageorgiou et. al., 2015); data coverage 1962-2014.
  - Quality index normalization: 90th percentile = 1.
- Historical trends:
  - Latin America improved export quality before 1980 at the same pace as East Asia & Pacific, but stalled since the mid-1980s.
  - Among LA5, Colombia and Brazil made most progress in quality upgrading, catching up to East Asia by mid-1980s; Colombia’s progress stalled in the 1990s and reversed after the Financial Crisis (composition effect due to oil).
- Sectoral quality profile:
  - Beverages & Tobacco at the 75th percentile of the quality ladder.
  - Crude materials exc. fuels and Mineral fuels rank low.
  - Quality at the median and relatively close to the frontier in Chemicals, Manufactured goods, and Machinery and Transport Equipment — identified as promising starting points for upgrading.
- Constraints and analysis needs:
  - The analysis does not quantify the cost of upgrading; distance-to-frontier could indicate required investment costs (e.g., refining capacity for mineral fuels).
  - Quality measures should be a starting point for granular sectoral analysis to identify feasible and efficient upgrading opportunities.

### Summary findings on diversification and policy implications
- Strategic guidance:
  - Colombia has substantial potential to diversify exports by building on existing comparative advantages and upgrading export quality rather than attempting to build industries from scratch.
  - It may not be cost-effective or timely to upgrade quality where the gap to the global frontier is too large.
  - Public policies are important to address market failures and improve business climate, including upgrading public goods (infrastructure) and tackling coordination failures (education and labor-skill alignment).
  - Government engagement with the private sector is important to identify development avenues and strengthen export diversification strategy.
  - Diversification requires improving firm productivity and resource allocation (see Chapter 1 on productivity); policies should level the playing field and remove distortions affecting firms of different sizes or sectors differentially.
  - Public resources alone are insufficient; policies should encourage private sector investment and participation.

### Methodology and macroeconomic effects of fiscal consolidation (selected empirical results)
- Estimation approach:
  - Local projection method (Jordà, 2005) used to estimate responses of the current account balance and the log REER to fiscal consolidation shocks measured as percent of GDP.
  - Benchmark specification includes country and time fixed effects and controls: lags of current account/GDP, lags of fiscal shocks, crisis and sudden stop dummies, contemporaneous growth rate, and lags of terms of trade.
- Estimated effects on the current account balance (response to a 1 percent of GDP fiscal consolidation):
  - Latin America and the Caribbean, Baseline:
    - h=0: 0.69*** (0.15)
    - h=1: 0.72*** (0.19)
    - h=2: 0.73** (0.33)
  - Advanced Economies, Baseline:
    - h=0: 0.48*** (0.14)
    - h=1: 0.61*** (0.14)
    - h=2: 0.57*** (0.12)
- Real exchange rate (REER) effects:
  - LAC: estimated depreciation of about 4 percent within two years.
  - AEs: estimated depreciation of around 1.0 percent within two years.
- Tax-based vs. spending-based consolidations:
  - Advanced Economies: tax-based point estimate 0.60 vs spending-based 0.52 for current account — difference not statistically significant.
  - LAC: tax-based consolidations appear larger and more persistent (raise current account by 2 percent of GDP after 4 years); spending-based consolidations: 0.76 percent of GDP on impact, then declining.
  - Caveat: low number and lack of persistence of spending-based episodes in LAC may affect estimates.
- Role of initial conditions:
  - Public debt ratio: no statistically significant differences between high and low debt cases for both AEs and LAC. Median public debt ratios: LAC 40 percent of GDP; AEs 60 percent of GDP.
  - International investment position (IIP): impact of fiscal consolidation on the current account is larger when the IIP ratio is lower. Median IIP: LAC −30 percent of GDP; AEs −11.5 percent of GDP.
- Robustness and alternative estimators:
  - Results robust to dropping outliers and restricting to consolidations only.
  - Augmented inverse propensity score weighting (AIPW) yields LAC AIPW estimate after two years 0.51 (0.33) vs baseline 0.73; AE AIPW estimate 1.0 percent vs baseline 0.6 percent (differences in transformation and comparability noted).
  - Selected alternative specification estimates (examples):
    - LAC Without Outliers: h=0 0.46* (0.26); h=1 0.63*** (0.17); h=2 0.66** (0.33)
    - AE AIPW Matching Estimator: h=0 0.18* (0.11); h=1 0.65*** (0.16); h=2 1.04*** (0.19)

- Preliminary policy-relevant conclusions:
  - Public and private savings: public savings tend to be negatively, but imperfectly, correlated with private savings; Colombia’s negative and statistically significant correlation implies an increase in public savings would only be partially offset by a decline in private savings.
  - Fiscal consolidations improve the external current account balance in both AEs and LAC, consistent with the “twin deficits” hypothesis.
  - REER depreciation typically accompanies external adjustment and is larger for LAC.
  - Both tax- and spending-based consolidations improve the current account; LAC spending-based estimates are sensitive to sample size.
  - Initial conditions: lower IIP appears to amplify the effects of fiscal consolidation.

*Prepared by Vu Chau and Marco Arena. Source: IMF staff chapter text.*

### References ____________________________________________________________________________ 10

### UNPACKING LOW PRODUCTIVITY IN COLOMBIA: EVIDENCE FROM FIRM-LEVEL DATA

### A. Stylized Facts about TFP Growth
- Colombia’s Total Factor Productivity (TFP) declined by more than 10 percent in the 1990s and has stagnated at that low level until 2019.
- Recent estimates suggest that TFP has not grown since 2019, as productivity growth in the post-pandemic years was just enough to offset the pandemic decline (DANE, 2023).
- Regional comparisons:
  - High-productivity-growth emerging markets (China, India, Korea) recorded productivity growth of some 20-40 percent.
  - Peers: Brazil and Mexico show similar declines; Peru and Chile fared better.
- Firm-level dataset and methodology:
  - Orbis database: 600,000 firms in Colombia; 150,000 firms have sufficient data to estimate TFP (2005-2019).
  - Firm-level productivity estimated using De Loecker and Warzynski (2012) and Ackerberg, Caves, and Frazer (2015) methodologies.
- Sectoral outcomes:
  - Mining saw the largest TFP decline, reaching a trough of -30 percent following the oil price collapse in late 2014, with some subsequent recovery.
  - Manufacturing TFP declined by around 7 percent between 2005-2019, with largest declines in manufacturing of concrete, electrical equipment, basic iron and steel, and special-purpose machinery.
  - TFP improved by 5-10 percent in construction, trade, and transport during 2014-2019, coinciding with implementation of the 4th Generation (4G) infrastructure agenda after 2014.

### B. Understanding the Nature of TFP Decline: A Decomposition
- Decomposition framework: dynamic Olley-Pakes decomposition (Melitz and Polanec, 2015) into four channels:
  - Technology: firm-level TFP changes holding market shares fixed.
  - Reallocation: changes because high (low) TFP firms gain (lose) market share.
  - Entry: new entrants being more (less) productive than incumbents.
  - Exit: exit of low (high) TFP firms.
- Formal expression: aggregate TFP for sector s is 푧_{s,t} = ∑_{f∈i} ω_{f,t} z_{f,t}; change decomposed into Technology + Reallocation + Entry − Exit (formula and notation provided in source).
- Empirical decomposition results (sectoral heterogeneity):
  - Mining: TFP decline mostly reflects the technology channel (firm-level productivity declines) rather than composition effects.
  - Manufacturing: between 2005-2015, positive reallocation (high TFP firms grew) improved TFP; after 2015, negative technology channel dominated causing overall TFP decline.
  - Agriculture: firm-level productivity decline dominates but partially mitigated by reallocation and exit of inefficient firms.
  - Construction: firm-level productivity improvements drove sector TFP increases.
- Business dynamism:
  - Less than 10 percent of firms in the sample are “new” firms; they are smaller and tend to have lower TFP than average incumbents.
  - Firm exits contribute significantly to aggregate TFP dynamics, while lack of productive entry is a notable constraint.

### C. Quantifying Misallocation and Its Growth Impact
- Misallocation measurement: Hsieh and Klenow (2009) methodology using dispersion in revenue productivity (TFPR) to infer distortions.
  - Revenue productivity TFPR_i definition and relation to implicit taxes τ_i^K and τ_i^Y provided in source.
  - Loss of TFP relative to first-best approximated by −(σ^2/2) var(log(TFPR_i)) under specified assumptions.
- Key numeric findings on misallocation:
  - Standard deviation of log revenue productivity in Colombia: around 0.7 in 2005.
  - Comparable benchmarks: China 0.63 and India 0.67 in 2005 (Hsieh and Klenow, 2009); United States around 0.45.
  - Using elasticity of substitution between goods σ = 3, reducing misallocation to U.S. level would increase TFP by 3/2 * (0.7^2 − 0.45^2) ≈43 percent.
  - Back-of-envelope scenario: reducing misallocation from 0.7 to 0.6 over two decades could result in 20 percent higher TFP, or almost 1 percent higher potential growth annually.
- Sectoral patterns:
  - Resource allocation has improved in most sectors since 2005, especially in agriculture, utilities, and mining.
  - Misallocation remains high in services, notably postal, accommodation, and telecommunication.
- Distortions affecting capital-labor input choice:
  - Two distortion types: τ^K (affecting capital-labor choice) and τ^Y (general firm-level distortions).
  - The ratio of the variance of τ^K to τ^Y has generally been larger than 1 and rising since 2005, indicating an increasing role of capital-labor input distortions.
  - Potential drivers to investigate: collateral constraints, financial market underdevelopment, and heterogeneous impacts of labor-cost reforms or measurement changes (discussion and caveats provided).

### D. Conclusion and Policy Implications
- Core conclusions:
  - Low productivity in Colombia is broad-based across many sectors and has weighed on growth for decades.
  - Resource misallocation is sizable and a meaningful contributor to low TFP.
  - Lack of entry by productive new firms is an important factor behind missing TFP growth.
- Policy directions highlighted:
  - Improve business dynamism: encourage firm entry, reduce barriers to entry, and facilitate firm growth and innovation.
  - Alleviate distortions affecting capital-labor choices and access to external financing for small firms (e.g., address financial market frictions and collateral constraints).
  - Complement TFP improvements with policies to increase labor force participation (e.g., integrate Venezuelan immigrants) and capital accumulation (encourage investment).
- Expected gains:
  - Reducing misallocation could substantially raise TFP and sustainable growth; specific numerical scenarios outlined above illustrate large potential payoffs.

*Prepared by Vu Chau and Marco Arena.*

### 13. Misallocation of factors of production appear to pose a sizable growth cost for

### 13. Misallocation of factors of production appear to pose a sizable growth cost for

### Misallocation: findings and scope
- The paper finds that misallocation of factors of production appears to pose a sizable growth cost for Colombia, though progress in reducing misallocation has been made in recent years.
- The paper does not identify the causes of misallocation; possible causes could include industry- and firm-specific bottlenecks such as taxes, adverse incentives, access to financing, labor market rigidities, red tape, or policy uncertainty.
- More work is needed to better understand how relative changes impact firms of different sizes and potential implications for misallocation, including potential relative changes from the current labor reform proposal that could raise wage and non-wage costs (see Annex I of the Staff Report).

### Policy principles and recommendations to address misallocation
- Policies should be derived from studies that explore the causes of misallocation and industry-specific problems.
- In general, policies should be based on the principles of leveling the playing field, encouraging investment, and removing distortions.
- Improving resource allocation would include:
  - Streamlining regulations;
  - Lowering labor market rigidity;
  - Providing more policy certainty (e.g., on regulated prices);
  - Removing bottlenecks that keep firms smaller and outside the recorded economy.

### Export concentration, energy transition risks, and fiscal exposure
- Since 2000, oil exports accounted for about 20-30 percent of annual total goods and services exports, except for the oil boom after the 2008 Global Financial Crisis when the oil share reached a peak of almost 50 percent in 2013.
- Coal export shares increased from 5.6 percent in 2000 to a peak of 16.8 percent in 2022.
- In 2023, oil and coal together account for about 47 percent of goods exports, the equivalent of around 5 percent of GDP.
- Oil-related revenues account for an average of 9 percent of the non-financial public sector’s revenue.
- International Energy Agency (IEA, 2023) projections cited:
  - Oil and natural gas consumption would peak in 2030 and decline slowly and only by 5 percent by 2050;
  - Coal consumption would peak over the next few years and decline by 40 percent by 2050.
- Boer et. al. (2023) suggests oil prices could be between 25 USD and 130 USD per barrel by 2030, depending on policy pathways.
- Proven reserves are only enough to cover 7.5 years of production at the current rate (National Hydrocarbons Agency, 2023).
- Annual oil production currently stands at around 275 million barrels, is expected to halve by 2030 and approaches zero by 2040.
- Oil reserves are expected to drop by 75 percent in this decade, though enhanced recovery and contingent reserves could moderate the pace of decline.

### Export diversification: historical trends and lessons
- Export diversification needs to be integral to Colombia’s development strategy given high concentration in energy; services account for less than one-third of total exports (IMF, 2023).
- Historical diversification (Theil index, IMF dataset covering 1962-2014):
  - Latin America significantly diversified exports between 1964-2000 but progress stalled since 2000.
  - Colombia reduced its Theil index by more than 2 points between 1962 and 2002, catching up to Chile, Peru, and Mexico by 2000.
  - The post-2008 oil boom reversed two decades of diversification progress; the reversal was larger in Colombia compared to LA5 peers (except Brazil).
  - The reversal was driven both by a booming mining sector and by dented growth in non-mining exports: farming and industrial export values doubled between 2000-2008, then fell and stagnated after 2008.
- Diversification historically occurred more through the intensive margin than the extensive margin:
  - More than ¾ of Colombia’s diversification gain during 1962-2000 was due to the intensive margin (producing a more balanced mix of existing products) rather than exporting new products.
  - This pattern is similar to other commodity exporters and emerging Asian manufacturing hubs.

### Comparative advantage dynamics and implications for diversification strategy
- Comparative advantages can evolve but exhibit high inertia.
- Using UN COMTRADE (1988-2022) and RCA calculations:
  - Industries with initial RCA > 1 tend to grow more and faster than industries with initial RCA < 1.
  - More than 60 percent of industries with initial RCA < 1 have speed lower than 0.05, implying these industries would require more than 2 decades (1/0.05 = 20 years) to develop comparative advantage from RCA = 0 to RCA = 1.
  - The red distribution (industries with initial RCA > 1) has a mean speed of 0.44, indicating much faster potential gains.
- Policy implication: diversification is more likely to succeed and be faster if built on existing comparative advantage.
- For Colombia, potential diversification opportunities identified where RCA > 1 but export share < 1 percent include:
  - Agriculture products (coffee extracts, live animals, coconut, sugar cane, soya-bean oil, etc.);
  - Plastic and polymer products;
  - Chemical manufacturing (inorganic chemicals/carbon, organic chemicals/acid, hygiene products);
  - Metal products (aluminum building materials, precious-metal ores, copper, steel pipe).

### Services, quality upgrading, and future avenues
- Services export share rose from an average of 13 percent in the 2000s to around 23 percent in 2023.
- Composition shifts within services exports (2000 to 2023):
  - Travel exports account for half of services exports; tourism’s share of services exports has not changed between 2000-2023.
  - “Other Business Services” grew from almost 4 percent in 2000 to 18 percent in 2023, led by a sharp growth in call center services.
  - The 14pp increase in export share of business services is roughly matched by a 10pp decline of transportation services and 3pp decline of financial services.
- Remote work and teleworkability present promising avenues for increasing services exports given Colombia’s labor cost advantage relative to advanced economies (Hausmann and Bustos, 2021).
- Quality upgrading remains an opportunity: Colombia still has room to improve export quality in most industries, which could raise exports of higher value-added goods and improve diversification.

*Source: IMF staff chapter text.*

### 15. One important avenue of growing non-traditional exports is upgrading export quality

### 15. One important avenue of growing non-traditional exports is upgrading export quality and increasing value added

### Quality upgrading: concept and measurement
- Export diversification can be achieved by increasing export share of non-traditional products through both larger volumes and higher unit values; increasing unit values can reflect quality upgrading and higher value added.
- Export quality is measured by the unit value of exports purged of contaminating factors (Papageorgiou et. al., 2015), with data available for 200 countries between 1962-2014.
- The quality index used is normalized with the 90th percentile = 1.

### Historical trends and regional comparison
- Latin America improved export quality before 1980 at the same pace as East Asia & Pacific, but has stalled since the mid-1980s while East Asia continued to improve quality and largely caught up with Europe.
- Among LA5 countries, Colombia and Brazil made the most progress in quality upgrading, catching up to East Asia by mid-1980s.
- Colombia’s quality progress stalled in the 1990s, followed by a sharp reversal after the Financial Crisis; this reversal is possibly a composition effect as the export share of oil (with lower quality) expanded — this decline was observed only in Colombia among the LA5.

### Colombia’s sectoral quality profile and opportunities
- Colombia’s export quality is relatively high in several industries:
  - Beverages & Tobacco is at the 75th percentile of the quality ladder across countries.
  - Crude materials exc. fuels and Mineral fuels rank low, possibly because Colombia exports more crude than refined petroleum.
  - Quality is at the median level and relatively close to the frontier in Chemicals, Manufactured goods, and Machinery and Transport Equipment — these are identified as promising starting points for quality upgrading.

### Costs, constraints, and need for granular analysis
- The analysis does not quantify the cost of upgrading quality to the frontier; distance-to-frontier could indicate required investment costs.
- Example: improving quality in crude materials and mineral fuels could require costly investments in refining capacity.
- The presented quality measures should serve as a starting point for more granular sectoral analysis to identify where quality could be upgraded most easily and efficiently.

### Summary findings on diversification and policy implications
- Colombia has substantial potential to diversify exports by building on existing comparative advantages and upgrading export quality rather than attempting to build industries from scratch.
- It may not be cost-effective, including in terms of time, to upgrade quality where the gap to the global frontier is too large.
- Public policies play an important role in fostering successful export diversification, especially policies that address market failures and improve business climate.
- Government engagement with the private sector and stakeholders is important to identify development avenues and strengthen export diversification strategy.
- Fostering diversification requires improving firm productivity and resource allocation (see Chapter 1 on productivity); policies should level the playing field (e.g., remove distortions affecting firms of different sizes or sectors differentially).
- Public policies can address market failures by upgrading public goods such as infrastructure and by tackling coordination failures (for example, ensuring education policies build labor skills needed in growth industries).
- Diversification policies should encourage private sector investment and participation, as public resources alone would not be sufficient (nor efficient) to meet Colombia’s ambitious diversification and climate goals.
- Coordination problems may include linkages between industries and the labor market; policies to improve human capital and reduce labor market rigidities could help facilitate export diversification.

### Key statistics and quantifications from the chapter
- Quality data coverage: 1962-2014.
- Quality index normalization: 90th percentile = 1.
- Beverages & Tobacco industry: at the 75th percentile of the quality ladder.
- Historical correlation figures and statistics referenced:
  - Correlation between public and private savings for emerging markets (EMs) estimated around -0.25; for 2000-2019 sample period about -0.23.
  - Colombia’s correlation between public savings and private savings is -0.61 (period 2000-2019).
  - Correlation for the period 1980-2023 is -0.41 (Annex VI of the Staff Report).
- Narrative fiscal policy sample and identified fiscal policy changes (Table 3):
  - Sample includes 30 economies (16 advanced economies and 14 Latin American and Caribbean countries).
  - Advanced Economies (1978-2019) — Tax-Based: Average 1.07; Median 0.79; Minimum 0.02; Maximum 5.23; Number of Policy Changes 85.
  - Advanced Economies (1978-2019) — Expenditure-Based: Average 1.04; Median 0.78; Minimum -0.31; Maximum 5.01; Number of Policy Changes 152.
  - Latin America and the Caribbean (1989-2019) — Tax-Based: Average 0.78; Median 0.62; Minimum -0.90; Maximum 4.10; Number of Policy Changes 56.
  - Latin America and the Caribbean (1989-2019) — Expenditure-Based: Average 1.27; Median 1.05; Minimum -0.50; Maximum 3.80; Number of Policy Changes 24.
- Cross-country literature estimates cited: a fiscal consolidation of one percent of GDP improves the current account balance between 0.6 and 0.8 percent of GDP (Bluedorn and Leigh, 2011; Furceri and Zdzienicka, 2018; Carriere-Swallow et al, 2021; IMF External Sector Report, 2021).
- Abbas et al (2011) cited: strengthening fiscal balance by 1 percentage point of GDP associated with a current account improvement of about 0.3 percentage point of GDP on average (stronger in emerging markets and low-income countries).
- Country-level study (Garrido-Tejada et al, 2021) finds that a shock of one percentage point of the external imbalance increases the public imbalance on average by 0.25 percentage points in the second year and 0.32 percentage points in the third year.

*Source: IMF staff chapter "15. One important avenue of growing non-traditional exports is upgrading export quality" (from the provided IMF PDF content).*

### 9. Following Carriere-Swallow et al (2021), we estimate the macroeconomic effects of

### 9. Following Carriere-Swallow et al (2021), we estimate the macroeconomic effects of

### Methodology
- Estimation approach: local projection method (LP) proposed by Jordà (2005).
- Benchmark specification for horizons h = 0, 1, 2 years:
  - y_{i,t+h} − y_{i,t−1} = α^h_i + γ^h_t + β_h Σ_{s=t}^{t+h} FC_{i,s} + δ X_{i,t} + ε_{i,t+h}
  - y is either the current account balance as a share of GDP or the log of the real effective exchange rate (REER).
  - FC denotes the measure of fiscal consolidation in percent of GDP.
  - X_t includes: two lags of the current account/GDP, two lags of fiscal shocks, crisis dummies (Laeven and Valencia, 2020), sudden stop dummies (David and Gonçalves, 2019), contemporaneous growth rate, and two lags of the terms of trade (Gruss and Kebhaj, 2019).
  - Time (γ^h_t) and country (α^h_i) fixed effects included.
  - β_h estimates the response of the macroeconomic variable to the cumulative fiscal shock over horizon h.

### Estimated effects on the current account balance
- Interpretation: reported impulse responses correspond to the response of the current account ratio to a fiscal policy change (shock) of 1 percent of GDP (i.e., a “multiplier” at each horizon).
- Advanced Economies (AEs):
  - Fiscal policy changes lead to a contraction in the current account ratio of about 0.6 percent with a 90 percent confidence interval of 0.4-0.8 percent.
- Latin America and the Caribbean (LAC):
  - Fiscal policy changes are estimated to lead to a contraction in the current account ratio of 0.7 percent after two years, with a 90 percent confidence interval of 0.2-1.3 percent.
  - Footnote: using commodity terms of trade (Gruss and Kebhaj, 2019) yields an estimated contraction of 0.8 percent for LAC; inclusion of growth rate of external demand does not change results.

- Key tabulated baseline estimates (Effect of a 1 Percent of GDP Fiscal Consolidation in Year t+h; Percent)
  - Latin America and the Caribbean, Baseline:
    - h=0: 0.69*** (0.15)
    - h=1: 0.72*** (0.19)
    - h=2: 0.73** (0.33)
  - Advanced Economies, Baseline:
    - h=0: 0.48*** (0.14)
    - h=1: 0.61*** (0.14)
    - h=2: 0.57*** (0.12)

### Real exchange rate effects (REER)
- Fiscal consolidation tends to depreciate the real effective exchange rate, facilitating current account adjustment.
- Magnitude:
  - LAC: estimated depreciation of about 4 percent within two years.
  - AEs: estimated depreciation of around 1.0 percent within two years.
- The larger LAC effect is related to adoption of flexible exchange rate regimes in the 1990s; result mainly driven by cases with flexible exchange rates (AREAER classification).

### Tax-Based vs. Spending-Based Consolidations
- Estimation extends baseline with indicator TB_{i,t} = 1 for tax-based consolidation; model:
  - y_{i,t+h} − y_{i,t−1} = α^h_i + γ^h_t + (β^e_h EB_{i,t} + β^t_h TB_{i,t}) Σ_{s=t}^{t+h} FC_{i,s} + δ X_{i,t} + ε_{i,t+h}
- Advanced Economies:
  - Point estimates: tax-based consolidations 0.60 versus spending-based 0.52 for the current account response; difference not statistically significant.
- Latin America and the Caribbean:
  - Tax-based consolidations appear larger and more persistent: raise the current account by 2 percent of GDP after 4 years.
  - Spending-based consolidations: 0.76 percent of GDP on impact, then declining over time.
  - Caveat: low number and lack of persistence of spending-based episodes in LAC could drive results (only 10 out of 24 LAC fiscal policy changes, 42 percent, happened in consecutive years vs. 75 percent in AEs).

### Role of initial conditions (public debt ratio and IIP)
- Split-sample estimation following Jordà and Taylor (2016) using state indicator S_{t-1} (above/below median).
- Public debt ratio:
  - No statistically significant differences in fiscal consolidation effects on the current account between high and low public debt ratio cases for both AEs and LAC.
  - Median public debt ratios: LAC 40 percent of GDP; AEs 60 percent of GDP.
- International investment position (IIP) ratio:
  - For both AEs and LAC, impact of fiscal consolidation on the current account is larger when the IIP ratio is lower.
  - Median IIP: LAC −30 percent of GDP; AEs −11.5 percent of GDP.

### Robustness checks
- Sensitivity to outliers:
  - Re-estimate after dropping largest and smallest 1 percent of fiscal policy changes; estimates without extreme outliers similar to baseline.
- Restricting to consolidations only (non-negative fiscal policy changes):
  - Results similar to baseline.
- Augmented inverse propensity score weighting (AIPW) estimator:
  - Addresses potential predictability of narrative consolidation episodes; first-stage probit uses lags of consolidations, lags of current account changes, terms of trade and lags (or commodity ToT), lagged debt-to-GDP, crisis and sudden stop dummies, and country fixed effects.
  - LAC results: statistically significant AIPW estimate after two years is 0.51 (0.33), lower than baseline 0.73; using commodity-terms of trade instead of terms of trade yields results similar to baseline.
  - AEs results: AIPW yields significant and slightly higher point estimate of 1.0 percent (compared with 0.6 percent baseline). Note: shock transformed into binary treatment variable; point estimates and SEs rescaled for comparability but not exactly comparable.

- Selected alternative specification estimates from Table 4 (Percent, with standard errors in parentheses):
  - Latin America and the Caribbean
    - Without Outliers: h=0 0.46* (0.26); h=1 0.63*** (0.17); h=2 0.66** (0.33)
    - Fiscal Policy Shocks (consolidations): h=0 0.72*** (0.15); h=1 0.74*** (0.19); h=2 0.74*** (0.14)
    - AIPW Matching Estimator: h=0 0.72*** (0.19); h=1 0.93*** (0.29); h=2 0.51* (0.33)
    - AIPW (using commodity ToT): h=0 0.71*** (0.21); h=1 0.93*** (0.30); h=2 0.67** (0.34)
  - Advanced Economies
    - Without Outliers: h=0 0.34** (0.11); h=1 0.52*** (0.13); h=2 0.55*** (0.13)
    - Fiscal Policy Shocks (consolidations): h=0 0.47** (0.14); h=1 0.60*** (0.14); h=2 0.57*** (0.12)
    - AIPW Matching Estimator: h=0 0.18* (0.11); h=1 0.65*** (0.16); h=2 1.04*** (0.19)
    - AIPW (using commodity ToT): h=0 0.21* (0.11); h=1 0.68*** (0.17); h=2 1.05*** (0.20)

### Preliminary conclusions
- Public and private savings:
  - For a large sample, public savings tend to be negatively, but imperfectly, correlated with private savings; an increase in public savings can improve the current account, but the strength varies across countries.
  - For Colombia, results suggest a negative and statistically significant correlation between public and private savings, implying that an increase in public savings would only be partially offset by a decline in private savings.
- Fiscal consolidations and current account imbalances:
  - Fiscal consolidations improve the external current account balance in both AEs and LAC, consistent with the “twin deficits” hypothesis.
  - REER depreciation typically accompanies external adjustment and is larger for LAC.
  - Both tax- and spending-based consolidations improve the current account. In LAC, the estimated impact of spending-based consolidations is affected by the low number of identified episodes.
  - Initial conditions: no statistically significant differences between high and low public debt ratio cases; lower IIP appears to amplify the effects of fiscal consolidation.

*Source: Authors’ elaboration (chapter content).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1colea2024002.pdf_