## wpiea2021092-print-pdf

## Source details

**Canonical URL:** [wpiea2021092-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021092-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021092-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021092-print-pdf.pdf.json)

---

### Korea’s recent growth trajectory and convergence
- Beginning in the early 1960’s, Korea’s income per capita was less than 10 percent of that in the United States.
- Per-capita income growth then exceeded 10 percent per year for some 10-year periods.
- By recent decades, Korea’s per capita income level rose to two-thirds of U.S. income.
- Growth has slowed since the Asian financial crisis (AFC) and global financial crisis (GFC), nearing the pace of advanced economy comparators (pre-COVID potential growth had already decelerated to about 2½ percent).

### Main historical growth drivers and recent headwinds
- Drivers:
  - Export-oriented manufacturing expansion.
  - High investment in physical capital and rising educational attainment.
  - A growing working-age population and deeper structural/institutional reforms.
  - Resulted in a high growth–high investment equilibrium with contributions from factor accumulation and some TFP.
- Recent headwinds:
  - Investment weakened after the AFC.
  - Pace of productivity growth declined; TFP level stagnated since early 2000s at less than two-thirds the U.S. level.
  - Favorable demographics turned into a headwind as population ages.
  - Product and labor market rigidities and structural features amplify adverse effects.

### COVID-19 shock and estimated impacts on potential output
- Estimated cumulative impact on potential output: 2.5 percent (also reported as about 3 percent in other estimates).
- Standard errors encompass an impact of between 1½ and 5 percent.
- Alternative estimates noted:
  - Using the impact on headline output directly yields a response of 3.5 percent.
  - Aggregate factor-based estimate cited a figure truncated in source but consistent with ~3 percent range.
- Principal channels:
  - Primary: lower investment rates over the medium term → slower capital accumulation.
  - Secondary: lower labor force participation.
- Episode-specific considerations:
  - Synchronized global shock could amplify effects.
  - Large policy response in Korea may buffer medium-term impact.
  - Scarring risks from firm destruction, worker-firm relationship losses, skills mismatch, worsening balance sheets.
  - Possible mitigating factors: smaller investment transmission than after AFC; potential gains from accelerated digitalization.

### Sector-level productivity, convergence, and structural transformation needs
- Aggregate convergence:
  - Labor productivity relative to the United States converged from 13 percent on average in the 1960s to about 55 percent in the 2010s.
- Manufacturing:
  - Convergence strongest: labor productivity in the 2010s ≈ 60 percent of U.S. level (from 7 percent in 1960s).
  - TFP growth in export-oriented manufacturing averaged 6 percent per year in the 1970s-1990s.
- Services:
  - Converged more slowly: reach 49 percent of U.S. level in the 2010s (from 27 percent in 1960s).
  - Market services with labor productivity < 40 percent of U.S.; specific services at or below 30 percent: wholesale and retail; transportation and storage; accommodation and food services; publishing and communications; information services.
  - Productivity closer to U.S. level in finance and insurance and professional, scientific, and technical services.
- Implication:
  - Need to rebalance toward services as manufacturing contribution may fade; services rebalancing could raise output and employment given higher labor intensity.

### Demographics and labor input projections
- Aggregate labor force participation:
  - Projected to increase through 2025 driven by rising educational attainment and younger female cohort participation.
  - After 2025, aggregate participation projected to decrease steadily as elderly share rises and educational attainment pace levels off.
- Cohort approach:
  - Uses labor force entry and exit rates by age and gender cohort and produces substantially higher aggregate participation than forecasts using constant age-gender participation rates.
- Labor quality:
  - Average years of schooling is high with little scope for large further increases; tertiary, secondary, primary series shown through 2050.

### Investment outlook and capital accumulation
- Private non-residential investment model:
  - Expanded accelerator linking investment-capital ratio to lagged output growth-capital stock ratio, expected growth over next five years, and lagged government investment.
  - Model explains investment rise in 1970s-1990s, drop after AFC, and gradual decline since.
  - Model forecast: private non-residential investment of about 19-20 percent of GDP over the next decade, slightly below recent levels.
  - With higher capital-output ratio, implies a slower rate of capital stock growth and a headwind for potential growth.

### Capacity utilization and capital services adjustment
- Manufacturing capacity utilization regression (median quantile):
  - Constant: 76.09 (Standard error: 1.05)
  - Cyclical component coefficient: 0.8 (Standard error: 0.2)
  - Mean dependent variable: 76.22; Observations: 40; Adjusted R-squared: 0.38
- Aggregate capacity utilization constructed as weighted average of observed manufacturing and estimated non-manufacturing, weights by capital stock shares.
- Adjustment rationale:
  - Capital services adjusted for capacity utilization to include utilization fluctuations in capital services contributions; raises estimated potential TFP growth in periods when utilization is below long-term averages once utilization normalizes.

### Potential output estimation approaches and pre-COVID baseline results
- Two approaches:
  1. Production function (Y = A K^{(1−a)} L^{a} with labor share α = 0.61) using detailed labor inputs, capacity-adjusted capital services, and H-P filter (lambda = 100).
     - Assumes TFP growth returns to its 2000-2019 average of 1.4 percent.
     - Implied convergence: from 63 percent of U.S. productivity currently to 72 percent in 2050, assuming U.S. productivity growth of 0.8 percent.
  2. Multivariate filter (MVF): Kalman filter conditioning on Okun’s law and the Phillips curve, includes Consensus Forecasts of real GDP and inflation and capacity utilization.
- Pre-COVID estimates:
  - Both approaches find recent potential growth in the mid-2 percent range.
  - Pre-COVID 2019: Production function 2.7 versus MVF 2.4.
  - Long-run (pre-COVID) projections:
    - Potential growth about 2.2-2.4 percent in the 2020s.
    - Declining to about 2 percent per year in the 2030s.
    - Declining to about 1.5 percent per year by 2050.

### Historical-recession evidence on scarring and COVID-19 scenario
- Local projections on previous recessions:
  - Previous recessions in Korea produced a sizable negative impact on potential output averaging 7 percent over five years.
  - Factor-level aggregated estimate: cumulative reduction of potential output estimated at -5.2 percent with majority from reduction in capital services.
  - TFP often resilient post-recession; main channels of scarring were lower investment and labor force participation.
- Applied to COVID-19:
  - Estimated that COVID-19 could lower level of output by about 3 percent (reported estimates range 2.5–3.5 percent).
  - Timing: largest effects on potential growth occur immediately after shock with slow normalization.
  - Standard errors for central estimate of 2.5 percent span 1½ to 5 percent.

### Structural reform scenario and quantified gains
- Scenario design:
  - Uses Kim and Loayza (2019) for productivity drivers and Dao and others (2014) for labor force participation/employment.
  - Assumes indicators raised to the 75th percentile of OECD where Korea lagged; where Korea exceeded benchmark, standing unchanged.
  - Labor market reforms assumed to close one third of the female-male participation gap and raise youth employment rates by ten percentage points.
- Long-term quantified impacts:
  - Potential output about 12 percent higher than baseline once effects are fully realized.
  - Income per capita lifted from around two-thirds of the U.S. level currently to roughly 80 percent.
  - Productivity-enhancing reforms account for about eight percentage points of total impact—six percentage points directly and the rest via higher investment and capital stock; labor market reforms explain the remainder.
  - Peak impact on potential growth is over half a percentage point per year.
- Component-specific illustrative estimates cited in chapter:
  - 4.7 percent: Employment protection and network sector regulations — impact on productivity only at 10-year horizon (Korea reaches top-3 average OECD).
  - 6-7 percent: Product markets, labor markets, training programs, childcare benefits — long-term impact with Korea reaching top-3 average OECD and tax shifts toward consumption taxes.

### Policy implications and priorities
- Urgency: Downward pressure on potential growth highlights the need for structural reforms to provide upward momentum, especially post-COVID.
- Priority reforms:
  - Relax product market rigidities; enhance competition in product markets and network sectors.
  - Relax labor market rigidities; increase flexibility, improve training and worker matching, reduce participation disincentives (focus on female and youth participation, reduce dualism).
  - Ensure competitive landscape to facilitate investment and activity in dynamic post-COVID sectors (digitalization and green transition).
- Korean New Deal (KND):
  - Five-year strategy with government funding KRW 114.1 trillion through 2025 (about 1 percent of GDP per year) across 28 projects in nine areas to strengthen digital capacity, accelerate low-carbon transition, and bolster social safety net.
  - Expected to mobilize private investment, broaden ICT use, and strengthen training/human capital.

### Methodology notes (accelerator, capacity utilization, MVF, local projections)
- Accelerator model (private non-residential investment ratio to capital stock) estimated on annual data with first two lags of output-to-capital ratio; adjusted R-squared: 0.88; Observations: 47.
  - Selected coefficient highlights:
    - Constant: 55.88 (Standard error: 7.48)
    - Private non-residential capital stock, inverse (lagged): -848.13 (Standard error: 130.89)
    - Expected real GDP growth (one-sided H-P filter, first lag): 1.81 (Standard error: 0.52)
    - Government investment to private non-residential capital stock (first lag): 2.21 (Standard error: 0.28)
- Capacity utilization estimated via median quantile regression for manufacturing and applied to non-manufacturing; used to adjust capital services for utilization fluctuations.
- Production function uses labor share α = 0.61; TFP trend assumed to return to 2000-2019 average of 1.4 percent.
- MVF augments production function with Phillips curve and Okun’s law, solved with Bayesian Maximum Likelihood and includes consensus forecasts.
- Local projections identify recessions as two consecutive quarters of GDP contraction and estimate five-year horizon effects; sample includes 1979-80, 1997-98, and 2008-9 episodes.

*Source: wpiea2021092-print-pdf - IMF staff calculations and chapter text.*

### INTRODUCTION _________________________________________________________________________________ 4

### INTRODUCTION

### Korea’s recent growth trajectory and convergence
- Beginning in the early 1960’s, Korea’s income per capita was less than 10 percent of that in the United States.
- Per-capita income growth then exceeded 10 percent per year for some 10-year periods.
- By recent decades, Korea’s per capita income level rose to two-thirds of U.S. income.
- Growth has slowed since the Asian financial crisis (AFC) and global financial crisis (GFC), nearing the pace of advanced economy comparators.

### Main drivers of historical growth
- Rapid expansion driven by increasingly competitive export-oriented manufacturing.
- Underpinned by:
  - High investment in physical capital.
  - Increasing educational attainment (human capital).
  - A growing working-age population.
  - Deeper structural policies, institutional improvements, sound macroeconomic policies, and an industrialization orientation toward export competitiveness.
- Resulted in a high growth–high investment equilibrium with contributions from factor accumulation and some contribution from total factor productivity (TFP).

### Recent headwinds and structural challenges
- Investment weakened after the AFC as firms corrected pre-AFC financial imbalances.
- Pace of productivity growth declined.
- Favorable demographics have turned into a headwind as the population ages.
- Shifts in demand patterns expose structural weaknesses that were less apparent during the high-growth era.
- Product and labor market rigidities and structural features may amplify adverse effects from shocks.

### COVID-19 shock and implications for potential output
- The pandemic adds to headwinds on potential output and may produce scarring through lasting reductions in physical and human capital accumulation and on productivity growth.
- Pre-COVID potential growth had already decelerated to about 2½ percent.
- The COVID-19 shock could lower potential output in the medium term by about three percent, with wide uncertainty.

### Assessment approach and key analytical elements
- Industry-level assessment of TFP convergence and levels across sectors, showing:
  - Stronger convergence in manufacturing.
  - Significant scope for convergence in most industries.
  - Especially low levels of TFP in services.
- Detailed analysis of demographics’ impact on labor force participation and a projection for private investment.
- Two models used to estimate a baseline path for potential output pre-COVID.
- Examination of COVID-19’s impact drawing on Korea’s experience after past downturns.
- An illustrative structural reform scenario to quantify possible gains.

### Policy implications and illustrative reform impact
- Downward pressure on potential growth highlights urgency of structural reforms to provide upward momentum.
- Priority areas identified:
  - Relax product market rigidities.
  - Relax labor market rigidities.
  - Ensure a competitive landscape that facilitates investment and activity in sectors likely to be dynamic post-COVID.
- An illustrative reform scenario shows sizable possible gains—sufficient to raise medium-term potential growth by about half a percent per year.

### Structure of subsequent analysis (as presented in the source)
- Detailed demographic and labor force participation analysis.
- Private investment projection inputs.
- Baseline pre-COVID potential output estimation using two models.
- COVID-19 impact on potential output with scenarios informed by past recessions.
- Policies to promote structural transformation and an illustrative reform scenario estimating potential output gains.

*Source: wpiea2021092-print-pdf - INTRODUCTION*

### Appendix for a list of comparators and description of selection criteria.

### Appendix for a list of comparators and description of selection criteria

### Comparison with comparator economies and growth drivers
- Plots compare Korea against up to 20 comparator economies showing median, 1st and 3rd quartiles, and 10th and 90th percentiles.  
- Key descriptive findings:
  - Korea's per capita income is now approaching the advanced economy median.
  - Growth was achieved through heavier reliance on accumulation of physical and human capital and at lower levels of productivity than comparators.
  - Capital services have been the main contributor to growth, with the pace slowing over time in parallel with the overall economy.
  - During the 1970s, the rising working-age population helped drive a sharp increase in hours worked; since then, hours worked have increased slowly with a decline in hours per worker offsetting much of the employment increase.
  - Rising education levels increased labor quality contributions in the 1980s through the 2000s.
  - TFP growth was strongest in the 1970s and 1980s before slowing subsequently.

### Productivity performance by sector
- Manufacturing:
  - TFP growth in export-oriented manufacturing averaged 6 percent per year in the 1970s-1990s.
  - High-tech industries (generally export-oriented) were primary drivers; slower TFP growth occurred in other manufacturing industries.
- Services and other goods:
  - TFP growth in the other goods sectors was negligible.
  - Services experienced positive but slow TFP growth overall.
  - Contribution of capital deepening varied less across sectors; hours worked increased more in services as services grew in importance with rising income.

### Recent convergence and productivity trends
- TFP level has stagnated since the early 2000s at less than two-thirds the U.S. level.
- TFP growth since 2010 has been at its slowest pace since the growth takeoff began.
- Reversion from the outsized growth in the 1970s and 1980s explains two to three percentage points of the slowdown in growth in recent decades.
- Korea’s growth in the 2010s remained at the top 10th percentile for countries of its income level.
- Structural challenge: shift from capital and human accumulation toward raising productivity and facilitating structural transformation, particularly as population and working-age share are set to decline.

### COVID-19 initial impacts (observed)
- Actual output contracted by over four percent in the first half of 2020.
- Employment declined by over three percent by early 2021.
- Reallocation across industries during COVID-19:
  - Overall reallocation higher than during the GFC, surpassed only by the AFC.
  - COVID-19 prompted new highs in reallocation within services; AFC mainly affected manufacturing.
  - Initial effects in 2020 were greatest in relatively low-wage industries (bubble chart relates Real GDP growth, 2020 - Q2 (y/y % change) to Average monthly wage by industry (million won, 2019)).

### Demographics and labor input projections
- Historical labor utilization:
  - Indexes (2000=100) show measures for Hours worked, Employment, Hours per worker.
- Key demographic projections and implications:
  - Total population, 25-64 years old, and 65 and older projected annual percent changes shown for 2020–2050.
  - Aggregate labor force participation:
    - Projected to increase through 2025 driven by rising educational attainment and increasing propensity of younger female cohorts to participate.
    - After 2025, aggregate participation projected to decrease steadily as the elderly share rises and educational attainment pace levels off.
  - Cohort approach to participation:
    - Uses labor force entry and exit rates by age and gender cohort.
    - Produces a substantially higher aggregate participation rate than forecasts using constant age-gender participation rates.
  - Labor quality (average years of schooling) is high with little scope for large further increases; tertiary, secondary, primary series shown through 2050.

### Investment outlook and capital accumulation
- Private non-residential investment modeled using an expanded accelerator linking investment-capital ratio to lagged output growth-capital stock ratio, adding expected growth over the next five years and lagged government investment.
- Model findings:
  - Estimation extends back to 1970 and explains the increase in investment in the 1970s-1990s, the drop after the AFC, and the gradual decline since then.
  - Model forecast suggests private non-residential investment of about 19-20 percent of GDP over the next decade, slightly below recent levels.
  - With a higher capital-output ratio now, this implies a slower rate of growth in the capital stock and a headwind for potential growth.

### Capacity utilization and capital services adjustment
- Capacity utilization:
  - Estimated relationship between manufacturing-sector GDP and capacity utilization applied to non-manufacturing to derive economy-wide estimate weighted by capital stock.
  - Manufacturing capacity utilization historically above non-manufacturing in 2004-2007 and 2010-11, more recently below non-manufacturing.
  - Aggregate capacity utilization peaked before the AFC, dipped to lowest levels since the AFC by 2019.
- Adjustment rationale:
  - Capital services are adjusted for capacity utilization to include fluctuations in capital utilization as contributions of capital services to output, mirroring how labor fluctuations are accounted for via hours worked.
  - This adjustment produces higher estimated potential TFP growth in periods when capacity utilization is below long-term averages once utilization returns to typical levels.

### Potential output estimation approaches and pre-COVID baseline results
- Two approaches used:
  - Production function:
    - Decomposes output into capital, labor, and TFP using income shares and H-P filter for trends.
    - Incorporates detailed labor inputs (cohort-based participation, hours per worker, labor quality) and capacity utilization–adjusted capital services.
    - Factors filtered using both historical and forecast values with forecasts as of January 2020 (pre-COVID).
  - Multivariate filter (MVF):
    - Kalman filter conditioning on Okun’s law and the Phillips curve, and including Consensus Forecasts of real GDP and inflation.
    - Includes capacity utilization as an observable factor.
- Pre-COVID estimates:
  - Both approaches find potential output growth recently declined to the mid-2 percent range.
  - MVF and production function broadly coincide, with potential growth decelerating from around 7 percent in the mid-1990s.
  - Pre-COVID estimate for 2019: Production function 2.7 versus MVF 2.4.
  - Production function without capacity utilization adjustment yields lower potential growth since 2012 due to not accounting for below-average capacity utilization (affecting potential TFP estimates).
  - Long-run projections (pre-COVID):
    - Potential growth about 2.2-2.4 percent in the 2020s.
    - Declining to about 2 percent per year in the 2030s.
    - Declining to about 1.5 percent per year by 2050.
  - Assumption: TFP growth returns to its average 2000-19 pace, implying some recovery relative to 2016-19.

### COVID-19 potential medium-term effects (scenario based on historical recessions)
- Historical-recession econometric evidence for Korea:
  - Potential output fell after previous recessions and remained below pre-recession level in the medium term.
  - Most of the impact on potential came through lower investment and labor force participation; post-recession TFP tended to be resilient.
- Applied to COVID-19:
  - Estimated that COVID-19 could lower the level of output by about 3 percent.
  - Using the impact on headline output directly yields a response of 3.5 percent; aggregating estimated responses across factors of production yields a response of (figure text truncated in source).
- Qualitative mechanisms highlighted for post-COVID effects:
  - Persistent reallocation across industries, especially services.
  - Greater initial impact on low-wage industries where adaptability may be lower.
  - Potential for long-term effects given reshaping of demand patterns and supply chains despite comprehensive fiscal, monetary, and financial policy responses.

*Source: IMF staff Appendix (comparators list and selection criteria, figures and analysis as provided in the source content).*

### 2.5 percent. There is uncertainty surrounding

### wpiea2021092-print-pdf - 2.5 percent. There is uncertainty surrounding

### Impact of COVID-19 on potential output
- Estimated cumulative impact on potential output: 2.5 percent.
- Standard errors encompass an impact of between 1½ and 5 percent.
- Timing and composition:
  - Largest effects on the growth rate of potential output occur in the immediate aftermath of the shock, with a slow normalization back toward the previous potential growth rate.
  - Principal channel: lower investment rates over the medium term leading to slower capital accumulation.
  - Secondary channel: lower labor force participation.
- Episode-specific factors affecting interpretation:
  - Synchronized, global nature of the shock could amplify impact since potential in other economies is also affected.
  - Large policy response in Korea may buffer the medium-term impact.
  - Sudden stop in cash flows for many firms due to shutdowns early in the outbreak.
  - Deeper structural shifts in supply chains and demand across sectors could cause scarring: destruction of firms and worker-firm relationships, skills mismatch, uncertainty and worsening balance sheets constraining investment.
  - Relevance for Korea: large proportion of SME debt where firms’ cash flows do not cover debt service; product and labor market rigidities may impede sectoral reallocation.
  - Possible mitigating factors: transmission through investment may be smaller than after the AFC; potential positive effects from accelerated digitalization.

### Prospects for structural transformation
- Challenge: boost potential growth given shrinking working-age population and COVID-19 drag via facilitating structural transformation.
- Approach in the chapter:
  - Measures labor productivity relative to the frontier at detailed industry level to assess scope for convergence.
  - Discusses barriers constraining growth and convergence and identifies reforms to promote structural transformation.
  - Provides a high-level illustrative scenario to quantify possible gains.

### Sector-level productivity and convergence
- Aggregate convergence:
  - Labor productivity relative to the United States converged from 13 percent on average in the 1960s to about 55 percent in the 2010s.
- Sector patterns:
  - Manufacturing convergence strongest: labor productivity in the 2010s reached about 60 percent of the U.S. level, from 7 percent in the 1960s.
  - Services converged more slowly: reach 49 percent of the U.S. level in the 2010s, from 27 percent in the 1960s.
- Industry variation (manufacturing):
  - Labor productivity ranges from below 50 percent of U.S. productivity in manufacturing of food and beverages and petroleum and coal products to over 80 percent in textile and leather manufacturing and production of basic and fabricated metals.
- Services productivity gaps:
  - Market services in which labor productivity is less than 40 percent of that in the United States.
  - Specific services at or below 30 percent of the U.S. level: wholesale and retail; transportation and storage; accommodation and food services; publishing and communications; information services.
  - Productivity closer to U.S. level in finance and insurance and in professional, scientific, and technical services.

### Shifting global demand and constraints on rebalancing
- Pre-COVID trends and COVID-accelerated shifts:
  - Overall growth in global trade has fallen; slowing import growth in Korea’s trading partners.
  - Global demand shifting toward services, where Korea’s market share is lower.
  - Producers seeking more localized supply chains and greater digitalization of economic activity may accelerate these trends.
  - Digitalization favors Korea’s tech-intensive export composition.
- Implication:
  - Need to rebalance toward services as manufacturing contribution may fade; services rebalancing could raise output and employment given services’ higher labor intensity.
- Korea’s reliance on manufacturing:
  - Korea’s manufacturing share deviated from the typical de-industrialization pattern and reversed since 2000 due to China’s integration and demand.

### Structural strengths and rigidities
- Strengths (competitiveness and innovation):
  - Ranks first in ICT adoption; well-developed high-tech industry, strong digital infrastructure, high digital penetration (with a large generational gap).
  - Strong ranking in innovation pillar: research and development and commercialization of innovation.
  - Infrastructure, business dynamism, institutions, and skills are in favorable ranges (top quintile for some).
- Product and labor market rigidities:
  - Product market regulation restrictive relative to OECD average; restrictive regulation of communication sector and professional services; high costs to start a business; shortcomings in competition due to market dominance; distortive taxes and subsidies; relatively high and complex tariff and non-tariff barriers.
  - Labor market rigidities: flexibility of wage determination, hiring and firing flexibility, redundancy costs, and labor-employer cooperation are problem areas; meritocracy and incentivization relatively competitive.
- Labor market dualism and underutilized capacities:
  - Dualism between “regular” and “non-regular” employment limits youth and female participation; female employment concentrated in non-regular positions or unpaid family work; substantial female-male wage gap.
  - Skills mismatch despite high educational attainment; need for improved vocational training and employer-education coordination.
  - Seniority-based wage system and mandatory retirement ages reduce elderly participation.

### Korean New Deal (KND)
- Objectives: five-year strategy to facilitate transformation toward a more digital and green economy by 1) strengthening digital capacity, 2) accelerating transition to a low-carbon economy, 3) strengthening the social safety net.
- Planned scope and financing:
  - 28 projects in nine key areas with estimated government funding of KRW 114.1 trillion through 2025, or about 1 percent of GDP per year.
- Expected channels to reinvigorate growth:
  - Mobilize private investment by creating new markets, stimulating private demand, and improving regulations.
  - Broaden ICT use to raise productivity.
  - Strengthen training and human capital investments to raise labor force participation and broaden sharing of growth benefits.

### A reform scenario and quantified gains
- Scenario construction:
  - Uses Kim and Loayza (2019) for productivity drivers (innovation, infrastructure, education, labor markets, institutions) and Dao and others (2014) for labor force participation and employment.
  - Assumes indicators are gradually raised to the 75th percentile of OECD countries where Korea lagged; where Korea exceeded the benchmark, standing remains unchanged.
  - Labor market reforms assumed to close one third of the gap between female and male labor force participation rates and raise youth employment rates by ten percentage points.
- Long-term quantified impacts:
  - Potential output about 12 percent higher than the baseline once effects are fully realized.
  - Income per capita lifted from around two-thirds of the U.S. level currently to roughly 80 percent.
  - Productivity-enhancing reforms account for about eight percentage points of the total impact—six percentage points directly and the rest through higher investment and capital stock—with labor market reforms explaining the remainder.
  - The productivity contribution is broadly balanced across model areas, largest from institutions (including regulatory quality).
  - Peak impact on potential growth is over half a percentage point per year.
- Comparative context:
  - Estimates broadly align with other studies: Bouis and Duval (2011) estimate 10 percent impact at 10-year horizon under top-3 OECD attainment; OECD (2018a) estimates 20 percent at 40-year horizon under top-5 OECD attainment. Table 2 in the chapter summarizes these comparisons.

*Source: IMF staff calculations and text from wpiea2021092-print-pdf.*

### 4.7 percent Employment protection and

### 4.7 percent Employment protection and network sector regulations

### Estimated reform impacts and assumptions
- 4.7 percent: Employment protection and network sector regulations — Impact on productivity only, at 10-year horizon; Korea assumed to reach top-3 average OECD (Zoli and others (2018)).
- 6-7 percent: Product markets, labor markets, training programs, childcare benefits — Long-term impact; Korea assumed to reach top-3 average OECD in labor and product markets; also assumes shift to consumption taxes from capital and income taxes.
- Scenario timing: The above estimates measure long-term impact on output, assuming measures begin to be implemented in the near term and evaluated over a long horizon (effects may take several years to be fully implemented and several more years to materialize).

### Medium-term potential growth scenario and COVID-19 interaction
- Under a scenario encompassing both the COVID-19 shock and prompt implementation of reforms, the economy could return to roughly the pre-COVID path of potential output by 2030, though this estimate is subject to uncertainty surrounding both the COVID shock and reform effects (Figure 24).
- The cumulative impact of reforms would only offset that of the COVID-19 shock after several years, highlighting the desirability of early reform implementation to avoid a more protracted return to the pre-COVID 19 path of potential output.
- The COVID-19 pandemic has been relatively large as a reallocation shock—on par with the Asian Financial Crisis to this point—and more concentrated in labor-intensive services sectors.

### Quantified COVID-19 scarring risk
- The COVID-19 pandemic could reduce potential output by 2.5-3.5 percent, mainly through lower investment and labor force participation.
- The COVID-19 shock has been smaller than previous recessions in Korea but still poses risks to medium-term output via reduced investment and labor force participation.

### Conclusions on Korea’s potential growth dynamics
- Potential growth has decelerated from about 7 percent in the mid-1990s to about 2½ presently, with contributions from all factors of production decreasing.
- Contributing factors to slower trend growth:
  - Slower labor force growth due to demographics (working-age population begins falling and educational attainment levels off).
  - Lower investment and slower growth in the capital stock.
  - Convergence to high-income status.
  - Shifts in external demand (growth in merchandise trade slows and service sectors rise).
  - Long-standing structural rigidities that have become more binding as the economy moves closer to the frontier.
- The export-oriented manufacturing sector has achieved notably high productivity growth, but there remains scope for convergence to frontier levels of productivity in all industries, especially services.

### Policy implications and priorities
- Urgency of reforms to boost labor force participation, productivity, and investment, which have become even more critical in the wake of COVID-19.
- Priority reforms include:
  - Competition-enhancing reforms to product markets.
  - Labor market reforms to increase labor inputs by increasing flexibility, improving training and worker matching, and reducing disincentives for participation.
  - Policies to facilitate restructuring and transformation to reduce long-term scarring from the pandemic.
- The Korean New Deal is noted as taking steps to help address these issues.
- Prompt implementation is essential because the effects of such reforms would occur only gradually and early action can buffer any slowdown in potential growth caused by COVID-19.

*Source: IMF staff calculations as presented in the chapter section.*

### 1. Wholesale & retail trade

### 1. Wholesale & retail trade

### Accelerator model: specification and estimation
- Estimation framework builds on the accelerator model in Oliner, Rudebusch, and Sichel (1995).
- Core equations:
  - Investment identity (A.2): IT = α + Σβi ΔK*_{t−i} + δ K_{t−1}
  - Estimation equation (A.3): IT / K_{t−1} = δ + α / K_{t−1} + Σ_{i=1..N} βi (ΔY_{t−i} / K_{t−i+1}) + y_{t−1}^r + III_{t−1} / K_{t−2}
    - y_{t−1}^r is expected real GDP growth extrapolated from the trend using a one-sided H-P filter.
    - III_{t−1} is lagged government investment.
- Data and frequency:
  - Dependent variable and capital stock include only private non-residential investment or capital.
  - Model estimated on annual data.
  - First two lags of the ratio of change in output to lagged capital stock are included; additional lags were rejected by standard tests.
- Estimation results (Table A2):
  - Dependent Variable: Private non-residential investment (ratio to private non-residential capital stock)
  - Coefficients and statistics:
    - Constant: 55.88; Standard error: 7.48; t-Statistic: 7.47; Probability: 0.00
    - Private non-residential capital stock, inverse (lagged): -848.13; Standard error: 130.89; t-Statistic: -6.48; Probability: 0.00
    - Output growth, ratio to private non-residential capital stock (first lag): 0.11; Standard error: 0.07; t-Statistic: 1.63; Probability: 0.11
    - Output growth, ratio to private non-residential capital stock (second lag): 0.09; Standard error: 0.09; t-Statistic: 0.98; Probability: 0.33
    - Expected real GDP growth, extrapolated from one-sided H-P filter (first lag): 1.81; Standard error: 0.52; t-Statistic: 3.50; Probability: 0.00
    - Government investment to private non-residential capital stock (first lag): 2.21; Standard error: 0.28; t-Statistic: 7.77; Probability: 0.00
  - Goodness of fit and sample:
    - Adjusted R-squared: 0.88
    - Durbin-Watson stat: 1.26
    - Observations: 47

### Economy-wide capacity utilization: measurement and estimation
- Theoretical relationship (A.4):
  - CCUi = α + β (yi − yi*) + Δyi
    - where yi − yi* is the cyclical component of output in sector i.
- Cyclical component estimated with H-P filter (lambda = 100).
- Estimation method: quantile regression (median) due to skewness of capacity utilization series.
- Application:
  - Parameters from manufacturing capacity utilization regression applied to non-manufacturing value added to estimate non-manufacturing capacity utilization.
  - Total economy capacity utilization computed as weighted average of actual manufacturing and estimated non-manufacturing figures, weights = sector shares in total capital stock.
- Manufacturing capacity utilization results (Table A3):
  - Dependent Variable: Capacity utilization in manufacturing (percent of total)
  - Method: Quantile Regression (Median)
  - Coefficients and statistics:
    - Constant: 76.09; Standard error: 1.05; t-Statistic: 72.55; Probability: 0.00
    - Real value added in manufacturing, log difference: 13.23; Standard error: 11.27; t-Statistic: 1.17; Probability: 0.25
    - Cyclical component of real value added in manufacturing: 0.8; Standard error: 0.2; t-Statistic: 4.02; Probability: 0.00
  - Summary statistics:
    - Mean dependent variable: 76.22
    - Quantile dependent variable: 77.60
    - Adjusted R-squared: 0.38
    - Observations: 40

### Production function and multivariate filter models: long-term projections
- Production function (A.5): Y = A K^{(1−a)} L^{a}
  - Y = output; A = total factor productivity; K = capital services; L = labor inputs; α = labor share.
  - Labor share used: 0.61 (matches Bank of Korea study reference).
- Data inputs:
  - Capital services from OECD (assuming constant capacity utilization), adjusted for capacity utilization.
  - Labor inputs include labor quality from Penn World Table version 9.1 and hours worked from OECD.
  - Historical TFP estimated as residual from A.5.
- Long-term projection assumptions:
  - Capital services:
    - Gross fixed capital formation projected using accelerator model results, yielding broadly stable ratio of investment as share of GDP.
    - Capital depreciation rate assumed to follow recent trends.
    - Capacity utilization assumed to return to its historical norm by 2025.
  - Labor inputs:
    - Population projections by age and gender from Statistics Korea.
    - Labor force participation projected using cohort approach described in main text.
    - Employment rate assumed to return to historical norm by 2025.
    - Hours per worker extrapolated from recent trends.
    - Labor quality change estimated combining projections of educational attainment in Barro and Lee (2013) with population and participation projections.
  - Total factor productivity:
    - Productivity growth assumed to gradually return to its 2000-2019 average of 1.4 percent.
    - Implied convergence: from 63 percent of the U.S. level currently to 72 percent in 2050, assuming U.S. productivity growth of 0.8 percent (U.S. 2000-2019 average).
- Trend estimation:
  - Trend value for each component estimated by applying Hodrick-Prescott filter (lambda = 100) on historical data and projections together.
  - Potential output calculated as sum of trend components according to equation A.5.

- Multivariate filter (MVF):
  - Extends production function by adding Phillips curve and Okun’s law relationships and includes consensus forecasts for real GDP and inflation and capacity utilization as observable.
  - Labor supply block includes working-age population and labor force participation rate.
  - Long-term assumptions for capital, labor inputs, and productivity same as production function (except labor quality and hours per worker not part of MVF).
  - Model solved using Bayesian Maximum Likelihood techniques.

### Local projections: impact of previous recessions on potential output
- Recession definition and sample:
  - Recessions defined as two consecutive quarters of contraction in real GDP.
  - Recessions in Korea identified in 1979-80 and 1997-98; 2008-9 included due to large decline in Q4-2008 and narrowly positive Q1-2009.
- Method:
  - Recession variable measuring growth surprise estimated in an autoregressive model including third through fifth lags of real GDP growth.
  - Effects estimated out to a five-year horizon using local projections (Jordà (2005); Teulings and Zubanov (2014)).
- Main econometric finding:
  - Previous recessions in Korea produced a sizable negative impact on potential output averaging 7 percent over five years (Figure A.1).
  - Effects largest in immediate aftermath but continue to materialize over time; potential growth only slowly converges to the no-shock path.
- Factor-level impacts (Figure A.2 and discussion):
  - Insignificant effects on productivity in some episodes:
    - Trend TFP slowed after 1979-1980 but not after 1997-1998; actual TFP rebounded sharply.
    - TFP was resilient during global financial crisis; trend slowdown occurred a few years later.
  - Sharp reduction in investment and slower growth of capital inputs:
    - Investment-GDP ratios fell immediately and remained depressed for several years, notably after the AFC (Asian Financial Crisis).
    - Results similar for investment-capital ratios, capital stock, and capital services.
  - Temporary reduction in growth of labor inputs:
    - Potential employment rates dipped after recessions then recovered in the medium term.
    - Recessions found to have long-lasting effects on labor force participation rates.
    - Little impact found on hours per worker.
  - Aggregated effect from factor-level estimates:
    - Cumulative reduction of potential output estimated at -5.2 percent using factor contributions, with majority of impact in medium term from reduction in capital services.
- Interpretation:
  - Findings confirm that large downturns can have long-lasting, potentially permanent, effects on output through scarring/hysteresis.
  - Number of observations small; results may not generalize to current episode.

### Structural reforms scenario: quantification approach
- Two-stage quantitative approach:
  1. Use parameters estimated in Kim and Loayza (2019) and Dao and others (2014) to quantify direct impact of assumed reforms on productivity and labor force participation, respectively.
  2. Feed direct effects into the production function model for potential output to estimate feedback to output and investment.
- Treatment of labor force participation effects:
  - Given persistence of past participation in cohort outcomes, reforms’ effects on participation applied via cohort model to incoming cohorts but not to cohorts already in the labor force (conservative assumption).
- Modeling caveats:
  - Aggregate structure and reduced-form empirical estimates imply no feedback to labor quality, hours worked, or equilibrium unemployment rate is assumed.
  - Effects through those channels likely smaller than incorporated channels.

*Source: IMF staff calculations (excerpt from wpiea2021092-print-pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021092-print-pdf.pdf_