## wpiea2020272-print-pdf

## Source details

**Canonical URL:** [wpiea2020272-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020272-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020272-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020272-print-pdf.pdf.json)

---

### Major findings on recessions and medium-term GDP effects
- On average, GDP was about 4¾ percent below trend three years after the start of a recession.
- After the third year, the impulse response flattens out: the growth rate of GDP returns to pre-recession values, though the level of GDP remains below trend in the medium-term.
- Large recessions (peak to trough decline in GDP in the top quartile, defined as a fall of 4.25 percent or more) have more persistent effects:
  - GDP continues falling below trend for several years after large recessions and is about 11 percent below trend in the medium term.
  - For other recessions (bottom three quartiles), output is about 2.5 percent below trend two years after the start of a recession, with catch up growth after that and GDP not being significantly below trend by the fifth year.
- Specific 2020Q2 observations: In Australia and New Zealand, the decline in GDP in 2020Q2 relative to peak was 7.2 and 13.4 percent, respectively, well above the threshold for a large recession.

### Drivers of persistent medium-term effects
- Total Factor Productivity (TFP) declines:
  - For large recessions, TFP is almost 6 percent below trend 5 years after the start of a recession.
- Capital stock declines:
  - For large recessions, the capital stock is about 7 percent below trend 5 years after the start of a recession.
- Investment effect:
  - Investment as a share of GDP remains about 1.5 percentage points below pre-recession levels in the medium term.
- Labor market effects:
  - Increase in the unemployment rate by about 3.5 percentage points; no significant average change in labor force participation (noting heterogeneity across episodes).
- Results are qualitatively similar when considering all recessions, though quantitatively smaller.

### Robustness and role of financial crises
- About 40 percent of the large recessions in the sample were accompanied by financial crises.
- Limiting the sample to large recessions without banking or currency crises, or excluding the period of the global financial crisis:
  - Reduces the estimate for the medium-term impact on GDP from over 11 percent to about 6 percent.
  - The impact remains significantly different from zero, indicating non-financial crises can also result in significant scarring.
- Results are robust to alternative recession definitions and to including year fixed effects in the local projection model.

### Prospects for medium-term potential output in Australia and New Zealand — pre-pandemic
- Potential output estimation approach:
  - Semi-structural multivariate filter model incorporating a Phillips curve and Okun’s law to identify unobservable slack and potential output from output, inflation, and unemployment data.
  - Growth accounting framework applied to estimated potential output growth.
- Pre-pandemic trends:
  - Australia: Low productivity growth and weak capital accumulation in both mining and non-mining sectors contributed to slowed potential output after the mining boom of the early 2010s.
  - New Zealand: Weak productivity growth was masked by strong working age population growth (driven by immigration) and a rising capital stock.

### Transmission channels of the pandemic on potential output
- Framework: Simple growth accounting examining impacts on TFP, capital, and labor; labor supply effects decomposed into NAIRU, labor force participation, and working-age population growth.
- Overall conclusion: The current crisis is likely to affect medium-term potential output significantly in Australia and New Zealand through TFP, capital, and labor channels; some transmission channels differ from previous recessions.

### NAIRU, sectoral reallocation, and labor-market impacts
- Sectoral reallocation:
  - Job losses concentrated in service sectors dependent on personal contact (e.g., accommodation, food services, recreational services).
  - The degree of sectoral reallocation in employment growth and stock returns during the current crisis is the largest in the last 20 years.
- Expected NAIRU impact:
  - Heightened sectoral reallocation is expected to increase skills mismatches and raise the NAIRU.
  - Structural VAR model estimates suggest unemployment rates in Australia and New Zealand will likely increase by around 0.6 and 0.8 percentage point in the medium term due to sectoral reallocation.
- Sectoral reallocation indices:
  - Stock-market based index: 25 industries for Australia, 12 industries for New Zealand; monthly FTSE data.
  - Labor-market based index: 19 industries for Australia, 16 industries for New Zealand; quarterly employment data.
  - For New Zealand, labor-market reallocation was masked by a large-scale wage subsidy program.

### Labor force participation and working-age population effects
- Labor force participation:
  - Initial large drop in labor force participation following lockdowns, particularly among female and young workers.
  - Baseline expectation: labor force participation rates in Australia and New Zealand could remain below pre-COVID-19 trends by about 0.2 percentage point in the medium term.
  - The lower labor force participation rates would reduce medium-term potential output by 0.1 percent.
  - IMF (2018) finds that a 10 percentage point change in the relative service employment share leads to 0.1 percent increase in the labor force participation rate.
- Working-age population and immigration:
  - Before the pandemic, net migration accounted for 60 percent and 70 percent of population growth in Australia and New Zealand, respectively.
  - With pandemic, immigration flows have collapsed; net migration is expected to remain at zero while travel restrictions are in place and to recover gradually, returning to pre-COVID-19 trends in 2024.
  - The interim loss of net migration has a permanent impact on the stock of immigrants.
  - Medium-term impact on working age population expected to be -2.0 percent for Australia and -2.8 percent for New Zealand.
  - This translates to a 1.2-1.7 percent loss of potential output in the medium term.

### Total Factor Productivity (TFP) and reallocation effects
- Baseline assumption: within-sector productivity slowdown is assumed to reduce TFP and potential output by two percent.
- Reallocation effects:
  - Labor reallocation in Australia during the current crisis is estimated to have a positive impact on productivity, reflecting expected labor shift from lower-productivity to higher-productivity sectors.
  - Positive impacts of compositional change are estimated at about 0.9 percent and are expected to partially offset the within-sector TFP slowdown and loss of potential output.
- Long-lasting TFP impacts are expected due to impaired balance sheets, reduced R&D and innovative investment, and reallocation to less-productive sectors in past recessions.

### Capital accumulation and firm-level determinants
- Medium-term capital effects:
  - Capital is expected to remain 5 percent and 6 percent below pre-COVID trend in Australia and New Zealand, respectively.
  - Lower capital would reduce medium-term potential output by 2.0-2.4 percent.
- Illustrative 2020 impacts:
  - Revenue losses leading to higher debt levels could reduce firms’ capital accumulation by about 0.2 percent in 2020 in Australia and New Zealand, with persistent medium-term effects.
  - Elevated uncertainty could reduce capital accumulation by 0.5 percent in 2020 in Australia and by 0.8 percent in New Zealand.
  - Taken together, deteriorated balance sheets and elevated uncertainty would lead to a 0.3 percent and 0.4 percent loss of potential output in 2020 for Australia and New Zealand, respectively.
  - Note: illustrative debt effect estimation does not incorporate fiscal support; in Australia government support was large enough to offset the initial impact on corporate balance sheets.
- Firm-level investment determinants (key regression coefficients):
  - Australia (1994-2018, 7,893 observations):
    - Cost of Debt (-1): -0.404*** (0.097)
    - Debt (-1): -0.213*** (0.027)
    - Liquidity (-1): 0.009*** (0.002)
    - Uncertainty (-1): -0.037*** (0.012)
    - Tobin's Q (-1): 0.019*** (0.005)
    - R-squared: 0.064
  - New Zealand (1990-2019, 1,443 observations):
    - Cost of Debt (-1): -0.124 (0.171)
    - Debt (-1): -0.157*** (0.046)
    - Liquidity (-1): 0.013* (0.007)
    - Uncertainty (-1): -0.069*** (0.034)
    - Tobin's Q (-1): 0.044*** (0.007)
    - R-squared: 0.058
  - Significance indicators: *** and * indicate statistical significance at 1 and 10 percent levels, respectively.

### Projected medium-term potential output: baseline and scenarios
- Baseline simulation results:
  - Australia’s potential output in 2025 is estimated to be 5 percent below the pre-pandemic potential output trend.
  - New Zealand’s potential output in 2025 is estimated to be 6 percent below the pre-pandemic trend.
  - Labor productivity of Australia and New Zealand are both estimated to be 2½ percent lower than pre-pandemic trend.
- Scenario types:
  - Downside: more sluggish productivity growth; lower labor force participation rate.
  - Upside: stronger immigration flows; stronger productivity growth.
- Scenario-specific 2025 deviations from pre-COVID trends:
  - Downside scenarios:
    - Australia: potential output 6–7¼ percent lower than pre-COVID trends.
    - New Zealand: potential output 7¼–8 ½ percent below pre-COVID trends.
  - Upside scenarios:
    - Australia: potential output 2¾-3¼ percent lower than the pre-COVID trend.
    - New Zealand: potential output 3¾-4 percent below its pre-COVID trend.
- Additional simulation assumptions and parameters:
  - Immediate TFP shock: within-sector productivity assumed to decline by 2 percent in 2020 in the stronger TFP growth scenario before recovering.
  - Sectoral reallocations decay by 10 percent each year.
  - Debt overhang effects decay by 20 percent each year.
  - Uncertainty effects decay by 50 percent each year.
  - Cobb-Douglas production function with capital share α set as 0.4 for both countries.
  - Capital persistence parameter γ set at 0.66 for both countries.
  - Trend labor force participation rate obtained by the Hodrick-Prescott filter with smoothing parameter λ=100.
  - Balanced growth path condition and capital law used to model capital dynamics.

### Policy implications and recommendations
- Maintain macroeconomic support and implement reforms:
  - Maintain vigorous fiscal and monetary stimulus as long as needed to support demand and limit medium-term scarring.
  - Accommodative monetary policy should be maintained for a significant amount of time.
  - Given substantial fiscal space, the fiscal deficit should only be lowered gradually to avoid steep fiscal cliffs as emergency programs (for example, Australia’s JobKeeper program) are wound down.
  - Additional discretionary measures can be especially helpful as fiscal multipliers are likely to be large during the recession and in a low-interest-rate environment.
- Labor-market and reallocation policies:
  - Gradually phase out exceptional wage subsidy support as health risks diminish to facilitate movement to new, more productive jobs.
  - Consider well-designed and targeted hiring subsidies to encourage firms to hire new workers.
  - Implement active labor market policies, including retraining programs, to reduce skill mismatch and support worker reallocation.
  - Ensure adequate safety nets and adequate unemployment insurance, potentially linked to regional economic conditions in the event of localized lockdowns.
  - Use well-designed tax and expenditure policies to boost labor force participation, particularly for women and older workers.
- Investment and structural reforms:
  - Redeploy fiscal resources freed from winding down emergency programs toward productive investments that spur medium-term growth, including infrastructure and a green investment push.
  - Promote investment in digital technologies and scale up research spending to facilitate innovation and technology adoption.
  - Implement structural reforms to simplify business processes, improve competition in product and service markets, and reduce regulatory and tax burdens to boost innovation, productivity, and investment.
  - Phase out liquidity and other government support gradually to allow non-viable firms to exit while providing residual support to ensure survival of viable firms and finance necessary restructuring.

*Source — wpiea2020272-print-pdf (excerpt).*

### Appendix 1 for details on the methodology). Our sample covers the period from 1970 to

### wpiea2020272-print-pdf - Appendix 1 for details on the methodology). Our sample covers the period from 1970 to

### Major findings on recessions and medium-term GDP effects
- On average, GDP was about 4¾ percent below trend three years after the start of a recession.
- After the third year, the impulse response flattens out: the growth rate of GDP returns to pre-recession values, though the level of GDP remains below trend in the medium-term.
- Large recessions (peak to trough decline in GDP in the top quartile, defined as a fall of 4.25 percent or more) have more persistent effects:
  - GDP continues falling below trend for several years after large recessions and is about 11 percent below trend in the medium term.
  - For other recessions (bottom three quartiles), output is about 2.5 percent below trend two years after the start of a recession, with catch up growth after that and GDP not being significantly below trend by the fifth year.
- Specific 2020Q2 observations: In Australia and New Zealand, the decline in GDP in 2020Q2 relative to peak was 7.2 and 13.4 percent, respectively, well above the threshold for a large recession.

### Drivers of persistent medium-term effects
- Persistent medium-term effects are driven by:
  - Total Factor Productivity (TFP) declines: For large recessions, TFP is almost 6 percent below trend 5 years after the start of a recession.
  - Capital stock declines: For large recessions, the capital stock is about 7 percent below trend 5 years after the start of a recession.
  - Investment effect: Investment as a share of GDP remains about 1.5 percentage points below pre-recession levels in the medium term.
  - Labor market effects: Increase in the unemployment rate by about 3.5 percentage points; no significant average change in labor force participation (noting heterogeneity across episodes).
- Results are qualitatively similar when considering all recessions, though quantitatively smaller.

### Robustness and role of financial crises
- About 40 percent of the large recessions in the sample were accompanied by financial crises.
- Limiting the sample to large recessions without banking or currency crises, or excluding the period of the global financial crisis:
  - Reduces the estimate for the medium-term impact on GDP from over 11 percent to about 6 percent.
  - The impact remains significantly different from zero, indicating non-financial crises can also result in significant scarring.
- Results are robust to alternative recession definitions and to including year fixed effects in the local projection model.

### Prospects for medium-term potential output in Australia and New Zealand — pre-pandemic
- Potential output estimation approach:
  - Semi-structural multivariate filter model incorporating a Phillips curve and Okun’s law to identify unobservable slack and potential output from output, inflation, and unemployment data.
  - Growth accounting framework applied to estimated potential output growth.
- Pre-pandemic trends:
  - In Australia: Low productivity growth and weak capital accumulation in both mining and non-mining sectors contributed to slowed potential output after the mining boom of the early 2010s.
  - In New Zealand: Weak productivity growth was masked by strong working age population growth (driven by immigration) and a rising capital stock.

### Transmission channels of the pandemic on potential output
- Framework: Simple growth accounting examining impacts on TFP, capital, and labor; labor supply effects decomposed into NAIRU, labor force participation, and working-age population growth.
- Overall conclusion: The current crisis is likely to affect medium-term potential output significantly in Australia and New Zealand through TFP, capital, and labor channels; some transmission channels differ from previous recessions.

### Non-Accelerating Inflation Rate of Unemployment (NAIRU) and sectoral reallocation
- The pandemic induced large sectoral reallocation with uneven labor market adjustment; job losses concentrated in service sectors dependent on personal contact (e.g., accommodation, food services, recreational services).
- Sectoral dispersion:
  - The degree of sectoral reallocation in employment growth and stock returns during the current crisis is the largest in the last 20 years.
- Expected NAIRU impact:
  - The heightened sectoral reallocation is expected to increase skills mismatches and raise the NAIRU.
  - Structural VAR model estimates suggest unemployment rates in Australia and New Zealand will likely increase by around 0.6 and 0.8 percentage point in the medium term due to sectoral reallocation.

*IMF staff calculations and analysis as presented in the source content.*

### 0.3 percent and 0.5 percent loss of

### wpiea2020272-print-pdf - 0.3 percent and 0.5 percent loss of

### Sectoral impacts and financial market reflection
- Stock prices by industry show uneven sectoral impact from end-January 2020 to end-June 2020 (Figure 7).
- The speed of economic reallocation across sectors is unprecedented (Figure 8), measured by Lilien (1982) sectoral reallocation index calculated from FTSE industry level stock price data (25 industries for Australia, 12 industries for New Zealand).
- Sectoral reallocation leads to a persistent increase in the unemployment rate: impulse responses of unemployment rate to 0.1 percent point reallocation shock are shown for Australia and New Zealand (Figure 9).

### Labor market policies and NAIRU
- A permanent increase in unemployment benefits is often associated with a higher NAIRU.
- In New Zealand, social benefits including unemployment benefits have been increased permanently by NZ$25 per month.
- OECD cross-country evidence (Gianella et al., 2008) suggests the expanded unemployment benefits would be expected to lead to a 0.2 percentage point permanent increase in unemployment in New Zealand, which translates to a 0.1 percent loss of potential output in 2020.
- Assumption: the Coronavirus Supplement in Australia and the COVID-19 Income Relief Payment in New Zealand are temporary and will not affect NAIRU.

### Labor force participation
- Labor force participation is expected to decline moderately in the medium term.
- Pre-pandemic trend: participation was rising in many advanced economies, including Australia and New Zealand, supported by expansion of service sectors.
- Initial large drop in labor force participation followed lockdowns, particularly among female and young workers.
- Baseline expectation: labor force participation rates in Australia and New Zealand could remain below pre-COVID-19 trends by about 0.2 percentage point in the medium term.
- The lower labor force participation rates would reduce medium-term potential output by 0.1 percent.
- Note: IMF (2018) finds that a 10 percentage point change in the relative service employment share leads to 0.1 percent increase in the labor force participation rate.

### Working age population and immigration
- Closed borders constrain population growth through immigration.
- Before the pandemic, net migration accounted for 60 percent and 70 percent of population growth in Australia and New Zealand, respectively.
- With pandemic, immigration flows have collapsed; net migration is expected to remain at zero while travel restrictions are in place and to recover gradually, returning to pre-COVID-19 trends in 2024.
- The interim loss of net migration has a permanent impact on the stock of immigrants.
- Medium-term impact on working age population expected to be -2.0 percent for Australia and -2.8 percent for New Zealand.
- This translates to a 1.2-1.7 percent loss of potential output in the medium term.

### Total Factor Productivity (TFP)
- Long-lasting effects on TFP are expected, driven by decline in within-sector TFP growth and reallocation to less-productive sectors in previous recessions.
- In the current recession, productivity in each sector is expected to decline as innovative investment, including R&D, is impeded by firms’ impaired balance sheets and heightened uncertainty.
- Baseline assumption: within-sector productivity slowdown is assumed to reduce TFP and potential output by two percent.
- Reallocation effects may partially offset TFP slowdown: labor reallocation in Australia during the current crisis is estimated to have a positive impact on productivity, reflecting expected labor shift from lower-productivity to higher-productivity sectors.
- Positive impacts of compositional change are estimated at about 0.9 percent and are expected to partially offset the within-sector TFP slowdown and loss of potential output.

### Capital accumulation
- Capital accumulation is expected to slow, following an endogenous balanced growth path determined by productivity and labor input.
- In the medium-term, capital is expected to remain 5 percent and 6 percent below pre-COVID trend in Australia and New Zealand, respectively.
- Lower capital would reduce medium-term potential output by 2.0-2.4 percent.
- Firm-level analysis (IMF Corporate Vulnerability Unit / Worldscope) shows firms’ investment behavior is determined by firms’ financial positions and firm-level uncertainty in addition to expected profitability (Figure 12).
- Illustrative scenario: revenue losses assumed to result in higher debt levels could reduce firms’ capital accumulation by about 0.2 percent in 2020 in Australia and New Zealand, with persistent medium-term effects (Figure 13).
- Elevated global and domestic uncertainty could further reduce capital accumulation by 0.5 percent in 2020 in Australia and by 0.8 percent in New Zealand, though these effects are likely to be short-lived compared with adverse balance sheet effects.
- Taken together, deteriorated balance sheets and elevated uncertainty would lead to a 0.3 percent and 0.4 percent loss of potential output in 2020 for Australia and New Zealand, respectively.
- Note: the illustrative debt effect estimation does not incorporate fiscal support; in Australia government support was large enough to offset the initial impact on corporate balance sheets.

### Projecting medium-term potential output: baseline and scenarios
- Baseline simulation: weaker productivity growth, a slowing pace of capital accumulation and lower labor input all contribute to lower medium-term potential output (Figure 15).
- Australia’s potential output in 2025 is estimated to be 5 percent below the pre-pandemic potential output trend.
- New Zealand’s potential output in 2025 is estimated to be 6 percent below the pre-pandemic trend, reflecting higher reliance on net migration flows.
- Labor productivity of Australia and New Zealand are both estimated to be 2½ percent lower than pre-pandemic trend.
- Scenario types considered:
  - Downside: more sluggish productivity growth; lower labor force participation rate.
  - Upside: stronger immigration flows; stronger productivity growth.
- More sluggish productivity growth scenario: within-sector productivity in the medium-term will be 3.5 percent lower than pre-COVID, nearly double the decline assumed in the baseline.
- Downside labor force participation scenario: recognizes that many workers have left the labor market and assumes labor force participation rates remain below baseline (text truncated in source).

*Italic: Source — wpiea2020272-print-pdf (excerpt).*

### 0.8 percent below pre-COVID trends in the medium term, as more workers remain on the

### wpiea2020272-print-pdf - 0.8 percent below pre-COVID trends in the medium term, as more workers remain on the

### Alternative scenarios and projected potential output
- Baseline expectation: COVID-19 will have a lasting impact on potential output through weaker productivity, slower capital accumulation, and lower labor inputs.
- Upside and downside scenario summaries:
  - Stronger immigration flow
    - Travel restrictions lifted earlier; immigration flows resume faster in 2021.
    - From 2022, net migration to Australia and New Zealand would reach levels above the pre-pandemic trend, fully offsetting the decrease during the border closure.
  - Stronger TFP growth
    - Following a 2 percent decline in 2020, within-sector productivity is assumed to grow faster than under the pre-COVID trend in the medium-term, recovering to near pre-COVID trend levels in the medium-term (0.5 percent below the pre-COVID trend in 2025).
  - More sluggish productivity growth, lower labor force participation, and other downside scenarios (figures referenced in the source).
- Scenario-specific projected deviations from pre-COVID trends in 2025:
  - Downside scenarios:
    - Australia: potential output 6–7¼ percent lower than pre-COVID trends.
    - New Zealand: potential output 7¼–8 ½ percent below pre-COVID trends.
  - Upside scenarios:
    - Australia: potential output 2¾-3¼ percent lower than the pre-COVID trend.
    - New Zealand: potential output 3¾-4 percent below its pre-COVID trend.
- Medium-term aggregate statement:
  - Labor productivity in Australia and New Zealand will remain ¼-4¾ below pre-COVID trends under alternative scenarios.
- Chart references (2015=100): pre-COVID projection, post-COVID baseline, and alternative scenario level paths for 2015–2025.

### Key quantitative findings and dynamics
- Immediate TFP shock: within-sector productivity assumed to decline by 2 percent in 2020 in the stronger TFP growth scenario before recovering.
- Recovery/decay assumptions in simulations:
  - Sectoral reallocations are assumed to decay by 10 percent each year.
  - Debt overhang effects are assumed to decay by 20 percent each year.
  - Uncertainty effects are assumed to decay by 50 percent each year.
- Production-function parameterization and accounting:
  - Cobb-Douglas production function used with constant capital share 훼 set as 0.4 for both Australia and New Zealand.
  - Capital persistence parameter γ set at 0.66 for both Australia and New Zealand.
  - Trend labor force participation rate obtained by the Hodrick-Prescott filter with smoothing parameter λ=100.
- Balanced growth path condition for capital accumulation (as in the source):
  - Δln ( K_t^BGP ) = Δln (z_t^*)/(1−α) + Δln( L_t^* )
- Capital law in simulation (as in the source):
  - ln (K_t) = (1−γ) ln (K_t^BGP) + γ ln (K_{t−1}) + D O_t + U_t

### Policy implications and recommendations
- Overarching message:
  - Maintaining macroeconomic policy support until the recovery is firmly entrenched and stepping up economic reforms is paramount to mitigate COVID-19’s fallout on medium-term output.
- Fiscal and monetary policy:
  - Maintain vigorous fiscal and monetary stimulus as long as needed to support demand and limit medium-term scarring.
  - Accommodative monetary policy should be maintained for a significant amount of time.
  - Given substantial fiscal space, the fiscal deficit should only be lowered gradually to avoid steep fiscal cliffs as emergency programs (for example, Australia’s JobKeeper program) are wound down.
  - Additional discretionary measures can be especially helpful as fiscal multipliers are likely to be large during the recession and in a low-interest-rate environment.
- Reallocation and labor-market policies:
  - Gradually phase out exceptional wage subsidy support as health risks diminish to facilitate movement to new, more productive jobs.
  - Consider well-designed and targeted hiring subsidies to encourage firms to hire new workers.
  - Implement active labor market policies, including retraining programs, to reduce skill mismatch and support worker reallocation.
  - Ensure adequate safety nets and adequate unemployment insurance, potentially linked to regional economic conditions in the event of localized lockdowns.
  - Use well-designed tax and expenditure policies to boost labor force participation, particularly for women and older workers.
- Investment and structural reforms:
  - Redeploy fiscal resources freed from winding down emergency programs toward productive investments that spur medium-term growth, including infrastructure and a green investment push.
  - Promote investment in digital technologies and scale up research spending to facilitate innovation and technology adoption.
  - Implement structural reforms to simplify business processes, improve competition in product and service markets, and reduce regulatory and tax burdens to boost innovation, productivity, and investment.
  - Phase out liquidity and other government support gradually to allow non-viable firms to exit while providing residual support to ensure survival of viable firms and finance necessary restructuring.

### Cross-country empirical framework (Appendix 1)
- Estimation approach:
  - Use standard local projection models (LPMs) to estimate medium-term dynamics around recession episodes.
  - Baseline LPM specification (as in source):
    - Y_{i,t+h} − Y_{i,t−1} = α_{i}^h + β_h RECESS_{i,t} + γ_h X_{i,t} + Ɛ_{i,t}
  - RECESS_{i,t} is a dummy for the start of a recession from Martin et al (2015); X_{i,t} includes lagged recession dummies and lagged differences of the dependent variable; α_{i}^h are country fixed effects.
  - Standard errors clustered at the country level.
  - Baseline excludes year fixed effects to capture the synchronized nature of the current shock; robustness to including year fixed effects is reported in the main paper.
- Sample and data:
  - Baseline uses recession start dates for 23 advanced economies (Martin et al., 2015), broadly covering 1970 to 2012.
  - Robustness check uses Dovern and Zuber (2020) recession dates (27 countries), time coverage 1990 to 2017.
  - Annual data: GDP, TFP, employment, capital stock from Penn World Table 9.1 (PWT); unemployment rate and labor force participation from the OECD.
  - Crisis episode data from Laeven and Valencia (2018) used to identify large recessions not accompanied by banking or currency crises.

### Estimation of potential output before the pandemic (Appendix 2)
- Semi-structural multivariate filter model incorporating a Phillips curve and Okun’s law.
- Definitions and key equations (as in source):
  - Output gap: y_t = Y_t − Y_t^*
  - Level of potential output dynamics:
    - Y_t^* = Y_{t−1}^* + G_t + Ɛ_t^{Y^*}
    - G_t = θ G_{ss} + (1−θ) G_{t−1} + Ɛ_t^G
    - y_t = Φ y_{t−1} + Ɛ_t^y
  - Phillips curve (hybrid):
    - π_t = λ π_{t+1} + (1−λ) π_{t−1} + β y_t + Ɛ_t^π
  - Okun’s law and NAIRU dynamics:
    - u_t = γ u_{t−1} + δ y_t + Ɛ_t^u
    - u_t = U_t^* − U_t
    - U_t^* = κ U_{ss} + (1−κ) U_{t−1}^* + μ_t + Ɛ_t^{U^*}
    - μ_t = η μ_{t−1} + Ɛ_t^μ
- Estimation details:
  - Observables used: real GDP, inflation, unemployment rate.
  - Annual data: 1990–2019 for Australia; 1995–2019 for New Zealand.
  - Parameters estimated with Bayesian estimation techniques; priors reported in Blagrave et al. (2015).
- Decomposition:
  - Estimated potential growth decomposed into capital, NAIRU, trend labor force participation rate, working-age population, and TFP based on Cobb-Douglas specification.

### Projecting medium-term growth and growth-accounting framework (Appendix 3)
- Growth accounting specification (as in source):
  - ln(Y_t^*) = ln(z_t^*) + α ln(K_t) + (1−α) ln( WAP_t × LFP_t^* × (1−U_t^*) )
  - α set at 0.4 for both countries.
- Medium-term projection method:
  - Projections expressed as deviations from pre-COVID potential output projections (January 2020 WEO vintage as benchmark).
  - Trajectories of each right-hand-side variable are computed as deviations from pre-COVID projections and aggregated via Cobb-Douglas function.
- Capital accumulation dynamics and assumptions:
  - Balanced growth path for capital (see Balanced growth equation above).
  - Capital converges gradually to balanced growth path; capital adjustment costs implied.
  - Capital accumulation equation includes debt overhang effects (DO_t) and uncertainty effects (U_t) alongside persistence parameter γ = 0.66.
- Labor reallocation and productivity effects:
  - Aggregate output growth decomposed into within-sector labor productivity growth, labor reallocation effects (R_t^L), and aggregate labor inputs:
    - Δ ln(Y_t^*) = Δ ln(LP_{t}^*, within) + R_t^L + Δ ln(WAP_t × LFP_t^* × (1−U_t^*))
  - Reallocation term (as in source):
    - R_t^L = Σ_{i=1}^n (LP_{i,t−1} / LP_{t−1}) × (L_{i,t}/L_t − L_{i,t−1}/L_{t−1})
  - Positive R_t^L when employment shifts from low-productivity to high-productivity sectors.
  - Reallocation effects included in total factor productivity in aggregate growth accounting.
- Specific calibration and data notes:
  - Reallocation effects in the 1990s for Australia calculated based on labor reallocation from 1990Q3 to 1993Q2.
  - Reallocation effects after COVID-19 calculated based on labor reallocation from March 14, 2020 to July 25, 2020 (weekly payroll data).
  - For New Zealand, labor reallocation effects in Australia are used due to masking by a large-scale wage subsidy program.

*Source: IMF staff calculations*

### APPENDIX 4: SECTORAL REALLOCATION AND ITS EFFECT ON UNEMPLOYMENT RATE

### APPENDIX 4: SECTORAL REALLOCATION AND ITS EFFECT ON UNEMPLOYMENT RATE

### Sectoral Reallocation Index
- Methodology:
  - Sectoral reallocation index computed using Lilien (1982) methodology.
  - Index formula (as presented):  
    휎휎 푡푡 = ��푆푆 푖푖 �훥훥훥훥훥훥 푥푥 푖푖,푡푡 −훥훥훥훥훥훥 푋푋 푡푡 �2 푖푖 푖푖=1 �1 2  (A4.1)
- Stock market specification:
  - 푆푆 푖푖 = market capitalization share of sector 푖푖.
  - 푥푥 푖푖,푡푡 = sectoral stock return at time 푡푡.
  - 푋푋 푡푡 = total stock market return at time 푡푡.
  - Data source: Financial Times Stock Exchange (FTSE) database.
  - Industry coverage: 25 industries for Australia, 12 industries for New Zealand.
  - Frequency: monthly.
  - Sectoral stock prices used to quantify sectoral dispersion at time 푡푡.
- Labor market specification:
  - 푆푆 푖푖 = share of employment in sector 푖푖.
  - 푥푥 푖푖,푡푡 = sectoral employment at time 푡푡.
  - 푋푋 푡푡 = total employment at time 푡푡.
  - Industry coverage: 19 industries for Australia, 16 industries for New Zealand.
  - Frequency: quarterly.
- Empirical observations reported:
  - For Australia: large sectoral reallocation observed in both stock market and labor markets.
  - For New Zealand: sharp increase in stock-implied sectoral reallocation in the second quarter, while labor market-implied sectoral reallocation remains low; labor-market reallocation masked by a large-scale wage subsidy scheme provided by the government.
- Figures:
  - Figure A4.1 displays sectoral reallocation indices estimated separately from stock markets and labor markets (left panel: FTSE industry-level stock price data; right panel: sector-level employment data).
  - Time series spans shown in figure captions include years from 1986 through 2020 for labor-market index and 1995 through 2020 for stock-market index (per figure axes).

### Effects of Reallocative Shocks
- Model specification:
  - Structural vector autoregression (SVAR) model:  
    푦푦 푡푡 = 훽훽 0 + �훽훽 푘푘 푦푦 푡푡−푘푘 푚푚 푘푘=1 + 푢푢 푡푡  (A4.2)
  - Vector 푦푦 푡푡 includes:
    - change in the unemployment rate, and
    - sectoral reallocation index based on stock market returns (from Equation A4.1).
  - 훽훽 푘푘 = coefficient matrix on kth lag; 훽훽 0 = constant terms; 푢푢 푡푡 = reduced-form error terms.
- Data and estimation details:
  - Australia:
    - Frequency: monthly.
    - Sample period: September 1995-May 2020.
    - Number of lags: 12.
  - New Zealand:
    - Frequency: quarterly (unemployment rate only available quarterly).
    - Sample period: 2000Q4-2020Q1.
    - Number of lags: 6.
- Identification:
  - Sectoral reallocation shocks identified using Cholesky decomposition.
  - Identification restriction: reallocation shocks do not affect the unemployment rate contemporaneously (following Campbell and Kuttner 1996, Tase 2019).
- Reported outputs:
  - Figure 9 of Section IV reports cumulative impulse responses of unemployment rate changes to identified reallocation shocks scaled to the magnitude in the COVID-19 episode.
  - Robustness note: Results are broadly unchanged if the change in terms of trade is included in the VAR.

### Determinants of Firm-level Investment Behavior (Appendix 5 summary relevant to reallocation context)
- Panel regression (Tobin’s Q augmented model) used to analyze effects of debt and uncertainty on firms’ investment:  
  퐼퐼 푖푖,푡푡 퐾퐾 푖푖,푡푡−1 = 훼훼 + 휏휏 푖푖 + 훿훿 푡푡 + 훽훽 푋푋 푖푖,푡푡 + Ɛ i,t  (A5.1)
  - 퐼퐼 푖푖,푡푡 = firm i’s capital expenditure at time t.
  - 퐾퐾 푖푖,푡푡−1 = firm i’s capital stock at previous period.
  - 푋푋 푖푖,푡푡 includes:
    - debt level (debt-to-asset ratio),
    - firm-level uncertainty (firm-level stock volatility),
    - cost of debt (interest rate expenditure-to-debt),
    - liquidity (current-asset-to-current-liability ratio),
    - Tobin’s Q (sum of market value of equity and book value debt divided by book value of asset).
  - Regression includes firm-level and time fixed effects; firm-clustered robust standard errors.
  - All explanatory variables included with a one-year lag.
  - Firm-level annual data from IMF Corporate Vulnerability Unit Database (based on Worldscope); financial sector firms excluded.
- Table A5.1: Key estimated coefficients and sample statistics (dependent variable: Investment-to-Lagged Capital)
  - Australia coefficients (standard errors in parentheses):
    - Cost of Debt (-1): -0.404*** (0.097)
    - Debt (-1): -0.213*** (0.027)
    - Liquidity (-1): 0.009*** (0.002)
    - Uncertainty (-1): -0.037*** (0.012)
    - Tobin's Q (-1): 0.019*** (0.005)
    - R-squared: 0.064
    - Sample Period: 1994-2018
    - Number of Observations: 7,893
  - New Zealand coefficients (standard errors in parentheses):
    - Cost of Debt (-1): -0.124 (0.171)
    - Debt (-1): -0.157*** (0.046)
    - Liquidity (-1): 0.013* (0.007)
    - Uncertainty (-1): -0.069*** (0.034)
    - Tobin's Q (-1): 0.044*** (0.007)
    - R-squared: 0.058
    - Sample Period: 1990-2019
    - Number of Observations: 1,443
  - Note: *** and * indicate statistical significance at 1 and 10 percent levels, respectively. Time fixed effects and firm fixed effects controlled.

### Effects of Debt Overhang
- Projection of increase in debt due to COVID-19:
  - Business survey data from the Australian Bureau of Statistics provide sectoral information on revenue losses (Figure A5.1).
  - Change in debt projected by equation (A5.2):  
    훥훥퐷퐷 푖푖,푡푡 = δ∗ (1−푇푇) ∗ 1 4 ∗ { 푆푆 푖푖,푡푡 −75% ∗ (−.75) ∗ R 푖푖,푡푡−1 + 푆푆 푖푖,푡푡 −50% ∗ (−.625) ∗ R 푖푖,푡푡−1 + 푆푆 푖푖,푡푡 −25% ∗ (−.375) ∗ R 푖푖,푡푡−1 + 푆푆 푖푖,푡푡 −퐿퐿25% ∗ (−.125) ∗ R 푖푖,푡푡−1 }  (A5.2)
  - Parameters and definitions:
    - 훥훥퐷퐷 푖푖,푡푡 = change in debt for firm i at time t.
    - 푇푇 = effective corporate tax (set at 0.3 in the simulation).
    - R i,t−1 = revenue in previous period.
    - 푆푆 푖푖,푡푡 푗푗 = share of firms in the sector that report revenue loss at range j (survey ranges).
    - δ = elasticity of debt to revenue loss, set as 0.8 (based on De Vito and Gomez, 2020).
  - Survey revenue-loss ranges: 0-25 percent, 25-50 percent, 50-75 percent, greater than 75 percent (푆푆 푖푖,푡푡 푗푗 determined by sector).
- Translating debt increase into investment impact:
  - Firm-level debt overhang effect estimated by (A5.3):  
    훥훥(퐼퐼/퐾퐾) 푖푖,푡푡 = 훽훽 푑푑푑푑푑푑 푡푡 훥훥퐷퐷 푖푖,푡푡 퐴퐴 푖푖,푡푡−1  (A5.3)
  - Definitions:
    - 훥훥(퐼퐼/퐾퐾) 푖푖,푡푡 = impact of debt on firm i’s investment-to-capital ratio (change).
    - 훽훽 푑푑푑푑푑푑 푡푡 = regression coefficient on debt level (debt-to-asset ratio) from Equation A5.1.
    - 퐴퐴 푖푖,푡푡−1 = firm i’s total assets at previous period (so 훥훥퐷퐷 푖푖,푡푡 퐴퐴 푖푖,푡푡−1 is estimated change in debt-to-asset ratio).
  - Aggregate impact:
    - Aggregate level of debt impact on capital accumulation obtained using a weighted average of firm-level impacts.
    - For New Zealand: aggregate debt effects assumed similar to Australia due to data availability issues, with adjustments for parameter differences reported in Table A5.1.

### Effects of Uncertainty
- Firm-level uncertainty impact specification:
  - 훥훥(퐼퐼/퐾퐾) 푖푖,푡푡 = 훽훽 휎휎 훥훥휎휎 푡푡  (A5.4)
  - Definitions:
    - 훥훥(퐼퐼/퐾퐾) 푖푖,푡푡 = uncertainty effect on firm i’s investment-to-capital ratio (change).
    - 훽훽 휎휎 = regression coefficient on uncertainty (firm-level stock volatility) from Equation A5.1.
    - 훥훥휎휎 푡푡 = change in uncertainty.
- Aggregate measurement and application:
  - Due to data limitations, aggregate change in uncertainty calculated based on S&P/ASX200 VIX index and applied to firm-level uncertainty.
  - Change in S&P/ASX200 VIX also applied to New Zealand data due to limited data availability.
  - Aggregate level effects obtained by applying Equation A5.4 and aggregating across firms.

*Source: APPENDIX 4 and APPENDIX 5 of the working paper (text and tables as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020272-print-pdf.pdf_