## wpiea2024092-print-pdf

## Source details

**Canonical URL:** [wpiea2024092-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024092-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024092-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024092-print-pdf.pdf.json)

---

### Prior distributions of the estimated parameters
- Beta distribution used for parameters bounded between 0 and 1, including:
  - all autoregressive coefficients,
  - interest rate smoothness ρm and ρ∗m,
  - correlations between identical types of shocks across the two countries,
  - strength of home bias in consumption, 1−γ.
- Prior mean is 0.6 for all autoregressive coefficients.
- Prior mean is 0.3 for cross-country shock correlations.
- Inverse gamma distribution applied for standard deviations of the shocks with all prior means set at 0.01.
- Normal distribution employed for unbounded parameters, with prior means adhering to conventional values in the literature.
- Identical priors used for all G7 countries.

### Estimation setup
- Model estimated with eight exogenous shocks and eight observables (matching VAR specification):
  - Observables: ∆nxt, ∆et, πt, π∗t, ct, c∗t, it, i∗t.
- Posterior distribution obtained using the Metropolis-Hastings algorithm.
- Note: A model cannot be estimated with fewer shocks than observables in Bayesian estimation (stochastic singularity).

### Main estimation results (parameters and posterior summaries)
- γ
  - Prior Mean: 0.07
  - Post. Mean: 0.0086
  - Mode: 0.0088
  - 90% HPD Interval: [0.0067,0.0109]
  - Prior: Beta
  - Prior stdev: 0.02
- κ
  - Prior Mean: 8.7
  - Post. Mean: 7.59
  - Mode: 7.517
  - 90% HPD Interval: [6.6445,8.4225]
  - Prior: Normal
  - Prior stdev: 0.5
- φπ (Home)
  - Prior Mean: 1.5
  - Post. Mean: 1.3485
  - Mode: 1.3653
  - 90% HPD Interval: [1.2063,1.5281]
  - Prior: Normal
  - Prior stdev: 0.1
- φ∗π (Foreign)
  - Prior Mean: 1.5
  - Post. Mean: 1.5402
  - Mode: 1.5422
  - 90% HPD Interval: [1.3788,1.6966]
  - Prior: Normal
  - Prior stdev: 0.1
- φy (Home)
  - Prior Mean: 0.5
  - Post. Mean: 0.4545
  - Mode: 0.4526
  - 90% HPD Interval: [0.3649,0.5364]
  - Prior: Normal
  - Prior stdev: 0.05
- φ∗y (Foreign)
  - Prior Mean: 0.5
  - Post. Mean: 0.4798
  - Mode: 0.4774
  - 90% HPD Interval: [0.3990,0.5556]
  - Prior: Normal
  - Prior stdev: 0.05
- χ2
  - Prior Mean: 0.001
  - Post. Mean: 0.0014
  - Mode: 0.0016
  - 90% HPD Interval: [0.0002,0.0028]
  - Prior: Normal
  - Prior stdev: 0.001
- Persistence (selected ρ parameters)
  - ρa: Prior Mean: 0.6; Post. Mean: 0.6757; Mode: 0.6662; 90% HPD Interval: [0.5517,0.7830]; Prior: Beta; Prior stdev: 0.1
  - ρ∗a: Prior Mean: 0.6; Post. Mean: 0.6817; Mode: 0.6804; 90% HPD Interval: [0.6047,0.7502]; Prior: Beta; Prior stdev: 0.1
  - ρψ: Prior Mean: 0.6; Post. Mean: 0.7146; Mode: 0.708; 90% HPD Interval: [0.6504,0.7743]; Prior: Beta; Prior stdev: 0.1
  - ρm: Prior Mean: 0.6; Post. Mean: 0.7606; Mode: 0.75; 90% HPD Interval: [0.7007,0.7994]; Prior: Beta; Prior stdev: 0.1
  - ρ∗m: Prior Mean: 0.6; Post. Mean: 0.7912; Mode: 0.7837; 90% HPD Interval: [0.7407,0.8275]; Prior: Beta; Prior stdev: 0.1
  - ρv: Prior Mean: 0.6; Post. Mean: 0.1576; Mode: 0.1733; 90% HPD Interval: [0.1055,0.2450]; Prior: Beta; Prior stdev: 0.1
  - ρ∗v: Prior Mean: 0.6; Post. Mean: 0.2171; Mode: 0.232; 90% HPD Interval: [0.1513,0.3135]; Prior: Beta; Prior stdev: 0.1
  - ρΩ: Prior Mean: 0.6; Post. Mean: 0.7042; Mode: 0.7032; 90% HPD Interval: [0.6311,0.7665]; Prior: Beta; Prior stdev: 0.1
  - ρ∗Ω: Prior Mean: 0.6; Post. Mean: 0.7084; Mode: 0.7008; 90% HPD Interval: [0.6495,0.7569]; Prior: Beta; Prior stdev: 0.1
  - ργ∗: Prior Mean: 0.6; Post. Mean: 0.8012; Mode: 0.8045; 90% HPD Interval: [0.7431,0.8720]; Prior: Beta; Prior stdev: 0.1
- Shock standard deviations (selected σ parameters; prior Mean: 0.01 for all)
  - σa: Post. Mean: 0.0191; Mode: 0.0194; 90% HPD Interval: [0.0154,0.0228]; Prior: Inverse Gamma; Prior stdev: Inf
  - σ∗a: Post. Mean: 0.0129; Mode: 0.013; 90% HPD Interval: [0.0112,0.0147]; Prior: Inverse Gamma; Prior stdev: Inf
  - σψ: Post. Mean: 0.0127; Mode: 0.013; 90% HPD Interval: [0.0101,0.0157]; Prior: Inverse Gamma; Prior stdev: Inf
  - σv: Post. Mean: 0.0054; Mode: 0.0056; 90% HPD Interval: [0.0047,0.0065]; Prior: Inverse Gamma; Prior stdev: Inf
  - σ∗v: Post. Mean: 0.0033; Mode: 0.0034; 90% HPD Interval: [0.0029,0.0038]; Prior: Inverse Gamma; Prior stdev: Inf
  - σΩ: Post. Mean: 0.020; Mode: 0.0201; 90% HPD Interval: [0.0183,0.0220]; Prior: Inverse Gamma; Prior stdev: Inf
  - σ∗Ω: Post. Mean: 0.0166; Mode: 0.0167; 90% HPD Interval: [0.0151,0.0181]; Prior: Inverse Gamma; Prior stdev: Inf
  - σγ∗: Post. Mean: 0.0016; Mode: 0.0016; 90% HPD Interval: [0.0015,0.0017]; Prior: Inverse Gamma; Prior stdev: Inf
- Cross-country correlations (prior Mean: 0.3)
  - ρa,a∗: Post. Mean: 0.3641; Mode: 0.3619; 90% HPD Interval: [0.2700,0.4641]; Prior: Beta; Prior stdev: 0.1
  - ρv,v∗: Post. Mean: 0.2267; Mode: 0.2333; 90% HPD Interval: [0.1341,0.3253]; Prior: Beta; Prior stdev: 0.1
  - ρΩ,Ω∗: Post. Mean: 0.3387; Mode: 0.338; 90% HPD Interval: [0.2475,0.4350]; Prior: Beta; Prior stdev: 0.1

### Key interpretative findings
- γ (trade openness-related parameter)
  - Posterior mean: 0.0086 (prior mean 0.07); 90% HPD interval: [0.0067,0.0109].
- Taylor rule coefficients (posterior means)
  - Home: φπ = 1.3485; φy = 0.4545.
  - Foreign: φ∗π = 1.5402; φ∗y = 0.4798.
- Capital adjustment cost κ: post. mean 7.59 (prior mean 8.7 lies outside its 90% HPD interval).
- Interest rate smoothing: ρm ≈ 0.7606 and ρ∗m ≈ 0.7912 (both around 0.8).
- ξ2 (χ2) aligns with prior means and falls within its 90% HPD interval.

### Persistence and shock characteristics
- Estimated persistence levels:
  - Capital flow shock: around 0.70 (ρΩ and related parameters indicate high persistence; ρΩ post. mean 0.7042).
  - Aggregated demand shock: around 0.70 (ρa and ρ∗a post. means 0.6757 and 0.6817).
  - Relative demand shock: around 0.8 (ργ∗ post. mean 0.8012).
  - Non-tradable goods persistence reported for comparison: 0.63.
  - Tradable goods persistence reported in Stockman and Tesar(1995): 0.15.
- Volatility (log deviation from steady state):
  - Relative demand shock volatility: around 19% (reported as "around 19%").
- Inter-country correlations for identical shock types:
  - Range: 0.3 to 0.4; consistent with calibrations used in Itskhoki and Mukhin(2021).
- Monetary policy shocks:
  - Display minimal persistence domestically and abroad, characterized by AR(1) coefficients near 0.2 (ρv post. mean 0.1576; ρ∗v post. mean 0.2171).

### Model fit (selected moments and comparisons)
- Overall: the model’s unconditional moments are quite close to actual data overall.
- Specific comparisons (selected exact model vs data moments from Table 3):
  - US std(∆nx_t): model 0.0034, data 0.0031
  - US std(∆e_t): model 0.0483, data 0.0411
  - US std(c_t): model 0.0125, data 0.011
  - US std(i_t): model 0.0072, data 0.0037
  - US std(π_t): model 0.0051, data 0.0048
  - US std(z_t): model 0.0399, data 0.0353
  - Correlation ρ(∆nx_t, ∆e_t) for US: model 0.1716, data 0.155
- Country-level fit notes:
  - Current account volatility: accurately captured for all countries examined.
  - Domestic consumption: model diverges slightly; high accuracy for US, UK, Germany; lower accuracy for France, Canada, Japan.
  - Foreign consumption: model-implied volatility very close to empirical moment for US; substantial discrepancies for other countries.
  - Investment volatility: model overestimates across the G7.
  - Exchange rate volatility: close for US; noticeable deviations for UK, France, Italy.
  - CPI inflation: domestic inflation matched well; slightly larger discrepancies for foreign inflation.
  - Interest rate volatility: model forecasts higher volatility for both domestic and foreign interest rates, most pronounced for Italy.
  - Correlation between current account balances and exchange rate changes: model substantially reduces the traditionally strong linkage by incorporating relative demand shocks; some country-specific underprediction (UK, Italy, Canada) and some overestimation for other countries.

### Which shocks matter (FEVD and IRF evidence)
- FEVD for US ∆nx_t:
  - Relative demand shock: more than 80% of US current account variations across horizons.
  - Capital flow shock: second largest, over 10 percent.
  - All other shocks: small remaining portion.
- FEVD for US ∆e_t:
  - Capital flow shocks: dominant driver across all horizons.
  - Relative demand and domestic and foreign monetary policy shocks together: approximately 10% of variance in ∆e_t.
- FEVD patterns across G7:
  - Relative demand shock accounts for the largest share of current account fluctuations by far; capital flow shock explains bulk of exchange rate fluctuations.
- IRF evidence:
  - Relative demand and monetary policy shocks are the only structural shocks showing a negative correlation between exchange rate and current account consistent with the empirical dominant CA shock at business-cycle frequency.
  - Monetary policy shocks produce much shorter-lived effects on the current account (~4 quarters) than the relative demand shock (~20 quarters for dominant CA shock from the SVAR).
  - Relative demand shock dynamics (qualitative):
    - Increases demand for domestic goods relative to foreign goods → improves domestic current account balance and appreciates the real exchange rate.
    - Home: pushes up domestic inflation and nominal interest rate; home consumption and investment decrease.
    - Foreign: PPI disinflation (π*_Ft < 0); import price inflation picks up (π*_Ht > 0) but overall CPI inflation π*_t decreases due to larger share of foreign goods in consumption basket; lower CPI inflation and lower detrended output (y*_t < 0) lead foreign central bank to lower interest rates and raise foreign investment and consumption.

### Comparing historical shocks (regression linking DSGE shocks to empirical dominant CA shock)
- Regression: dominant CA shock_t = Σ_i β_i * structural DSGE shock_i,t + u_t (estimated separately for each country).
- Two structural shocks identified as critical across G7:
  - Capital flow shock: statistically significant across all G7 countries.
  - Relative demand shock: largest and highly significant across all G7 countries.
- Selected regression coefficients (exact values from Table 4):
  - Capital Flow coefficients:
    - US 0.278*** (0.063)
    - UK 0.213*** (0.053)
    - DE 0.191*** (0.071)
    - FR 0.136* (0.077)
    - IT 0.296*** (0.095)
    - CA 0.246*** (0.066)
    - JP 0.238*** (0.063)
  - Relative Demand coefficients:
    - US 4.818*** (0.479)
    - UK 1.394*** (0.101)
    - DE 2.021*** (0.182)
    - FR 2.507*** (0.199)
    - IT 1.877*** (0.230)
    - CA 3.277*** (0.289)
    - JP 2.183*** (0.156)
  - R^2 values by country:
    - US 0.726
    - UK 0.874
    - DE 0.783
    - FR 0.803
    - IT 0.712
    - CA 0.791
    - JP 0.843
- Additional heterogeneity:
  - TFP shock integral to dominant CA shocks in UK, France, Italy, Canada.
  - Foreign TFP shock integral in US, UK, Italy, Canada.
  - Aggregate demand and monetary policy shocks show country-specific importance.

### Max-share SVAR on model-simulated data (role of relative demand shock)
- Procedure:
  - Apply max-share SVAR to model-simulated data (1,000 periods) generated under different combinations of structural shocks.
- Findings when all shocks allowed:
  - Simulated dominant CA driver:
    - Current account increases on impact; domestic consumption and investment decrease.
    - Foreign consumption exhibits a short-term increase.
    - TFP trajectory: initially declining then rising (insignificant), aligning with empirical counterpart.
    - Nominal exchange rate initially depreciates over three quarters (transitory), then shows persistent appreciation; initial depreciation differs from empirical findings and is mainly due to a negative capital flow shock that is highly correlated with the dominant empirical CA shock.
- Findings when only the relative demand shock allowed:
  - Simulated dominant CA shock IRFs are qualitatively similar to the full-shock model and the empirical dominant CA shock; nominal exchange rate appreciates immediately after the shock.
- Findings when relative demand shock excluded:
  - Simulated dominant CA shock displays significant short-run expenditure switching, contradicting the data and the full-shock model; consumption and investment display irregular dynamics not observed empirically.

### Discussion and implications
- Core insight:
  - Relative demand shocks play an important role in accounting for current account fluctuations at business-cycle frequency for the US and other G7 countries.
  - The dominant CA shock is characterized by an increase in the current account balance and an exchange rate appreciation, implying a preference shift from foreign to domestic goods that can offset expenditure-switching effects from other shocks (e.g., capital flow shocks).
- Comparison with conventional shocks:
  - Conventional aggregate shocks operate through expenditure switching and thus do not generate the observed comovement between current account and exchange rate that characterizes the dominant CA shock.
  - The importance of relative demand shocks echoes Stockman and Tesar(1995) that taste shocks are needed to generate data-consistent comovements in open-economy RBC models.
- Long-run considerations:
  - The role of relative demand shocks seems to diminish in the long run with heterogeneity across countries.
  - For the US, the long-run dominant CA shock yields an increase in current account balance and exchange rate depreciation, suggesting expenditure switching resurfaces in the long run (possible causes: low persistence of relative demand shock, lagged supply response, or both).

### Conclusion and avenues for future research
- Summary conclusion:
  - The paper documents dominant CA shocks at business cycle frequency and over the long run using max-share identification and finds that relative demand shocks are pivotal in driving the empirically observed dominant CA shock across G7 economies.
  - Dominant CA shocks often coincide with exchange rate appreciation or stability and near- to medium-term reductions in consumption and investment, with cross-country heterogeneity.
- Suggested future research directions:
  - Strengthen the model’s data-matching ability and explanatory power by adding additional structures (e.g., consumption habits and non-tradables) or by deconstructing relative demand shocks into more primitive shocks.
  - Develop models better suited for long-run analyses to interpret the dominant long-run CA shock.
  - Examine emerging markets (commodity exporters, countries with active foreign exchange interventions) where different factors might underlie the dominant CA shock.

*wpiea2024092-print-pdf*

### 0.75    Conventional value

### 0.75    Conventional value

### Prior distributions of the estimated parameters
- Beta distribution used for parameters bounded between 0 and 1, including:
  - all autoregressive coefficients,
  - interest rate smoothness ρm and ρ∗m,
  - correlations between identical types of shocks across the two countries,
  - strength of home bias in consumption, 1−γ.
- Prior mean is 0.6 for all autoregressive coefficients.
- Prior mean is 0.3 for cross-country shock correlations.
- Inverse gamma distribution applied for standard deviations of the shocks with all prior means set at 0.01.
- Normal distribution employed for unbounded parameters, with prior means adhering to conventional values in the literature.
- Identical priors used for all G7 countries.

### Estimation setup
- Model estimated with eight exogenous shocks and eight observables (matching VAR specification):
  - Observables: ∆nxt (current account), ∆et (nominal exchange rate), πt (domestic CPI inflation), π∗t (foreign CPI inflation), ct (domestic consumption, log-deviation), c∗t (foreign consumption, log-deviation), it (domestic nominal interest rate), i∗t (foreign nominal interest rate).
- Posterior distribution obtained using the Metropolis-Hastings algorithm.
- Note: A model cannot be estimated with fewer shocks than observables in Bayesian estimation (stochastic singularity).

### Main estimation results (Table 2 parameters and posterior summaries)
- γ
  - Prior Mean: 0.07
  - Post. Mean: 0.0086
  - Mode: 0.0088
  - 90% HPD Interval: [0.0067,0.0109]
  - Prior: Beta
  - Prior stdev: 0.02
- κ
  - Prior Mean: 8.7
  - Post. Mean: 7.59
  - Mode: 7.517
  - 90% HPD Interval: [6.6445,8.4225]
  - Prior: Normal
  - Prior stdev: 0.5
- φπ (Home)
  - Prior Mean: 1.5
  - Post. Mean: 1.3485
  - Mode: 1.3653
  - 90% HPD Interval: [1.2063,1.5281]
  - Prior: Normal
  - Prior stdev: 0.1
- φ∗π (Foreign)
  - Prior Mean: 1.5
  - Post. Mean: 1.5402
  - Mode: 1.5422
  - 90% HPD Interval: [1.3788,1.6966]
  - Prior: Normal
  - Prior stdev: 0.1
- φy (Home)
  - Prior Mean: 0.5
  - Post. Mean: 0.4545
  - Mode: 0.4526
  - 90% HPD Interval: [0.3649,0.5364]
  - Prior: Normal
  - Prior stdev: 0.05
- φ∗y (Foreign)
  - Prior Mean: 0.5
  - Post. Mean: 0.4798
  - Mode: 0.4774
  - 90% HPD Interval: [0.3990,0.5556]
  - Prior: Normal
  - Prior stdev: 0.05
- χ2
  - Prior Mean: 0.001
  - Post. Mean: 0.0014
  - Mode: 0.0016
  - 90% HPD Interval: [0.0002,0.0028]
  - Prior: Normal
  - Prior stdev: 0.001
- ρa
  - Prior Mean: 0.6
  - Post. Mean: 0.6757
  - Mode: 0.6662
  - 90% HPD Interval: [0.5517,0.7830]
  - Prior: Beta
  - Prior stdev: 0.1
- ρ∗a
  - Prior Mean: 0.6
  - Post. Mean: 0.6817
  - Mode: 0.6804
  - 90% HPD Interval: [0.6047,0.7502]
  - Prior: Beta
  - Prior stdev: 0.1
- ρψ
  - Prior Mean: 0.6
  - Post. Mean: 0.7146
  - Mode: 0.708
  - 90% HPD Interval: [0.6504,0.7743]
  - Prior: Beta
  - Prior stdev: 0.1
- ρm
  - Prior Mean: 0.6
  - Post. Mean: 0.7606
  - Mode: 0.75
  - 90% HPD Interval: [0.7007,0.7994]
  - Prior: Beta
  - Prior stdev: 0.1
- ρ∗m
  - Prior Mean: 0.6
  - Post. Mean: 0.7912
  - Mode: 0.7837
  - 90% HPD Interval: [0.7407,0.8275]
  - Prior: Beta
  - Prior stdev: 0.1
- ρv
  - Prior Mean: 0.6
  - Post. Mean: 0.1576
  - Mode: 0.1733
  - 90% HPD Interval: [0.1055,0.2450]
  - Prior: Beta
  - Prior stdev: 0.1
- ρ∗v
  - Prior Mean: 0.6
  - Post. Mean: 0.2171
  - Mode: 0.232
  - 90% HPD Interval: [0.1513,0.3135]
  - Prior: Beta
  - Prior stdev: 0.1
- ρΩ
  - Prior Mean: 0.6
  - Post. Mean: 0.7042
  - Mode: 0.7032
  - 90% HPD Interval: [0.6311,0.7665]
  - Prior: Beta
  - Prior stdev: 0.1
- ρ∗Ω
  - Prior Mean: 0.6
  - Post. Mean: 0.7084
  - Mode: 0.7008
  - 90% HPD Interval: [0.6495,0.7569]
  - Prior: Beta
  - Prior stdev: 0.1
- ργ∗
  - Prior Mean: 0.6
  - Post. Mean: 0.8012
  - Mode: 0.8045
  - 90% HPD Interval: [0.7431,0.8720]
  - Prior: Beta
  - Prior stdev: 0.1
- σa
  - Prior Mean: 0.01
  - Post. Mean: 0.0191
  - Mode: 0.0194
  - 90% HPD Interval: [0.0154,0.0228]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σ∗a
  - Prior Mean: 0.01
  - Post. Mean: 0.0129
  - Mode: 0.013
  - 90% HPD Interval: [0.0112,0.0147]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σψ
  - Prior Mean: 0.01
  - Post. Mean: 0.0127
  - Mode: 0.013
  - 90% HPD Interval: [0.0101,0.0157]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σv
  - Prior Mean: 0.01
  - Post. Mean: 0.0054
  - Mode: 0.0056
  - 90% HPD Interval: [0.0047,0.0065]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σ∗v
  - Prior Mean: 0.01
  - Post. Mean: 0.0033
  - Mode: 0.0034
  - 90% HPD Interval: [0.0029,0.0038]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σΩ
  - Prior Mean: 0.01
  - Post. Mean: 0.020
  - Mode: 0.0201
  - 90% HPD Interval: [0.0183,0.0220]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σ∗Ω
  - Prior Mean: 0.01
  - Post. Mean: 0.0166
  - Mode: 0.0167
  - 90% HPD Interval: [0.0151,0.0181]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- σγ∗
  - Prior Mean: 0.01
  - Post. Mean: 0.0016
  - Mode: 0.0016
  - 90% HPD Interval: [0.0015,0.0017]
  - Prior: Inverse Gamma
  - Prior stdev: Inf
- ρa,a∗ (cross-country correlation)
  - Prior Mean: 0.3
  - Post. Mean: 0.3641
  - Mode: 0.3619
  - 90% HPD Interval: [0.2700,0.4641]
  - Prior: Beta
  - Prior stdev: 0.1
- ρv,v∗ (cross-country correlation)
  - Prior Mean: 0.3
  - Post. Mean: 0.2267
  - Mode: 0.2333
  - 90% HPD Interval: [0.1341,0.3253]
  - Prior: Beta
  - Prior stdev: 0.1
- ρΩ,Ω∗ (cross-country correlation)
  - Prior Mean: 0.3
  - Post. Mean: 0.3387
  - Mode: 0.338
  - 90% HPD Interval: [0.2475,0.4350]
  - Prior: Beta
  - Prior stdev: 0.1

### Key interpretative findings
- The posterior mean for the trade openness-related parameter, γ, is approximately 0.009, significantly below its prior mean of 0.07.
  - The 90% HPD interval for γ is [0.0067,0.0109], indicating a narrow posterior.
- Taylor rule coefficients (Home): posterior means
  - φπ ≈ 1.35 (prior mean 1.5)
  - φy ≈ 0.45 (prior mean 0.5)
- Taylor rule coefficients (Foreign): posterior means
  - φ∗π ≈ 1.54 (prior mean 1.5)
  - φ∗y ≈ 0.48 (prior mean 0.5)
- Capital adjustment cost coefficient κ estimated at approximately 7.6 (prior mean 8.7 is outside its 90% HPD interval).
- Interest rate smoothing parameters ρm and ρ∗m are both around 0.8.
- ξ2 estimate aligns with its prior means and falls within its 90% HPD intervals.
- Shocks persistence:
  - Monetary policy shocks display minimal persistence domestically and abroad, characterized by AR(1) coefficients near 0.2.
  - Other types of shocks exhibit notably higher persistence.
  - Persistence of TFP shocks is identified at approximately (text cut off in source).

*wpiea2024092-print-pdf*

### 0.68 for each country. This value is close to the persistence found in non-tradable goods

### 0.68 for each country. This value is close to the persistence found in non-tradable goods (0.63) and markedly surpasses that of tradable goods (0.15) in Stockman and Tesar(1995).

### Persistence and shock characteristics
- Estimated persistence levels:
  - Capital flow shock: around 0.70
  - Aggregated demand shock: around 0.70
  - Relative demand shock: around 0.8
  - Non-tradable goods persistence reported: 0.63 (for comparison)
  - Tradable goods persistence reported in Stockman and Tesar(1995): 0.15
- Volatility (measured as log deviation from steady state):
  - Relative demand shock volatility: around 19%
- Inter-country correlations for identical shock types:
  - Range: 0.3 to 0.4
  - Consistent with calibrations used in Itskhoki and Mukhin(2021)

### Model fit (summary of Table 3 and text)
- The model’s unconditional moments are quite close to actual data overall.
- Specific model-data comparisons (high-level):
  - Current account volatility: accurately captured for all countries examined.
  - Domestic consumption: model diverges slightly from data; accuracy varies across countries.
    - High accuracy: US, UK, Germany
    - Lower accuracy: France, Canada, Japan
  - Foreign consumption: model-implied volatility very close to empirical moment for US; substantial discrepancies for other countries.
  - Investment volatility: model overestimates across the G7 (investment dynamics not a matching target).
  - Exchange rate volatility: close for US; noticeable deviations for UK, France, Italy.
  - CPI inflation: domestic inflation matched well; slightly larger discrepancies for foreign inflation.
  - Interest rate volatility: model forecasts higher volatility for both domestic and foreign interest rates, with most pronounced overprediction for Italy.
  - Correlation between current account balances and exchange rate changes:
    - Model substantially reduces the traditionally strong linkage by incorporating relative demand shocks.
    - Some country-specific underprediction (UK, Italy, Canada) and some overestimation for other countries.
    - Improvement relative to prior literature that finds correlation over 0.95 when all shocks induce expenditure-switching.

- Select exact model vs data moments from Table 3 (examples preserved exactly):
  - US std(∆nx_t): model 0.0034, data 0.0031
  - US std(∆e_t): model 0.0483, data 0.0411
  - US std(c_t): model 0.0125, data 0.011
  - US std(i_t): model 0.0072, data 0.0037
  - US std(π_t): model 0.0051, data 0.0048
  - US std(z_t): model 0.0399, data 0.0353
  - Correlation ρ(∆nx_t, ∆e_t) for US: model 0.1716, data 0.155
  - (Table 3 provides analogous model and data pairs for UK, DE, FR, IT, CA, JP across listed variables.)

### Which shocks matter (FEVD and IRF evidence)
- Forecast error variance decomposition (FEVD) for US ∆nx_t:
  - Relative demand shock: more than 80% of US current account variations across horizons.
  - Capital flow shock: second largest, over 10 percent.
  - All other shocks: small remaining portion.
- FEVD for US nominal exchange rate ∆e_t:
  - Capital flow shocks: dominant driver across all horizons.
  - Relative demand and domestic and foreign monetary policy shocks together: approximately 10% of variance in ∆e_t.
- FEVD patterns are similar across other G7 countries: relative demand shock accounts for the largest share of current account fluctuations by far; capital flow shock explains bulk of exchange rate fluctuations.
- Impulse response function (IRF) evidence:
  - Relative demand and monetary policy shocks are the only structural shocks showing a negative correlation between exchange rate and current account consistent with the empirical dominant CA shock at business-cycle frequency.
  - Monetary policy shocks produce much shorter-lived effects on the current account than the relative demand shock.
  - Relative demand shock dynamics:
    - Increases demand for domestic goods relative to foreign goods → improves domestic current account balance and appreciates the real exchange rate.
    - Expansionary effect for home country: pushes up domestic inflation and nominal interest rate, generating higher inflation and interest rate differentials in the short to medium term.
    - Home consumption and investment decrease.
    - Foreign country: PPI disinflation (π*_Ft < 0); import price inflation picks up (π*_Ht > 0) but overall CPI inflation π*_t decreases due to larger share of foreign goods in consumption basket; lower CPI inflation and lower detrended output (y*_t < 0) lead foreign central bank to lower interest rates and raise foreign investment and consumption.
  - Monetary policy shocks affect the current account for around four quarters, much shorter than around 20 quarters for the dominant CA shock estimated from the SVAR.

### Comparing historical shocks (regression linking DSGE shocks to empirical dominant CA shock)
- Regression specification (estimated separately for each country):
  - dominant CA shock_t = Σ_i β_i * structural DSGE shock_i,t + u_t
- Two structural shocks identified as critical contributors across G7:
  - Capital flow shock: statistically significant coefficients across all G7 countries.
  - Relative demand shock: largest and highly significant coefficients across all G7 countries.
- Selected regression coefficients (preserved exactly from Table 4):
  - Capital Flow coefficients:
    - US 0.278*** (0.063)
    - UK 0.213*** (0.053)
    - DE 0.191*** (0.071)
    - FR 0.136* (0.077)
    - IT 0.296*** (0.095)
    - CA 0.246*** (0.066)
    - JP 0.238*** (0.063)
  - Relative Demand coefficients:
    - US 4.818*** (0.479)
    - UK 1.394*** (0.101)
    - DE 2.021*** (0.182)
    - FR 2.507*** (0.199)
    - IT 1.877*** (0.230)
    - CA 3.277*** (0.289)
    - JP 2.183*** (0.156)
  - R^2 values by country:
    - US 0.726
    - UK 0.874
    - DE 0.783
    - FR 0.803
    - IT 0.712
    - CA 0.791
    - JP 0.843
- Additional heterogeneity:
  - TFP shock is integral to dominant CA shocks in UK, France, Italy, Canada.
  - Foreign TFP shock plays an integral role in US, UK, Italy, Canada.
  - Aggregate demand and monetary policy shocks show country-specific importance.

### Max-share SVAR on model-simulated data (role of relative demand shock)
- Procedure:
  - Apply max-share SVAR to model-simulated data (1,000 periods) generated under different combinations of structural shocks.
  - Compare model-based dominant CA shocks with empirical dominant CA shock.
- Findings from simulated data allowing all shocks:
  - Simulated dominant CA driver:
    - Current account increases on impact; domestic consumption and investment decrease.
    - Foreign consumption exhibits a short-term increase.
    - TFP trajectory: initially declining then rising (insignificant), aligning with empirical counterpart.
    - Nominal exchange rate initially depreciates over three quarters (transitory), then shows persistent appreciation; initial depreciation differs from empirical findings.
      - The transitory depreciation is mainly due to a negative capital flow shock, which is highly correlated with the dominant empirical CA shock.
- Findings from simulated data allowing only the relative demand shock:
  - The max-share SVAR on these simulations yields a simulated dominant CA shock whose IRFs are qualitatively similar to the full-shock model and the empirical dominant CA shock.
  - Nominal exchange rate appreciates immediately after the shock in this scenario.
- Findings from simulated data excluding relative demand shock:
  - Simulated dominant CA shock displays significant short-run expenditure switching, contradicting the data and the full-shock model.
  - Consumption and investment display irregular dynamics not observed in empirical dominant CA shock or simulated shocks with all shocks or only the relative demand shock.

### Discussion
- Core insight:
  - Relative demand shocks play an important role in accounting for current account fluctuations at business cycle frequency for the US and other G7 countries.
  - The dominant CA shock is characterized by an increase in the current account balance and an exchange rate appreciation, implying a preference shift from foreign to domestic goods that can offset expenditure switching effects from other shocks (e.g., capital flow shocks).
- Implications vis-à-vis conventional shocks:
  - Conventional aggregate shocks (demand or supply) operate through expenditure switching and thus do not generate the observed comovement between current account and exchange rate that characterizes the dominant CA shock.
  - The importance of relative demand shocks echoes Stockman and Tesar(1995) argument that taste shocks are needed to generate data-consistent comovements in open-economy RBC models.
- Long-term considerations:
  - The role of relative demand shocks seems to diminish in the long run with heterogeneity across countries.
  - For the US, the long-run dominant CA shock yields an increase in current account balance and exchange rate depreciation, suggesting the expenditure switching effect resurfaces in the long run.
  - Possible reasons: low persistence of relative demand shock, lagged supply response that enables expenditure switching to resurface, or both.

### Conclusion and avenues for future research
- Summary conclusion:
  - The paper documents dominant CA shocks at business cycle frequency and over the long run using max-share identification and finds that relative demand shocks are pivotal in driving the empirically observed dominant CA shock across G7 economies.
  - Dominant CA shocks often coincide with exchange rate appreciation or stability and near- to medium-term reductions in consumption and investment, with cross-country heterogeneity.
- Future research directions suggested:
  - Strengthen the model’s data-matching ability and explanatory power by adding additional structures (e.g., consumption habits and non-tradables) or by deconstructing relative demand shocks into more primitive shocks.
  - Develop models better suited for long-run analyses to interpret the dominant long-run CA shock.
  - Examine emerging markets (commodity exporters, countries with active foreign exchange interventions) where different factors might underlie the dominant CA shock.

*Content based on the provided IMF PDF excerpt.*

### References

### References

### Theoretical frameworks on current account, intertemporal models, and trade
- Alexander, S. S. (1952).  Effects of a devaluation on a trade balance.Staff  Papers-International Monetary Fund, 2(2):263–278.
- Buiter, W. H. (1981).  Time preference and international lending and borrowing in an overlapping-generations model.Journal of political economy, 89(4):769–797.
- Hahn, F. H. (1959). The balance of payments in a monetary economy.The Review of Economic Studies, 26(2):110–125.
- Obstfeld, M. and Rogoff, K. (1995). The intertemporal approach to the current account. Handbook of international economics, 3:1731–1799.
- Sachs, J. D., Cooper, R. N., and Fischer, S. (1981). The current account and macroeconomic adjustment in the 1970s.Brookings papers on economic activity, 1981(1):201–282.
- Sheffrin, S. and Woo, W. (1990). Present value tests of an intertemporal model of the current account.Journal of International Economics, 29(3-4):237–253.
- Obstfeld, M. (2012). Does the current account still matter?American Economic Review: Papers & Proceedings, 102(3):1–23.
- Chinn, M. D. and Prasad, E. E. (2003). Medium-term determinants of current accounts in industrial and developing countries: an empirical exploration.Journal of International Economics, 59:47–76.
- Ghosh, A. R. and Ostry, J. D. (1995). The current account in developing countries: A perspective from the consumption-smoothing approach.The  World  Bank  Economic Review, 9(2):305–333.

### Exchange rates, risk premia, and asset prices
- Bergin, P. and Sheffrin, S. (2000). Interest rates, exchange rates and present value models of the current account.Economic Journal, 110(463):535–58.
- Bergin, P. R. (2006). How well can the new open economy macroeconomics explain the exchange rate and current account?Journal  of  International  Money  and  Finance, 25:675–701.
- Gabaix, X. and Maggiori, M. (2015). International liquidity and exchange rate dynamics. The Quarterly Journal of Economics, 130(3):1369–1420.
- Engel, C. (2016). Exchange rates, interest rates, and the risk premium.American Economic Review, 106(2):436–74.
- Engel, C. and Rogers, J. H. (2006). The U.S. current account deficit and the expected share of world output.Journal of Monetary Economics, 53(5):1063–1093.
- Pavlova, A. and Rigobon, R. (2007). Asset prices and exchange rates.The Review of Financial Studies, 20(4):1139–1180.
- Chahrour, R., Cormun, V., Leo, P. D., Guerron-Quintana, P., and Valchev, R. (2021). Exchange rate disconnect revisited. Boston College Working Papers in Economics 1041, Boston College Department of Economics.
- Its h kok i, O. and Mukhin, D. (2021). Exchange rate disconnect in general equilibrium. Journal of Political Economy, 129(8):000–000.
- Chinn, M. D. and Ito, H. (2022).  A requiem for blame it on beijing interpreting rotating global current account surpluses.Journal of International Money and Finance, 121:102510. Current Account (Im-)Balances: Determinants and Policy Implications.

### Macroeconomic models, DSGE, VARs, and identification
- Angeletos, G.-M., Collard, F., and Dellas, H. (2020). Business-cycle anatomy.American Economic Review, 110(10):3030–70.
- Smets, F. and Wouters, R. (2007). Shocks and frictions in us business cycles: A bayesian dsge approach.American economic review, 97(3):586–606.
- Schmitt-Groh ́e, S. and Uribe, M. (2012). What’s news in business cycles.Econometrica, 80(6):2733–2764.
- Rabanal, P. and Rubio-Ram ́ırez, J. F. (2005). Comparing new keynesian models of the business cycle: A bayesian approach.Journal of Monetary Economics, 52(6):1151–1166.
- Giannone, D., Lenza, M., and Primiceri, G. (2015). Prior selection for vector autoregressions.The Review of Economics and Statistics, 97(2):436–451.
- Lenza, M. and Primiceri, G. E. (2022).  How to estimate a vector autoregression after march 2020.Journal of Applied Econometrics, 37(4):688–699.
- Stock, J. H. S. and Wastson, M. W. (2018). Identification and estimation of dynamic causal effects in macroeconomics using external instruements.Economic Journal, 128:917–948.
- Faust, J. (1998). The robustness of identified var conclusions about money.Carnegie-Rochester Conference Series on Public Policy, 49:207–244.
- Uhlig, H. (2003). What moves gnp? Technical report, Unpublished.
- Uhlig, H. (2005). What are the effects of monetary policy on output? Results from an agnostic identification procedure.Journal of Monetary Economics, 52(2):381–419.
- Pfeifer, J. (2014). A guide to specifying observation equations for the estimation of dsge models.Research series, pages 1–150.

### Productivity, demographics, and capital flows
- Ferrero, A. (2010). A structural decomposition of the u.s. trade balance: Productivity, demographics and fiscal policy.Journal of Monetary Economics, 57:478–490.
- Barany, Z., Coueurdacier, N., and Guibaud, S. (2023). Captial flows in an aging world. Journal of International Economics, 140:1037.
- Backus, D., Cooley, T., and Henriksen, E. (2014). Demography and low-frequency capital flows.Journal of International Economics, 92:S94–S102.
- Caballero, R. J., Farhi, E., and Gourinchas, P.-O. (2008). An equilibrium model of ”global imbalances” and low interest rates.American Economic Review, 98(1):358–93.
- Miyamoto, W., Nguyen, T. L., and Oh, H. (2023). In Search of Dominant Drivers of the Real Exchange Rate.The Review of Economics and Statistics, pages 1–50.
- Mullen, M. M. and Woo, S. K. (2024). Real Exchange Rate and Net Trade Dynamics: Financial and Trade Shocks.
- Darvas, Z. (2021). Timely measurement of real effective exchange rates. Working paper 2021/15, Bruegel.
- Feenstra, R. C., Luck, P., Obstfeld, M., and Russ, K. N. (2018). In search of the armington elasticity.Review of Economics and Statistics, 100(1):135–150.
- Fernald, J. G. (2014). A quarterly, utilization-adjusted series on total factor productivity. Technical report.
- Cao, S. (2021). Estimates of seasonally-adjusted quarterly total factor productivity for the canadian business sector. Technical report.

### Regional and policy-focused assessments
- Lee, J., Licci, L., Milesi-Ferretti, G.-M., Ostry, J., and Prati, A. (2008). Exchagne rate assessments: Cger methodologies. IMF Occasional Paper 261, International Monetary Fund.
- Coutinho, L., Turrini, A., and Zeugner, S. (2022). Assessing the euro area current account.Journal  of  International  Money  and  Finance, 121:102512.  Current Account (Im-)Balances: Determinants and Policy Implications.
- Allen, C., Casas, C., Ganelli, G., Juvenal, L., Leigh, D., Rabanal, P., Rebillard, C., Rodriguez, J., and Jalles, J. (2023). 2022 update of the external balance assessment methodology. IMF Working Papers 2023/047, International Monetary Fund.
- Blanchard, O., Giavazzi, F., and Sa, F. (2005). International investors, the u.s. current account, and the dollar.Brookings Papers on Economic Activity, 36(1):1–66.
- Goldstein, M. and Khan, M. S. (1985). Income and price effects in foreign trade.Handbook of international economics, 2:1041–1105.

*Dominant Drivers of Current Account Dynamics Working Paper No. WP/2024/092*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024092-print-pdf.pdf_