## EXCHANGE RATE PASS-THROUGH TO INFLATION IN SINGAPORE

## Source details

**Canonical URL:** [EXCHANGE RATE PASS-THROUGH TO INFLATION IN SINGAPORE](https://www.imf.org/-/media/files/publications/cr/2024/english/1sgpea2024002-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2024/english/1sgpea2024002-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2024/english/1sgpea2024002-print-pdf.pdf.json)

---

### Background and recent inflation and monetary policy developments
- The Monetary Authority of Singapore (MAS) operates a basket, band, and crawl (BBC) exchange rate-based monetary policy framework that manages the nominal effective exchange rate (S$NEER) against an undisclosed basket of currencies.
- The exchange rate affects prices through both the “imported inflation” channel and the “derived demand” channel.
- Inflation and policy chronology:
  - Headline inflation peaked at 7.5 percent in September 2022 and moderated to 3.7 percent in December 2023.
  - MAS core inflation fell to 3.3 percent in December 2023 from 5.5 percent in February 2023.
  - MAS responded with five rounds of consecutive tightening and has remained on pause since April 2023.
- Labor market: “The labor market remains tight compared to its pre-pandemic level.” Job vacancy to unemployed person ratio used as a measure of labor market tightness.

### Empirical identification of plausibly exogenous S$NEER changes (two-stage strategy)
- Stage 1: identify plausibly exogenous changes in S$NEER using MAS meeting-level data and estimate:
  - Equation (1): ∆S$NEER_m = β0 + β1 S$NEER_{m−1} + Σ δ_k π_{m,t+k} + Σ δ_k y_{m,t+k} + η_m_NEER
  - η_m_NEER is the residual from equation (1) and interpreted as an unexpected or surprise change in S$NEER at meeting m.
- Data and estimation details:
  - Estimates cover the period 2000m1 to 2023m10.
  - Monetary policy shocks series are set to zero in months without a monetary policy meeting and converted to monthly frequency following Romer and Romer (2004) and Holm et al. (2021).
  - The model explains about 30 percent of the variation in the change of S$NEER (R-squared 0.296).
- Selected Table 1 estimates (exactly as reported):
  - S$NEER_{m−1}: 0.036** (0.016)
  - y_{m,t}: 0.001* (0.000)
  - y_{m,t+1}: 0.003** (0.001)
  - π_{m,t}: -0.002 (0.001)
  - π_{m,t+1}: 0.001* (0.000)
  - β0: -0.188** (0.078)
  - N 50
  - R-squared 0.296
- Examples of identified surprises from η_m_NEER:
  - April 2008 upward recentering → surprise appreciation ≈ 1 percent.
  - March 2020 downward recentering → surprise depreciation ≈ 0.6 percent.
  - July 2022 off-cycle upward recentering → surprise appreciation ≈ 0.4 percent.

### Estimating pass-through and role of labor market tightness (local projections)
- Stage 2: estimate pass-through of identified S$NEER shocks to price indices using local projections (Jordà, 2005) over 2000m1–2020m3.
  - Equation (2) (local projection specification): Y_{t+h} − Y_{t−1} = α^0_h + θ_h η_t_NEER + λ_h (η_t_NEER × vac_ratio_t) + Σ φ^j_h X^D_{t−j} + Σ μ^j_h X^G_{t−j} + ξ_{t+h}
  - Y_t = log price index of interest (CPI headline, MAS core, Goods CPI, Services CPI).
  - vac_ratio_t = job vacancy to unemployed person ratio (quarterly series).
  - Domestic controls X^D include output gap, lagged inflation, lagged NEER shocks, job vacancy ratio, lagged changes in log manufacturing PPI (with 12 lags).
  - Global controls X^G include global output gap, global supply chain pressure index, changes in log global food price and oil price indices.
- Response interpretation:
  - θ_h captures percent response of prices to a 1 percentage point appreciation in the S$NEER shock at horizon h months, absent labor market interaction.
  - Cumulative impact accounting for labor market tightness computed as θ_h + λ_h × p75_vac, where p75_vac is the 75th percentile of the job vacancy ratio series.
- Methodological notes:
  - Local projection framework chosen for robustness to misspecification and flexibility for state-dependent analyses.
  - Monthly output gap obtained by HP filtering monthly industrial production; global output gap obtained similarly using global industrial production.
  - Job vacancy ratio available only quarterly and used accordingly.

### Main empirical findings on exchange rate pass-through (exact reported magnitudes)
- Pass-through magnitude:
  - A 1 percent appreciation shock in the S$NEER leads to a cumulative 2 percentage points reduction in headline inflation in about 9 months.
  - For an initial inflation rate of 3 percent, a 1 percent appreciation shock in the S$NEER would lead to an inflation rate of about 1 percent in 9 months.
  - For MAS core inflation, a 1 percent appreciation shock leads to about 1 percentage point cumulative reduction over 9 months.
  - For comparison, recent estimates for advanced economies show a pass-through coefficient about 0.1 percentage point cumulatively over 12 months.
- Endogeneity matters:
  - Exchange rate pass-through, in absolute terms, is larger when using the plausibly exogenous monetary policy–induced exchange rate shocks (η_m_NEER) than when using raw monthly changes in log S$NEER.
  - Example statement: “the average exchange rate pass-through is about twice as large when using the plausibly exogenous exchange rate shocks” (headline inflation comparison).
- Role of labor market tightness:
  - Labor market tightness significantly weakens the S$NEER pass-through to inflation:
    - The marginal impact under tight labor market conditions represents only about fourth and half of the impact previously discussed for headline and MAS core inflation, respectively.
  - Weaker pass-through during tight labor markets may reflect that labor markets typically tighten more in response to domestic shocks, introducing substantial time lags in exchange rate transmission.
  - Inclusion of the output gap among controls mitigates demand-side pressures, but other demand-side factors not accounted for could play a role.
  - Despite being lower than normal conditions for Singapore, the marginal impact under tight labor market conditions remains large compared to estimates for advanced economies on average.
- Goods versus services inflation:
  - Pass-through is larger and quicker for goods than for services:
    - Goods CPI responds to a S$NEER shock in a statistically significant manner two months ahead of services CPI.
    - For headline CPI: a 1 percent appreciation is likely to reduce goods inflation by 2 percent cumulatively in 7 months, and reduce services inflation by about 1 percent.
    - Under tight labor market conditions:
      - The pass-through to goods CPI is about halved.
      - The pass-through to services CPI represents about fourth of the average effect.
    - Overall, exchange rate pass-through to services CPI is likely to be severely weakened under tight labor market conditions.

### Robustness, persistence, and sensitivity
- Robustness checks undertaken (Appendix II) include:
  - Adding controls for certificate of entitlement (COE) quotas.
  - Controlling for GST hikes via a yearly dummy for GST rate hikes.
  - Redefining the goods CPI basket to include “Utilities and other fuels”.
  - Employing an alternative, broader measure of labor market tightness (job vacancy rate defined as total number of job vacancies divided by the total demand for manpower).
- Results are broadly robust to:
  - Accounting for COE quotas.
  - Controlling for episodes of GST hikes.
  - Redefining the CPI basket (including “Utilities and other fuels”).
  - Using the job vacancy rate instead of the job-vacancy-to-unemployed ratio.
- Using the job vacancy rate shows a more significant impact of labor market tightness on exchange rate pass-through but remains consistent with the finding of subdued pass-through under tight labor market conditions.
- Horizon and persistence of effects:
  - Extending impulse-response horizon to 24 months:
    - Transmission of exchange rate shocks persists mainly only over about 14 months, except for goods CPI where the impact can persist up to 20 months.
    - The extension beyond the standard 12-month horizon could weaken the quality of statistical inference within the local projection framework.

### Policy implications and recommendations
- Exchange rate pass-through to inflation in Singapore is strong, driven mainly by goods components of the CPI basket, but can be weakened under tight labor market conditions—especially for services inflation.
- Monetary policy implications:
  - Monetary policy transmission via the exchange rate can be weakened under tight labor market conditions because labor market tightness dampens exchange rate pass-through.
  - “Overall, the results suggest that monetary policy should be more vigilant under a tight labor market condition.”
  - Implication for MAS: When the labor market is tight, exchange-rate-based tightening may have reduced ability to lower inflation via pass-through channels, so policymakers should account for labor market conditions when assessing the likely impact of S$NEER adjustments on inflation.
- Structural policy recommendations to support monetary policy effectiveness:
  - Policies to ease structural labor market tightness could help support monetary policy in ensuring price stability:
    - Addressing persistent matching frictions in the labor market (e.g., shortage of technology skills, shifts in the foreign workforce affecting sectors such as health and social services).
    - Enhancing existing policies such as Career Conversion Programs to equip workers with skills and ease matching frictions.

*Source: IMF staff prepared section “EXCHANGE RATE PASS-THROUGH TO INFLATION IN SINGAPORE” (Content unit: 1. Inflation Developments).*

### 1. Inflation Developments ______________________________________________________________ 4

### EXCHANGE RATE PASS-THROUGH TO INFLATION IN SINGAPORE

### Background and recent inflation/monetary policy developments
- The Monetary Authority of Singapore (MAS) operates a basket, band, and crawl (BBC) exchange rate-based monetary policy framework that manages the nominal effective exchange rate (S$NEER) against an undisclosed basket of currencies.
- The exchange rate affects prices through both the “imported inflation” channel and the “derived demand” channel.
- Inflation and policy chronology:
  - Headline inflation peaked at 7.5 percent in September 2022 and moderated to 3.7 percent in December 2023.
  - MAS core inflation fell to 3.3 percent in December 2023 from 5.5 percent in February 2023.
  - MAS responded with five rounds of consecutive tightening and has remained on pause since April 2023.
- Labor market: “The labor market remains tight compared to its pre-pandemic level.” Job vacancy to unemployed person ratio used as a measure of labor market tightness.

### Empirical identification of plausibly exogenous S$NEER changes (two-stage strategy)
- Stage 1: Identify plausibly exogenous changes in S$NEER using MAS monetary policy meeting-level data and estimate:
  - Equation (1): ∆S$NEER_m = β0 + β1 S$NEER_{m−1} + Σ δ_k π_{m,t+k} + Σ δ_k y_{m,t+k} + η_m_NEER
  - η_m_NEER is the residual from equation (1) and interpreted as an unexpected or surprise change in S$NEER at meeting m.
- Data and estimation details:
  - Estimates cover the period 2000m1 to 2023m10.
  - Monetary policy shocks series are set to zero in months without a monetary policy meeting and converted to monthly frequency following Romer and Romer (2004) and Holm et al. (2021).
  - The model explains about 30 percent of the variation in the change of S$NEER (R-squared 0.296).
- Selected Table 1 estimates (exactly as reported):
  - S$NEER_{m−1}: 0.036** (0.016)
  - y_{m,t}: 0.001* (0.000)
  - y_{m,t+1}: 0.003** (0.001)
  - π_{m,t}: -0.002 (0.001)
  - π_{m,t+1}: 0.001* (0.000)
  - β0: -0.188** (0.078)
  - N 50
  - R-squared 0.296
- Examples of identified surprises from η_m_NEER:
  - April 2008 upward recentering → surprise appreciation ≈ 1 percent.
  - March 2020 downward recentering → surprise depreciation ≈ 0.6 percent.
  - July 2022 off-cycle upward recentering → surprise appreciation ≈ 0.4 percent.

### Estimating pass-through and role of labor market tightness (local projections)
- Stage 2: Estimate pass-through of identified S$NEER shocks to price indices using local projections (Jordà, 2005) over 2000m1–2020m3.
  - Equation (2) (local projection specification): Y_{t+h} − Y_{t−1} = α^0_h + θ_h η_t_NEER + λ_h (η_t_NEER × vac_ratio_t) + Σ φ^j_h X^D_{t−j} + Σ μ^j_h X^G_{t−j} + ξ_{t+h}
  - Y_t = log price index of interest (CPI headline, MAS core, Goods CPI, Services CPI).
  - vac_ratio_t = job vacancy to unemployed person ratio (quarterly series).
  - Domestic controls X^D include output gap, lagged inflation, lagged NEER shocks, job vacancy ratio, lagged changes in log manufacturing PPI (with 12 lags).
  - Global controls X^G include global output gap, global supply chain pressure index, changes in log global food price and oil price indices.
- Response interpretation:
  - θ_h captures percent response of prices to a 1 percentage point appreciation in the S$NEER shock at horizon h months, absent labor market interaction.
  - Cumulative impact accounting for labor market tightness computed as θ_h + λ_h × p75_vac, where p75_vac is the 75th percentile of the job vacancy ratio series.

### Key empirical findings
- Endogeneity matters:
  - Exchange rate pass-through, in absolute terms, is larger when using the plausibly exogenous monetary policy–induced exchange rate shocks (η_m_NEER) than when using raw monthly changes in log S$NEER.
  - Example statement: “the average exchange rate pass-through is about twice as large when using the plausibly exogenous exchange rate shocks” (headline inflation comparison).
- Labor market tightness dampens pass-through:
  - Labor market tightness (measured by vacancy ratio) weakens exchange rate pass-through; tight labor market conditions reduce the extent to which S$NEER appreciations lower inflation.
  - When accounting for labor market tightness and not addressing exchange-rate endogeneity, the analysis fails to identify pass-through—i.e., pass-through can be obscured without isolating exogenous S$NEER changes.
- Components of CPI:
  - The pattern of larger pass-through using exogenous shocks and damping by labor market tightness holds for CPI components (goods and services) as well.
- Robustness and methodology rationale:
  - Local projection framework chosen for robustness to misspecification and flexibility for state-dependent analyses.
  - Monthly output gap obtained by HP filtering monthly industrial production; global output gap obtained similarly using global industrial production.
  - Job vacancy ratio available only quarterly and used accordingly.

### Policy implications and recommendations
- Monetary policy transmission via the exchange rate can be weakened under tight labor market conditions because labor market tightness dampens exchange rate pass-through.
- “Overall, the results suggest that monetary policy should be more vigilant under a tight labor market condition.”
- Implication for MAS: When the labor market is tight, exchange-rate-based tightening may have reduced ability to lower inflation via pass-through channels, so policymakers should account for labor market conditions when assessing the likely impact of S$NEER adjustments on inflation.

*Source: IMF staff prepared section “EXCHANGE RATE PASS-THROUGH TO INFLATION IN SINGAPORE” (Content unit: 1. Inflation Developments).*

### 9.      We find that the S$NEER pass-through to inflation is strong (Figure 4). The results show

### 1sgpea2024002-print-pdf - 9.      We find that the S$NEER pass-through to inflation is strong (Figure 4). The results show

### Main empirical findings on exchange rate pass-through
- A 1 percent appreciation shock in the S$NEER leads to a cumulative 2 percentage points reduction in headline inflation in about 9 months.
- For an initial inflation rate of 3 percent, a 1 percent appreciation shock in the S$NEER would lead to an inflation rate of about 1 percent in 9 months.
- For MAS core inflation, a 1 percent appreciation shock leads to about 1 percentage point cumulative reduction over 9 months.
- For comparison, recent estimates for advanced economies show a pass-through coefficient about 0.1 percentage point cumulatively over 12 months.

### Role of labor market tightness
- Labor market tightness significantly weakens the S$NEER pass-through to inflation:
  - The marginal impact under tight labor market conditions represents only about fourth and half of the impact previously discussed for headline and MAS core inflation, respectively.
  - Weaker pass-through during tight labor markets may reflect that labor markets typically tighten more in response to domestic shocks, introducing substantial time lags in exchange rate transmission.
  - Inclusion of the output gap among controls mitigates demand-side pressures, but other demand-side factors not accounted for could play a role.
  - Despite being lower than normal conditions for Singapore, the marginal impact under tight labor market conditions remains large compared to estimates for advanced economies on average.

### Goods versus services inflation
- Pass-through is larger and quicker for goods than for services:
  - Goods CPI responds to a S$NEER shock in a statistically significant manner two months ahead of services CPI.
  - For headline CPI: a 1 percent appreciation is likely to reduce goods inflation by 2 percent cumulatively in 7 months, and reduce services inflation by about 1 percent.
  - Under tight labor market conditions:
    - The pass-through to goods CPI is about halved.
    - The pass-through to services CPI represents about fourth of the average effect (i.e., much more severely weakened).
  - Overall, exchange rate pass-through to services CPI is likely to be severely weakened under tight labor market conditions.

### Robustness checks and sensitivity
- Robustness checks undertaken (Appendix II) include:
  - Adding controls for certificate of entitlement (COE) quotas.
  - Controlling for GST hikes via a yearly dummy for GST rate hikes.
  - Redefining the goods CPI basket to include “Utilities and other fuels”.
  - Employing an alternative, broader measure of labor market tightness (job vacancy rate defined as total number of job vacancies divided by the total demand for manpower).
- Results are broadly robust to:
  - Accounting for COE quotas.
  - Controlling for episodes of GST hikes.
  - Redefining the CPI basket (including “Utilities and other fuels”).
  - Using the job vacancy rate instead of the job-vacancy-to-unemployed ratio.
- Using the job vacancy rate shows a more significant impact of labor market tightness on exchange rate pass-through but remains consistent with the finding of subdued pass-through under tight labor market conditions.

### Horizon and persistence of effects
- Extending impulse-response horizon to 24 months:
  - Transmission of exchange rate shocks persists mainly only over about 14 months, except for goods CPI where the impact can persist up to 20 months.
  - The extension beyond the standard 12-month horizon could weaken the quality of statistical inference within the local projection framework.

### Conclusions and policy implications
- Exchange rate pass-through to inflation in Singapore is strong, driven mainly by goods components of the CPI basket, but can be weakened under tight labor market conditions—especially for services inflation.
- Exchange rate-based monetary policy serves Singapore well, but it needs to be more vigilant when the labor market is tight.
- Policies to ease structural labor market tightness could help support monetary policy in ensuring price stability:
  - Addressing persistent matching frictions in the labor market (e.g., shortage of technology skills, shifts in the foreign workforce affecting sectors such as health and social services).
  - Enhancing existing policies such as Career Conversion Programs to equip workers with skills and ease matching frictions.

*Source: IMF staff calculations.*

### 1.       GDP growth is expected to slow in

### 1.       GDP growth is expected to slow in

### Medium-term outlook for China and implications for Singapore
- Growth in China is expected to slow to just over 3 percent in the medium term.
- IMF staff have revised growth projections to suggest this transition will happen earlier than previously expected.
- The moderation in GDP growth is projected to be broad-based, with consumption and investment growth slowing.
- China’s declining trend growth, driven by an aging population and low productivity growth, will likely impact other Asia-Pacific countries, including Singapore.
- Much of China’s role in regional trade stems from its insertion into global value chains (GVCs), which has shifted from assembly of foreign inputs to China now being a key supplier of production inputs to many Asian countries.

### GVC exposure and sectoral linkages between China and Singapore
- Eora Global Supply Chain database coverage and construction:
  - c=189 countries, s=26 intermediate good sectors, and f=6 final demand components over the years 1990-2022.
  - The 6 final demand components are private consumption, public consumption, non-profit institutions serving households, gross fixed capital formation, changes in inventories, and acquisitions less disposals of valuables.
- Trade and value-added statistics between Singapore and China:
  - In 2022, bilateral exports of GVC goods to China were equal to over 75 percent of Singapore’s aggregate exports to China.
  - Only 29 percent of Singapore’s aggregate bilateral exports to China contains value-added that originates in Singapore.
  - Over 9 percent of Singapore’s value-added in 2022 was ultimately absorbed by China.
- Concentration of Singapore’s value added absorbed by China:
  - Financial intermediation and business services, electrical and machinery manufacturing, and petroleum and chemicals industries account for about 60 percent of Singapore’s value added that is ultimately absorbed by China (5.7 percentage points of the 9 percent).
  - In business and financial services and in petroleum, chemical and non-metallic mineral sectors, approximately half the value added goes to Chinese final consumption and half to Chinese final investment demand.
  - In electrical and machinery manufacturing, demand comes primarily from Chinese final investment demand.
- China’s sectors that absorb Singapore’s exports:
  - Singapore’s exports to China are absorbed mainly in three Chinese destination sectors: electrical and machinery equipment, construction, and financial and business services.
  - Except for construction, a large share of Singapore value added absorbed in China remains within the same sector.

### Hypothetical scenario: 10 percent decline in Chinese final demand and sectoral impact on Singapore
- Scenario construction:
  - A hypothetical 10 percent reduction in both Chinese final consumption and final investment demand across all country-sectors origins is imposed.
  - Intermediate goods demand and final demand matrices are reduced by 10 percent (denoted 풀* and 푻*), and the corresponding hypothetical value added originating in Singapore that is absorbed in China is calculated using a Leontief framework (equations (2) and (3) in the source).
- Key sectoral impacts (percent point differences in share of Singapore’s value added absorbed by China):
  - The main sectors impacted by the decline in Chinese consumption and investment are petroleum, electrical and machinery, and financial intermediation sectors.
  - Calculated declines in value added absorbed by Chinese final consumption and investment demand:
    - Petroleum: decline of 0.15 percentage points.
    - Electrical and machinery: decline of 0.25 percentage points.
    - Financial intermediation and business services: decline of 0.25 percentage points.
  - Total decline across all sectors equals 1.1 percentage points.
- Source-of-slowdown heterogeneity:
  - A slowdown in Chinese final consumption would hit services and household goods (food, clothing, textiles, education etc.) relatively more than an investment slowdown.
  - These consumption-sensitive sectors represent less than half of Singapore’s value added exported to China.
- Caveats on the static exercise:
  - The exercise does not consider behavioral responses (e.g., reorienting exports to alternative destinations) that could offset the calculated decline.
  - It also does not capture possible greater fragmentation in GVCs and lower demand for goods and services that cannot be represented in static input-output tables.

### Aggregate empirical approach to estimating spillovers to ASEAN and Singapore
- Identification of China-domestic growth shocks:
  - The China-domestic growth shock series, 휗t_CHN, is obtained from regression ∆logYt = 훽0 + 훽1 Xt + 휗t_CHN, where ∆logY is the first difference of log China real GDP (source: World Development Indicators) and X is a vector of global factors.
  - Global control variables included contemporaneously: change in log global oil price, global metal price, the US long term bond yield (10 years), and real GDP growth of the G7 economies.
  - The residual 휗t_CHN is interpreted as the China domestically originated growth shock, purging global-factor-driven changes in Chinese real GDP.
- Data sources and notes:
  - US long term (10 years) bond yield is taken from the Federal Reserve Bank; other control variables are sourced from the IMF World Economic Outlook (WEO).
  - The approach follows Ahmed et al. (2022) and is similar in spirit to Furceri et al. (2017).
  - The source notes potential estimation caveats: omission of global variables could bias the shock estimate upward, while endogeneity if China’s GDP changes drive control variables could bias it downward.
- Presentation:
  - The China-domestic growth shocks estimated by equation (4) are shown in Figure 7 of the source (series labeled as Domestic growth shock).

*Prepared by Kodjovi Eklou, Shujaat Khan, and Margaux MacDonald. Source: IMF staff calculations using Eora Global Supply Chain Database, IMF World Economic Outlook, and World Development Indicators.*

### 12.      The impact of China’s growth shock on ASEAN countries’ trend growth is then

### 12.      The impact of China’s growth shock on ASEAN countries’ trend growth is then

### Methodology
- Sample: 10 ASEAN economies, period 1990-2019.
- Empirical framework: Local projection framework (Jordà, 2005).
- Baseline specification (equation (5)):
  - Dependent variable: Y_ct+h − Y_ct−1 where Y_ct is the logarithm of trend GDP.
  - Regressor of interest: 휗_t_CHN (China-specific or China-domestic growth shock).
  - Controls Z_ct include lags of China-specific growth shocks, G_ct, lags of trend GDP, and lags of trade openness.
  - Country fixed effects: 훼_c.
  - Horizon h up to five years.
- Heterogeneity specification (equation (6)):
  - Adds interaction term γ_h (휗_t_CHN × G_ct) where G_ct indicates high/low trade linkage (above/below sample median) following a semi-parametric approach (Cloyne et al., 2023).
- Trend GDP construction: HP filter applied to real GDP series.
- Trade openness: total trade as share of GDP from World Development Indicators.
- Inference: Kraay and Driscoll standard errors; local projection inference used for robustness to misspecifications (Montiel Olea and Plagborg-Møller, 2021).
- Caveat: Short time dimension (about 30 years) may bias impulse responses over longer horizons; mitigated by short horizon of projections.

### Empirical findings — average spillovers
- Magnitude:
  - A 1 percentage point decline in China’s domestic growth → cumulative decline of about 1 percentage point in ASEAN average trend growth after five years.
  - Equivalent to about 0.2 percent per year.
  - Comparable to Dizioli et al (2016) estimate: elasticity of Chinese growth on ASEAN-5 ≈ 0.2 percent (1981Q1–2013Q1, GVAR).
- Heterogeneity by trade linkage:
  - Spillovers are driven by strong trade linkages: countries with stronger trade ties to China experience larger spillovers.
  - Spillovers are twice as large for countries with stronger trade linkage to China (measured via gross trade or value-added terms).
  - Singapore-specific: cumulative impact of 2.1 percentage points over five years in Singapore (given dependence on Chinese final demand), versus 1.4 percentage points for the average ASEAN country in the sample (period 1990-2019).

### Channels: investment, employment, and TFP
- Cobb-Douglas decomposition used: g_y = g_TFP + (1−α) g_K + α g_L.
- Data: TFP, capital stock, employment (persons employed and average hours worked) from Penn World Tables.
- Estimates (Figure 11 results):
  - A 1 percent decline in China domestic growth shock → cumulative decrease in capital stock in ASEAN economies of 1 percentage point over 4 years, moderating to about 0.8 percentage points in year 5.
  - Impact on employment: about 0.5 percentage points cumulatively over five years.
  - No meaningful impact on hours worked and on TFP.
- Interpretation: Growth spillovers from China to ASEAN trend output are likely driven particularly strongly through the impact on investment (capital stock), and to a lesser extent through employment; TFP largely unaffected in estimates.

### Robustness checks
- Alternative China-specific shock measure (Furceri et al., 2017 approach; equation (7)):
  - ∆logY_Ct = β_C + τ_t + ε_ct estimated in panel; ε_ct for c=China used as China-specific shock.
  - Results similar to baseline:
    - A 1 percent increase in China’s growth → cumulative increase of about 0.8 percentage points in ASEAN trend growth over five years.
    - Translates to average annual growth impact of 0.16 percent (close to 0.2 percent in baseline).

### General equilibrium analysis (GIMF augmented with GVCs)
- Model: IMF’s Global Integrated Monetary and Fiscal model (GIMF) augmented with global value chains (GVCs); multi-regional DSGE with 10 regions and bilateral trade tracking.
- GVC augmentation: disaggregates tradable goods into non-GVC tradable and GVC tradable goods; allows round-about trade and intermediate use of GVC goods.
- Calibration:
  - Data source: OECD Inter-Country Input-Output Database of 2018.
  - Southeast Asia (excluding Indonesia): exports and imports ≈ 60 percent of GDP.
  - GVC tradable exports ≈ 40 percent of aggregate exports; GVC tradable imports ≈ 50 percent of aggregate imports.
  - Assumes relatively inelastic demand for goods in GVC tradable sector.
  - Note: Model groups other Southeast Asia (Brunei, Cambodia, Hong Kong SAR, Lao PDR, Malaysia, Myanmar, Philippines, Singapore, Thailand, Vietnam) without disaggregation.
- Simulation scenario:
  - China aggregate annual productivity grows about 1 percentage point higher relative to baseline for 15 years.
  - In China’s GVC sector, productivity grows twice as fast as in non-tradables.
- Simulation results (Figure 14 summary):
  - China’s GDP level in the long-run increases by over 20 percent relative to baseline.
  - GDP level in Southeast Asia (excl. Indonesia) increases by over 4 percent relative to baseline.
  - Mechanisms: higher productivity → greater investment and capital stock → higher labor demand, labor income, and consumption → stronger external demand for Southeast Asia. Indirect technological spillovers and GVC integration amplify effects.
- Asymmetric note: Positive productivity shock illustrated; an equivalent negative shock would likely produce opposite-signed results but may not be exactly symmetric.

### Summary and policy implications
- IMF staff projection context:
  - China’s GDP growth projected to moderate to around 3.5 percent by 2028 due to ageing population and slowing productivity growth; moderation arises from declines in both investment and consumption.
- Key quantitative implications:
  - 1 percentage point decline in Chinese domestic growth → cumulative ≈ 1 percentage point decline in ASEAN average trend growth over five years (≈ 0.2 percent per year).
  - Singapore’s cumulative decline could be as high as 2.1 percentage points over five years given high exposure to Chinese final demand.
  - Primary transmission through decline in capital stock (investment) and secondarily through employment; TFP not meaningfully affected in estimates.
- Economic interpretation:
  - Trade channel is the primary transmission mechanism; stronger trade linkages yield larger spillovers.
  - Deep integration in GVCs amplifies spillovers via both demand and indirect technological channels.
- Policy-relevant considerations (implied by analysis):
  - Economies with strong trade and GVC linkages to China should prepare for investment-driven effects from Chinese slowdowns.
  - Diversification of trade links and measures to support domestic investment could mitigate long-run trend growth losses from adverse Chinese growth shocks.
  - Monitoring of capital formation and employment trends is critical for assessing medium-term spillover risks.

*Source: 1sgpea2024002-print-pdf*

### References

### References

### Key findings on AI and Singapore’s labor market
- Singapore is well-prepared for AI adoption, scoring highly on the IMF’s AI Preparedness Index (AIPI) across digital infrastructure, human capital and labor market policies, innovation and economic integration, and regulation and ethics.
- Singapore’s labor market is highly exposed to AI: about 77 percent of Singapore’s employed resident workers are highly exposed to AI.
- Comparative exposure benchmarks: 40 percent for emerging market economies (EMs), 26 percent for low-income countries (LICs), and 60 percent for advanced economies.
- Only about 23 percent of employed resident workers in Singapore are in low-skilled jobs.
- Within the highly exposed group:
  - 38.9 percent of employed are in occupations with high AI complementarity.
  - 38.6 percent of employed are in occupations with low AI complementarity.
- Demographic differentials:
  - Female workers: about 49 percent of female workers are employed in occupations with high AI exposure and low AI complementarity, compared to 29 percent of male workers.
  - Occupation share examples driving gender differences: clerical support roles (15 percent for women, 4 percent for men), business and administration associate professionals (14 percent for women, 9 percent for men), and sales workers (6 percent for women, 3 percent for men).
  - Men are more represented in high-exposure, high-complementarity occupations: 44 percent of men versus 33 percent of women.
  - Specific male-dominated occupations with higher complementarity: managers (20 percent for men, 16 percent for women) and science and engineering associate professionals (7 percent for men, 2 percent for women).
  - Age groups: about 6 percent of workers are between 15 to 24 years old; half of these younger workers are employed in occupations with high AI exposure and low AI complementarity, and only one-fifth of young workers are employed in occupations with high AI exposure and high AI complementarity.

### Methodology and definitions
- Occupational exposure to AI (AIOE) follows Felten et al. (2021): links 10 AI applications to 52 workplace abilities (cognitive, physical, psychomotor, sensory) using relatedness scores x_ij for AI application i ∈ {1,...,10} and ability j ∈ {1,...,52}. Ability-level AI exposure A_j = sum_{i=1}^{10} x_ij. Occupational exposure combines ability-level exposure with O*NET weights for ability prevalence and importance.
- The 10 AI applications considered by Felten et al. (2021) are: abstract strategy games; real-time video games; image recognition; visual question answering; image generation; reading comprehension; language modeling; translation; speech recognition; and instrumental track recognition.
- AI complementarity follows Pizzinelli et al. (2023): uses O*NET data on work contexts (social and physical dimensions, decision-making importance, consequence of mistakes) and skills (education, on-the-job training, professional experience) to estimate how well AI can complement human labor in an occupation.
- Classification rule: an occupation is high (low) in AI exposure or complementarity if its score is above (below) the median across occupations.
- Focus is on high-exposure occupations because complementarity is less relevant for low-exposure roles.

### Occupational patterns
- High-skilled occupations (managers, professionals, technicians and associate professionals) tend to have higher AIOE than the median.
- Low-skilled occupations (elementary occupations; plant and machine operators and assemblers; craft and related trade workers; skilled agriculture, forestry and fishery workers) tend to have lower AIOE.
- Semi-skilled roles such as clerical support workers are estimated to be highly exposed to AI.
- Services and sales workers show mixed AI exposure depending on specific jobs.
- Significant within-category variation exists in AI complementarity (for example, health professionals versus information and communication technology professionals within the “professionals” category).

### Implications and risks
- Workers in high-exposure, high-complementarity occupations (managers; science and engineering, health, and legal professionals; associate professionals; teaching professionals) stand to gain productivity improvements from appropriate AI integration, conditional on access to infrastructure and necessary skills.
- Workers in high-exposure, low-complementarity occupations (business and administration professionals; information and communications technology professionals; clerical support workers; some services and sales workers) face higher substitution risk from AI.
- Female and younger workers are disproportionately represented in roles with high exposure and low complementarity, which could exacerbate income inequality without policy intervention.
- Emerging research (Noy and Zhang, 2023; Peng et al., 2023; Brynjolfsson et al., 2023) suggests generative AI may also complement inexperienced and lower-skilled workers in routine tasks (writing, coding, call center operations), potentially mitigating some displacement risks. The report notes limitations from broad age categories and that 15–24 year-olds in the data are often students in casual or part-time jobs.

### Policy recommendations highlighted
- Targeted training policies, leveraging the existing SkillsFuture program, can harness AI’s potential.
- Focused upskilling can mitigate the disruptive impact of AI on vulnerable workers.
- Ensure access to required infrastructure and training so workers in high-exposure, high-complementarity occupations can realize productivity gains.
- Tailor policies to address the higher exposure and lower complementarity observed among women and younger workers to prevent widening gender and age-related inequalities.

*Source: References and chapter "IMPACT OF AI ON SINGAPORE’S LABOR MARKET" (extracts from the provided PDF content).*

### 13.      AI technologies offer the potential to boost productivity across sectors, but realizing

### 13.      AI technologies offer the potential to boost productivity across sectors, but realizing

### Potential productivity gains from AI
- AI can boost productivity across sectors if appropriate policies are implemented.
- A significant share of Singapore’s workforce—particularly managers, science and engineering professionals and associate professionals, health and legal experts, and educators—could see productivity gains from AI, provided they receive the right training.
- High complementarity with AI does not automatically imply that workers in those occupations will see productivity increases or are immune to AI-induced displacement; relevant training and skills development are essential.

### Existing frameworks and initiatives for skills development
- Singapore has existing frameworks that provide training geared towards AI technologies, including:
  - SkillsFuture program
  - Career Conversion Programmes (CCPs)
- Initiatives to raise awareness of the impact of key technologies such as AI on jobs and skills through Jobs Transformation Maps can encourage adoption of AI to transform existing operations and processes.
- The Forward Singapore initiative sets policy objectives to foster a culture of continuous learning and innovation.

### Policy recommendations to realize AI benefits
- Strengthen and enhance existing training programs (SkillsFuture, CCPs) to focus on AI-relevant skills.
- Government and private sector collaboration to identify emerging skills gaps and develop targeted training programs.
- Encourage adoption of AI within firms to transform operations and realize productivity gains.
- Foster a culture of continuous learning and innovation to keep the workforce competitive and adaptable.

### Mitigating disruptive impacts of AI
- Many workers in Singapore are in jobs with low AI complementarity, increasing their risk of being replaced by AI.
- Roles at particular risk include clerical support, business and administration, and sales; upskilling and reskilling programs are critical to facilitate their transition to new employment opportunities.
- The planned introduction of a scheme offering temporary support for the involuntarily unemployed can mitigate potential disruptive impacts on affected workers.

### Gaps and extensions in current policy coverage
- The SkillsFuture Level-Up Program offers promising support but predominantly targets mid-career individuals.
- Expanding such initiatives to include younger workers, who are at higher risk of displacement by AI advancements, can mitigate the impact of AI.
- Incorporating AI literacy and digital skills training into the education system can proactively prepare future generations for the evolving job market.
- Singapore’s National Digital Literacy Programme aims to improve digital literacy and equip the workforce with necessary skills for an AI-driven economy.

*From: 1sgpea2024002-print-pdf - 13.      AI technologies offer the potential to boost productivity across sectors, but realizing*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2024/english/1sgpea2024002-print-pdf.pdf_