## 1. Drivers of Private Consumption (Percent of GNDI)

## Source details

**Canonical URL:** [1. Drivers of Private Consumption (Percent of GNDI)](https://www.imf.org/-/media/files/publications/selected-issues-papers/2026/english/sipea2026075.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/selected-issues-papers/2026/english/sipea2026075.pdf.md)
- [Structured JSON version](/-/media/files/publications/selected-issues-papers/2026/english/sipea2026075.pdf.json)

---

### Introduction
- Singapore’s private consumption as a share of gross national disposable income (GNDI) was 38 percent in 2024, down from 46 percent in 1990.
- Sample: 31 advanced economies, 1990–2024.
- Novel controls added to prior literature:
  - Time-varying financial center intensity: gross external assets and liabilities relative to GDP (log).
  - Interaction between financial center intensity and small geographic size.
  - Time-varying control for pension systems (net contributions).

### Methodology
- Empirical approach: two-way fixed effects regressions with country fixed effects (훿i) and year fixed effects (휃t).
- Dependent variable: private consumption as a share of GNDI (ci,t).
- Controls include:
  - Demographics: population growth rate; old-age dependency ratio; population aging speed (difference between old-age dependency ratio 20 years ahead and in year t).
  - Structural: labor income share; lagged real GDP per capita (log).
  - Cyclical: real GDP growth; household credit/GDP; real short-term deposit rate; change in terms of trade; unemployment rate; income uncertainty.
  - Policy: government consumption/GDP; defined benefit pension net contributions as a share of GDP.
  - Financial center intensity (log of gross assets and liabilities/GDP) and interaction with small geographic size.
- Estimation details:
  - Robust standard errors.
  - Dependent variable uses GNDI as denominator.
- Caveats:
  - Measurement error in constructed proxies (income uncertainty, financial center intensity).
  - Pension control captures net contributions for redistributive schemes; set to zero for defined contribution schemes.
  - Results identify partial correlations, not full general-equilibrium causal effects.

### Key Findings
- Overall:
  - Structural factors dominate cyclical factors in explaining cross-country differences in private consumption ratios.
- Demographics:
  - Old-age dependency ratio: preferred specification (Column 8 of Table 1) — a one percentage point (ppt) increase in the old-age dependency ratio is associated with a 0.41 ppt increase in the consumption ratio.
  - Population growth: a 1 ppt increase is associated with a roughly 0.23 ppt lower consumption ratio (negative and significant in most specifications).
  - Population aging speed: positive sign but not statistically significant.
- Labor share and development:
  - Labor income share: a one ppt increase is associated with a 0.44 ppt higher consumption-to-GNDI ratio (significant).
  - Lagged real GDP per capita (log): in the preferred specification, a one log unit increase is associated with a 4.4 ppt lower consumption ratio (significant).
- Cyclical and policy-related factors:
  - Real GDP growth: a one ppt increase is associated with a 0.15 ppt lower consumption ratio (preferred specification).
  - Household credit/GDP: a one ppt increase is associated with about a 0.04 ppt higher consumption ratio.
  - Real short-term deposit rate: a one ppt increase is associated with a 0.17 ppt lower consumption ratio (preferred specification).
  - Change in terms of trade: improvements lower consumption (negative and significant).
  - Income uncertainty: tends to reduce consumption, but effect becomes statistically insignificant once broader controls are included.
  - Unemployment rate and government consumption/GDP: not statistically significant.
- Pensions:
  - Defined benefit/pay-as-you-go net contributions: a one ppt increase is associated with about a 1.1 ppt lower consumption-to-GNDI ratio (control is a rough proxy).
  - Pension-toolkit analysis: Singapore’s defined contribution pension system can help explain up to 1.6 percent of GDP of its current account surplus.
- Financial center intensity:
  - A one log unit increase in the log financial-center measure is associated with about a 1.9 ppt lower consumption ratio (significant).
  - Interaction with small geographic size: interaction term positive and significant and fully offsets the negative base effect of financial center intensity — small geographic financial hubs have higher consumption rates relative to larger financial hubs, conditional on intensity.
  - Robustness: interaction with population share produces a net positive result (interaction term 2.64 offsets base term -2.36; both significant at the 5 percent level); interaction with output share is insignificant.
- Singapore-specific decomposition (2024, preferred specification):
  - Major downward contributors to Singapore’s consumption-to-GNDI relative to other advanced economies:
    - Old-age dependency ratio (Singapore’s old-age dependency ratio is lower than other advanced economies’ average in 2024 and is the most important driver in the decomposition).
    - Labor share (Singapore’s labor share is about 10 ppts below the sample average).
    - Higher lagged real GDP per capita (log).
    - Financial center status.
    - Lower household credit/GDP.
  - Changes over time (2002 vs. 2024) attributed to higher development, falling labor share, growing financial hub prominence, and year fixed effects (notably negative around the Covid-19 pandemic); the increase in Singapore’s old-age dependency ratio partially offsets the decline.

### Quantitative Estimates (selected coefficients from Table 1, preferred specifications highlighted)
- Income uncertainty: example coefficients -0.133*, -0.087*, -0.139***, -0.114**.
- Population growth rate: example coefficients -0.761**, -0.745**, -0.629**, -0.500*, -0.447*, -0.221*, -0.233.
- Old-age dependency ratio: example coefficients 0.273**, 0.179, 0.195**, 0.222**, 0.319**, 0.395***, 0.414***, 0.441***.
- Labor income share: example coefficients 0.440***, 0.399***, 0.428***, 0.481***, 0.448***, 0.437***, 0.375***.
- Lag Real GDP per capita (Log): example coefficients -5.121**, -5.640***, -5.254**, -4.443**, -4.361**, -6.220***.
- Real GDP growth: example coefficients -0.099, -0.126**, -0.169***, -0.141***, -0.151***, -0.163***.
- Household credit/GDP: example coefficients 0.016, 0.027, 0.042**, 0.044**, 0.052***.
- Real deposit interest rate: example coefficients 0.045, 0.004, -0.164*, -0.173*, -0.215**.
- Change in terms of trade: example coefficients -0.040***, -0.033**, -0.037**, -0.039***.
- Defined Benefit Pension Adj./GDP: example coefficients -1.261**, -1.125**, -0.875***.
- Financial Center Intensity: example coefficients -1.865**, -2.464***.
- Financial Center Intensity x Small Geographic Size: 2.978**.
- Table sample sizes and fit: Observations range across specifications (e.g., 998 to 707); Countries range 31 to 24; R-squared (within) increases from 0.239 to 0.656 across columns.

### Conclusion and Policy Implications
- Primary conclusion: Singapore’s low private consumption-to-GNDI ratio is primarily structural — demographics, lower labor share, high development level, and financial center characteristics depress consumption relative to other advanced economies; cyclical factors are less important.
- Policy implications:
  - Slow-moving nature of drivers implies medium-term structural policy levers are most relevant to reducing Singapore’s external surpluses over the medium and long term.
  - Suggested areas for future work to better inform policy:
    - Microeconomic analysis using household-level data to identify heterogeneity in saving and consumption across demographic groups.
    - Deeper analysis of the pension system and broader social protection as determinants of saving behavior.
    - More precise study of precautionary motives linked to financial center status and the role of multinational enterprises in shaping external imbalances.

### Annex I. Data Definitions and Sources
- Private consumption as a share of gross national disposable income (GNDI): IMF WEO, CEIC, OECD.
- World Uncertainty Index (in logs): World Uncertainty Index.
- Income uncertainty: author’s calculations using real GDP per capita data from IMF WEO; preferred measure is the conditional standard deviation of annual real GDP per capita growth from country-specific GARCH(1,1) models following IMF 2021; if GARCH fails, ARCH(1); if both fail, used 5-year rolling variance.
- Old-age dependency ratio; Population aging speed: UN population data; author’s calculations.
- Population growth rate: IMF WEO.
- Population share (% of world population); Output share (% of world GDP): World Bank.
- Labor income share: Penn World Table; share of labor compensation in GDP at current national prices; for 2024, the 2023 estimates are used.
- Log real GDP per capita; Growth of real GDP per capita; Real GDP growth; Real deposit interest rate; Annual change in the terms of trade; Unemployment rate: IMF WEO.
- Household credit to GDP: BIS; credit to the non-financial sector, Households & NPISHs.
- General government consumption expenditure as a share of GDP: World Bank.
- Pension adjustment as a share of GDP for defined benefit systems: OECD; IMF; author’s calculations; set to zero for countries whose core mandatory pension pillar is funded defined contribution.
- Financial center intensity (in logs): External Wealth of Nations Dataset; author’s calculations; gross external assets plus gross liabilities relative to GDP.
- Small geographic size dummy: author’s classification using geographic size for smallest three economies in the sample; equals one for Singapore, Hong Kong SAR, and Luxembourg.

*Prepared by Sandile Hlatshwayo (with research support from Aishwary Trivedi); analysis covers 31 advanced economies over 1990–2024.*

### 1. Drivers of Private Consumption (Percent of GNDI) ________________________________ 8

### 1. Drivers of Private Consumption (Percent of GNDI)

### Introduction
- Singapore’s private consumption as a share of gross national disposable income (GNDI) was 38 percent in 2024, down from 46 percent in 1990.
- The sample for cross-country analysis covers 31 advanced economies over the period from 1990 to 2024.
- Prior work identifies high development, rapid aging, and low labor share as contributors to low consumption ratios; this analysis builds on that literature and adds three novel controls:
  - A time-varying control for financial center intensity, measured using gross external assets and liabilities relative to GDP.
  - An interaction between financial center intensity and small geographic size.
  - A time-varying control for pension systems (net contributions).

### Methodology
- Empirical approach: two-way fixed effects regressions with country fixed effects (훿i) and year fixed effects (휃t).
- Baseline estimating equation components:
  - Dependent variable: private consumption as a share of GNDI (ci,t).
  - Demographics: population growth rate; old-age dependency ratio; population aging speed (difference between old-age dependency ratio 20 years ahead and in year t).
  - Structural controls: labor income share; lagged real GDP per capita (log).
  - Cyclical factors: real GDP growth, household credit/GDP, real short-term deposit rate, change in terms of trade, unemployment rate, income uncertainty.
  - Policy-related controls: government consumption/GDP; defined benefit pension net contributions as a share of GDP.
  - Financial center intensity: log of gross assets and liabilities/GDP.
- Robust standard errors used; dependent variable uses GNDI as denominator (chosen for sample coverage though imperfect relative to household personal disposable income).
- Caveats noted:
  - Measurement error in constructed proxies (income uncertainty, financial center intensity).
  - Pension control captures net contributions for redistributive schemes (defined benefit/pay-as-you-go); set to zero for defined contribution schemes.
  - Results identify partial correlations, not full general-equilibrium causal effects.

### Key Findings
- Structural factors dominate over cyclical factors in explaining cross-country differences in private consumption ratios.
- Demographics:
  - Old-age dependency ratio: in the preferred specification (Column 8 of Table 1), a one percentage point (ppt) increase in the old-age dependency ratio is associated with a 0.41 ppt increase in the consumption ratio.
  - Population growth: a 1 ppt increase is associated with a roughly 0.23 ppt lower consumption ratio (negative and significant in most specifications).
  - Population aging speed: positive sign but not statistically significant across specifications.
- Labor share:
  - A one ppt increase in the share of labor compensation in national income is associated with a 0.44 ppt higher consumption-to-GNDI ratio (significant).
- Income/development:
  - Lagged real GDP per capita (log): in the preferred specification, a one log unit increase is associated with a 4.4 ppt lower consumption ratio (significant).
- Cyclical and policy-related factors:
  - Real GDP growth: a one ppt increase is associated with a 0.15 ppt lower consumption ratio (preferred specification).
  - Household credit/GDP: a one ppt increase is associated with about a 0.04 ppt higher consumption ratio.
  - Real short-term deposit rate: a one ppt increase is associated with a 0.17 ppt lower consumption ratio (preferred specification).
  - Change in terms of trade: improvements lower consumption (negative and significant).
  - Income uncertainty: tends to reduce consumption, but effect becomes statistically insignificant once broader controls included.
  - Unemployment rate and government consumption/GDP: not statistically significant.
- Pensions:
  - Defined benefit/pay-as-you-go net contributions: a one ppt increase is associated with about a 1.1 ppt lower consumption-to-GNDI ratio (control is a rough proxy).
  - Separate pension-toolkit analysis: Singapore’s defined contribution pension system can help explain up to 1.6 percent of GDP of its current account surplus.
- Financial center intensity:
  - A one log unit increase in the log financial-center measure is associated with about a 1.9 ppt lower consumption ratio (significant).
  - Interaction with small geographic size: the interaction term is positive and significant and fully offsets the negative base effect of financial center intensity, implying that small geographic financial hubs have higher consumption rates relative to larger financial hubs, conditional on intensity.
  - Robustness: interaction with population share produces a net positive result (interaction term 2.64 offsets base term -2.36; both significant at the 5 percent level); interaction with output share is insignificant.
- For Singapore specifically (2024 decomposition, preferred specification):
  - Major downward contributors to Singapore’s consumption-to-GNDI relative to other advanced economies:
    - Old-age dependency ratio (Singapore’s old-age dependency ratio is lower than other advanced economies’ average in 2024, and this is the most important driver in the decomposition).
    - Labor share (Singapore’s labor share is about 10 ppts below the sample average).
    - Higher lagged real GDP per capita (log).
    - Financial center status.
    - Lower household credit/GDP.
  - The decomposition attributes other changes over time (2002 vs. 2024) to higher development, falling labor share, growing financial hub prominence, and year fixed effects (notably negative around the Covid-19 pandemic), while the increase in Singapore’s old-age dependency ratio partially offsets the decline.

### Quantitative Estimates (selected coefficients from Table 1, preferred specifications highlighted)
- Income uncertainty: coefficient estimates range (examples) -0.133*, -0.087*, -0.139***, -0.114** (standard errors reported in table).
- Population growth rate: example coefficients -0.761**, -0.745**, -0.629**, -0.500*, -0.447*, -0.221*, -0.233.
- Old-age dependency ratio: example coefficients 0.273**, 0.179, 0.195**, 0.222**, 0.319**, 0.395***, 0.414***, 0.441***.
- Labor income share: example coefficients 0.440***, 0.399***, 0.428***, 0.481***, 0.448***, 0.437***, 0.375***.
- Lag Real GDP per capita (Log): example coefficients -5.121**, -5.640***, -5.254**, -4.443**, -4.361**, -6.220***.
- Real GDP growth: example coefficients -0.099, -0.126**, -0.169***, -0.141***, -0.151***, -0.163***.
- Household credit/GDP: example coefficients 0.016, 0.027, 0.042**, 0.044**, 0.052***.
- Real deposit interest rate: example coefficients 0.045, 0.004, -0.164*, -0.173*, -0.215**.
- Change in terms of trade: example coefficients -0.040***, -0.033**, -0.037**, -0.039***.
- Defined Benefit Pension Adj./GDP: example coefficients -1.261**, -1.125**, -0.875***.
- Financial Center Intensity: example coefficients -1.865**, -2.464***.
- Financial Center Intensity x Small Geographic Size: 2.978** (offsets base negative effect).
- Table sample sizes and fit: Observations range across specifications (e.g., 998 to 707); Countries range 31 to 24; R-squared (within) increases from 0.239 to 0.656 across columns.

### Conclusion and Policy Implications
- Singapore’s low private consumption-to-GNDI ratio is primarily structural: demographics, lower labor share, high development level, and financial center characteristics depress consumption relative to other advanced economies; cyclical factors are less important.
- The slow-moving nature of these drivers implies medium-term structural policy levers are most relevant to reducing Singapore’s external surpluses over the medium and long term.
- Suggested areas for future work to better inform policy:
  - Microeconomic analysis using household-level data to identify heterogeneity in saving and consumption across demographic groups.
  - Deeper analysis of the pension system and broader social protection as determinants of saving behavior.
  - More precise study of precautionary motives linked to financial center status and the role of multinational enterprises in shaping external imbalances.

*Prepared by Sandile Hlatshwayo (with research support from Aishwary Trivedi); analysis covers 31 advanced economies over 1990–2024.*

### Annex I. Data Definitions and Sources

### Annex I. Data Definitions and Sources

### Variable definitions and data sources
- Private consumption as a share of gross national disposable income (GNDI)
  - IMF WEO, CEIC, OECD
- World Uncertainty Index (in logs)
  - World Uncertainty Index
- Income uncertainty
  - Author’s calculations using real GDP per capita data from IMF WEO.
  - Preferred measure is the conditional standard deviation of annual real GDP per capita growth from country-specific GARCH(1,1) models following IMF 2021; if GARCH fails, ARCH(1); if both fail, used 5-year rolling variance
- Old-age dependency ratio
  - UN population data
- Population aging speed
  - UN population data; author’s calculations; difference between the old-age dependency ratio 20 years ahead and its current value
- Population growth rate
  - IMF WEO
- Population share (% of world population)
  - World Bank
- Output share (% of world GDP)
  - World Bank
- Labor income share
  - Penn World Table; share of labor compensation in GDP at current national prices; for 2024, the 2023 estimates are used
- Log real GDP per capita
  - IMF WEO; GDP per capita at constant PPP prices
- Growth of real GDP per capita
  - IMF WEO
- Real GDP growth
  - IMF WEO
- Household credit to GDP
  - BIS; credit to the non-financial sector, Households & NPISHs
- Real deposit interest rate
  - IMF WEO
- Annual change in the terms of trade
  - IMF WEO
- Unemployment rate
  - IMF WEO
- General government consumption expenditure as a share of GDP
  - World Bank
- Pension adjustment as a share of GDP for defined benefit systems
  - OECD; IMF; author’s calculations.
  - In line with the 2008 System of National Accounts, the adjustment for the change in pension entitlements is defined as the total value of the actual and imputed social contributions payable into pension schemes, plus the total value of contribution supplements payable out of the property income attributed to pension fund beneficiaries, minus the value of the associated service charges, minus the total value of the pensions paid out as social insurance benefits by pension schemes.
  - The data on the pension adjustments are sourced from the OECD and are divided by nominal GDP. The indicator is set to zero for countries whose core mandatory pension pillar is funded defined contribution
- Financial center intensity (in logs)
  - External Wealth of Nations Dataset; author’s calculations; gross external assets plus gross liabilities relative to GDP
- Small geographic size dummy
  - Author’s classification using geographic size for smallest three economies in the sample; equals one for Singapore, Hong Kong SAR, and Luxembourg.

### Annex II — comparison note and figure description
- Title: Key Drivers of Differences in Private Consumption to GNDI for Singapore and Other Financial Centers, 2023
- Figure note:
  - Middle bars show coefficients' estimated contribution for various drivers of Singapore's 2023 consumption rate relative to the 2023 average consumption rate of other financial centers.
  - The comparison group of financial center economies includes Hong Kong SAR, Ireland, Luxembourg, the Netherlands, and Switzerland.
  - This figure uses 2023 instead of 2024 to allow for more data coverage of the financial centers across the various indicators.

*Source: sipea2026075 - Annex I. Data Definitions and Sources*

---


_Source: https://www.imf.org/-/media/files/publications/selected-issues-papers/2026/english/sipea2026075.pdf_