## 1. Aging and TFP Growth in G7 Countries

## Source details

**Canonical URL:** [1. Aging and TFP Growth in G7 Countries](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16237.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16237.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16237.pdf.json)

---

### Introduction and context
- Japan: declining working age population since early 1990s; total population peaked in 2010 at 127 million.
- UN projections cited in source: overall population expected to reach 119 million by 2030, the same level as in 1984.
- Japan’s old age dependency ratio has more than doubled since 1990 and is expected to exceed 50 percent by 2030.
- Paper assesses contribution of Japan’s aging and shrinking population to low TFP growth and inflation using prefectural panel data for 1990–2007.

### Macroeconomic and demographic outcomes for G7 economies (1990–2007) — key figures
- Annual Inflation:
  - Canada: 2.3
  - France: 2.0
  - Germany: 2.3
  - Italy: 3.9
  - Japan: 0.4
  - U.K.: 2.6
  - U.S.: 3.3
  - Average: 2.4
- Annual Real GDP Growth:
  - Canada: 2.6
  - France: 1.9
  - Germany: 1.7
  - Italy: 1.5
  - Japan: 1.3
  - U.K.: 2.9
  - U.S.: 3.0
  - Average: 2.1
- Contributions to Real GDP Growth — Labor:
  - Canada: 0.9
  - France: 0.4
  - Germany: 0.2
  - Italy: 0.3
  - Japan: 0.4
  - U.K.: 0.3
  - U.S.: 0.8
  - Average: 0.5
- Contributions to Real GDP Growth — Capital stock:
  - Canada: 1.1
  - France: 0.8
  - Germany: 0.7
  - Italy: 0.7
  - Japan: 0.8
  - U.K.: 1.0
  - U.S.: 1.3
  - Average: 0.9
- Contributions to Real GDP Growth — Total factor productivity:
  - Canada: 0.9
  - France: 0.7
  - Germany: 0.8
  - Italy: 0.5
  - Japan: 0.2
  - U.K.: 1.7
  - U.S.: 1.0
  - Average: 0.8
- Output per Worker (labor productivity):
  - Canada: 1.5
  - France: 1.2
  - Germany: 1.4
  - Italy: 1.0
  - Japan: 0.8
  - U.K.: 2.5
  - U.S.: 1.8
  - Average: 1.4
- Contributions to Output per Worker — Capital/labor ratio:
  - Canada: 0.6
  - France: 0.5
  - Germany: 0.6
  - Italy: 0.5
  - Japan: 0.6
  - U.K.: 0.8
  - U.S.: 0.8
  - Average: 0.6
- Demographic outcomes:
  - Aging (change in old age dependency ratio):
    - Canada: 2.8
    - France: 4.1
    - Germany: 8.6
    - Italy: 8.3
    - Japan: 15.8
    - U.K.: -0.9
    - U.S.: -0.4
    - Average: 5.5
  - Annual Population growth:
    - Canada: 1.1
    - France: 0.5
    - Germany: 0.2
    - Italy: 0.3
    - Japan: 0.2
    - U.K.: 0.3
    - U.S.: 1.11
    - Average: 0.5

### Mechanisms from literature (selected)
- Inverted U-shaped productivity profile by age (Feyrer 2007): 40–49 most productive.
- Older workers: higher productivity via experience; younger workers: health, processing speed, adaptability, entrepreneurship → innovation.
- Aging can shift demand toward services (more labor intensive, less productive).
- Population size/density influences productivity via agglomeration, networks, learning by doing.
- Ambiguous effects of demographics on inflation: demand vs supply channels, expectations, dissaving by retirees, repatriation of foreign savings, political-economy preferences for lower inflation.
- Secular stagnation link: demographics may lower equilibrium real interest rate; Gagnon et al (2016) estimate a 1.25 percentage point decline in U.S. equilibrium real interest rate since 1980 from demographic and related changes.

### Empirical approach and data
- Prefectural panel for Japan, 1990–2007, to exploit inter-regional variation and avoid cross-country heterogeneity.
- TFP model (based on Feyrer 2007):
  - log TFP_{r,t} = α1 W20_{r,t} + α2 W30_{r,t} + α3 W50_{r,t} + α4 W60_{r,t} + β1 OADDR_{r,t} + β2 log DWA_{r,t} + β3 SER_{r,t} + δ_r + θ_t + u_{r,t}
  - W40 excluded (reference); W20, W30, W50, W60 are shares of 10-year age groups (20–69).
  - OADDR = population aged 70+ as share of working age population.
  - DWA = density of working age population.
  - SER = size of service sector as share of nominal prefectural GDP.
- Inflation model:
  - π_{r,t} = α1 ΔOADDR_{r,t} + α2 Δlog Pop_{r,t} + β OPG_{r,t} + γ π_{r,t−1} + δ_r + θ_t + u_{r,t}
  - π_{r,t}: annual inflation; Δlog Pop: change in prefectural population; OPG: prefectural output gap (HP-filtered).
- Estimation issues and methods:
  - Serial correlation and potential unit roots → first differences and lagged dependent variable models; Arellano-Bond GMM estimator used.
  - Robust standard errors clustered by prefecture.
  - Working age population (20–69) used instead of workforce to mitigate endogeneity from labor force participation.
- Data sources and sample:
  - Sample period: 1990 to 2007.
  - Prefectural real growth, capital stock, labor input: Regional-Level Japan Industrial Productivity (R-JIP).
  - Prefectural TFP: derived via simple growth accounting (Kyoji, Tatsuji and Joji (2015)).
  - Prefectural inflation and age distributions: Portal Site of Official Statistics of Japan.
  - Population data available in 5-year age groups; analysis uses 10-year age groups.

### Inter-prefectural variation (examples, 1990–2007)
- Akita: working age population drop by over 13 percent; old age dependency ratio increase by 20 percentage points.
- Saitama: working age population rise by 16 percent; old age dependency ratio increase by 9 percentage points.
- Kagoshima: TFP grew by 23 percent.
- Nara: TFP contracted by 11 percent.
- Aomori: CPI level rose by 14 percent.
- Saga: CPI level rose by 4 percent.

### Main empirical findings on TFP and age distribution
- Significant relationship between changes in working age distribution and level of TFP.
- Inverted U-shaped productivity profile across age groups; ages 40–49 most productive.
- Prefectures aging faster tend to have relatively lower overall inflation; prefectures with higher population growth typically have higher relative inflation.
- Contributions to Japan’s historical TFP performance:
  - Changes in distribution of working age population depressed annual TFP growth by approximately 0.7–0.9 percent between 1990 and 2007, primarily reflecting a shrinking share of the 40-year-age group.
  - Deflationary pressures from aging and population growth over 1990–2007 were relatively mild.
  - Extending results to 2035 suggests deflationary pressures from shrinking population—as opposed to aging—will become more prominent.

### Selected TFP regression coefficients (Table 2, reproduced exactly)
- Age group 20-29:
  - First difference: -1.174*** (standard error (0.320))
  - Feyrer (2008): -2.225*** (standard error (0.544))
  - Lagged dependent variable: -3.169** (standard error (1.297))
- Age group 30-39:
  - First difference: -4.382*** (standard error (0.988))
  - Feyrer (2008): -1.039 (standard error (0.833))
  - Lagged dependent variable: -3.828*** (standard error (1.206))
- Age group 50-59:
  - First difference: -1.256 (standard error (0.884))
  - Feyrer (2008): -0.475 (standard error (0.563))
  - Lagged dependent variable: -2.120** (standard error (1.051))
- Age group 60-69:
  - First difference: -2.221** (standard error (0.989))
  - Feyrer (2008): -1.593* (standard error (0.885))
  - Lagged dependent variable: -2.066* (standard error (1.211))
- Old-age dependency ratio:
  - First difference: -0.788 (standard error (0.600))
  - Feyrer (2008): -0.220 (standard error (0.675))
  - Lagged dependent variable: 0.478 (standard error (1.001))
- Pop. density:
  - First difference: -0.0554 (standard error (0.331))
  - Feyrer (2008): 0.141 (standard error (0.263))
- Size of service sector:
  - First difference: -2.245*** (standard error (0.239))
  - Feyrer (2008): -1.595*** (standard error (0.249))
- Lagged TFP (Feyrer (2008) specification): 0.758*** (standard error (0.0336))
- Time fixed effects: Yes, Yes, Yes (first two); Observations: 799, 799, 499; Prefectures: 47, 47, 87.
- Note: Arellano-Bond estimator used with one lag of dependent variables; robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Inflation regression results (Table 3 — key coefficients reproduced exactly)
- Sample: Observations: 752; No. prefectures: 47.
- Δ Old-age dep. ratio (coefficients by CPI category):
  - Overall inflation: -0.110*** (standard error (0.0401))
  - Food and beverages: -0.385*** (standard error (0.0845))
  - Housing: 0.0179 (standard error (0.114))
  - Fuel, light and water charges: -0.144 (standard error (0.139))
  - Furniture and household utensils: 0.306* (standard error (0.165))
  - Clothes and footwear: 0.0229 (standard error (0.168))
  - Medical care: -0.192** (standard error (0.0902))
  - Transportation and communication: 0.0829* (standard error (0.0455))
  - Education: 0.105 (standard error (0.0853))
  - Reading and recreation: 0.255*** (standard error (0.0786))
- Pop. Growth (coefficients by CPI category):
  - Overall inflation: 0.304** (standard error (0.143))
  - Food and beverages: 0.517** (standard error (0.216))
  - Housing: 0.897*** (standard error (0.275))
  - Fuel, light and water charges: -1.693** (standard error (0.685))
  - Furniture and household utensils: -0.0214 (standard error (0.534))
  - Clothes and footwear: -0.892* (standard error (0.469))
  - Medical care: 0.685** (standard error (0.309))
  - Transportation and communication: -0.367* (standard error (0.199))
  - Education: 0.192 (standard error (0.387))
  - Reading and recreation: -0.0773 (standard error (0.379))
- Regional output gap (coefficients by CPI category):
  - Overall inflation: 0.0506*** (standard error (0.00793))
  - Food and beverages: 0.0652*** (standard error (0.0160))
  - Housing: -0.0189 (standard error (0.0224))
  - Fuel, light and water charges: 0.168*** (standard error (0.0277))
  - Furniture and household utensils: 0.116*** (standard error (0.0299))
  - Clothes and footwear: 0.181*** (standard error (0.0368))
  - Medical care: -0.0525*** (standard error (0.0120))
  - Transportation and communication: 0.0489*** (standard error (0.00771))
  - Education: -0.00533 (standard error (0.0165))
  - Reading and recreation: 0.0272** (standard error (0.0119))
- Lagged inflation rate (coefficients by CPI category):
  - Overall inflation: 0.233*** (standard error (0.0353))
  - Food and beverages: -0.000274 (standard error (0.0419))
  - Housing: 0.343*** (standard error (0.0379))
  - Fuel, light and water charges: 0.0941*** (standard error (0.0283))
  - Furniture and household utensils: 0.141*** (standard error (0.0502))
  - Clothes and footwear: 0.229*** (standard error (0.0403))
  - Medical care: -0.271*** (standard error (0.0265))
  - Transportation and communication: 0.149*** (standard error (0.0363))
  - Education: 0.325*** (standard error (0.0341))
  - Reading and recreation: 0.163*** (standard error (0.0263))
- Note: Arellano-Bond estimator used with one lag of dependent variable; robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Dynamic projections and interpretation
- Historical contribution:
  - "Shifts in the age distribution reduced Japan’s annual TFP growth by around 0.7–0.9 percent between 1990 and" (text ends in source).
  - Magnitude observed is similar to the productivity gap between Japan and U.S. circa 2005 (source text context).
- Age composition and TFP growth projections (1995–2035 references):
  - Large negative impact on TFP growth mainly due to decline in the 40-year age group between 1990 and 2005.
  - Subsequent increase in 40-year age group after 2010 should have a positive but declining effect on TFP growth.
  - After 2025, shifts in the age distribution projected to again have a negative impact on TFP growth.
- Inflation dynamics (1995–2035 references):
  - Deflationary pressure from aging more than offset by population growth in first half of 1990s.
  - In 2000s, slowing population growth plus increased speed of aging began to exert overall deflationary pressure.
  - UN-based projections suggest shrinking population will overtake aging as the dominant deflationary demographic source going forward.
- Caveat: projected impacts beyond 2005 subject to concern that coefficients may be specific to sample period.

### Interpretation of mechanisms and monetary policy interaction
- Age composition of working age population helps explain low TFP growth in Japan, 1990–2007.
- Sharp increase in speed of aging and slow population growth contributed to deflationary pressures.
- Monetary policy could in principle offset such effects on national inflation, but:
  - Short-term nominal interest rates were close to or at their lower effective bound during this period, constraining Bank of Japan’s ability to stimulate the economy.
  - If aging and slower population growth lowered the real neutral interest rate, demographic forces further hampered monetary policy’s ability to reflate the economy.
- Medical care price puzzle: medical care prices in Japan largely regulated by the government — possible explanation for unexpected coefficient signs in medical care CPI results.

### Policy implications and recommendations
- Policies to move the Japanese economy beyond a "muddle-through" status must address underlying demographic headwinds.
- Structural reform agenda under Prime Minister Abe’s administration gives demographics a high priority.
- More research needed to identify and better understand mechanisms through which demographic factors affect productivity and inflation.

*Source: _wp16237 - 1. Aging and TFP Growth in G7 Countries (excerpt), IMF staff calculations, figures and tables as provided in the source content.*

### 1. Aging and TFP Growth in G7 Countries ...............................................................................4

### 1. Aging and TFP Growth in G7 Countries

### Introduction and context
- Japan has been experiencing a declining working age population since the early 1990s; total population peaked in 2010 at 127 million.
- UN projections cited in the source: overall population expected to reach 119 million by 2030, the same level as in 1984.
- Japan’s old age dependency ratio has more than doubled since 1990 and is expected to exceed 50 percent by 2030.
- The paper assesses the contribution of Japan’s aging and shrinking population to low TFP growth and inflation using prefectural panel data for 1990–2007.

### Macroeconomic and demographic outcomes for G7 economies (1990–2007) — key figures from Table 1
- Annual Inflation:
  - Canada: 2.3
  - France: 2.0
  - Germany: 2.3
  - Italy: 3.9
  - Japan: 0.4
  - U.K.: 2.6
  - U.S.: 3.3
  - Average: 2.4
- Annual Real GDP Growth:
  - Canada: 2.6
  - France: 1.9
  - Germany: 1.7
  - Italy: 1.5
  - Japan: 1.3
  - U.K.: 2.9
  - U.S.: 3.0
  - Average: 2.1
- Contributions to Real GDP Growth — Labor:
  - Canada: 0.9
  - France: 0.4
  - Germany: 0.2
  - Italy: 0.3
  - Japan: 0.4
  - U.K.: 0.3
  - U.S.: 0.8
  - Average: 0.5
- Contributions to Real GDP Growth — Capital stock:
  - Canada: 1.1
  - France: 0.8
  - Germany: 0.7
  - Italy: 0.7
  - Japan: 0.8
  - U.K.: 1.0
  - U.S.: 1.3
  - Average: 0.9
- Contributions to Real GDP Growth — Total factor productivity:
  - Canada: 0.9
  - France: 0.7
  - Germany: 0.8
  - Italy: 0.5
  - Japan: 0.2
  - U.K.: 1.7
  - U.S.: 1.0
  - Average: 0.8
- Output per Worker (labor productivity):
  - Canada: 1.5
  - France: 1.2
  - Germany: 1.4
  - Italy: 1.0
  - Japan: 0.8
  - U.K.: 2.5
  - U.S.: 1.8
  - Average: 1.4
- Contributions to Output per Worker — Capital/labor ratio:
  - Canada: 0.6
  - France: 0.5
  - Germany: 0.6
  - Italy: 0.5
  - Japan: 0.6
  - U.K.: 0.8
  - U.S.: 0.8
  - Average: 0.6
- Demographic outcomes:
  - Aging (change in old age dependency ratio):
    - Canada: 2.8
    - France: 4.1
    - Germany: 8.6
    - Italy: 8.3
    - Japan: 15.8
    - U.K.: -0.9
    - U.S.: -0.4
    - Average: 5.5
  - Annual Population growth:
    - Canada: 1.1
    - France: 0.5
    - Germany: 0.2
    - Italy: 0.3
    - Japan: 0.2
    - U.K.: 0.3
    - U.S.: 1.11
    - Average: 0.5

### Empirical approach and regional variation
- The study exploits prefectural panel data for Japan covering 1990–2007 to avoid cross-country institutional heterogeneity.
- Inter-regional variation is used to estimate the impact of changes in the age distribution of the working age population on TFP and inflation.

### Main empirical findings on TFP and age distribution
- A significant relationship exists between changes in age distribution of the working age population and the level of total factor productivity.
- There is an inverted U-shaped productivity profile across age groups, with those aged 40–49 being the most productive.
- Prefectures aging at a faster pace tend to have relatively lower overall inflation; prefectures with higher population growth typically have higher relative inflation.

### Quantitative impacts and projections
- Using past and projected national data on demographics, changes in the distribution of the working age population depressed annual total factor productivity growth by approximately 0.7–0.9 percent between 1990 and 2007, primarily reflecting a shrinking share of the 40-year-age group.
- Deflationary pressures from aging and population growth over 1990–2007 were relatively mild.
- Extending results to 2035 suggests that deflationary pressures from the shrinking population—as opposed to aging—will become more prominent in the future.

### Interpretation and implications
- Japan’s weak growth performance in output per worker appears rooted in depressed TFP growth rather than capital accumulation.
- A strong negative relationship between aging and TFP growth is apparent across G7 countries for 1990–2007, with Japan as an outlier at the extreme end.
- The findings support the notion that demographic headwinds have and will play an important role with respect to total factor productivity and deflationary pressures in Japan.

*Source: _wp16237 - 1. Aging and TFP Growth in G7 Countries (excerpt), IMF staff calculations, figures and tables as provided in the source content.*

### Section VI concludes.

### _wp16237 - Section VI concludes.

### II. Linking demographics with TFP growth and inflation — key literature and mechanisms
- Demographics ↔ productivity mechanisms:
  - Older workers: higher productivity via accumulation of work experience.
  - Younger workers: better health, higher processing speed, ability to adjust to rapid technological changes, and greater entrepreneurship → more innovation.
  - Net effect may produce an inverted U-shaped pattern between age and productivity.
  - Aging may increase relative demand for services (e.g., health care), causing a sectorial shift toward more labor intensive and less productive services.
  - Population size/density may affect productivity via technological change, agglomeration and network effects, learning by doing, and innovative activity.
- Selected empirical findings:
  - Feyrer (2007): inverted U-shaped relationship; age-group 40–49 most productive.
  - Aiyar et al (2016): increase in share of workers aged 55–64 leads to a significant and permanent increase in TFP growth (European countries).
  - Mestas et al (2016): using U.S. states, two-thirds of the GDP per capita impact of old age dependency is due to slower productivity across the entire age distribution and one-third due to slower labor force growth.
  - Ciccone and Hall (1996) and Ciccone (2002): strong positive relationship between productivity and employment density for selected European countries.
- Demographics ↔ inflation channels and evidence:
  - Ambiguous net effect: population growth raises demand but also supply via greater labor availability; inflationary pressure depends on supply responsiveness (land, housing, natural resources).
  - Expectations channel (Shirakawa): agents anticipate lower future growth → cut current consumption/investment → deflationary pressures.
  - Alternative channel (Juselius and Takátsin, 2015): aging constrains production but dissaving by retirees keeps demand stable → excess demand → inflationary pressure.
  - Countervailing force (Andersson, Botman and Hunt, 2014): repatriation of foreign savings by retirees → currency appreciation → lower inflation (Japan example).
  - Political economy: older cohorts prefer lower inflation and may gain political influence (Bullard et al, 2012), potentially affecting monetary policy outcomes.
  - Empirical evidence mixed:
    - Yoon et al (2014): declining and aging population → significant deflationary impact (OECD cross-country).
    - Juselius and Takats (2015): larger share of dependents → higher inflation; larger share of working age cohorts → lower inflation (panel of 22 advanced countries).
  - Distinction emphasized: whether an older population is associated with a higher price level or with a higher inflation level drives divergent results.
- Secular stagnation nexus:
  - Secular stagnation (Hansen 1938; Summers 2013): structural excess savings, lack of investment → very low real neutral interest rate → monetary policy constrained → economy below potential → low inflation and low growth.
  - Demographics may influence saving/investment propensities: aging and negative population growth may reduce corporate willingness to invest; increased longevity, uncertainty over public debt and social security, and bequest motives may increase savings.
  - Gagnon et al (2016): changes in U.S. population growth, family composition, life expectancy, and labor market activity accounted for a 1.25 percentage point decline in the equilibrium real interest rate since 1980 in the U.S.

### III. Methodology — empirical specifications
- Total factor productivity model (based on Feyrer 2007):
  - Equation (1):
    - log TFP_{r,t} = α1 W20_{r,t} + α2 W30_{r,t} + α3 W50_{r,t} + α4 W60_{r,t} + β1 OADDR_{r,t} + β2 log DWA_{r,t} + β3 SER_{r,t} + δ_r + θ_t + u_{r,t}
  - Definitions and choices:
    - TFP_{r,t}: level of total labor force productivity for prefecture r at time t.
    - W20, W30, W50, W60: share of 10-year age groups of the working age population (ages 20–69) in prefecture r at time t. W40 is reference group and excluded.
    - OADDR: old age dependency ratio = population aged 70+ as share of working age population.
    - DWA: density of the working age population (number of people per unit of area).
    - SER_{r,t}: size of service sector as share of nominal prefectural GDP.
    - Includes prefecture fixed effects δ_r and common time trend θ_t.
  - Estimation concerns and approaches:
    - Serial correlation and potential unit roots → estimate in first differences and include a lagged dependent variable.
    - Standard errors robust to heteroskedasticity and clustered by prefecture.
    - Use of working age population (20–69) instead of workforce to avoid endogeneity from differential labor force participation responses to productivity shocks.
    - Endogeneity from migration acknowledged; net migration between 1990 and 2007 averaged between 0.1 and 0.4 percent across prefectures; 18–19 percent of migrants between 1996–2006 moved for work reasons. Arellano-Bond GMM estimator used for lagged dependent model.
- Inflation model:
  - Equation (2):
    - π_{r,t} = α1 ΔOADDR_{r,t} + α2 Δlog Pop_{r,t} + β OPG_{r,t} + γ π_{r,t−1} + δ_r + θ_t + u_{r,t}
  - Definitions:
    - π_{r,t}: annual inflation for prefecture r at time t.
    - ΔOADDR: change in old age dependency ratio.
    - Δlog Pop: change in prefectural population.
    - OPG_{r,t}: prefectural output gap (HP-filtered real prefectural output), capturing regional excess demand.
    - π_{r,t−1}: lagged inflation to capture persistence.
  - Estimation:
    - Equation (2) interpreted as a regional Phillips curve with backward-looking expectations.
    - Also estimated for CPI subcomponents.
    - Arellano-Bond GMM estimator used; standard errors robust and clustered by prefecture.

### IV. Data
- Sample period: 1990 to 2007.
- Sources:
  - Prefectural real growth, capital stock, labor input (hours worked): Regional-Level Japan Industrial Productivity (R-JIP).
  - Prefectural TFP: derived via simple growth accounting following Kyoji, Tatsuji and Joji (2015).
  - Prefectural inflation and components, annual prefectural age distributions: Portal Site of Official Statistics of Japan.
  - Population data available in 5-year age groups; analysis uses 10-year age groups.
- Inter-prefectural variation examples (1990–2007):
  - Akita: working age population drop by over 13 percent; old age dependency ratio increase by 20 percentage points.
  - Saitama: working age population rise by 16 percent; old age dependency ratio increase by 9 percentage points.
  - Kagoshima: TFP grew by 23 percent.
  - Nara: TFP contracted by 11 percent.
  - Aomori: CPI level rose by 14 percent.
  - Saga: CPI level rose by 4 percent.
- Data patterns:
  - Median prefectural inflation shows general downward trend with temporary deflation episodes; spike in 1997 reflects consumption tax hike.
  - TFP growth fell significantly in early 1990s (negative) then returned to positive in 2000s.
  - Old age dependency ratio steadily increased; population growth slowed; several prefectures experienced negative population growth.
  - Age distribution shifts: fall in 20-year age group starting in 1995; rise in 60-year age group in the 2000s.
- Figure 3 (described): shows prefectural variation and trends for TFP growth, inflation, old-age dependency ratio, population growth, and aging of working age population; shaded areas reflect minimum and maximum across prefectures and over time.

### V. Results — TFP and inflation regressions and dynamic projections
- Total factor productivity regressions (Table 2) — principal findings:
  - The age distribution of the working age population had a significant impact on TFP.
  - First difference model: clear inverted U-shaped productivity pattern; excluded age group 40–49 is the most productive.
    - Point estimates (interpretation examples in text):
      - "A one percentage point shift from the 30-year age group to the 40-year age group increases the level of total factor productivity by close to 4.4 percent."
      - "A similar shift from the 40 to the 50-year age group decreases productivity by 1.3 percent."
  - Lagged dependent variable model: weaker but still shows negative statistical significance for the 20-year and 60-year groups relative to 40-year age group.
  - Some estimates roughly similar to Feyrer (2007), particularly for 60-year age group.
  - Old age dependency ratio: estimated economic impact negative and relatively large but statistically insignificant.
  - Population density: no evidence that density matters for TFP.
  - Size of service sector: prefectures with larger service sectors have on average lower overall TFP (consistent across specifications).
- Table 2 (selected coefficients reproduced exactly as in source table):
  - Age group 20-29: -1.174*** (First difference), -2.225*** (Feyrer (2008)), -3.169** (Lagged dependent variable) with standard errors (0.320), (0.544), (1.297).
  - Age group 30-39: -4.382***, -1.039, -3.828*** with standard errors (0.988), (0.833), (1.206).
  - Age group 50-59: -1.256, -0.475, -2.120** with standard errors (0.884), (0.563), (1.051).
  - Age group 60-69: -2.221**, -1.593*, -2.066* with standard errors (0.989), (0.885), (1.211).
  - Old-age dep. ratio: -0.788, -0.220, 0.478 with standard errors (0.600), (0.675), (1.001).
  - Pop. density: -0.0554, 0.141, ... with standard errors (0.331), (0.263), ...
  - Size of service sector: -2.245***, -1.595***, ... with standard errors (0.239), (0.249), ...
  - Lagged TFP: ..., 0.758***, ... with standard error (0.0336).
  - Time fixed effects: Yes, Yes, Yes (first two); no. observations: 799, 799, 499; no. prefectures: 47, 47, 87.
  - Note: The Arellano-Bond estimator is used using one lag of dependent variables as covariates. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
- Inflation regressions (Table 3) — principal findings:
  - Aging and population growth have significant impacts on overall inflation:
    - "A one percent increase in the old age dependency ratio reduces the inflation rate by 0.1 percentage points."
    - "A one percent increase in population growth increases inflation by 0.3 percentage points."
  - Time fixed effects control for national inflation trends and uniform BoJ monetary policy effects across prefectures.
  - CPI subcomponents: results are not consistent across categories:
    - Increase in old age dependency ratio is negatively related to cost of food and beverages and negatively related to cost of medical care (contrary to simple expectations).
    - Population growth positively related to cost of housing and food and beverages, but negatively related to energy costs, clothes and footwear, and transportation and communication.
- Dynamic projections using UN data (1990–2035) and estimated coefficients:
  - Apply estimated coefficients from first difference (FD) and lagged dependent variable (LDV) models to realized and forecasted 5-year changes in working population distribution.
  - Figure 4 and Figure 5 depict projected impacts on annual TFP growth and inflation, respectively.
  - Caveat: projected impacts beyond 2005 subject to usual concern that coefficients may be specific to sample period.
  - Historical contribution estimate:
    - "Shifts in the age distribution reduced Japan’s annual TFP growth by around 0.7–0.9 percent between 1990 and" (text ends here).

*Italic source attribution: _wp16237 - Section VI concludes.*

### 2005. Interestingly, the magnitude is similar to the productivity gap between Japan and U.S

### _wp16237 - 2005. Interestingly, the magnitude is similar to the productivity gap between Japan and U.S

### Age composition and TFP growth
- Large negative impact on TFP growth was mainly due to the decline in the 40-year age group—the most productive—between 1990 and 2005.
- The subsequent increase in the 40-year age group after 2010 should have a positive but declining effect on TFP growth.
- After 2025, shifts in the age distribution will again have a negative impact on TFP growth.
- Figure reference points:
  - Change in 40-year age group shown for 1995–2035.
  - Annual TFP growth presented as (FD) and (LDV) in Figure 4.

### Demographic impact on inflation dynamics
- Deflationary pressure from aging was more than offset by population growth in the first half of the 1990s.
- In the 2000s, the combination of slowing population growth and the increased speed of aging started to exert overall deflationary pressure.
- Going forward, based on UN projections, the shrinking of the population will overtake aging as the most prominent deflationary demographic source.
- Figure reference points:
  - Old age dependency ratio, Population growth, Lagged inflation, and Annual impact on inflation displayed for 1995–2035 in Figure 5.

### Empirical results (Table 3: Effect of Aging and Population Growth on Inflation)
- Estimation method: The Arellano-Bond estimator is used using one lag of dependent variables as covariates. Robust standard errors in parentheses.
- Sample dimensions:
  - Observations: 752
  - No. prefectures: 47
- Key coefficient estimates (dependent variable columns correspond to: Overall inflation; Food and beverages; Housing; Fuel, light and water charges; Furniture and household utensils; Clothes and footwear; Medical care; Transportation and communication; Education; Reading and recreation):
  - Δ Old-age dep. ratio:
    - Overall inflation: -0.110***
    - Food and beverages: -0.385***
    - Housing: 0.0179
    - Fuel, light and water charges: -0.144
    - Furniture and household utensils: 0.306*
    - Clothes and footwear: 0.0229
    - Medical care: -0.192**
    - Transportation and communication: 0.0829*
    - Education: 0.105
    - Reading and recreation: 0.255***
    - Robust standard errors (Overall inflation): (0.0401); (Food and beverages): (0.0845); (Housing): (0.114); (Fuel...): (0.139); (Furniture...): (0.165); (Clothes...): (0.168); (Medical care): (0.0902); (Transportation...): (0.0455); (Education): (0.0853); (Reading...): (0.0786)
  - Pop. Growth:
    - Overall inflation: 0.304**
    - Food and beverages: 0.517**
    - Housing: 0.897***
    - Fuel, light and water charges: -1.693**
    - Furniture and household utensils: -0.0214
    - Clothes and footwear: -0.892*
    - Medical care: 0.685**
    - Transportation and communication: -0.367*
    - Education: 0.192
    - Reading and recreation: -0.0773
    - Robust standard errors: (0.143), (0.216), (0.275), (0.685), (0.534), (0.469), (0.309), (0.199), (0.387), (0.379)
  - Regional output gap:
    - Overall inflation: 0.0506***
    - Food and beverages: 0.0652***
    - Housing: -0.0189
    - Fuel, light and water charges: 0.168***
    - Furniture and household utensils: 0.116***
    - Clothes and footwear: 0.181***
    - Medical care: -0.0525***
    - Transportation and communication: 0.0489***
    - Education: -0.00533
    - Reading and recreation: 0.0272**
    - Robust standard errors: (0.00793), (0.0160), (0.0224), (0.0277), (0.0299), (0.0368), (0.0120), (0.00771), (0.0165), (0.0119)
  - Lagged inflation rate:
    - Overall inflation: 0.233***
    - Food and beverages: -0.000274
    - Housing: 0.343***
    - Fuel, light and water charges: 0.0941***
    - Furniture and household utensils: 0.141***
    - Clothes and footwear: 0.229***
    - Medical care: -0.271***
    - Transportation and communication: 0.149***
    - Education: 0.325***
    - Reading and recreation: 0.163***
    - Robust standard errors: (0.0353), (0.0419), (0.0379), (0.0283), (0.0502), (0.0403), (0.0265), (0.0363), (0.0341), (0.0263)
- Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

### Interpretation and mechanisms
- The age composition of Japan’s working age population helps explain the low TFP growth between 1990 and 2007.
- The sharp increase in the speed of aging and slow population growth contributed to deflationary pressures.
- Monetary policy in principle could offset such effects on national inflation, but:
  - Short-term nominal interest rates were close to or at their lower effective bound during this period, severely constraining the Bank of Japan’s ability to stimulate the economy.
  - If aging and slower population growth lowered the real neutral interest rate, these demographic forces further hampered the ability of monetary policy to reflate the economy.
- One possible reason for the puzzling result regarding medical care prices is that medical care prices in Japan are largely regulated by the government.

### Policy implications and recommendations
- Policies designed to push the Japanese economy out of its current muddle-through status quo must address underlying demographic headwinds.
- Structural reform agenda under Prime Minister Abe’s administration gives demographics a high priority.
- To enhance these efforts, more research is needed to identify and better understand the mechanism through which demographic factors affect productivity and inflation.

*Source: Excerpt from IMF working paper content provided in the supplied PDF text.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16237.pdf_