## _wp13191

## Source details

**Canonical URL:** [_wp13191](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13191.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13191.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13191.pdf.json)

---

### Introduction and central hypothesis
- Context:
  - Inflation targeting emerged as the dominant monetary framework prior to the global financial crisis; many central banks formally adopted IT since the mid-1990s.
  - Some central banks have formalized 2 percent inflation objectives.
- Empirical puzzle:
  - Recent studies document that monetary policy has had a more benign impact on unemployment and inflation since the mid-1980s.
  - IMF (2013) finds inflation remains stuck around 2 percent throughout business cycles in the U.S. and is not responsive to changes in the output gap.
- Central hypothesis:
  - A demographic shift to an older society ("graying") can weaken monetary policy effectiveness via life-cycle behavior: older societies have a larger share of creditor households and are less sensitive to interest rate changes; younger societies have more debtors and higher sensitivity.
- Research approach:
  - Bayesian estimation of time-varying impacts of monetary policy on unemployment and inflation for the U.S., Canada, Japan, U.K., and Germany.
  - Panel co-integration tests between ageing and weakening monetary policy effectiveness.
  - Dynamic panel OLS to attribute weakening effectiveness to demographic changes.
- Policy implication preview:
  - "To obtain the same impact on output and inflation, ceteris paribus, interest rate changes will have to become larger in older societies than in younger ones."

### Transmission channels and theory
- Channels through which ageing may alter monetary effectiveness (ambiguous net effect a priori):
  - Interest Rate Channel:
    - Life-cycle hypothesis: debt rises then falls over the life-cycle; young households (debtors) more sensitive to interest rate changes; old households (creditors) less sensitive.
    - Practical caveat: poor older households are often credit constrained and banks reluctant to lend.
  - Credit Channel:
    - External finance premium inversely related to borrower net worth.
    - In ageing societies, greater net worth among the elderly implies more self-financing and lower external finance premia, reducing credit channel sensitivity.
    - Old age poverty and credit rationing can further weaken the credit channel.
  - Wealth Effect Channel:
    - Wealth tends to be concentrated among the elderly in OECD countries; older households hold more interest-sensitive fixed-income products.
    - Graying raises relative importance of the wealth effect channel, which could increase monetary effectiveness via larger wealth changes when rates move.
  - Risk-Taking Channel:
    - More important in young societies; less important in old societies due to higher risk aversion among older adults, reducing monetary policy potency.
  - Expectation Channel:
    - Conditional on central bank credibility; survey evidence suggests inflation expectations rise with age, implying older adults are more sensitive to inflation expectations.
    - The expectation channel could gain importance as ageing societies may push central banks to emphasize price stability.
  - Exchange Rate Channel:
    - Ambiguous: ageing may reduce private savings and worsen public finances; net effect on current account and exchange rate is country specific.

### Empirical strategy and data
- Model:
  - Time-varying coefficient vector autoregressive model (TVC-VAR) with stochastic volatility (SV).
  - Variables: inflation (π), unemployment rate (u), policy rate (r).
  - Time-varying constant vector and coefficient matrix; heteroskedastic shocks with SV.
  - Parameters follow random walks; diagonal components of volatility follow geometric random walks.
  - Identification: ordering π, u, r; monetary policy shock identified by assuming contemporaneous changes in the policy rate do not immediately impact inflation or unemployment; Cholesky decomposition used.
- Estimation:
  - Bayesian MCMC with Gibbs Sampler following Primiceri (2005).
  - Priors: data driven normal-inverse-Wishart conjugate priors; diffuse/uninformative per conventions.
  - Computational details: 10,000 runs of the Gibbs Sampler, discarding the first 2,000; only every other draw saved.
- Data:
  - Countries: U.S., U.K., Canada, Germany, Japan.
  - Frequency: quarterly.
  - Inflation: annual growth rate in the CPI.
  - Unemployment: civilian unemployment rate (seasonally adjusted).
  - Interest rate proxies:
    - U.S., U.K., Canada: yield on three-month treasury bills (policy rate used in robustness checks).
    - Germany, Japan: three-month money market rates.
  - Estimation sample: 1963Q1 to 2007Q1; Germany sample ends in 1999; Japan sample ends in 1995 (to avoid contamination).
- Dependent-variable proxies for monetary policy effectiveness:
  - (i) cumulative response of inflation and unemployment over the subsequent five years following a one percentage point shock to the interest rate;
  - (ii) corresponding maximum response over the same horizon.

### Time-varying empirical results — weakening effectiveness
- General patterns:
  - Point estimates show unemployment generally increases for all countries during the first 10 quarters after a one percentage point unanticipated increase in the interest rate.
  - Cross-country sensitivity differences:
    - The U.K. and Canada have unemployment rates up to four times more sensitive to interest rate changes than Japan.
  - Persistence differences:
    - Unemployment response more protracted in Canada, the U.K. and Japan—on average, more than 40 percent of the peak responses still remaining after 20 quarters—than in the U.S. and Germany.
- Decline in impact over time:
  - Noticeable decrease in the impact of monetary policy on unemployment for all countries (less visible for Canada).
  - Peak responses decline by around 0.2 percentage points for the U.S. and the U.K.; larger maximum decrease observable for Germany.
  - Japan and Canada exhibit smaller decreases in peak responses (circa 0.05-0.01).
  - Timing:
    - Conspicuous decrease starts around 1980 for the U.S. and Germany.
    - For the U.K., Canada and Japan, the decrease starts about a decade later.
- Probabilities that the maximum impulse response has decreased from peak to trough along the same sample path:
  - U.S.: 92 percent
  - U.K.: 72 percent
  - Germany: 81 percent
  - Canada: 70 percent
  - Japan: 95 percent
- Conclusion:
  - Fairly strong evidence for a decrease in the effectiveness of monetary policy across the sample of countries.

### Evidence on inflation responses
- Decline in impact on inflation:
  - Peak responses decrease by about 0.2 percentage points for Canada and Germany; decrease somewhat less for the remaining countries.
- Probability that the minimum impulse response has increased (comparing estimates along the same sample path):
  - U.S.: 63 percent
  - U.K.: 77 percent
  - Germany: 68 percent
  - Canada: 67 percent
  - Japan: 83 percent
- Observed timing of large changes in inflation responses corresponds roughly to dates identified for unemployment, suggesting a common factor driving time-variation in both responses.

### Panel regression and long-run estimates attributing decline to ageing
- Panel regression specification (equation (16)):
  - Impact of monetary policy on variable j at time t in country i modeled as function of:
    - Old-age dependency ratio (people above age 65 relative to those between 15 and 64)
    - Share of manufacturing in total production (Share of Manu.)
    - Degree of economic openness (Openness = total trade/GDP)
    - Share of small firms (employment in firms with <50 employees)
    - Private sector credit (Credit-to-GDP)
- Time-series properties:
  - Panel unit-root tests indicate series are I(1); first-differences achieve stationarity.
  - Kao (1999) panel co-integration test used.
- Panel unit-root test p-values (asymptotic reported) for series:
  - Cum_U: Levin, Lin & Chu t-stat 0.20; Im, Pesaran & Shin W-stat 0.64; ADF - Fisher Chi-square 0.68; PP - Fisher Chi-square 1.00
  - Max_U: 0.11; 0.20; 0.20; 0.99
  - Cum_Pi: 0.21; 0.40; 0.48; 0.98
  - Max_Pi: 0.12; 0.36; 0.40; 0.95
  - Old-Age: 0.33; 0.86; 0.13; 0.33
  - Openness: 0.32; 0.76; 0.89; 0.90
  - Manufacturing: 0.39; 0.23; 0.04; 0.29
  - Credit: 0.38; 0.23; 0.33; 0.80
- Long-run estimates (D-OLS with one lead and one lag; Sieve bootstrapped standard errors; Number of Obs. = 173):
  - Kao (1999) CI-test statistics and p-values (P-value in squared brackets):
    - Equation (1): -2.122** [0.02]
    - Equation (2): -2.820*** [0.00]
    - Equation (3): -3.217** [0.00]
    - Equation (4): -2.388** [0.00]
- Selected estimated D-OLS coefficients (standard errors in parentheses). Dependent variables: Equation (1) Cum π; Equation (2) Cum u; Equation (3) Max π; Equation (4) Max u.
  - Constant:
    - Equation (1): -25.271*** (3.175)
    - Equation (2): -25.431*** (3.176)
    - Equation (3): -0.423* (0.176)
    - Equation (4): -1.688*** (0.210)
  - Old-age Ratio:
    - Equation (1): 0.073* (0.044)
    - Equation (2): -0.350*** (0.045)
    - Equation (3): 0.013*** (0.002)
    - Equation (4): -0.019*** (0.003)
  - Share of Manu.:
    - Equation (1): 3.634*** (0.695)
    - Equation (2): 10.129*** (0.697)
    - Equation (3): -0.037 (0.039)
    - Equation (4): 0.659*** (0.046)
  - Openness:
    - Equation (1): 3.414*** (0.520)
    - Equation (2): 0.381 (0.521)
    - Equation (3): 0.001 (0.029)
    - Equation (4): 0.021 (0.034)
  - Credit-to-GDP:
    - Equation (1): -0.114 (0.195)
    - Equation (2): 0.958*** (0.195)
    - Equation (3): 0.010 (0.010)
    - Equation (4): 0.071*** (0.013)
  - Target Change:
    - Equation (1): 0.290 (0.205)
    - Equation (2): 0.444** (0.205)
    - Equation (3): 0.029** (0.011)
    - Equation (4): 0.026** (0.014)
- Interpretation of ageing coefficients:
  - All else equal, an increase in the old-age dependency ratio of one point lowers (in absolute terms) the cumulative impact of a monetary policy shock on:
    - Inflation by 0.10 percentage points
    - Unemployment by 0.35 percentage points
  - The corresponding maximum impact of monetary policy is lowered by:
    - Inflation: 0.01 percentage points
    - Unemployment: 0.02 percentage points
  - Example implication: a projected 10 point rise in the old-age dependency ratio in Germany over the next decade would imply substantially weaker monetary transmission to inflation and unemployment.
- Other variable interpretations:
  - Manufacturing share: larger manufacturing shares associated with greater responsiveness of unemployment to monetary policy; mixed effects on inflation responsiveness.
  - Openness: associated with weaker effect on inflation; impact on unemployment statistically insignificant.
  - Credit-to-GDP: higher credit associated with greater real-sector responsiveness to monetary policy (equation (2): 0.958***).

### Robustness checks and caveats
- Robustness:
  - Alternative specifications tested: use of GDP deflator rather than CPI inflation; spare capacity rather than unemployment; policy rate rather than 3-month rates.
  - Result: IRFs to a 100 basis point increase in interest rates are robust; impact on inflation and unemployment found to be weakening for most countries.
  - Exception: U.K. results are an exception, potentially due to numerous policy regime changes affecting credibility over the last four decades.
- Caveats:
  - The study focuses on net aggregate effects rather than isolating each individual transmission channel.
  - Demographic change is one of several explanations (others include institutional changes in credit markets and better-anchored expectations); demographic effects can coexist with these factors.
  - Estimation addresses some biases by using a time-varying constant term to control for changes in real interest rates and implied inflation targets.

### Policy implications and recommendations
- Main policy implications:
  - Monetary policy will have to become more "activist" in ageing societies, with larger interest rate changes required to achieve the same effects on output and inflation.
  - With weakened monetary transmission, fiscal and macroprudential policy may need to assume a larger role in stabilizing the economy and the financial system.
- Specific conjectures:
  - Optimal policy trade-offs may change:
    - Older households' larger asset holdings increase aversion to unexpected inflation, which could, ceteris paribus, pull optimal inflation targets lower.
    - A declining real interest rate (a consequence of ageing) could work in the opposite direction; central banks must weigh these forces.
  - Conventional small policy-rate steps (e.g., 25bps) may need to be amplified in graying societies to obtain the same real effects.
  - Non-monetary or structural responses:
    - Policies that influence demographics (e.g., higher fertility or migration) could enhance monetary policy effectiveness but are neither risk-free nor panaceas.

### Broader implications and demographic context
- Macro effects of ageing:
  - Graying is expected to reduce real interest rates and diminish long-term growth potential.
  - Ageing may put downward pressure on prices and help explain deflationary tendencies in graying societies.
  - Increased demand for services by the elderly (where prices are stickier) implies prices may be less flexible but potentially less inflationary.
- EMs and LICs:
  - Advanced Economies (AEs) will be the first to experience rapid ageing; EMs and LICs will age more gradually.
  - EMs and LICs often age without equivalent rises in wealth; wealth concentration among the elderly is less pronounced, so the dominance of wealth effects over credit-channel effects may be less important there.
  - Implication: monetary policy effectiveness in EMs and LICs may not weaken as much as in AEs, or may operate through different channels.
- Demographic drivers:
  - Longevity gains driven by lifestyle changes, better nutrition, medical advances, and improved education.
  - The cohort of old (65+) and very old (80+) is growing faster than the young in AEs.
  - Given persistent low fertility and restrictive immigration policies in many AEs, significant graying going forward is considered largely unavoidable.

*Source: WP/13/191, "Shock from Graying: Is the Demographic Shift Weakening Monetary Policy Effectiveness", Prepared by Patrick Imam, September 2013.*

### Section 1

### _wp13191 - Section 1

### Introduction: motivation and central hypothesis
- Context:
  - Prior to the global financial crisis, inflation targeting (IT) emerged as the dominant monetary framework; trend since the mid-1990s among many central banks was to formally adopt IT.
  - Some central banks have formalized 2 percent inflation objectives (e.g., the U.S. Fed in early 2012; the Bank of Japan in early 2013).
- Empirical puzzle:
  - Recent studies (survey by Boivin and others, 2010) document that monetary policy has had a more benign impact on unemployment and inflation since the mid-1980s.
  - IMF (2013) finds inflation remains stuck around 2 percent throughout business cycles in the U.S. and is not responsive to changes in the output gap.
- Additional explanation proposed:
  - Demographic shift to an older society ("graying") can weaken monetary policy effectiveness via life-cycle behavior: older societies have a larger share of creditor households and are less sensitive to interest rate changes; younger societies have more debtors and higher sensitivity.
- Research approach summarized:
  - Use Bayesian estimation methods to estimate the time-varying impact of monetary policy on unemployment and inflation for the five biggest advanced economies with independent monetary policies: U.S., Canada, Japan, U.K., and Germany.
  - Test for a panel co-integration relationship between ageing and weakening monetary policy effectiveness.
  - Use dynamic panel OLS techniques to attribute weakening effectiveness to demographic changes.
- Key policy implication preview:
  - To obtain the same impact on output and inflation, ceteris paribus, interest rate changes will have to become larger in older societies than in younger ones.

### Literature review: theory and prior evidence
- Theoretical background and ambiguity:
  - Miles (2002), using an overlapping generation model, highlights ambiguity: ageing could raise monetary effectiveness via wealth concentration among elderly or reduce it via reduced credit-constraint and weaker credit channel.
  - Kara and von Thadden (2010) calibrate a small-scale DSGE with demographics for the euro area and conclude demographic changes are too slow to materially affect monetary policy effectiveness over typical policymaking horizons.
  - Bean (2004) suggests the "glacial nature of demographic change" implies modest implications for monetary policy.
  - Fujiwara and Teranishi (2007), using a dynamic new Keynesian model with life-cycle behavior, find monetary policy effectiveness depends on demographic structure but do not resolve the net effect.
- Empirical gap:
  - Little empirical evidence exists on differential effects of monetary policy across cohorts or on whether monetary policy effectiveness changes in ageing societies.

### Transmission channels through which ageing may alter monetary effectiveness
- Overview:
  - Several channels can change as societies age; some channels strengthen, others weaken, producing ambiguous net effects a priori.
- Interest Rate Channel:
  - Life-cycle hypothesis (Modigliani, 1970): individuals accumulate assets during working life and dissave in retirement; debt rises then falls over the life-cycle.
  - Young households (typically debtors) are more sensitive to interest rate changes; old households (typically creditors) are less sensitive.
  - Practical note from the source: theoretically an unexpected temporary change in interest rate spreads over fewer periods for older people, implying potentially larger per-period impact if consumption smoothing is present; but in practice, poor older households are often credit constrained and banks reluctant to lend (see Seidman and Lewis, 1999).
- Credit Channel:
  - External finance premium is inversely related to borrower net worth (Bernanke and Gertler, 1989).
  - In ageing societies, older households with greater net worth rely more on self-financing; lower external finance premia reduce the sensitivity of the credit channel to monetary policy.
  - Old age poverty and credit rationing can limit borrowing for some elderly, reinforcing weaker credit channel sensitivity.
- Wealth Effect Channel:
  - Demographic shifts affect asset prices through life-cycle behavior (table introduced the channel and signals it as relevant; full elaboration continues beyond provided excerpt).
- Other channels summarized in the source table format:
  - Risk Taking Channel: More important in young societies; less important in old societies due to higher risk aversion among older adults.
  - Expectation Channel: Less important in young societies; more important in old societies because older adults are more sensitive to inflation expectations.
  - Exchange Rate Channel: Not clear—direction ambiguous across demographics.

### Empirical findings reported in Section 1 (preview and methodology)
- Economies analyzed:
  - U.S., Canada, Japan, U.K., and Germany.
- Methods previewed:
  - Time-varying vector autoregressive model with Bayesian estimation to obtain impulse response functions over time.
  - Panel co-integration tests to check for long-run relationship between ageing and monetary policy weakening.
  - Dynamic panel OLS to attribute weakening to demographic changes and run robustness checks.
- Key empirical claim from Section 1:
  - Using these techniques, the study finds a general weakening of the impact of monetary policy on unemployment and inflation over time, with both unemployment and inflation less responsive to interest rate changes as society grays.
- Quantitative policy-relevant statement:
  - "To obtain the same impact on output and inflation, ceteris paribus, interest rate changes will have to become larger in older societies than in younger ones."

### Implications and interpretation highlighted in Section 1
- Interaction with other explanations:
  - Other explanations for weakening monetary transmission include institutional changes in credit markets (e.g., securitization, shadow banking) and better-anchored expectations allowing "open mouth operations".
  - The paper offers demographic change as an additional, previously under-explored, explanation that can coexist with these factors.
- Broader note:
  - The global financial crisis reversed some institutional changes in the credit market and challenged central bank credibility in some countries; thus institutional explanations may not fully account for future dynamics—demographic effects can provide a persistent, structural contributor to weaker monetary transmission.
- Net-effect focus:
  - The paper focuses on testing the net effect of demographic change on monetary policy effectiveness; disentangling the relative importance of each individual channel is beyond its scope.

*Source: WP/13/191, "Shock from Graying: Is the Demographic Shift Weakening Monetary Policy Effectiveness", Prepared by Patrick Imam, September 2013.*

### Section 2

### _wp13191 - Section 2

### Wealth, demographic concentration, and monetary transmission
- Young individuals typically possess few assets, while older ones own many.
- When a household has acquired substantial assets, the impact of interest rate changes on its wealth are larger than when a household has few assets (see Iacoviello, 2005).
- In graying societies, wealth effects gain importance because wealth tends to be concentrated among the elderly (at least in OECD countries).
- Older households have portfolios typically more heavily made up of interest-sensitive fixed-income products rather than equities, accentuating sensitivity to interest rate changes via the wealth effect channel.
- The demographic shift tends to raise the relative importance of the wealth effect channel, increasing the effectiveness of monetary policy.
- Not all asset prices are equally impacted: riskier assets may be more sensitive to the demographic shift, and the effect on housing prices depends on the extent to which housing demand remains strong among the elderly.

### Other transmission channels affected by ageing
- Risk-Taking Channel:
  - Monetary policy influences risk-taking by encouraging the “search for yield” (see Borio and Zhu, 2008).
  - Financial entities have been found to take on more risk when interest rates fall (and less when interest rates rise).
  - In an ageing society, the time to recoup losses is lower (or less) than in a younger society, which may lead to more risk-averse households and less overall risk-taking.
  - Result: the risk-taking channel is likely less potent in a graying society, thus reducing monetary policy effectiveness.
- Expectation Channel:
  - The expectation channel should not be impacted as much by the demographic shift because it is conditional on central bank credibility.
  - Survey evidence suggests, ceteris paribus, inflation expectations rise with age, implying higher concern and risk aversion to inflation in a graying society (Blanchflower and MacCoille, 2009).
  - Older households may put more weight on worse inflation outcomes due to larger impact on them given their creditor status.
  - In practice, central banks may place greater emphasis on price stability, implying the expectation channel could gain importance as ageing societies lead central banks to employ monetary policy more aggressively to combat inflation.
- Exchange Rate Channel:
  - From the life-cycle hypothesis, an ageing population is expected to exert a negative impact on private savings: older cohorts lead to more dis-saving than saving, implying a decline in population savings.
  - Fiscal positions (public savings) should worsen with ageing due to declining revenues and higher expenditures.
  - Both public and private investments are expected to decline as the population ages.
  - There is no theoretical a priori on whether declines in savings or investment dominate; the effect on the current account and exchange rate is ambiguous and country specific.

### Caveats and broader macro effects of ageing
- The estimation that follows focuses on the aggregate effect rather than isolating individual channels.
- A graying population is expected to:
  - Reduce real interest rates.
  - Diminish long-term growth potential.
  - Impact international capital flows (e.g. Bean, 2004; Bloom et al., 2011; Poterba, 2004; Shirakawa, 2012).
- Ageing may have a direct impact on price developments:
  - Population ageing, especially with population decline, is likely to impact aggregate demand negatively, putting downward pressure on prices and explaining deflationary tendencies in graying societies (see Shirakawa, 2011).
  - Increased demand for services by an ageing population—where price changes tend to be stickier—implies prices are less flexible, but potentially also less inflationary.
- Estimation approach addresses some biases by using a time-varying constant term to control for changes in real interest rates and implied inflation targets.

### Model specification for time-varying impact of monetary policy
- Aim: estimate changing impact of monetary policy using a time-varying coefficient vector autoregressive model (TVC-VAR) with stochastic volatility (SV).
- Variables: inflation (π), unemployment rate (u), policy rate (r).
- Model elements:
  - Time varying constant vector.
  - Time-varying coefficient matrix.
  - Heteroskedastic shocks with covariance matrix modeled with stochastic volatility.
- Parameter dynamics:
  - Time-varying parameters are assumed to follow random walks.
  - Diagonal components of volatility follow geometric random walks.
- Identification:
  - Dependent variable ordering: π, u, r.
  - Monetary policy shock identified by assuming contemporaneous changes in the policy rate do not immediately impact inflation or unemployment.
  - Unemployment has no contemporaneous impact on inflation in the non-policy block.
  - Cholesky decomposition used to obtain structural shocks.

### Estimation technique and data
- Estimation method: Bayesian MCMC methods with Gibbs Sampler following Primiceri (2005).
- Priors: data driven normal-inverse-Wishart conjugate priors, diffuse/uninformative per literature conventions.
- Computational choices:
  - Simulate distribution of unknown parameters using MCMC.
  - Gibbs Sampler carried out in four steps (Carter and Kohn simulation smoother; Kim et al. method for volatility; inverse-Wishart draws for V).
- Data and sample:
  - Countries: U.S., U.K., Canada, Germany, Japan (five largest economies with independent monetary policy).
  - Frequency: quarterly.
  - Inflation measure: annual growth rate in the CPI.
  - Unemployment: civilian unemployment rate (seasonally adjusted).
  - Interest rate:
    - U.S., U.K., Canada: yield on three-month treasury bills (preferred for longer availability; policy rate used in robustness checks without significant change).
    - Germany, Japan: three-month money market rates (due to data limitations).
  - Estimation sample: 1963Q1 to 2007Q1.
  - To avoid contamination:
    - Germany sample ends in 1999.
    - Japan sample ends in 1995.
  - MCMC specifics: 10,000 runs of the Gibbs Sampler, discarding the first 2,000; to mitigate serial correlation, only every other draw is saved.
- Dependent-variable proxies for monetary policy effectiveness:
  - (i) cumulative response of inflation and unemployment over the subsequent five years following a one percentage point shock to the interest rate;
  - (ii) corresponding maximum response over the same horizon.

### Empirical results: time variation and weakening effectiveness
- General patterns:
  - Point estimates show unemployment generally increases for all countries during the first 10 quarters after a one percentage point unanticipated increase in the interest rate.
  - Noticeable cross-country differences in sensitivity:
    - The U.K. and Canada have unemployment rates up to four times more sensitive to interest rate changes than Japan.
  - Persistence differences:
    - Unemployment response is more protracted in Canada, the U.K. and Japan—on average, more than 40 percent of the peak responses still remaining after 20 quarters—than in the U.S. and Germany.
- Time-series variation and weakening:
  - For all countries (less visible for Canada), there is a noticeable decrease in the impact of monetary policy on the unemployment rate.
  - Peak responses decline by around 0.2 percentage points for the U.S. and the U.K.
  - An even larger maximum decrease is observable for Germany.
  - Japan and Canada exhibit smaller decreases in peak responses (circa 0.05-0.01).
  - Timing of decreases varies by country:
    - Conspicuous decrease starts around 1980 for the U.S. and Germany.
    - For the U.K., Canada and Japan, the decrease is more recent, starting about a decade later.
- Probabilities that the maximum impulse response has decreased from its peak to its trough along the same sample path:
  - U.S.: 92 percent
  - U.K.: 72 percent
  - Germany: 81 percent
  - Canada: 70 percent
  - Japan: 95 percent
- Conclusion: All else being equal, there is fairly strong evidence for a decrease in the effectiveness of monetary policy across the sample of countries.

*Source: _wp13191 - Section 2*

### Section 3

### _wp13191 - Section 3

### Evidence of a decline in monetary policy effectiveness
- Point estimates indicate fairly large changes in the effectiveness of monetary policy across time, confirming evidence in Boivin and others (2011).
- Changes in interest rates across the sample of five countries have less impact on inflation and unemployment today than they used to.
- The decrease in impact on inflation: peak responses decrease by about 0.2 percentage points for Canada and Germany, while decreasing somewhat less for the remaining countries.
- Probability that the minimum impulse response has increased (comparing estimates along the same sample path):
  - U.S.: 63 percent
  - U.K.: 77 percent
  - Germany: 68 percent
  - Canada: 67 percent
  - Japan: 83 percent
- Observation: years with large changes in the impact on inflation correspond roughly to dates identified for unemployment, suggesting a common factor driving time-variation in both responses.
- Footnote note: sharp decrease in Japan’s unemployment impact in the late 1980s/early 1990s may reflect the crisis that resulted in the policy rate reaching the zero lower bound (ZLB) in 1995.

### Time-varying impulse responses (unemployment and inflation)
- Median impulse response set-up:
  - Measure: response to a 1 percentage point unanticipated shock to the monetary policy instrument.
  - Model: TVC-BVAR with SV.
  - Impact measured in percentage points.
- Finding: median IRFs for unemployment and inflation have weakened over time in most countries (see Figures referenced in source).

### Panel regression to explain decline: specification and data
- Panel regression (equation (16)) models the impact of monetary policy on variable j at time t in country i as a function of:
  - Old-age dependency ratio (ageing of society i)
  - Share of manufacturing in total production (Share of Manu.)
  - Degree of economic openness (Openness = total trade/GDP)
  - Share of small firms (share of employment in firms with <50 employees)
  - Private sector credit (Credit-to-GDP)
- Old-age dependency ratio defined as: number of people above age 65 relative to those between 15 and 64.
- Stationarity and cointegration:
  - Panel unit-root tests indicate series are I(1); first-differences achieve stationarity.
  - Under cointegration assumption, standard least squares estimates are consistent.
  - Kao (1999) panel co-integration test employed to test for cointegration.
- Panel unit-root test p-values (asymptotic reported) for series:
  - Cum_U: Levin, Lin & Chu t-stat 0.20; Im, Pesaran & Shin W-stat 0.64; ADF - Fisher Chi-square 0.68; PP - Fisher Chi-square 1.00
  - Max_U: 0.11; 0.20; 0.20; 0.99
  - Cum_Pi: 0.21; 0.40; 0.48; 0.98
  - Max_Pi: 0.12; 0.36; 0.40; 0.95
  - Old-Age: 0.33; 0.86; 0.13; 0.33
  - Openness: 0.32; 0.76; 0.89; 0.90
  - Manufacturing: 0.39; 0.23; 0.04; 0.29
  - Credit: 0.38; 0.23; 0.33; 0.80

### Long-run estimates (D-OLS) — main coefficients and diagnostics
- Estimation device: Dynamic OLS (D-OLS) with one lead and one lag; Sieve bootstrapped standard errors used.
- Sample: Number of Obs. = 173 for all specifications.
- Kao (1999) CI-test statistics and p-values (P-value in squared brackets):
  - Equation (1): -2.122** [0.02]
  - Equation (2): -2.820*** [0.00]
  - Equation (3): -3.217** [0.00]
  - Equation (4): -2.388** [0.00]
- Table 3 estimated D-OLS coefficients (standard errors in parentheses). Dependent variables:
  - Equation (1) Cum π
  - Equation (2) Cum u
  - Equation (3) Max π
  - Equation (4) Max u

- Constant:
  - Equation (1): -25.271*** (3.175)
  - Equation (2): -25.431*** (3.176)
  - Equation (3): -0.423* (0.176)
  - Equation (4): -1.688*** (0.210)

- Old-age Ratio:
  - Equation (1): 0.073* (0.044)
  - Equation (2): -0.350*** (0.045)
  - Equation (3): 0.013*** (0.002)
  - Equation (4): -0.019*** (0.003)

- Share of Manu.:
  - Equation (1): 3.634*** (0.695)
  - Equation (2): 10.129*** (0.697)
  - Equation (3): -0.037 (0.039)
  - Equation (4): 0.659*** (0.046)

- Openness:
  - Equation (1): 3.414*** (0.520)
  - Equation (2): 0.381 (0.521)
  - Equation (3): 0.001 (0.029)
  - Equation (4): 0.021 (0.034)

- Credit-to-GDP:
  - Equation (1): -0.114 (0.195)
  - Equation (2): 0.958*** (0.195)
  - Equation (3): 0.010 (0.010)
  - Equation (4): 0.071*** (0.013)

- Target Change:
  - Equation (1): 0.290 (0.205)
  - Equation (2): 0.444** (0.205)
  - Equation (3): 0.029** (0.011)
  - Equation (4): 0.026** (0.014)

- Significance notation: *,**,*** indicates rejection of zero at the 10%, 5% and 1% level, respectively.

### Interpretation of regression results — ageing and other factors
- Ageing effects (old-age dependency ratio):
  - All else equal, an increase in the old-age dependency ratio of one point lowers (in absolute terms) the cumulative impact of a monetary policy shock on:
    - Inflation by 0.10 percentage points
    - Unemployment by 0.35 percentage points
  - The corresponding maximum impact of monetary policy is lowered by:
    - Inflation: 0.01 percentage points
    - Unemployment: 0.02 percentage points
  - These imply a strong negative long-run effect of population ageing on monetary policy effectiveness (example: projected 10 point rise in the old-age dependency ratio in Germany over the next decade).
- Comparison to literature:
  - Results contrast with Miles (2002) (which predicts increased potency due to wealth effects) but confirm conjectures in Bean (2004) and Fujiware and Terakashi (2008) that negative channels (e.g., credit channel) can dominate.
- Manufacturing share:
  - Countries with larger manufacturing shares tend to have real sectors more responsive to monetary policy (notably strong for responsiveness of unemployment).
  - Mixed effects on inflation responsiveness: positive and significant in equation (1) but negative in equation (3); possible explanation: manufacturing exhibits less nominal price rigidity than services.
- Openness:
  - Appears associated with a weaker effect on inflation; estimated impact on unemployment is positive but statistically insignificant in relevant equations.
- Credit-to-GDP:
  - Countries with more credit tend to have real sectors more responsive to monetary policy; large positive and significant effect on unemployment responsiveness (equation (2): 0.958***).

### Robustness checks
- Alternative specifications tested:
  - Use deflator rather than inflation
  - Use spare capacity rather than unemployment
  - Use policy rate rather than 3 month interest rates
- Result: IRFs to a 100 basis point increase in interest rates are robust; impact on inflation and unemployment found to be weakening for most countries.
- Exception: U.K. results are an exception (potentially due to numerous policy regime changes affecting credibility over the last four decades).

### Conclusion and policy implications
- Demographic changes in five large advanced economies with independent monetary policy have contributed to the observed decrease in monetary policy effectiveness.
- Mechanism: societies dominated by young households are more sensitive to interest rate changes than graying societies; monetary policy becomes less potent as populations age.
- Policy conjectures and implications:
  - Changes in optimal conduct of monetary policy:
    - Relative preference between inflation versus output stabilization may change as older households have larger asset holdings and more to lose from unexpected inflation.
    - Increasing aversion to inflation may, ceteris paribus, lead to a lower optimal inflation target; a declining real interest rate (a consequence of ageing) may work in the opposite direction. Central banks will need to weigh these trade-offs.
  - More aggressive monetary policy may be needed:
    - If monetary policy is less effective in a graying society, a larger change in the policy rate will be needed to achieve the same real effects; traditional 25bps adjustments may need amplification.
  - Non-monetary or structural responses:
    - Reversing population ageing trends (e.g., through policies encouraging higher fertility or migration) could enhance monetary policy effectiveness but carries risks and is not a panacea.

*Source: _wp13191 - Section 3*

### Section 4

### _wp13191 - Section 4

### Monetary policy in ageing societies: main finding
- Monetary policy will have to become more “activist” in ageing societies, with higher variation in interest rates possible going forward.
- If graying societies reduce monetary effectiveness, the burden to stabilize the economy and the financial system may increasingly be borne by other policy tools.
- With monetary policy effectiveness marginally reduced, the relative role of fiscal and macroprudential policy as a means to stabilize the economy may become more important.

### Role of other policy tools
- Monetary policy is a key tool to stabilize the economy and proved powerful during the global crisis; however, demographic change could weaken some transmission channels.
- Fiscal and macroprudential policies will likely assume greater importance if monetary transmission weakens.

### Monetary policy in Emerging Markets (EMs) and Low-Income Countries (LICs)
- The paper focuses on Advanced Economies (AEs), which will be the first to experience the demographic shift; EMs and LICs will age more gradually.
- EMs and LICs are in a peculiar situation: they will go through a demographic transition without having grown rich.
- Because wealth is not as skewed toward older generations in EMs/LICs, and older people are often poor and supported by active family members, factors such as the increasing dominance of the wealth effect over the credit channel effect are likely to be less important.
- Implication: Monetary policy effectiveness in EMs and LICs may not weaken as much as in AEs, or may manifest through other channels.

### Demographic shift — causes and long-term implications (Appendix 1)
- Longevity gains have been driven by lifestyle changes, better nutrition, medical advances, and improved education.
- In the first half of the 20th century, mortality improvements in AEs mainly benefited those under 35 due to public health initiatives (immunization, penicillin, other drugs).
- In the second half of the 20th century, life expectancy gains accrued to the elderly (see Towers Watson, 2011), aided by lifestyle changes (e.g., less smoking and drinking), better education, and medical advances (new diagnostics and treatments).
- Recent lifestyle counterbalances include less exercise, worse dieting, and fewer marriages, which tend to reduce life expectancy.
- An ageing society is a worldwide phenomenon caused by a combination of declining fertility and increasing longevity.
- The cohort of old (65+) and very old (80+) people in AEs is growing much stronger than the young, with the population of old people also aging.
- Because demographic changes take a long time to work through age structures, significant graying going forward is unavoidable.
- Since fertility is unlikely to rise for the foreseeable future, and immigration policy remains restrictive in most AEs, the current trend of ageing populations appears irreversible.
- Developing countries are still younger, but demographic transition is starting to increasingly encompass EMs and LICs; rapid catching-up in developing countries is expected to happen in the course of this century (because fertility has declined only recently).
- Life expectancy comparisons and projections are presented for 2010 and 2050; regional time series are shown for 1950-2100 (Total Fertility by Region, Life Expectancy at Birth by Region, Proportion of Age Groups, Old-Age Dependency Ratio).

### Consequence highlighted: old-age dependency
- The demographic shift leads to an increasing old-age dependency ratio — an increasing number of old people that have to be supported by younger cohorts.
- Increases in longevity are easier to manage if fertility keeps pace with longevity increases, or if migration augments the share of the young; otherwise the rising number of old people must be supported by more working-age persons.
- Note on forecast uncertainty: errors in forecasting longevity have typically been large, reflecting unanticipated influences such as the AIDS epidemic, economic depressions, social changes, climate change (heat waves), and political collapses (e.g., the Soviet Union) that have impacted longevity (see GFSR, 2012). (Footnote references in text: 15, 16.)

### Data sources (Appendix 2) — key series and providers
- Inflation: BLS (U.S.), ONS (U.K.), Stat. Bundesamt (Germany), Min. of Internal Affairs and Com. (Japan), Statistics Canada (Canada).
- Unemployment: BLS (U.S.), ONS (U.K.), Stat. Bundesamt (Germany), Min. of Internal Affairs and Com. (Japan), Statistics Canada (Canada).
- Interest Rate: Fed. Reserve Board (U.S.), BoE (U.K.), Haver Analytics (Germany), Nikkei (Japan), Bank of Canada (Canada).
- Manufacturing Share, Openness, Share of Small Firms, Old-Age Dependency Ratio: BEA/ONS/Stat. Bundesamt/Cabinet Office of Japan/Statistics Canada depending on series and country.
- Credit: IFS (International Financial Statistics) for the listed countries.
- VAR data and regression data are indicated as used.

*Source: _wp13191 - Section 4*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13191.pdf_