## 1idnea2023003

## Source details

**Canonical URL:** [1idnea2023003](https://www.imf.org/-/media/files/publications/cr/2023/english/1idnea2023003.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2023/english/1idnea2023003.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2023/english/1idnea2023003.pdf.json)

---

### Major themes and scope
- Focus: Drivers, macro-financial implications, and policy lessons from shifts in the investor base for Indonesia’s local-currency (LC) sovereign debt.
- Data and methods:
  - Monthly data for Indonesia over 2002M12–2022M12; cross-country panel for 24 EMs over 2004Q1–2022Q2.
  - Methods include ARDL cointegration (Pesaran et al. 2001) and local projections (Jordà 2005).
- Key investor groups analyzed: Nonresidents (NR), Bank of Indonesia (BI), domestic banks, and domestic nonbank residents (institutional investors).

### Drivers of investor holdings (empirical findings)
- Global vs domestic determinants:
  - NR holdings of LC debt are driven predominantly by global factors, including global commodity prices, MOVE (bond volatility), and the ten-year US bond yield (yieldUS_10y).
  - Domestic investor holdings are mostly associated with domestic LC debt issuances (issuances).
- Role of BI:
  - BI increases holdings with higher debt issuances and is positively correlated with a worsening in the global factors considered—indicating BI acts as a residual financier under adverse conditions.
- Robustness:
  - Results hold under specifications adding exchange rate (IDR/USD), country-risk (ICRG inverse), VIX, yield spreads, different maturities, real yields, alternative dependent variable definitions, and varied sample periods.

### Quantitative shifts in Indonesia’s investor base and public debt
- Public debt and NR share:
  - Public debt: "40.1 percent of GDP in 2022" vs "30.6 percent in 2019".
  - NR share of LC debt: "39 percent in 2019Q4" → "about 14 percent at end-2022".
  - Decline magnitude: "largest decline in the NR share of LC debt, of roughly 25 percentage points until 2022Q4".
  - The decline in NR holdings of LC debt represented "almost 3 percent of GDP".
- NR concentration and index weight:
  - NR holdings remain mostly concentrated in long-dated securities.
  - Indonesia had "a weight of 10 percent in the GBI-EM index as of end-December 2021".

### Macro-financial implications (panel local-projection results)
- Calibration: Effects reported for a one-standard deviation increase in the sample NR share ("13.8 percent").
- Sovereign borrowing costs:
  - A one-standard deviation increase in the NR share is associated with a decline in ten-year LC bond yields of "0.2 percentage points after one year" and "over 0.4 percentage points after three years".
- Domestic credit:
  - Cumulative credit growth to the private nonfinancial sector increases by "2.5 percentage points after one year" and by "roughly 5 percentage points after three years".
- Volatility trade-off:
  - NR holdings support the bond market and domestic credit, but are associated with higher FX volatility and increased bond-yield volatility over time.
- Interpretation caveat:
  - Exercise is predictive/correlational; correlations do not necessarily imply causality.

### NR shock definition, frequency, and effects
- NR shock definition and frequency:
  - NR shock: NR share of LC debt declines by one standard deviation over two years—an "8.3 percentage-point" drop.
  - The NR shock binary variable accounts for "almost 7 percent" of all observations.
- Interaction with financial stress:
  - Stress dummy includes the 2007-09 GFC, the 2013 Taper Tantrum, and the initial quarters of the COVID-19 pandemic.
  - These episodes account for "25.6 percent" of all NR shock periods (and "16.7 percent" of all observations).
- Empirical effects outside financial stress (γ):
  - NR shocks outside financial stress correlate with higher FX volatility and weaker credit growth over the medium term; also associated with currency depreciation in the EM sample.
  - No robust evidence for effects on borrowing costs and bond-yield volatility in these episodes.
- Amplification during financial stress (γ + δ):
  - When NR shocks coincide with financial stress, negative economic effects are amplified—likely driven by larger capital outflows and concomitant adverse shocks.
  - Policy implication: abrupt transitions to a lower NR share pose greater risks if occurring during financial stress.

### Who absorbs new LC sovereign debt (marginal investor)
- Method:
  - Regress change in holdings by investor group j on change in stock of debt (both scaled by initial stock); β1 = percent of incremental debt absorbed by investor j; β2 = additional absorption during NR shocks.
- Key results (24 EMs, 2004Q1-2022Q2):
  - In absence of NR shocks, domestic banks and nonbanks together absorb about "70 percent" of new issuances of LC debt in the EM sample.
  - During large and rapid declines in the NR share, domestic nonbanks play a key role in absorbing new debt supply (β2 positive for domestic nonbanks).
- Policy implication:
  - Deepening the domestic nonbank sector and the retail investor base—by improving regulatory frameworks that encourage investment in long-term securities—supports market diversification and stability.

### Indonesia: NRIR, monetary context, and policy risks
- Estimated NRIR:
  - Estimated NRIR for Indonesia falls between "1 and 2 percent" and has remained broadly stable since the Global Financial Crisis.
- Recent monetary policy and key numbers:
  - BI increased its policy rate by a cumulative "225 basis points" during August 2022-January 2023 to "5.75 percent".
  - The policy rate "5.75 percent" is above the pre-pandemic level of "5 percent at end-2019".
  - BI withdrew excess banking system liquidity to bring the overnight interbank money market rate closer to the policy rate level ("5.6 percent").
  - Ex-ante real policy rate was "1.8 percent as of March 2023", up from "-0.5 percent in September 2022".
  - As of March 2023, the real policy rate is estimated at about "0.8-2.8 percent" based on different inflation measures; ex-ante one-year-ahead consensus forecast-based real policy rate at "1.8 percent", which falls within the estimated neutral zone of "1-2 percent".
- Interpretation and risks:
  - NR holdings can support domestic bond market depth and domestic credit growth but are associated with higher exchange rate volatility, particularly during financial turbulence.
  - Rapid and large declines in the NR share lead to greater market volatility; Indonesia has experienced a trend toward a lower NR share since the beginning of the COVID shock.
  - Analysis remains predictive/descriptive and is silent about causality.

### NRIR estimation methods and key quantitative estimates
- HP-filtered real policy rate measures (2000Q1-2022Q4; HP smoothing 1,600):
  - End-2022 NRIR: "about 1.1-1.3 percent", depending on inflation measure.
  - Sample period (2000Q1-2022Q4) NRIR range: "1.4 to 2.4 percent".
  - Sample average NRIR: "1.9 percent".
  - Average ex-post real policy rate: "1.7 percent".
  - Table 1 (End-2022 / Sample Avg. / Std. Dev.):
    - one-year-ahead consensus forecast: "1.3 / 2.4 / 2.3"
    - 4-quarter moving average (t-1 to t-4): "1.2 / 1.7 / 1.0"
    - year/year: "1.1 / 1.4 / 0.7"
    - quarter/quarter, s.a.a.r.: "1.1 / 2.0 / 1.4"
- HLW semi-structural model (2001Q1-2022Q2; COVID adjustment):
  - Two scenarios:
    - Model 1 (median λg and λz, “High lambda_z”): λg = "0.073", λz = "0.0216". Final state r*: "1.210" / sample avg "1.286" (std. error "6.760"). Standard error for r* at end-2022: "6.760".
    - Model 2 (median λg and lower bound λz, “Low lambda_z”): λg = "0.073", λz = "0.001". Final state r*: "1.200" / sample avg "1.285" (std. error "0.393"). Standard error for r* at end-2022: "0.393".
  - Model 2 implies a "95-percent confidence interval for r* of 0.9-1.6 percent".
- Yield curve model (Basdevant, Björksten, and Karagedikli, 2004; monthly since January 2005):
  - Estimated NRIR stable at "1.3-2.1 percent" since 2015; NRIR ranges "1.3-1.5 percent as of end-2022".
  - Sensitivity to smoothing parameter small across λ values considered: "0.0028, 0.0053, 0.0083".
- Consumption-based CAPM / Euler equation:
  - Calibration: average annual real GDP per capita growth g = "3.6 percent" (2000-2022); habit persistence ϕ = "0.95".
  - Model-implied NRIR for Indonesia ranges between "0.9 and 1.8 percent".
  - Average NRIR from consumption model: "1.34 percent". Std. dev: "0.36 percent".
  - NRIR estimate insensitive to ϕ in "0.93-0.97" (ranges between "1.32 and 1.35 percent").

### Consolidated NRIR, stability, and long-term outlook
- Consolidated estimates:
  - Different methods indicate Indonesia’s NRIR stands approximately at "1-2 percent".
  - Summary of estimates:
    - HP Filter latest (end-2022): "1.1 - 1.3"; sample avg "1.4 - 2.4"; std. dev "0.7 - 2.3".
    - HLW (2021) latest: "1.1 - 1.2"; sample avg "1.3"; std. dev "0.1 - 0.09".
    - BBK (2004) yield curve latest: "1.3 - 1.5"; sample avg "1.7"; std. dev "0.3 - 0.5".
    - Consumption model latest: "0.9 - 1.8"; sample avg "0.4".
  - Note: Latest data point is "2022Q2" for HLW and "2022Q4" for HP filter and BBK estimates.
- Stability and pandemic assessment:
  - NRIR broadly stable, fluctuating within a narrow band of "1.2-1.3 percent" since the Global Financial Crisis per model estimates.
  - Original HLW without COVID adjustment shows a downshift in NRIR at the pandemic onset (from "1.2 percent in 2019Q4" to "0.8-0.9 percent from 2020Q2"); COVID-adjusted HLW suggests this may reflect exceptional COVID-19 shock characteristics rather than a structural change.
  - Overall evidence supports that the pandemic has not led to a significant change in the NRIR to date.
- Long-term trajectory and uncertainty:
  - NRIR more likely to trend downward due to potential:
    - Pandemic scarring on education/training → lower productivity growth → reduced investment demand.
    - Favorable demographic trend slowdown (aging speed slowed from "0.31 year in 2000" to "0.19 in 2021") that may not persist.
    - Changes in inequality: success in reducing inequality would affect NRIR; higher inequality tends to increase savings and lower r*.
  - Upward pressures could include higher public debt in advanced economies and post-pandemic digitalization boosting productivity.
  - On balance, evidence points to greater likelihood of downward pressure on Indonesia’s NRIR, consistent with other emerging markets.

### Policy implications and recommendations
- Deepen domestic investor base:
  - Further develop domestic nonbank institutional investors and the retail investor base to support market depth and reduce volatility—domestic nonbanks are key absorbers of new LC debt during NR outflows.
- Risk management of public debt:
  - Balance benefits of higher NR participation (lower term premia, longer maturities, deeper markets) against costs (higher rollover and exchange-rate risks, potential for higher volatility and sudden reversals).
- Central bank role:
  - Recognize BI’s contingent role in primary market purchases under stress; policy should aim to reduce reliance on central bank financing over the medium term by broadening investor composition.
- Macroeconomic resilience:
  - Strengthen macroeconomic stability, fiscal and external buffers, and institutional quality to enhance resilience when capital flows reverse.

*Source: 1idnea2023003*

### References _____________________________________________________________________________ 15

### 1idnea2023003 - References _____________________________________________________________________________ 15

### Major themes and scope
- Focus: Drivers, macro-financial implications, and policy lessons from shifts in the investor base for Indonesia’s local-currency (LC) sovereign debt.
- Periods and data: Monthly data for Indonesia over 2002M12–2022M12; cross-country panel for 24 EMs over 2004Q1–2022Q2. Methods include ARDL cointegration (Pesaran et al. 2001) and local projections (Jordà 2005).
- Key investor groups analyzed: Nonresidents (NR), Bank of Indonesia (BI), domestic banks, and domestic nonbank residents (institutional investors).

### Drivers of investor holdings (empirical findings)
- Global vs domestic factors:
  - NR holdings of LC debt are driven predominantly by global factors, including global commodity prices, MOVE (bond volatility), and the ten-year US bond yield (yieldUS_10y).
  - Domestic investor holdings are mostly associated with domestic LC debt issuances (issuances).
- Role of BI:
  - BI increases holdings with higher debt issuances and is positively correlated with a worsening in the global factors considered—indicating BI acts as a residual financier under adverse conditions.
- Robustness:
  - Results hold under multiple specifications, including adding exchange rate (IDR/USD), country-risk (ICRG inverse), VIX, yield spreads, different maturities, real yields, alternative dependent variable definitions, and varied sample periods.

### Quantitative shifts in Indonesia’s investor base and debt
- Public debt and NR share:
  - Public debt: "40.1 percent of GDP in 2022" vs "30.6 percent in 2019".
  - NR share of LC debt: "39 percent in 2019Q4" → "about 14 percent at end-2022".
  - Decline magnitude: "largest decline in the NR share of LC debt, of roughly 25 percentage points until 2022Q4".
  - The decline in NR holdings of LC debt represented "almost 3 percent of GDP".
- NR concentration: NR holdings remain mostly concentrated in long-dated securities.
- Index weight: Indonesia had "a weight of 10 percent in the GBI-EM index as of end-December 2021".

### Macro-financial implications (panel local-projection results)
- Calibration: Effects reported for a one-standard deviation increase in the sample NR share ("13.8 percent").
- Sovereign borrowing costs:
  - A one-standard deviation increase in the NR share is associated with a decline in ten-year LC bond yields of "0.2 percentage points after one year" and "over 0.4 percentage points after three years".
- Domestic credit:
  - Cumulative credit growth to the private nonfinancial sector increases by "2.5 percentage points after one year" and by "roughly 5 percentage points after three years".
- Volatility trade-off:
  - NR holdings support the bond market and domestic credit, but are associated with higher FX volatility and increased bond-yield volatility over time.
- Interpretation caveat:
  - Exercise is predictive/correlational; correlations do not necessarily imply causality.

### Lessons from the COVID-era and related episodes
- Indonesia experienced an unprecedented fall in the NR share of LC debt during COVID-19:
  - Two phases: large NR outflows in 2020–2021 (COVID shock and BI primary market purchases under BI-MoF burden-sharing) followed by continued decline in 2022 amid global monetary tightening (Fed tightening).
  - NR share reduction in LC debt exceeded the decline in NR share of foreign-denominated debt during 2020–22.
- Comparative context:
  - Indonesia’s decline in NR share since COVID-19 was much larger than the average EM; a similar pattern is noted for ASEAN countries.

### Policy implications and recommendations (drawn from findings)
- Deepen domestic investor base:
  - Further develop domestic nonbank institutional investors to support market depth and reduce volatility, given domestic nonbanks’ role in absorbing new LC debt, particularly during episodes of large NR outflows.
- Risk-management of public debt:
  - Balance benefits of higher NR participation (lower term premia, longer maturities, deeper markets) against costs (higher rollover and exchange-rate risks, potential for higher volatility and sudden reversals).
- Central bank as residual financier:
  - Recognize the contingent role of BI in primary market purchases under stress; policy should aim to reduce reliance on central bank financing over the medium term by broadening investor composition.

### Methods and diagnostics
- Time-series model for investor holdings:
  - ARDL (long-run in levels; short-run as error-correction in differences) applied to investor holdings regressions with domestic and global controls; optimum lag selection by information criteria.
- Cross-country dynamics:
  - Local projections for horizons h = 0,...,12 quarters, with country and time fixed effects, lagged dependent variables, and controls (GDP growth, CPI inflation, current account %, policy rate, FX volatility, government debt % of GDP, foreign holdings of foreign-denominated sovereign debt).
- Data sources referenced: BIS, Bloomberg, CEIC, Haver; investor shares from Arslanalp and Tsuda (2014).

*Prepared by IMF staff; contents drawn from the referenced IMF chapter/section.*

### 15.      This section examines the implications for countries that experience fast and large

### 1idnea2023003 - 15.      This section examines the implications for countries that experience fast and large

### NR shock: definition, frequency, and identification
- NR shock defined as periods when the NR share of LC debt declines by one standard deviation over two years—amounting to an 8.3 percentage-point drop in the NR share.
- The NR shock binary variable accounts for almost 7 percent of all observations in the estimation sample.
- The regression specification isolates:
  - γ: dynamics of the real economy during NR shocks outside of financial stress periods.
  - δ: additional effect when the NR shock coincides with periods of financial stress.
- The Stress dummy includes the 2007-09 GFC, the 2013 Taper Tantrum, and the initial quarters of the COVID-19 pandemic. These episodes of financial stress account for 25.6 percent of all NR shock periods (and 16.7 percent of all observations in the sample).

### Empirical findings — NR shocks outside financial stress
- Rapid and large declines in the NR share of LC debt that occur outside financial stress periods correlate with:
  - Higher FX volatility (cumulative responses indicate positive γ for FX volatility).
  - Weaker credit growth over the medium term (the figure shows the γ for each horizon).
  - Currency depreciation in the EM sample (unreported results).
- No robust evidence found for effects on borrowing costs and on bond yields volatility in these episodes.

### Amplification during financial stress
- When an NR shock coincides with financial stress, negative economic effects are amplified:
  - The combined effect is captured by γ + δ (red lines in the referenced figure).
  - Amplification likely driven by larger capital outflows and the presence of other adverse shocks during such episodes.
- Policy-relevant implication: abrupt transitions to a lower NR share pose greater risks if they occur during financial stress.

### Who is the marginal investor of LC sovereign debt?
- Method: Following Fang et al. (2022), regress the change in the stock of LC debt holdings for each investor group j on the change in the stock of debt, both scaled by the initial stock of debt; include country and time fixed effects. Coefficient β1 indicates the percent of incremental debt absorbed by investor j; β2 (interaction with NR shock) captures additional absorption during NR shock periods.
- Data: Arslanalp and Tsuda (2014) database covering 2004Q1-2022Q2 for 24 EMs; six investor types analyzed: domestic central banks, domestic banks, domestic nonbanks, foreign official sector, foreign banks, and foreign nonbanks. (Nonbanks include NBFIs, nonfinancial corporations, and households.)
- Key quantitative findings:
  - In the absence of NR shocks, domestic banks and nonbanks together absorb about 70 percent of new issuances of LC debt in the EM sample.
  - During episodes of large and rapid declines in the NR share of LC debt, domestic nonbanks play a key role in absorbing new debt supply (β2 positive for domestic nonbanks).
- Policy implication: deepening the domestic nonbank sector and the retail investor base—by improving regulatory frameworks that encourage investment in long-term securities—supports market diversification and stability.

### Concluding remarks and policy context for Indonesia
- Estimated NRIR for Indonesia falls between 1 and 2 percent and has remained broadly stable since the Global Financial Crisis; apparently unaffected by the COVID-19 pandemic thus far.
- Recent monetary policy context and key numbers:
  - BI increased its policy rate by a cumulative 225 basis points during August 2022-January 2023 to 5.75 percent.
  - The policy rate 5.75 percent is above the pre-pandemic level of 5 percent at end-2019.
  - BI withdrew excess banking system liquidity to bring the overnight interbank money market rate closer to the policy rate level (5.6 percent).
  - Ex-ante real policy rate was 1.8 percent as of March 2023, up from -0.5 percent in September 2022.
- Interpretation and risks:
  - NR holdings can support domestic bond market depth and domestic credit growth but are associated with higher exchange rate volatility, particularly during financial turbulence.
  - New findings indicate greater market volatility following rapid and large declines in the NR share of LC debt; this is especially important for countries experiencing a trend toward a lower NR share, as in Indonesia since the beginning of the COVID shock.
  - Analysis is predictive/descriptive and remains silent about causality.
- Policy recommendations for Indonesia:
  - Further deepen the domestic retail investor base in line with the authorities’ medium-term debt strategy to support market depth and reduce volatility.
  - Strengthen macroeconomic stability, fiscal and external buffers, and institutional quality to enhance resilience when capital flows reverse.

*Prepared by IMF staff based on the chapter text provided.*

### 4.      Before estimating the models, we first examine the range of plausible values of the

### 4.      Before estimating the models, we first examine the range of plausible values of the

### HP-filtered real policy rate measures and NRIR (HP filter approach)
- Method:
  - Deflate nominal policy rate series using four inflation measures: (i) year/year inflation in period t; (ii) quarter/quarter, s.a.a.r. inflation in period t; (iii) four-quarter moving average inflation for period t-1 to t-4; and (iv) one-year-ahead consensus forecast in period t.
  - Use IMF staff forecasts for the policy rate and headline inflation up to 2024Q4 to mitigate end-point problems with the HP filter.
  - Hodrick-Prescott filter with a smoothing parameter value of 1,600 applied to quarterly data from 2000Q1 to 2022Q4 (with IMF forecasts until 2024Q4).
- Key statistics (HP filter estimates):
  - End-2022 NRIR: about 1.1-1.3 percent, depending on the inflation measure.
  - Sample period (2000Q1-2022Q4) NRIR range: 1.4 to 2.4 percent.
  - Sample average NRIR: 1.9 percent.
  - Average ex-post real policy rate: 1.7 percent.
- Table 1 summary (End-2022 / Sample Avg. / Std. Dev.):
  - one-year-ahead consensus forecast: 1.3 / 2.4 / 2.3
  - 4-quarter moving average (t-1 to t-4): 1.2 / 1.7 / 1.0
  - year/year: 1.1 / 1.4 / 0.7
  - quarter/quarter, s.a.a.r.: 1.1 / 2.0 / 1.4

### Holston-Laubach-Williams (HLW) semi-structural model and COVID adjustment
- Model structure:
  - Semi-structural macro model with IS equation linking output gap to NRIR and a Phillips curve linking inflation to the output gap.
  - Variables: y and y* (log real GDP and unobserved potential output); r and r* (real policy rate and NRIR); inflation π (q/q, s.a.a.r.); output gap y_gap.
  - NRIR defined as sum of growth rate of potential output (g) and an unobserved variable (z).
  - Stochastic error terms ε1 and ε2 assumed white noise with variances σπ2 and σy2.
- COVID-19 augmentation:
  - Include Oxford COVID-19 Government Response Tracker indicator d (range 0 to 100) to capture direct effects of pandemic-related mobility restrictions.
  - Replace potential output y* with y* + φ·d/100 and output gap with (y − y*) − φ·d/100; estimate φ along with other parameters.
  - Pre-pandemic values set equal to zero.
- Estimation details:
  - Kalman filter estimation using Indonesia quarterly data 2001Q1‒2022Q2.
  - Inflation π constructed using seasonally adjusted core inflation series from Haver (annualized) since December 2007 and splicing headline inflation since 2001Q1.
  - Real policy rate: BI’s 7-day reverse repo rate minus Consensus Forecast one-year-ahead expected inflation rate.
  - Use Stock and Watson’s (1998) median unbiased estimator to obtain ratios λg ≡ σg/σy* and λz ≡ σz/σy*, then impose these ratios in next-stage estimation to address the “pile-up” problem.
- Estimation scenarios and results (Table 2):
  - Two models presented:
    - Model 1: median values of λg and λz (“High lambda_z”) — (λg = 0.073, λz = 0.0216).
    - Model 2: median λg and lower bound of 90 percent interval for λz (“Low lambda_z”) — (λg = 0.073, λz = 0.001).
  - Coefficient signs consistent with expectations: β1 > 0 (IS output gap), β3 < 0 (Phillips curve real interest rate gap), though large standard errors.
  - Standard error for r* at end-2022 (2022Q2):
    - Model 1: 6.760
    - Model 2: 0.393
  - Model 2 implies a 95-percent confidence interval for r* of 0.9-1.6 percent.
  - Using the upper bound of the 90 percent interval for λz (0.166) yields implausibly volatile NRIR ranging from -1.7 to 3.5 percent over the sample.
  - Final state / sample averages (selected):
    - Model 1 final state r*: 1.210 / sample avg 1.286 (std. error 6.760)
    - Model 2 final state r*: 1.200 / sample avg 1.285 (std. error 0.393)
  - Sample period for HLW estimation: 2001Q1-2022Q2.

### Yield curve model (Basdevant, Björksten, and Karagedikli, 2004)
- Method:
  - Interpret NRIR as a common stochastic trend between short and long-term government bond yields.
  - Estimate a system using Kalman filter with nominal 3-month and 10-year government bond yields, yield curve spread α (term premium), one-year-ahead inflation forecast πe, and r*.
  - Monthly data since January 2005; 3-month JIBOR used as proxy for 3-month T-bill prior to August 2016.
  - Impose restriction on standard deviation of disturbance term ε1 to ensure NRIR is “sufficiently” smooth relative to short-term rate.
  - Examine sensitivity via smoothing parameter λ ≡ σr* / σs; values considered: 0.0028, 0.0053, 0.0083 (corresponding to ratios of σs2/σr*2 equal to 129600, 36000, 14400).
- Key results:
  - Estimated NRIR stable at 1.3-2.1 percent since 2015, after moderate decline following the Global Financial Crisis.
  - Estimates not sensitive to smoothing parameter choice; NRIR ranges 1.3-1.5 percent as of end-2022.
- Caveats:
  - Indonesia’s secondary bond market relatively shallow, limiting monetary policy transmission along the yield curve.
  - Since the start of the pandemic, BI’s secondary market interventions (“triple intervention” and “operation twist”) may moderate long-term rates; such moderation would tend to underestimate NRIR.

### Consumption-based CAPM / Euler equation approach
- Theory and formulas:
  - Representative agent with CRRA utility u(ct) = ct^(1−γ)/(1−γ), γ>0; discount factor β.
  - Approximate first-order condition (Fuentes and Gredig, 2007):
    - ln(1 + r*) ≅ −ln β + γ Et(g) − (1/2) γ^2 vart(g)
  - With habit persistence ϕ, first-order condition approximates:
    - ln(1 + r*) ≅ −ln β + γ Et(g) − (1/2) γ (1−ϕ)
- Calibration and results:
  - Use Indonesia average annual real GDP per capita growth g = 3.6 percent (period 2000-2022).
  - Habit persistence parameter ϕ = 0.95.
  - Using typical literature values for β and γ, model-implied NRIR for Indonesia ranges between 0.9 and 1.8 percent (Table 3).
  - Average NRIR from consumption model: 1.34 percent.
  - Std. dev: 0.36 percent.
  - NRIR estimate not sensitive to chosen ϕ between 0.93 and 0.97 (ranges between 1.32 and 1.35 percent).

### Main findings, comparisons, and policy implications
- Consolidated NRIR estimates:
  - Different methods indicate Indonesia’s NRIR stands approximately at 1-2 percent (Table 4).
  - HP Filter latest (end-2022): 1.1 - 1.3; sample avg 1.4 - 2.4; std. dev 0.7 - 2.3.
  - HLW (2021) latest: 1.1 - 1.2; sample avg 1.3; std. dev 0.1 - 0.09.
  - BBK (2004) yield curve latest: 1.3 - 1.5; sample avg 1.7; std. dev 0.3 - 0.5.
  - Consumption model latest: 0.9 - 1.8; sample avg 0.4.
  - Note: Latest data point is 2022Q2 for HLW (2021) and 2022Q4 for HP filter and BBK (2004) estimates.
- Stability and pandemic assessment:
  - NRIR broadly stable, fluctuating within a narrow band of 1.2-1.3 percent since the Global Financial Crisis (GFC) per model estimates.
  - Original HLW model without COVID adjustment shows a level downshift in NRIR at the pandemic onset (1.2 percent in 2019Q4 to 0.8-0.9 percent from 2020Q2); the COVID-adjusted model suggests this downshift may reflect exceptional COVID-19 shock characteristics rather than fundamental structural change.
  - Overall evidence supports that the pandemic has not led to a significant change in the NRIR to date.
- Monetary policy stance:
  - BI’s monetary policy has been responsive to inflation pressures since 2000 and is now in a neutral position.
  - Since 2001, four major episodes of markedly high inflation gap (headline inflation exceeding the mid-point of BI’s target band by 3 percentage points or more): 2001Q3-2002Q4, 2005Q4-2006Q3, 2008Q2-2009Q1, 2013Q3-2014Q2; BI increased the policy rate aggressively in these episodes.
  - Real policy rate increased significantly in 2022Q4 amid an average inflation gap of 2.5 percent.
  - As of March 2023, the real policy rate is estimated at about 0.8-2.8 percent, based on different measures of inflation:
    - Ex-ante one-year-ahead consensus forecast-based real policy rate at 1.8 percent, which falls within the estimated neutral zone of 1-2 percent.
- Long-term trajectory and uncertainty:
  - Long-term NRIR trajectory uncertain but appears more likely to trend downward.
  - Potential downward pressures:
    - Pandemic scarring effects on education/training → lower productivity growth → reduced investment demand.
    - Favorable demographic trend slowdown (aging speed slowed from 0.31 year in 2000 to 0.19 in 2021) may not persist, potentially raising savings and lowering investment.
    - Success in reducing inequality would affect NRIR; higher inequality tends to increase savings and lower r*.
  - Potential upward pressures:
    - Pandemic-induced increase in public debt in advanced economies.
    - Potential post-pandemic acceleration in productivity-enhancing digitalization.
  - On balance, Indonesia’s NRIR appears more likely to face downward pressures, consistent with other emerging market economies.

*Source: 1idnea2023003*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2023/english/1idnea2023003.pdf_