## _wp09228

## Source details

**Canonical URL:** [_wp09228](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09228.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09228.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09228.pdf.json)

---

### Study scope and context
- Sample period: January 10, 2005 to March 29, 2007.
- Dataset: anonymous individual daily bank-by-bank data; 543 observations.
- Policy context: Central Bank of Jordan (CBJ) operated a currency peg to the U.S. dollar (USD).
- CBJ practices during sample:
  - Set the level of its overnight deposit facility equal to the level of the Fed Funds target, imposing a positive spread over the U.S. rate.
  - Typically intervened bi-weekly in the money market through open market operations.
  - Maintained a spread of at least 250 bps between its short-term deposit and credit facilities.
- Operating implications: CBJ did not use the overnight rate as its operating target; day-to-day variations of the overnight rate were much larger than in the U.S. and largely reflected local liquidity conditions.

### Identification strategy and model specification
- Identification of exogenous daily reserve shocks:
  - Use the difference between morning projections by the CBJ and evening realizations of nonborrowed excess reserves as an exogenous source of variations in the daily change in nonborrowed excess reserves (following Carpenter and Demiralp (2006) approach).
  - Individual bank data are daily morning (pre-trade) projections of excess reserves for each bank.
- Variable of interest: daily change of the spread between the overnight rate and the rate of the overnight deposit facility.
- Econometric model (two equations estimated simultaneously):
  - Mean equation: ARMA(1,1)-X for the first difference of the spread.
  - Variance equation: EGARCH(1,1)-X for the volatility of the first difference of the spread.
- Extended specification: includes day-of-maintenance-period effects (10-day maintenance period), quadratic liquidity effects by day.

### Monetary framework and data features (Jordan)
- Exchange rate and reserves:
  - JD pegged to the USD since 1997 (parity fixed at 0.709 Jordanian Dinar per USD with a band of +/- 0.15 percent around the central parity).
  - Gross usable international reserves: USD 6.1 billion (or 11 times short-term external debt at remaining maturity) as of end-December 2006.
- Operational tools during January 10, 2005 to March 29, 2007:
  - reserve requirements (local and foreign currency);
  - overnight deposit facility;
  - 7-day refinancing facility (maturity reduced to overnight in May 2007);
  - bi-weekly open-market operations (central bills with maturities of 3 months and 6 months).
- Reserve maintenance:
  - maintenance period: 14 days (Monday to the Sunday of the following week);
  - banks required to maintain 65 percent of reserve requirements at all times and could average remaining 35 percent over period;
  - required reserves in local currency: 8 percent of daily average balances (rules in place since September 6, 2004);
  - reserve requirements set on the 5th of each month.
- Market structure and sample:
  - analysis excludes two Islamic banks; banking sector for this study: 21 banks, all private.
  - top three banks had combined market share over 40 percent as of end-2006.
  - banking system assets: JD 24.2 billion, or 240 percent of GDP, as of end-2006.
- Liquidity patterns:
  - system generally in excess liquidity; deposits on overnight window on last day of each maintenance period were large.
  - until mid-2006, excess liquidity ≈ fifty percent of required reserves; declined by half on average in most recent period.
  - daily volatility of excess reserves declined somewhat after mid-2006.
  - interbank trading volume increased as excess liquidity declined.
- Interbank spread and arbitrage:
  - spread between interbank rate and overnight deposit window ranged between 12 and 192 bps.
  - episodes of incomplete arbitrage: spike Oct 19-20, 2005 followed by drop of 170 bps on Oct 21 (day 7 of maintenance period) when a large amount of central bank bills matured.
- Distribution measures:
  - two normalized Herfindahl indices constructed from morning projections: one for “suppliers” (projected positive excess reserves) and one for “demanders” (projected negative excess reserves).
  - Fluctuations typically small, with occasional spikes; supplier index variations larger in second half of sample when aggregate excess liquidity was lower.

### Empirical strategy and dependent variables
- Dependent variable: first difference of the spread between the interbank rate and the deposit window rate (Δs).
- Key liquidity measure: non-borrowed excess reserves normalized by required reserves (RER).
- Additional controls:
  - cumulative sum variable: cumulative sum of closing balance minus required reserves since first day of maintenance period normalized by cumulative required reserves since first day.
  - normalized Herfindahl indices for suppliers and demanders and their first differences.
  - dummies: each day of maintenance period, beginning/end of month, beginning/end of quarter, day before/after holiday, day of and days surrounding change in overnight window rate.
- Identification instrument: CBJ morning forecast error (Shock) normalized by required reserves.
- Model specification:
  - ARMA(1,1)-EGARCH(1,1)-X with mean and variance equations as specified; stationarity conditions |φ| < 1 and |β| < 1.

### Baseline specification — selected estimates and summary statistics
- Regression summary:
  - R-squared 0.13; Adjusted R-squared 0.08; S.E. of regression 0.11; Log likelihood 963.57.
- Mean equation (selected coefficients):
  - (Δs)t-1 = 0.62 *** (St.dev. 0.05, p-value 0.00)
  - Shock = -0.33 *** (0.05, 0.00)
  - Shock*RER = 0.91 *** (0.21, 0.00)
  - Shock*(RER)² = -0.65 *** (0.22, 0.00)
  - ∆(Herf. Suppliers) = 0.52 *** (0.13, 0.00)
  - ∆(Herf. Suppliers) * RER = -1.10 ** (0.48, 0.02)
  - ∆(Herf. Demanders) = -0.27 ** (0.13, 0.04)
  - ∆(Herf. Demanders) * RER = 0.65 * (0.37, 0.08)
  - Dummy (policy change) = -0.12 *** (0.02, 0.00)
  - Dummy (policy change + 1) = 0.05 *** (0.01, 0.00)
  - Dummy (policy change - 1) = 0.05 *** (0.01, 0.00)
  - MA(1) = -0.31 *** (0.07, 0.00)
- Variance equation (selected coefficients):
  - Constant = -1.33 *** (0.10, 0.00)
  - Alpha: ARCH(1) = 0.96 *** (0.07, 0.00)
  - Gamma: Asymmetry parameter = 0.31 *** (0.05, 0.00)
  - Beta: GARCH(1) = 0.87 *** (0.02, 0.00)
  - Excess Reserves = -0.81 *** (0.09, 0.00)
  - Cumulative Sum = -4.43 *** (0.97, 0.00)
  - Dummy end-of-month = 1.39 *** (0.34, 0.00)
  - Dummy (holiday + 1) = 2.35 *** (0.15, 0.00)

### Key empirical findings — liquidity effect
- The liquidity effect is robustly negative and convex in RER:
  - Positive supply shocks reduce the spread; effect is stronger when RER is low.
- Quantified example:
  - a positive supply shock representing 26 percent of required reserves (one standard deviation) would decrease the spread by about 10 bps when RER is close to minimum sample value, but would not move when liquidity conditions are loose.
  - First-order autocorrelation 0.62 implies additional smaller impacts over next 1–2 days.
- Accumulated reserves over maintenance period:
  - Weak direct effect of accumulated reserves surplus/deficit on overnight rate.
  - Stronger indirect effect through its impact on the vigor of the liquidity effect: when banking system is behind in reserve satisfaction, liquidity effect is more powerful.
- Within-maintenance-period:
  - interbank rate appears to peak near end of first week of maintenance period; banks deposit larger amounts at overnight window as maintenance period progresses.
  - ratio of cumulative excess reserves to cumulative required reserves falls as maintenance period progresses.

### Key empirical findings — distributional/concentration effects
- Construction and effects of concentration measures:
  - Two normalized Herfindahl indices (suppliers and demanders) based on morning projections.
  - More concentrated supply pushes the rate up.
  - More concentrated demand pulls the rate down.
- Interaction with liquidity:
  - The vigor of distribution effects is a negative and convex function of daily liquidity conditions—stronger when total excess reserves are low.
- Magnitudes:
  - Supply-side effect roughly twice as strong as demand-side effect.
  - One-standard-deviation shock to Herfindahl suppliers (resp. demanders) moves spread by about 4 bps (resp. 2 bps) when RER is at sample minimum, and no impact when ample liquidity.
  - A shift of a Herfindahl index from sample mean to maximum could impact spread by 20 to 25 bps on same day under low RER.

### Key empirical findings — volatility
- Variance equation supports EGARCH specification.
- Volatility of the overnight interest rate:
  - Clearly stronger when liquidity is low.
  - Especially stronger when the banking system has accumulated a reserve deficit since the start of the maintenance period.
- Additional volatility drivers:
  - End-of-month and post-holiday periods show volatility bouts.
  - Changes in policy rates associated with increased volatility beforehand and decreased volatility after implementation.

### Extended specification (by maintenance day) — selected findings
- Method: replace Shock (and interactions) by Shock multiplied by each day dummy to estimate day-by-day liquidity effects.
- Day-specific liquidity effects:
  - Highly significant liquidity effects on days 4, 7, and 11 of maintenance period.
  - Particularly strong effect on day 7 when liquidity is scarce.
  - No significant liquidity effect on last day of maintenance period (day 14).
- Distribution effects:
  - Remain present; demand-side effect becomes stronger and significantly convex in extended specification.
- Reserve accumulation pattern:
  - Cumulative sum now significantly and strongly affects the liquidity effect.
- Regression summary (extended):
  - R-squared 0.15; Adjusted R-squared 0.05; S.E. of regression 0.11; Log likelihood 937.82; Durbin-Watson stat 2.25.

### Operational and policy implications
- Monitoring and indicators:
  - Monitor daily total excess reserves and pattern of reserve accumulation within the maintenance period to anticipate strength of liquidity effect and volatility in overnight rate.
  - Track concentration measures (normalized Herfindahl indices for suppliers and demanders) as additional signals on interbank price formation:
    - Higher supplier concentration signals upward pressure on rates.
    - Higher demander concentration signals downward pressure on rates.
- Implications for central bank operations when open market operation frequency is limited and overnight rate is not the operating target:
  - Account for time-varying excess liquidity conditions.
  - Account for distributional characteristics of surplus/deficit across banks.
  - Account for reserve accumulation trajectories during the maintenance period.
- Practical identification tool:
  - Use of morning pre-trade projections and comparison with evening realizations can serve to identify exogenous daily liquidity shocks.
- Policy evaluation:
  - CBJ forecast errors can quantify the degree of control a central bank can exert on the overnight interbank rate via open market operations.
  - Distributional changes are largely outside day-to-day central bank control but provide an easy-to-compute daily indicator and a way to assess how to neutralize distributional impacts through operations.

*Source: _wp09228 - 1.         Summary         Statistics...................................................................................*

### 1.         Summary         Statistics...................................................................................

### _wp09228 - 1.         Summary         Statistics...................................................................................

### Study scope and context
- Sample period: January 10, 2005 to March 29, 2007.
- Dataset: anonymous individual daily bank-by-bank data; 543 observations.
- Policy context: Central Bank of Jordan (CBJ) operated a currency peg to the U.S. dollar (USD).
- CBJ practices during sample:
  - Set the level of its overnight deposit facility equal to the level of the Fed Funds target, imposing a positive spread over the U.S. rate.
  - Typically intervened bi-weekly in the money market through open market operations.
  - Maintained a spread of at least 250 bps between its short-term deposit and credit facilities.
- Operating implications: CBJ did not use the overnight rate as its operating target; day-to-day variations of the overnight rate were much larger than in the U.S. and largely reflected local liquidity conditions.

### Identification strategy and model specification
- Identification of exogenous daily reserve shocks:
  - Use the difference between morning projections by the CBJ and evening realizations of nonborrowed excess reserves as an exogenous source of variations in the daily change in nonborrowed excess reserves (following Carpenter and Demiralp (2006) approach).
  - Individual bank data are daily morning (pre-trade) projections of excess reserves for each bank.
- Variable of interest: daily change of the spread between the overnight rate and the rate of the overnight deposit facility.
- Econometric model (two equations estimated simultaneously):
  - Mean equation: ARMA(1,1)-X for the first difference of the spread.
  - Variance equation: EGARCH(1,1)-X for the volatility of the first difference of the spread.
- Extended specification: includes day-of-maintenance-period effects (10-day maintenance period), quadratic liquidity effects by day.

### Key empirical findings — liquidity effect
- The vigor of the liquidity effect is a negative and convex function of the liquidity conditions of the day:
  - On days when excess liquidity is scarcer, the liquidity effect gets stronger.
- Evidence of a quadratic liquidity effect by day of the maintenance period:
  - Clear evidence of a strong effect on 3 days out of the 10 in each maintenance period (despite loss of identification power in the extended specification).
- Accumulated reserves over the maintenance period:
  - Weak evidence that an accumulated reserves surplus or deficit over the maintenance period has a small direct effect on the overnight rate.
  - Stronger evidence of an indirect effect through its impact on the vigor of the liquidity effect: when the banking system is behind in the satisfaction of reserve requirements, the liquidity effect appears more powerful.
- Implication: both the liquidity conditions of the day and the pattern of reserve accumulation over the maintenance period affect determination of the overnight rate.

### Key empirical findings — distributional/concentration effects
- Construction of two normalized Herfindahl indices:
  - One for banks with projected positive excess reserves on that day (the “suppliers”).
  - One for banks with negative projected excess reserves on that day (the “demanders”).
- Statistically significant effects:
  - More concentrated supply pushes the rate up.
  - More concentrated demand pulls the rate down.
- Interaction with liquidity conditions:
  - The vigor of these distribution effects is a negative and convex function of daily liquidity conditions — stronger on days when total excess reserves of the banking system are low.

### Key empirical findings — volatility
- Variance equation results validate the EGARCH specification.
- Volatility of the overnight interest rate:
  - Clearly stronger when liquidity is low.
  - Especially stronger when the banking system has accumulated a reserve deficit since the start of the maintenance period.

### Operational and policy implications (derived from findings)
- Monitoring daily total excess reserves and the pattern of reserve accumulation within the maintenance period is critical to anticipate strength of liquidity effect and volatility in the overnight rate.
- Tracking concentration measures (normalized Herfindahl indices for suppliers and demanders) provides additional signal on price formation in the interbank market:
  - Higher supplier concentration signals upward pressure on rates.
  - Higher demander concentration signals downward pressure on rates.
- In contexts with limited open market operation frequency (e.g., bi-weekly interventions) and non-targeting of overnight rates, central banks should account for:
  - Time-varying excess liquidity conditions.
  - Distributional characteristics of surplus/deficit across banks.
  - Reserve accumulation trajectories during the maintenance period.
- Use of morning pre-trade projections and comparison with evening realizations can serve as a practical identification tool for exogenous daily liquidity shocks.

*Source: _wp09228 - 1.         Summary         Statistics...................................................................................*

### Section III provides further background on the monetary framework in Jordan and contains a

### _wp09228 - Section III provides further background on the monetary framework in Jordan and contains a

### Literature context and methodological choices
- Hamilton (1997) identifies a daily-frequency liquidity effect in the U.S. overnight funds market and uses forecast errors of changes in the Treasury balance to instrument exogenous liquidity shocks.
- Carpenter and Demiralp (2006) extend Hamilton with the Fed Board’s forecast errors and find:
  - a liquidity effect on most days of the maintenance period;
  - nonlinearity: large changes in reserves have a more significant effect than small changes;
  - the aggregate level of total reserve balances affects the magnitude of the liquidity effect.
- Comparative and country studies cited: Thornton (2006, 2007), Bindseil and Seitz (2001), Angelini (2002), Moschitz (2004), Wurtz and Krylova (2004), Uesugi (2002), Prati, Bartolini, and Bertola (2003), Gaspar, Quirós, and Mendizábal (2004), Jurgilas (2006).
- The paper emphasizes:
  - analyzing first differences (daily changes) rather than levels;
  - a richer set of explanatory variables and allowing a moving average term in the mean equation;
  - adopting an ARMA(1,1)-EGARCH(1,1)-X specification for mean and volatility.

### The monetary framework and CBJ operations (Jordan)
- Exchange rate regime and reserves:
  - Jordanian dinar (JD) pegged to the USD since 1997 (parity fixed at 0.709 Jordanian Dinar per USD with a band of +/- 0.15 percent around the central parity).
  - Gross usable international reserves: USD 6.1 billion (or 11 times short-term external debt at remaining maturity) as of end-December 2006.
- Operational tools during January 10, 2005 to March 29, 2007:
  - reserve requirements (local and foreign currency);
  - overnight deposit facility;
  - 7-day refinancing facility (maturity reduced to overnight in May 2007);
  - bi-weekly open-market operations (central bills with maturities of 3 months and 6 months).
- Reserve maintenance:
  - maintenance period: 14 days (Monday to the Sunday of the following week);
  - banks required to maintain 65 percent of reserve requirements at all times and could average remaining 35 percent over period;
  - required reserves in local currency: 8 percent of daily average balances (rules in place since September 6, 2004);
  - reserve requirements set on the 5th of each month.
- Market structure:
  - analysis excludes two Islamic banks; banking sector for this study: 21 banks, all private.
  - top three banks had combined market share over 40 percent as of end-2006.
  - banking system assets: JD 24.2 billion, or 240 percent of GDP, as of end-2006.

### Data and descriptive features of the overnight market
- Sample period and variables:
  - period: January 10, 2005 to March 29, 2007;
  - aggregate daily data: interbank overnight interest rate, volume of transactions, rates of two CBJ facilities, end-of-day reserves position;
  - bank-level daily data (anonymous): deposits on overnight window, borrowings from refinancing window, required reserves, CBJ morning projection of evening reserves.
- Liquidity situation and patterns:
  - system generally in excess liquidity; deposits on overnight window on last day of each maintenance period were large.
  - until mid-2006, excess liquidity ≈ fifty percent of required reserves; declined by half on average in most recent period.
  - daily volatility of excess reserves declined somewhat after mid-2006.
  - interbank trading volume increased as excess liquidity declined.
- Interbank spread and arbitrage:
  - spread between interbank rate and overnight deposit window remained strictly positive, ranging between 12 and 192 bps.
  - episodes of incomplete arbitrage: spike Oct 19-20, 2005 followed by drop of 170 bps on Oct 21 (day 7 of maintenance period) when a large amount of central bank bills matured.
  - positive spread coexisting with surplus liquidity possibly related to end-of-day reserve volatility and payments system features.
- Distribution measures:
  - constructed two normalized Herfindahl indices based on morning projections: one for banks with projected reserves exceeding required reserves (“suppliers”) and one for banks with projected reserves below required reserves (“demanders”).
  - Fluctuations typically small, with occasional spikes of concentration; supplier index variations larger in second half of sample when aggregate excess liquidity was lower.

### Within-maintenance-period behavior
- Theoretical expectation: overnight rate should be a martingale under rational expectations.
- Empirical observations:
  - interbank rate appears to peak near end of first week of maintenance period, but differences across days are small.
  - banks tend to deposit larger amounts at the overnight window as the maintenance period progresses (consistent with a system in surplus where banks build buffers early and deposit excess later).
  - ratio of cumulative excess reserves to cumulative required reserves falls as maintenance period progresses.

### Empirical strategy and model
- Dependent variable:
  - first difference of the spread between the interbank rate and the deposit window rate (Δs).
- Key liquidity measure:
  - non-borrowed excess reserves (evening closing balance minus required reserves plus amount held at overnight window net of repo borrowings outstanding) normalized by required reserves (RER = excess to required reserves).
- Additional controls:
  - cumulative sum variable: cumulative sum of closing balance minus required reserves since first day of maintenance period normalized by cumulative required reserves since first day.
  - normalized Herfindahl indices for suppliers and demanders and their first differences.
  - dummies: each day of maintenance period, beginning/end of month, beginning/end of quarter, day before/after holiday, day of and days surrounding change in overnight window rate.
- Identification of exogenous shocks:
  - CBJ morning forecast error (difference between actual evening closing balances and CBJ morning projections) normalized by required reserves used as instrument (“Shock”).
- Model specification:
  - ARMA(1,1)-EGARCH(1,1)-X:
    - mean: (Δs)t = φ(Δs)t-1 + m·Xt + εt + ψ εt-1
    - variance: log ht = ω + β log ht-1 + ν·Yt + α |εt-1/√ht-1| + γ εt-1/√ht-1
    - εt = √ht ut, ut ~ N(0,1)
  - Stationarity conditions: |φ| < 1 and |β| < 1.

### Baseline specification key estimates and summary statistics (selected)
- Regression summary:
  - R-squared 0.13; Adjusted R-squared 0.08; S.E. of regression 0.11; Log likelihood 963.57.
- Mean equation (selected coefficients, significance noted):
  - (Δs)t-1 = 0.62 *** (St.dev. 0.05, p-value 0.00)
  - Shock = -0.33 *** (0.05, 0.00)
  - Shock*RER = 0.91 *** (0.21, 0.00)
  - Shock*(RER)² = -0.65 *** (0.22, 0.00)
  - ∆(Herf. Suppliers) = 0.52 *** (0.13, 0.00)
  - ∆(Herf. Suppliers) * RER = -1.10 ** (0.48, 0.02)
  - ∆(Herf. Demanders) = -0.27 ** (0.13, 0.04)
  - ∆(Herf. Demanders) * RER = 0.65 * (0.37, 0.08)
  - Dummy (policy change) = -0.12 *** (0.02, 0.00)
  - Dummy (policy change + 1) = 0.05 *** (0.01, 0.00)
  - Dummy (policy change - 1) = 0.05 *** (0.01, 0.00)
  - MA(1) = -0.31 *** (0.07, 0.00)
- Variance equation (selected coefficients):
  - Constant = -1.33 *** (0.10, 0.00)
  - Alpha: ARCH(1) = 0.96 *** (0.07, 0.00)
  - Gamma: Asymmetry parameter = 0.31 *** (0.05, 0.00)
  - Beta: GARCH(1) = 0.87 *** (0.02, 0.00)
  - Excess Reserves = -0.81 *** (0.09, 0.00)
  - Cumulative Sum = -4.43 *** (0.97, 0.00)
  - Dummy end-of-month = 1.39 *** (0.34, 0.00)
  - Dummy (holiday + 1) = 2.35 *** (0.15, 0.00)

### Baseline empirical findings (level and distribution effects)
- Liquidity effect:
  - Robust negative liquidity effect: positive supply shock reduces spread; effect is negative and convex in the level of excess reserves (RER).
  - Quantified example: a positive supply shock representing 26 percent of required reserves (one standard deviation) would decrease the spread by about 10 bps when RER is close to minimum sample value, but would not move when liquidity conditions are loose.
  - First-order autocorrelation 0.62 implies additional smaller impacts over next 1–2 days.
- Distribution effects:
  - Significant distribution effects on both supplier and demander sides, contingent on RER and convex in RER.
  - Supply-side effect roughly twice as strong as demand-side effect.
  - A one-standard-deviation shock to Herfindahl suppliers (resp. demanders) increases (resp. decreases) the spread by about 4 bps (resp. 2 bps) when RER is at sample minimum, and no impact when ample liquidity.
  - A shift of a Herfindahl index from sample mean to maximum could impact spread by 20 to 25 bps on same day under low RER.
- Special day effects:
  - Only maintenance day dummy statistically significant: day 8 (small effect).
  - Spread declines by 12 bps on days of policy rate adjustment (unadjusted for lag), climbs by 5 bps the day before and 5 bps the day after; adjustment of spread to new policy rate is gradual over three business days.
- Volatility:
  - Higher level of excess reserves decreases volatility of spread.
  - Accumulated reserves since beginning of reserve period has an even stronger volatility-reducing effect.
  - Volatility bouts at end-of-month and after a holiday; changes in policy rates associated with increased volatility beforehand and decreased volatility after implementation.

### Extended specification (by maintenance day) — selected findings
- Approach: replace Shock (and interactions) by Shock multiplied by each day dummy (10 day dummies) to estimate day-by-day liquidity effects.
- Key day-specific liquidity effects:
  - Highly significant liquidity effects on days 4, 7, and 11 of maintenance period.
  - Particularly strong effect on day 7 when liquidity is scarce.
  - No significant liquidity effect on the last day of maintenance period (day 14), possibly because reserve requirements nearly fully met by end of day 13.
  - Shock*RER*Dummy day 7 coefficient indicates a large sensitivity of liquidity shocks on day 7.
- Distribution effects remain present; demand-side effect becomes stronger and significantly convex in extended specification.
- Reserve accumulation pattern (cumulative sum) now significantly and strongly affects the liquidity effect.
- Regression summary (extended):
  - R-squared 0.15; Adjusted R-squared 0.05; S.E. of regression 0.11; Log likelihood 937.82; Durbin-Watson stat 2.25.

### Summary conclusions and operational implications
- Main empirical conclusions:
  - A robust negative relationship exists between daily changes in excess reserves and daily changes in the spread between the interbank rate and CBJ overnight deposit facility rate.
  - The liquidity effect is much stronger when daily excess reserves are low and when the banking system has accumulated a reserve deficit since the start of the maintenance period.
  - Quantified example reiterated: on a day with tight liquidity and no surplus in accumulated reserves, a supply shock equivalent to 26 percent of required reserves would affect the spread by about 10 basis points on that day, with additional smaller impacts over the next 1–2 days.
  - Distribution of excess reserves matters: one-standard-deviation shock to supplier (resp. demander) index moves rate by about 3 to 4 (resp. 2 to 3) basis points under tight liquidity; negligible under average liquidity.
- Policy and operational implications:
  - Using CBJ forecast errors as exogenous shocks provides a practical means to quantify the degree of control a central bank can exert on the overnight interbank rate via open market operations.
  - Changes in the distribution of excess reserves are largely out of day-to-day central bank control, but:
    - provide an easy-to-compute daily indicator to monitor distribution of excess reserves; and
    - offer a practical way to measure distributional impacts on the market for reserves and assess how to neutralize them through monetary operations if desired.
- Generalizability:
  - Estimates are obtained within a specific operational framework and may not be perfectly robust to changes in that framework, but the approach is adaptable and replicable in other operational settings and in other emerging market countries.

*Source: IMF Working Paper — Section III and related sections (pages excerpted).*

### References

### _wp09228 - References

### Literature on Liquidity Effects and Central Bank Operations
- Angelini, P., 2002, “Liquidity and Announcement Effects in the Euro Area,” Banca d’Italia Temi di Discussione No. 451.
- Bindseil, U., and F. Seitz, 2001, “The Supply and Demand for Eurosystem Deposits the First 18 Months,” ECB Working Paper No. 44.
- Bindseil, U., 2004, Monetary Policy Implementation, Oxford University Press.
- Carpenter, S., and S. Demiralp, 2006, “The Liquidity Effect in the Federal Funds Market: Evidence from Daily Open Market Operations,” Journal of Money, Credit and Banking, 38(4).
- Gaspar, V., G. P. Quirós, and H. R. Mendizábal, 2004, “Interest Rate Determination in the Interbank Market,” ECB Working Paper, No. 351.
- Hardy, D., 1996, “Market Reactions to Changes in German Official Interest Rates,” Economic Research Group of the Deutsche Bundesbank, Discussion Paper 4/96.
- Hardy, D., 1997, “Market Information and Signaling in Central Bank Operations, or, How Oftern Should a Central Bank Intervene?” Staff Papers, International Monetary Fund, Vol. 44(April), pp. 510-33.
- Mayes, D., and J. Toporowski, 2007, Open Market Operations and Financial Markets, Routledge.
- Moschitz, J., 2004, “The Determinants of the Overnight Interest Rate in the Euro Area,” ECB Working Paper No. 393.
- Prati, A., L. Bartolini, and G. Bertola, 2003, “The Overnight Interbank Market: Evidence from the G-7 and the Euro Zone,” Journal of Banking and Finance, 27, pp. 2045-83.
- Thornton, D.L., 2001, “Identifying the Liquidity Effect at the Daily Frequency,” Federal Reserve Bank of St. Louis Review, 83(4), pp. 59-78.
- Thornton, D.L., 2006, “The Daily Liquidity Effect,” Federal Reserve Bank of Saint Louis Working Paper 2006-020.
- Thornton, D.L., 2007, “The Daily and Policy-Relevant Liquidity Effects,” Federal Reserve Bank of Saint Louis Working Paper 2007-001A.
- Wurtz, F.R., and E. Krylova, 2004, “The Liquidity Effect in the Euro Area,” paper presented at the January 20-21, 2005 ECB Conference on Monetary Policy Implementation: Lessons from the Past and Challenges Ahead.

### Methodological and Econometric Sources
- Hamilton, J. D., 1994, “Time Series Analysis,” Princeton University Press, Princeton, New Jersey.
- Hamilton, J. D., 1996, “The daily Market for Federal Funds,” Journal of Political Economy, 104(1), 26-56.
- Hamilton, J. D., 1997, “Measuring the Liquidity Effect,” American Economic Review, 87(1), 80-97.
- He, C., T. Teräsvirta, and H. Malmsten, 2002, “Moment Structure of a Family of First-Order Exponential GARCH Models,” Econometric Theory, 18, pp. 868-85.
- Karanasos, M., and J. Kim, 2003, “Moments of the ARMA-EGARCH Model,” Econometrics Journal, 6, pp. 146-166.
- Nelson, D. B., 1991, “Conditional Heteroscedasticity in Asset Returns: A New Approach,” Econometrica, 59(2), pp. 347-70.

### Regional and Market-Specific Studies
- Jurgilas, M., 2006, “Interbank Markets under Currency Boards,” University of Connecticut Department of Economics Working Paper Series No. 2006-19.
- Uesugi, I., 2002, “Measuring the Liquidity Effect: The Case of Japan,” Journal of the Japanese and International Economies, 16(3), pp. 289-316.

*Source: _wp09228 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09228.pdf_