## _wp12143

## Source details

**Canonical URL:** [_wp12143](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12143.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12143.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12143.pdf.json)

---

### I. INTRODUCTION
- In most low-income countries (LICs), fiscal policy has not traditionally been a viable instrument for macroeconomic stabilization and has often behaved procyclically and/or aggravated exogenous shocks.
- Central bank autonomy in LICs has increased de jure and de facto, while fiscal institution reform has usually lagged—shifting primary stabilization responsibility to central banks.
- The effectiveness of central banks depends on whether their instruments can influence aggregate demand via monetary transmission, which may be significantly weaker in LICs than in advanced and emerging economies.
- The financial structure of LICs suggests the bank lending channel is likely the dominant channel, conditional on:
  - the domestic institutional context;
  - the structure of the banking system;
  - the intrinsic stability of the domestic macroeconomic environment.
- Weak and unreliable monetary transmission counsels restraint in monetary policy and suggests placing primary stabilization responsibility on central banks may be misguided.
- Empirical focus: ultimate effects of central bank actions on aggregate demand using impulse response functions (IRFs) from estimated vector autoregressions (VARs).
- Organization: theoretical rationale (Section II); VAR methodology and identification (Section III); regional evidence across five geographic regions (Sections IV–IX); summary and implications (Section X).

### II. MONETARY TRANSMISSION AND THE BANK LENDING CHANNEL
- Standard transmission channels in advanced economies: interest rate, asset, credit (bank lending and balance sheet channels), and exchange rate channels.
- Constraints in LICs that weaken non-bank channels:
  - absence of well-functioning markets for fixed-income securities, equities, and real estate;
  - very imperfect links with private international capital markets;
  - heavy central bank intervention in foreign exchange markets.
- Implication: limited scope for conventional interest rate, asset, or exchange rate channels; dominance of the bank lending channel where any transmission exists.
- Two critical links required for bank-dominated transmission:
  1. Link between monetary policy actions and availability/cost of bank credit.
  2. Link between availability/cost of bank credit and aggregate demand.
- Reasons the first link may be weak:
  - Noncompetitive banking industry: changes in banks’ costs of funds may be absorbed in profit margins rather than passed to lending supply.
  - Poor institutional environment: increases cost of bank lending and induces lending restrictions that weaken policy effects on loan supply.
- Appendix 1 model: central bank actions have weak effects on commercial bank lending rates when:
  - marginal cost of new lending is sharply increasing in loan volume;
  - degree of competition in banking sector is weak.
- Reliability of the bank lending channel depends on stability of factors determining banks’ intermediation costs; in LICs such factors may be non-structural and frequently changing → transmission may be both weak and unreliable. Empirical assessment required.

### III. EMPIRICAL METHODOLOGY
- Objective: measure empirical effect of monetary policy on aggregate demand via VAR-derived IRFs.
- Key methodological issue: identifying an empirically observable indicator of the stance of monetary policy (intermediate target), which may be:
  - a monetary aggregate;
  - a short-term interest rate;
  - the exchange rate;
  - a combination (e.g., a monetary conditions index).
- Misidentification risk: wrong intermediate target may yield correlations driven by common third factors rather than causal monetary policy effects; such correlations are likely sample-specific.
- Two identification approaches in VARs:
  1. A priori identification based on known operating procedures of monetary authorities.
  2. Inclusion of candidate intermediate targets (typically short-term interest rate and narrow monetary aggregate) in the VAR and imposing restrictions on reduced-form VAR innovations to extract structural monetary policy innovations.
- Example framework (Gordon and Leeper, 1994): structural modeling of reserves market combined with VAR residuals to extract monetary policy shock es from residuals.

### 2. Identifying exogenous monetary policy shocks
- Problem: VAR innovations in a policy instrument may combine exogenous policy shocks and endogenous contemporaneous policy responses to other variables in the system.
- Extracting the exogenous component requires departing from atheoretic VARs and is controversial.
- Identification methods described:
  i. Choleski decompositions (recursiveness assumption)
  - Assumes recursive contemporaneous ordering of variables; implementation depends critically on variable ordering (examples: monetary policy variable ordered first or last).
  - Limitations: ignores possibility that policy reacts contemporaneously to other macro variables in policymakers’ information sets.
  ii. Non-recursive (simultaneous) identification
  - When non-predetermined variables are in the central bank’s information set, additional a priori restrictions equal to the number of such variables are required; if rank condition satisfied, shock can be identified in a fully simultaneous system.
  - Sims and Zha (1998) example: augmenting systems requires imposing exclusion restrictions (examples provided) so some shocks solved simultaneously.
  - Kim and Roubini (2000) small open-economy model: seven-variable system with assumptions on which variables are observed contemporaneously by the central bank; R may be contemporaneously unaffected by Y and P.
- Estimation notes:
  - When monetary policy shock is not orthogonal to variables in the central bank’s information set, it cannot be extracted from a simple OLS regression of the policy variable on that information set.
  - If VAR contains variables predetermined with respect to monetary policy, the monetary policy shock can be extracted with an IV regression using predetermined variables as instruments.
  - When the central bank does not react to the exchange rate, the simultaneous scheme simplifies to solving money demand and monetary policy shocks simultaneously.
  - A third approach relying on long-run restrictions is noted but not reviewed for LIC literature here.

### 3. Exploring channels of transmission: methodology
- Two principal VAR-based approaches to identify channels (e.g., bank lending channel):
  i. Test whether a monetary policy shock significantly affects the candidate channel variable in the predicted direction (e.g., contraction → contraction in bank credit or increase in lending rates). An affirmative finding indicates a statistically well-defined role but not quantitative strength.
  ii. Ramey (1993) simulation: compare IRFs of aggregate demand when the channel variable is endogenous versus treated as exogenous. The difference measures the quantitative strength of the channel.

### Regional evidence and empirical findings (selected highlights)
- General patterns:
  - Most VAR studies for developing and low-income countries use short-term interest rates as policy indicator; many employ recursive identification though structural schemes are also used.
  - Across regions, evidence is frequently weak or mixed for statistically significant effects of monetary policy shocks on real output and the price level.
  - Two recurring interpretations:
    - methodological shortcomings (misspecified open-economy features, arbitrary identification, ignoring regime changes);
    - facts on the ground (financial structure and institutional features weaken or destabilize transmission).
  - Stronger evidence appears in relatively more prosperous and institutionally developed developing countries (some CEE economies, Morocco, Tunisia).

- Sub-Saharan Africa
  - Financial-structure features:
    - Banks hold large liquid buffers; tendencies for banks to hold "30-50 percent" of deposits as reserves at the central bank and in short-term foreign assets (Sacerdoti (2005)).
    - Saxegaard (2006) estimated excess reserves amounted to "over 13 percent of deposits on average in sub-Saharan banking systems in 2004."
    - Domestic bond and stock markets poorly developed; small formal financial sectors; larger net interest margins and higher concentration ratios; poor institutional environment.
  - Empirical findings (mixed methods, Table 2 studies):
    - Many recursive-identified studies find no unambiguous large and statistically significant effects of monetary policy innovations on prices and output.
    - Cheng (2006, Kenya): policy-driven interest rates had considerable impact on the price level and exchange rate but not on real output.
    - Saxegaard (2006): in Nigeria and Uganda, monetary shocks have weaker effects on output and inflation in high excess-reserve regimes; in CEMAC countries, weak effects under both regimes.
    - Lungu (2008): monetary innovations affected bank lending and deposit rates but did not affect output or prices (Ramey-style test).
    - Abradu-Otoo et al. (2003, Ghana): no statistically significant effects with either M2 or the T-bill rate as policy variable (1969–2002); point estimates sometimes inconsistent with theory.
    - Davoodi, Dixit, Pinto (2012, five EAC countries): generally no consistent statistically significant effects on output and prices using base money or T-bill rate across Burundi, Kenya, Rwanda, Tanzania, Uganda.
    - Montiel et al. (2012, Tanzania): small but statistically significant effects on the price level over short horizons with monetary base as policy instrument, but no effect on output.
    - Structural studies largely confirm weak evidence: Cheng (2006) structural results similar to recursive; Ngalawa (2009, Malawi) finds expected-direction effects but IRFs not statistically significant; Mugume (2011, Uganda) finds statistically insignificant effects on output and price level.
  - Conclusion for region: "very weak evidence at best" for effective monetary transmission regardless of policy indicator or identification method.

- Central and Eastern Europe (CEE)
  - More advanced than sub-Saharan LICs, but less than advanced economies.
  - Findings:
    - Ganev et al. (2002): some evidence that policy rates pass to commercial bank lending rates, but almost no evidence of lending rates affecting aggregate demand.
    - Egert and Macdonald (2009): short-end pass-through reasonably complete; transmission to longer-maturity rates weak; bank lending channel evidence limited and mixed. Methodological choices affect results substantially.
    - Country studies heterogeneous and sometimes counterintuitive (price puzzles, nonstandard output responses).

- Central Asia
  - Resemble sub-Saharan Africa: low external integration, heavy FX intervention, small securities markets, concentrated banking, weak institutions.
  - Findings:
    - Isakova (2008, Kazakhstan, Kyrgyz Republic, Tajikistan): policy rates pass to money market rates, but weak responses of prices and output → bank lending channel unimportant.
    - Georgia: mixed results; evidence points to transmission not primarily via bank lending but possibly via currency in circulation.

- Middle East and North Africa (MENA)
  - On average similar to sub-Saharan Africa and Central Asia, with exceptions (Bahrain, Oman, Tunisia).
  - Findings:
    - Ziaei (2009): negative association between policy rates and bank lending in cointegrating vectors for ten MENA countries; second-stage effects on aggregate demand not assessed.
    - Neaime (2008): IRF point estimates generally in expected direction for Lebanon and Tunisia, but statistically significant effects not present in LIC cases.
    - Boughrara (2008): Ramey-style test finds lending-channel evidence emerging after 12 quarters in Morocco and more immediate support for Tunisia.
    - Egypt studies: ambiguous or negligible effects on real output and prices; some puzzling result that public-sector credit played a larger transmission role than private-sector credit.

- Asia-Pacific
  - LICs show underdeveloped markets; India is relatively more developed.
  - Findings:
    - Pakistan (Agha et al. (2005), Alam and Waheed (2006)): mixed results; small or statistically insignificant aggregate output responses; sectoral effects sometimes significant.
    - India (Mallick (2009)): recursive VAR shows weak but statistically significant reduction in real output in quarters 2–3 after contractionary shock; price-level response counterintuitively positive; Sims-Zha structural identification yields no statistically significant output effect and a positive price puzzle.
    - Fiji and Papua New Guinea (Ahmad (2008)): variance-decomposition evidence suggests differing roles of reserves/deposits/loans in output variation; statistical significance not tested.

- Latin America and the Caribbean
  - LICs more integrated internationally; exchange rate regimes not usually very flexible; larger stock markets but less active trading; banking systems larger with larger net interest margins.
  - Findings:
    - Caribbean (Kendall (2001)): heterogeneous and weak lending-rate responses to policy across countries.
    - Costa Rica (Durán-Víquez and Esquivel-Monge (2008)): policy rate Granger-causes deposit and loan rates; almost complete long-run pass-through to deposit and lending rates; loan and deposit rates take "9.4" and "5" months respectively to fully pass a policy shock.
    - Jamaica (Robinson and Robinson (1997)): recursive VAR suggests counterintuitive short-run output increases to tightening and a sustained price puzzle.
    - Trinidad and Tobago (Ramlogan (2007), structural): required-reserves shocks lead to expected decreases in bank credit, slower growth, and lower inflation.

### Summary and policy implications
- Overall assessment: evidence for effective monetary transmission in LICs is weak and inconsistent; no cases where multiple careful studies confirmed effects similar to advanced-economy consensus were identified.
- Likely drivers of weak evidence:
  - Methodological shortcomings in empirical studies:
    - Omission of open-economy features and external variables that policymakers respond to.
    - Identification assumptions not grounded in independent evidence about domestic policy practice.
    - Limited attention to within-sample regime changes (financial reforms, exchange rate regime shifts, capital account liberalization).
  - Facts on the ground:
    - Structural features (small formal financial sectors, illiquid securities markets, concentrated banking systems, large bank liquidity holdings, poor institutional environment) plausibly weaken bank lending and other channels.
    - Stronger transmission in relatively more prosperous and institutionally developed developing countries (some CEE economies, Morocco, Tunisia) suggests structural factors matter.
- Research implication: distinguishing methodological deficiencies from structural constraints requires careful country-specific studies incorporating open-economy features, realistic policy-grounded identification, and attention to regime changes.
- Policy implication: if structural features materially weaken transmission, the stabilization challenge in developing countries is acute; identifying ways to enhance monetary policy effectiveness is a priority for policy and research.

### Appendix: model linking institutional quality, bank competition, and the bank lending channel (selected elements)
- Banking-sector setup and identities:
  - Balance identity: C = L + B + R – D (equation (1)).
  - Loan demand: L = L(i_L), L’ < 0 (equation (2a)).
  - Deposit supply: D = D(i_D), D’ > 0 (equation (2b)).
  - Intermediation costs: c = c(L), with c’ > 0, c’’ > 0 (equation (3)). Institutional deterioration → larger c’’.
  - Liquidity premium on reserves: ρ = ρ(R/D), with ρ’ < 0, ρ’’ > 0 (equation (4)).
  - Central bank credit constraint: C ≤ C̄ (equation (5)), central bank charges i_C and may ration credit.
- Key first-order and behavioral relations:
  - If central bank credit constraint binds (λ > 0), requires i_B > i_C (implication of equation (6c)).
  - Bank demand for reserves: R = h(i_B) D with h’ < 0 (equation (7)).
  - Optimal lending rate as markup:
    - i_L = (1 - 1/ξ_L)^{-1} (i_B + c’(L)) where ξ_L = -L’i_L / L > 1 (equation (8)).
  - Deposit rate:
    - i_D = (1 + 1/ξ_D)^{-1} [ i_B (1 - h) + ρ h ] (equation (9)).
- Pass-through and comparative-statics (specific nonlinear cost example):
  - di_L / di_B = 1 / (2 + α_1 g’’) > 0 for L ≥ L* (equation (10)); pass-through falls as α_1 increases and as g’’ (curvature of intermediation costs) increases.
  - dL / di_B = -1 / (2/α_1 + g’’) < 0 (equation (11)); lending volume response weakens when banking monopoly power is stronger (α_1 small) and when intermediation costs rise sharply (g’’ large).
- Key qualitative implications:
  - Uncompetitive banking (greater monopoly power) and deficient institutional environments (steeply rising intermediation costs) reduce pass-through from policy rate i_B to banks’ lending rates and weaken policy effects on loan volumes.
  - Any factor that alters banks’ intermediation cost curve g’’ (macro instability, policy regime change, institutional shifts) will change both the strength and reliability of pass-through and lending responses.

*Source: _wp12143 - References (IMF working paper excerpt).*

### References .............................................................................................................

### _wp12143 - References

### I. INTRODUCTION
- In the vast majority of low-income countries (LICs), fiscal policy has not traditionally represented a viable instrument for macroeconomic stabilization and has often behaved procyclically and/or aggravated the effects of exogenous shocks.
- Recent economic reforms have increased central bank autonomy in LICs, both de jure and de facto, while fiscal institution reform has usually lagged—shifting primary responsibility for stabilization to central banks.
- The effectiveness of central banks in LICs depends on whether their policy instruments can influence aggregate demand via the monetary transmission mechanism, which may be significantly weaker in LICs than in advanced and emerging economies.
- The financial structure of LICs suggests the bank lending channel is likely to be the dominant channel, but its effectiveness depends on:
  - the domestic institutional context;
  - the structure of the banking system;
  - the intrinsic stability of the domestic macroeconomic environment.
- Weak and unreliable monetary transmission would counsel restraint in the use of monetary policy and suggest that placing primary stabilization responsibility on central banks may be misguided.
- The paper’s empirical focus is on the ultimate effects of central bank monetary policy actions on aggregate demand, primarily using impulse response functions (IRFs) from estimated vector autoregressions (VARs).
- Organization: theoretical rationale for bank lending dominance and weaknesses (Section II); VAR methodology and identification strategies (Section III); regional evidence organized across five geographic regions (Sections IV-IX); summary and implications (Section X).

### II. MONETARY TRANSMISSION AND THE BANK LENDING CHANNEL IN DEVELOPING COUNTRIES
- In advanced economies, monetary transmission typically operates via four mechanisms: the interest rate channel, the asset channel, the credit channel (including the bank lending and balance sheet channels), and the exchange rate channel.
- Mechanism summaries:
  - Interest rate channel: central bank actions affect short-term money-market rates, arbitrage transmits effects to long-term bond returns, affecting demand for capital goods.
  - Asset channel: arbitrage between bonds and equities/real assets affects stock and real estate values and household wealth.
  - Exchange rate channel: arbitrage between domestic and foreign currency assets affects the real exchange rate and the composition of demand.
  - Credit channel: credit market frictions create borrowers reliant on bank credit; changes in banking system reserves affect loan terms (bank lending channel) and external finance premiums (balance sheet channel).
- Characteristics of LICs that weaken non-bank channels:
  - absence of well-functioning markets for fixed-income securities, equities, and real estate;
  - very imperfect links with private international capital markets;
  - heavy central bank intervention in foreign exchange markets.
- Implication: limited scope for conventional interest rate, asset, or exchange rate channels; bank lending channel (with a derivative balance sheet channel acting as financial accelerator) is likely dominant where monetary transmission exists.
- Two critical links for bank-dominated transmission:
  1. Link between monetary policy actions and availability/cost of bank credit.
  2. Link between availability/cost of bank credit and aggregate demand.
- Reasons the first link may be weak:
  - Noncompetitive banking industry: changes in banks’ costs of funds may be reflected in bank profit margins rather than in lending supply.
  - Poor institutional environment: increases cost of bank lending and induces lending restrictions that weaken policy effects on loan supply.
- Appendix 1: simple model of bank lending behavior shows central bank actions will have weak effects on commercial bank lending rates when:
  - marginal cost of new lending is sharply increasing in loan volume;
  - degree of competition in banking sector is weak.
- Reliability of the bank lending channel depends on stability of factors determining banks’ intermediation costs; in LICs such factors may be non-structural and frequently changing → transmission may be both weak and unreliable. Empirical assessment is required.

### III. EMPIRICAL METHODOLOGY
- Literature focus: measuring empirical effect of monetary policy on aggregate demand via VAR-derived impulse response functions (IRFs); substantial body of work exists for the United States, advanced countries, emerging economies, and an expanding literature for LICs.
- Key methodological issue: identifying an empirically observable indicator of the stance of monetary policy (the central bank’s intermediate target), which may be:
  - a monetary aggregate;
  - a short-term interest rate;
  - the exchange rate;
  - a combination (e.g., a monetary conditions index).
- Misidentification risk: if the intermediate target is misidentified, correlations between the chosen indicator and aggregate demand may reflect common third-factor influences rather than causal monetary policy effects; such correlations are likely sample-specific and therefore uninformative about true policy effects.
- Two approaches to identifying the intermediate target in VARs:
  1. A priori identification based on known operating procedures of monetary authorities.
  2. Inclusion of candidate intermediate targets (typically a short-term interest rate and a narrow monetary aggregate) in the VAR and imposition of restrictions on reduced-form VAR innovations to extract structural monetary policy innovations.
- Example (Gordon and Leeper, 1994): monthly VAR with seven variables—unemployment rate, real output, price level, long-term interest rate, commodity prices, stock of reserves, federal funds rate—combined with a structural model of the reserves market:
  - M = a1 R + a2 P + a3 Y + ed (demand for reserves)
  - R = a4 M + a5 R10 + a6 CP + es (supply of reserves)
  - Notation: Y, P, R10, CP, M, and R denote innovations in reduced-form VAR equations for real output, the price level, the long-term interest rate, commodity prices, the stock of reserves, and the federal funds rate; ed and es are structural shocks to demand and supply of reserves.
  - Estimating these from VAR residuals allows extraction of es as the relevant monetary policy shock.
- Similar approaches: Bernanke and Mihov (1998) used a different reserves-market model to let data identify the intermediate target rather than imposing it a priori.

*Content based solely on the provided source excerpt.*

### 2. Identifying exogenous monetary policy shocks

### 2. Identifying exogenous monetary policy shocks

### Problem statement: endogeneity of VAR innovations
- Innovations in a reduced-form VAR for a short-term domestic interest rate do not necessarily represent exogenous monetary policy shocks because monetary policy may respond contemporaneously to other variables in the system.
- If the residual (innovation) in the monetary policy instrument equation is correlated with innovations in other macroeconomic variables, the residual may combine:
  - an exogenous monetary policy shock, and
  - an endogenous contemporaneous response of monetary policy to innovations in other variables.
- Example: interest rates may be unexpectedly raised because the monetary authorities observe an innovation in aggregate demand.
- The challenge is to remove the effects of such contemporaneous responses to identify the exogenous component (the monetary policy shock).
- Extracting the exogenous component requires departing from the “atheoretic” VAR stance and is the most controversial aspect of the methodology.
- This issue has been addressed in several ways (outlined below).

### i. Choleski decompositions (recursiveness assumption)
- Under a Choleski decomposition the relationship among reduced-form innovations is assumed recursive:
  - the reduced-form innovation in the first ordered variable is structural;
  - the second innovation is structural plus a contemporaneous response to the first structural shock;
  - the third innovation is structural plus contemporaneous responses to the first two structural shocks; and so on.
- Christiano, Eichenbaum and Evans (1999) refer to this as the “recursiveness assumption.”
- Crude implementations (Sims’ original “atheoretic approach”) often order the monetary policy variable first, implicitly assuming innovations in this variable are exogenous but may affect others contemporaneously.
  - Example ordering in a three-variable system: (M, Y, P).
  - Identifying assumptions for (M, Y, P): monetary policy does not react contemporaneously to innovations in Y and P; innovations in P do not affect Y contemporaneously.
- Bernanke and Blinder (1992) proposed ordering the monetary policy variable last on the assumption that endogenous macro variables in the VAR can be observed contemporaneously by policymakers, but policy variables affect macro variables with a lag.
  - Alternative ordering example: Y, P, R.
- Peersman and Smets (2001) used quarterly VAR with Choleski ordering (Y, P, R, RER):
  - Assumptions: ECB looks at real GDP and consumer prices but not at the exchange rate when setting policy; exchange rate is affected contemporaneously by preceding variables; neither R nor RER affects Y and P within the quarter.
  - This approach assumes the monetary policy shock is orthogonal to variables in the central bank’s information set (the shock neither is influenced contemporaneously by them nor influences them contemporaneously).
- Limitation: this neglects possibility that policy reacts contemporaneously to other macro variables that are part of policymakers’ information set.

### ii. Non-recursive (simultaneous) identification
- When non-predetermined variables enter the central bank’s information set, the recursiveness assumption fails and additional a priori restrictions are required equal in number to the number of non-predetermined variables included.
- If those additional restrictions satisfy the rank condition, the monetary policy shock can be identified in a fully simultaneous system.
- Sims and Zha (1998) approach (described in a simplified setting):
  - Example: augment Peersman and Smets four-variable system with money stock to yield a five-variable VAR.
  - The order condition requires 10 restrictions in this case.
  - Plausible exclusion restrictions example:
    - real output not affected contemporaneously by any other variables;
    - price level affected only by real output (adds 7 exclusions);
    - innovation in money stock interpreted as money demand innovation allows excluding exchange rate from that equation (eighth restriction);
    - if central bank observes exchange rate and relevant monetary stock contemporaneously, but not real output and price level, this yields the final two exclusion restrictions required for identification in the interest rate equation.
  - In this framework:
    - exchange rate may respond contemporaneously to all other variables;
    - shocks to real output and the price level can be solved for recursively;
    - shocks to monetary policy, money demand, and the exchange rate must be solved for simultaneously.
- Kim and Roubini (2000) structural model for a small open economy (generalization of Sims and Zha):
  - Seven variables: world commodity prices CP, external interest rate R*, domestic real output Y, domestic price level P, money stock M, domestic short-run interest rate R, nominal exchange rate S.
  - Assumptions:
    - CP treated as exogenous (small country assumption), unaffected by other shocks.
    - R* is affected by CP, and Y is affected by CP and R*; these two variables are unaffected by any other variables in the model on impact.
    - Output is affected in the short run only by supply shocks (with supply shocks driven by other than commodity prices identified separately).
    - P is affected in Phillips Curve fashion by CP and Y.
    - With a money demand interpretation, M is unaffected contemporaneously by CP, R*, or S.
    - Central bank can observe CP, R*, S, and the relevant monetary stock contemporaneously, but not Y and P; thus R is contemporaneously unaffected by Y and P.
    - As in Sims and Zha (1998), S may be contemporaneously affected by all other variables.

### Identification implications and estimation notes
- When monetary policy shock is not orthogonal to variables in the central bank’s information set, the shock cannot be extracted from a simple OLS regression of the policy variable on the information set.
- If the VAR contains variables predetermined with respect to monetary policy, the monetary policy shock can be extracted with an IV regression of the policy variable on the central bank’s information set using predetermined variables as instruments.
- When the central bank does not react to the exchange rate, the simultaneous scheme simplifies (becoming similar to Gordon and Leeper (1994) and Bernanke and Mihov (1998)), requiring only money demand and monetary policy shocks to be solved for simultaneously.
- A third approach relying on long-run restrictions on policy effects is noted but not reviewed in the low-income country literature covered here.

*Source: _wp12143 - 2. Identifying exogenous monetary policy shocks*

### 3. Exploring channels of transmission

### _wp12143 - 3. Exploring channels of transmission

### Methodological approaches to identify transmission channels
- Two principal approaches require inclusion in the VAR of an indicator variable for the channel of interest (example: stock of bank credit and/or bank lending rate for the bank lending channel).
- Approach (i): Test whether a monetary policy shock has a statistically significant effect (IRF) on the candidate channel variable in the theoretically predicted direction (e.g., a monetary contraction → contraction in bank credit and/or an increase in lending rates). An affirmative result indicates a statistically well-defined role but not quantitative strength.
- Approach (ii) (Ramey (1993) simulation): Compare IRFs of the aggregate demand indicator (output or prices) to a monetary shock when the channel variable is allowed to respond endogenously versus when it is treated as exogenous. The difference between the two IRFs measures the quantitative strength of the channel.

### Evidence synthesis: general patterns
- Most recent VAR studies for developing and low-income countries use short-term interest rates as the monetary policy indicator; many also employ recursive identification (Bernanke-Blinder or Peersman-Smets) though structural schemes are used in some studies.
- Across regions, evidence is frequently weak or mixed for statistically significant effects of monetary policy shocks on both aggregate demand indicators (real output and the price level).
- Two recurring interpretations: (a) methodological shortcomings in empirical studies (misspecified open-economy features, arbitrary identification assumptions, limited attention to within-sample regime changes), and (b) "facts on the ground"—features of financial structure and institutional environment that weaken or destabilize transmission.
- The strongest evidence for conventional transmission emerges in relatively more prosperous and institutionally developed developing countries (examples in the text: some CEE transition economies, Morocco, Tunisia), suggesting facts on the ground likely matter substantially.

### Sub-Saharan Africa: evidence and mechanisms
- Financial-structure features relevant for weak transmission:
  - Banks hold large liquid buffers: reported tendencies for banks to hold "30-50 percent" of deposits as reserves at the central bank and in short-term foreign assets (Sacerdoti (2005)).
  - Saxegaard (2006) estimated excess reserves amounted to "over 13 percent of deposits on average in sub-Saharan banking systems in 2004."
  - Domestic bond and stock markets poorly developed; small formal financial sectors; larger net interest margins and higher concentration ratios; poor institutional environment.
- Empirical findings (Table 2 studies, mixed methods):
  - Many recursive-identified studies find no unambiguous large and statistically significant effects of monetary policy innovations on both prices and output.
  - Cheng (2006, Kenya): policy-driven interest rates had considerable impact on the price level and exchange rate but not on real output.
  - Saxegaard (2006): in Nigeria and Uganda, monetary shocks have weaker effects on output and inflation in high excess-reserve regimes; in CEMAC countries, weak effects under both regimes.
  - Lungu (2008): Ramey-style tests—monetary innovations affected bank lending and deposit rates but did not affect either output or prices.
  - Abradu-Otoo et al. (2003, Ghana): no statistically significant effects with either M2 or the T-bill rate as policy variable (1969–2002), with point estimates sometimes inconsistent with theory.
  - Davoodi, Dixit, Pinto (2012, five EAC countries): mixed results; generally no consistent statistically significant effects on output and prices using base money or T-bill rate as policy variables across Burundi, Kenya, Rwanda, Tanzania, Uganda.
  - Montiel et al. (2012, Tanzania): small but statistically significant effects on the price level over short horizons with monetary base as policy instrument, but no effect on output.
  - Structural studies (Panel B) largely confirm weak overall evidence: Cheng (2006) structural results similar to recursive; Ngalawa (2009, Malawi) finds expected-direction effects but IRFs not statistically significant; Mugume (2011, Uganda) finds statistically insignificant effects on output and the price level.
- Conclusion for region: "very weak evidence at best" for effective monetary transmission in sub-Saharan Africa regardless of policy indicator (interest rate or monetary base) or identification method.

### Transition economies in Central and Eastern Europe (CEE)
- These countries are generally more advanced than sub-Saharan LICs but still show less integration and development than advanced economies.
- Literature overview:
  - Ganev et al. (2002): some evidence that policy rates pass through to commercial bank lending rates, but almost no evidence of lending rates affecting aggregate demand.
  - Egert and Macdonald (2009): pass-through at the short end of the maturity spectrum is reasonably complete; transmission to longer-maturity rates weak; evidence for bank lending channel and links from lending rates to aggregate demand remains limited and mixed. They note that methodological choices (recursive vs. structural, sample splits, time-varying coefficients) substantially affect results.
  - Country studies produce heterogenous and sometimes counterintuitive results (price puzzles, nonstandard output responses). Examples:
    - Lang and Krznar (2004, Croatia): contraction reduces real GDP short-run and improves current account but counterintuitively increases the price level.
    - Elbourne and de Haan (2009): structural identification gives more theory-consistent results; large theory-consistent effects found mainly for higher-income Czech Republic and Poland; Slovak Republic shows small and/or counterintuitive effects.

### Transition economies in Central Asia
- These economies resemble sub-Saharan Africa in low international financial integration, heavy foreign exchange intervention, small securities markets, concentrated banking systems, and weak institutional environments.
- Empirical findings:
  - Isakova (2008, Kazakhstan, Kyrgyz Republic, Tajikistan): policy rates pass effectively to money market rates, but weak responses of prices and output suggest the bank lending channel has been unimportant.
  - Georgia: mixed results—Bakradze and Billmeier (2007) (currency in circulation as policy variable) find positive and statistically significant effects on output only for first two quarters and no significant effect on prices; Samkharadze (2008, monthly data, recursive and structural) finds significant short-horizon effects of monetary aggregates on prices and output, but transmission appears not to operate primarily through bank lending (instead via currency in circulation).

### Middle East and North Africa (MENA)
- On average, MENA LICs resemble sub-Saharan Africa and central Asia in features that weaken conventional transmission, though some countries (Bahrain, Oman, Tunisia) have more favorable characteristics.
- Selected empirical findings (Table 8):
  - Ziaei (2009, multi-country): negative association between policy rates and bank lending in cointegrating vectors for ten MENA countries (policy-rate increases move bank lending opposite to the rate change in the short run), but second-stage effects on aggregate demand not assessed.
  - Neaime (2008): four-variable VARs for Lebanon and Tunisia find IRF point estimates generally in expected direction, but statistically significant effects on prices and output not present in those LIC cases.
  - Boughrara (2008, Tunisia and Morocco): no effects of monetary innovations on exchange rate or asset prices; Ramey-style test finds significant loan-endogeneity differences in Morocco (long-horizon lending-channel pattern emerges after 12 quarters) and more immediate lending-channel support for Tunisia.
  - Egypt studies (Moursi et al. (2007); Al-Mashat and Billmeier (2007)): monetary policy shock variables align with episodes of tightening but effects on real output and prices are ambiguous or negligible; expanded systems show bank interest rates move appropriately but only deposit-rate IRFs attain statistical significance; Ramey tests find puzzlingly larger role for public-sector credit than private-sector credit in transmitting policy.

### Asia-Pacific
- LICs in the region show the familiar pattern of underdeveloped markets and less integration than advanced economies; India stands out as relatively more developed.
- Empirical findings:
  - Pakistan (Agha et al. (2005)): Ramey approach suggests a positive shock to the T-bill rate reduces output but increases prices; confidence bands not reported.
  - Pakistan (Alam and Waheed (2006)): recursive VAR with confidence bands—aggregate output response to a monetary shock has the right sign but is small and statistically insignificant; sectoral statistically significant effects limited to manufacturing and wholesale/retail trade.
  - India (Mallick (2009)): recursive VAR shows a weak but statistically significant reduction in real output in quarters 2–3 after a contractionary policy shock; price-level response is counterintuitively positive; Sims-Zha structural identification yields no statistically significant effect on output and a positive price puzzle.
  - Fiji and Papua New Guinea (Ahmad (2008)): variance-decomposition evidence (no IRFs reported) suggests bank reserves and deposits explain output variation in Fiji, while bank loans dominate in Papua New Guinea; statistical significance not tested.

### Latin America and the Caribbean
- LICs here tend to be more integrated with international markets; exchange rate regimes not usually very flexible; larger stock markets but less active trading; banking systems larger but with larger net interest margins.
- Empirical findings:
  - Caribbean multi-country (Kendall (2001)): heterogeneous and weak responses of lending rates to policy changes across Bahamas, Barbados, Belize, Guyana, Jamaica, Trinidad; required reserve shocks had surprising negative effects on lending rates in most countries.
  - Costa Rica (Durán-Víquez and Esquivel-Monge (2008)): policy rate Granger-causes deposit and loan rates; almost complete long-run pass-through to deposit and lending rates; loan and deposit rates take "9.4" and "5" months respectively to fully pass a policy shock.
  - Jamaica (Robinson and Robinson (1997)): recursive VAR suggests counterintuitive short-run output increases to tightening and a sustained price puzzle; confidence intervals not reported.
  - Trinidad and Tobago (Ramlogan (2007), structural): required-reserves shocks lead to expected decreases in bank credit, slower growth, and lower inflation; confidence intervals not reported.

### Summary and main conclusions
- Overall assessment: evidence for effective monetary transmission in low-income countries is weak and inconsistent. The review did not identify instances where multiple careful studies confirmed effects similar to the consensus in advanced economies.
- Likely drivers of weak evidence:
  - Methodological shortcomings in many empirical studies:
    - Omission of open-economy considerations despite policymakers reacting to external variables (exchange rate interactions can mis-specify monetary shocks).
    - Identification assumptions often not grounded in independent evidence about how monetary authorities operate in-country; lags and information sets frequently borrowed from industrial-country literature.
    - Limited attention to within-sample regime changes (financial reforms, exchange rate regime shifts, capital account liberalization).
  - Facts on the ground:
    - Structural features (small formal financial sectors, illiquid securities markets, concentrated banking systems, large bank liquidity holdings, poor institutional environment) plausibly weaken the bank lending channel and other conventional channels.
    - Evidence that stronger monetary transmission occurs in relatively more prosperous and institutionally developed developing countries (some CEE economies, Morocco, Tunisia) supports the view that facts on the ground matter.
- Research implication: distinguishing between methodological deficiencies and structural constraints requires careful country-specific studies that incorporate open-economy features, realistic identification grounded in policy practice, and attention to regime changes.
- Policy implication highlighted: if the conjecture that structural features materially weaken transmission is valid, the stabilization challenge in developing countries is acute and identifying means to enhance effectiveness of monetary policy is a priority for policy and research.

### Appendix summary: model linking institutional quality, bank competition, and the bank lending channel
- Banking-sector setup:
  - Balance identity: C = L + B + R – D (equation (1)).
  - Loan demand: L = L(i_L), L’ < 0 (equation (2a)).
  - Deposit supply: D = D(i_D), D’ > 0 (equation (2b)).
  - Intermediation costs: c = c(L), with c’ > 0, c’’ > 0 (equation (3)). Institutional deterioration → larger c’’ (costs rise faster with lending volume).
  - Liquidity premium on reserves: ρ = ρ(R/D), with ρ’ < 0, ρ’’ > 0 (equation (4)).
  - Central bank credit constraint: C ≤ C̄ (equation (5)), with central bank charging i_C and rationing credit.
- First-order conditions and implications (selected):
  - If central bank credit constraint binds (λ > 0), requires i_B > i_C (equation (6c) implication).
  - Bank demand for reserves R = h(i_B) D with h’ < 0 (equation (7)).
  - Optimal lending rate expressed as a markup: i_L = (1 - 1/ξ_L)^{-1} (i_B + c’(L)) where ξ_L = -L’i_L / L > 1 (equation (8)).
  - Deposit rate: i_D = (1 + 1/ξ_D)^{-1} [ i_B (1 - h) + ρ h ] (equation (9)).
- Pass-through and comparative-static results:
  - For a specific nonlinear cost example (linear loan demand, c(L) piecewise with steeply rising costs beyond L*), pass-through derivative derived:
    - di_L / di_B = 1 / (2 + α_1 g’’) > 0 for L ≥ L* (equation (10)); pass-through falls as α_1 increases (α_1 is loan demand slope) and as g’’ (curvature of intermediation costs) increases.
    - dL / di_B = -1 / (2/α_1 + g’’) < 0 (equation (11)); lending volume response weakens when banking monopoly power is stronger (α_1 small) and intermediation costs rise sharply (g’’ large).
  - Key qualitative implications:
    - Uncompetitive banking (greater monopoly power) and deficient institutional environments (steeply rising intermediation costs) reduce pass-through from policy rate (i_B) to banks’ lending rates and weaken the effect of policy on loan volumes.
    - Any factor that alters banks’ intermediation cost curve (g’’), including macro instability, policy regime change, or institutional shifts, will alter both the strength and reliability of pass-through and lending responses.

*Source: _wp12143 - 3. Exploring channels of transmission (IMF PDF chapter).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12143.pdf_