## _wp11259

## Source details

**Canonical URL:** [_wp11259](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11259.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11259.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11259.pdf.json)

---

### Introduction and research objectives
- Focus: transmission mechanism of monetary policy and effects of risk-premium shocks in Hungary during the inflation targeting period.
- Sample period: June 2001 until September 2010.
- Methodology: a large Bayesian Vector Autoregression (large BVAR) using information from "more than 100 macroeconomic and financial series".
- Main aims:
  - "look inside the 'black box' of monetary transmission" by using a comprehensive dataset;
  - analyze how a monetary policy shock and a risk-premium shock feed through the economy;
  - evaluate forecasting performance of the large BVAR compared with smaller models.

### Stylized facts about the Hungarian economy (context and vulnerabilities)
- Small open economy with strong exposure to international shocks via trade and capital flows.
- Exchange rate regimes and history:
  - Pre-inflation targeting: crawling-peg with allowed deviations ±2.5 percent.
  - After 2001: horizontal bands of ±15 percent (removed in February 2008).
  - IMF-supported program introduced in November 2008 allowing intervention.
- Key macro-financial vulnerabilities (figures refer to 2010Q3):
  - General government debt: close to 90 percent of GDP.
  - Gross external debt: over 160 percent of GDP.
  - Deposits in foreign currency: about one fifth of total deposits.
  - Share of foreign currency liabilities of households: exceeded 60 percent.
  - Share of foreign currency liabilities of corporations: around 25 percent.
  - Portfolio investments held by nonresidents: close to 50 percent of GDP (about half of which was government debt).
  - Banking sector foreign ownership: foreign bank subsidiaries holding about two thirds of total banking assets.
  - Banks funded from wholesale markets: loans exceeded deposits by about 30 percent.
- Policy environment implications:
  - Government bond yields and country risk premium systematically higher than regional peers (CE-3).
  - Policy rate systematically higher and more responsive to government bond yields than in regional peers.
  - International risk premia correlate with government yields and policy rates and pass through to private lending rates.

### Modeling approach and estimation strategy
- Rationale: large BVAR reduces omitted variable bias common to small VARs; literature finds factor models and large BVARs outperform small VARs in forecasting.
- Prior structure:
  - Minnesota prior augmented with an inverse Wishart prior for residual covariance and a sum-of-coefficients prior (Normal–Inverted Wishart conjugate prior).
  - Overall tightness hyperparameter λ chosen so that large model matches in-sample fit of a benchmark small VAR for four variables (industrial production, net core inflation, nominal effective exchange rate, monetary policy rate).
  - Sum-of-coefficients prior hyperparameter τ set to τ = 10λ (relative tightness).
- Data:
  - Monthly series June 2001–September 2010 from MNB and official sources.
  - Dataset: 98 domestic indicators and 13 foreign variables (international commodity prices, Germany/world industrial production and imports, sentiment, MSCI indices, Euro area policy rate).
  - Net core inflation used (indirect taxes and excises deducted).
- Implementation details preserved: dummy-observation implementation, Normal–Inverted Wishart posterior, and shrinkage design following Banbura et al. (2010) logic.

### Identification of structural shocks
- Two shocks analyzed: a monetary policy shock and a risk-premium shock — both calibrated as a one-time increase of 100 basis points.
- Identification method: recursive identification via Cholesky decomposition with exclusion restrictions on contemporaneous responses.
- Ordering for monetary policy shock:
  - External (slow) block first; domestic slow-moving variables (real activity, prices, wages, confidence, monetary/credit aggregates); policy rate next; financial (fast-moving) variables last (interest rates, exchange rates, stock index, country risk-premium).
  - Assumption: slow-moving variables do not respond contemporaneously to monetary policy shocks; financial variables do.
- Ordering for risk-premium shock:
  - Slow-moving variables first; EMBI sovereign spread next (the shock); fast-moving variables third with policy interest rate placed right after the shock in benchmark calibration.
  - Assumption: monetary policy may respond contemporaneously to a risk-premium shock; risk-premium shock may affect fast-moving variables within the period.
  - Robustness: placing the monetary policy rate prior to the risk-premium has no significant impact on results.

### Impulse responses to a 100 basis points monetary policy tightening — select quantitative findings
- Shock: 100 basis points.
- Policy persistence: about four months.
- Policy rate below equilibrium after easing: about 25 basis points.
- Interest rate channel:
  - 3-month treasury bills: +75 basis points immediately.
  - 10-year treasury bonds: +10 basis points immediately; −25 basis points after a quarter.
  - Corporate short-term borrowing (domestic currency): +0.6 percent (peak at ~3 months; back to zero ~1 year).
  - Corporate long-term borrowing (domestic currency): +0.5 percent (peak at ~3 months; back to zero ~1 year).
  - Consumption credit rates: +25 basis points.
  - Mortgage loan rates: +15 basis points.
- Exchange rate channel:
  - Forint nominal effective exchange rate appreciates by about 1 percent after three months; appreciation reverses after about six months.
- Balance sheet and credit channels:
  - Stock of forex loans: −1 percent for the first three months.
  - Forint loans to corporates: −20 percent immediately; forint lending recovers in < six months.
  - Foreign currency lending: increases initially (substitution), reverts in ~ nine months.
  - Net loans to households: −5 percent for both forint-denominated and foreign-currency lending.
- Asset prices and expectations:
  - Stock market (BUX): declines by around 4 percent for about one quarter.
  - Country risk (EMBI) increase: 20 basis points; elevated for ~ three quarters.
  - Economic activity expectations fall temporarily (about two quarters), followed by medium-term improvement.
- Real activity, labor market and prices:
  - Industrial production and manufacturing: −1 to −2 percent (hump-shaped), trough after ≈ one quarter, revert to trend after ≈ one year.
  - Unemployment: +0.05 percent.
  - Hours worked: −0.5 percent (revert after one year).
  - Wages: − up to 0.5 percent (about one year, result not significant).
  - Core CPI and net core CPI: − approximately 0.1 percent after six months.
- Additional numeric highlights from Appendix III (selected impulse-response rows, horizons 0, 3, 6, 12, 24):
  - Industrial Production: Impulse Response: 0 -1.03 -0.83 0.18 0.30.
  - Loans to Non-Financial Corporations: Transactions, FX: 0 -9.67 -15.89 9.01 -0.57.
  - Monetary Policy Interest Rate: 1.00 0.19 -0.18 -0.23 0.03.
  - JPM EMBI Global Hungary - blended spread: 22.54 7.59 5.67 -8.08 1.16.
  - BSE: Market Index (BUX): -4.23 -0.37 0.45 4.27 -2.76.
  - Forint/Euro exchange rate: -0.04 -0.97 -0.13 0.14 0.29.

### Impulse responses to a 100 basis points risk-premium shock — select quantitative findings
- Identification: unexpected rise in Hungary’s external sovereign risk-premium (JP Morgan’s EMBI Global bond spread).
- Persistence: about nine months (more persistent than monetary policy shocks).
- Monetary policy response: immediate tightening by about 65 basis points; rate kept high for one quarter; policy begins easing about half a year after onset.
- Yields and term structure:
  - 3-month rate increases about 45 basis points versus long-term yield increase of about 2 basis points (term structure flattens/inverts).
- Private lending rates:
  - Corporate lending: short-term +35 basis points; long-term +55 basis points.
  - Mortgage interest rates: +15 basis points.
  - Consumption lending rates: +50 basis points.
- Exchange rate and balance-sheet effects:
  - Forint depreciates by almost 2 percent on impact; depreciation persists ≈ nine months.
  - Forint value of foreign debt (corporate and household loan stocks): increase by about 2 percent.
- Credit dynamics:
  - New forint corporate loans: initial +10 percent, then −10 percent; recover after 2–3 quarters.
  - New foreign-currency corporate loans: fall on impact by about 30 percent; recover after 2–3 quarters.
  - Net loans to households: HUF loans −7.5 percent; FX loans −12 percent; both do not recover before one year.
  - Household lending recovery slower than corporate lending.
- Financial and confidence:
  - Economic confidence declines more persistently (up to twelve months).
  - Stock market falls by about 4 percent; decline lasts ≈ one year.
- Real activity, labor market and prices:
  - Industrial production: − about 2 percent, lasting one year; manufacturing − about 2.5 percent.
  - Unemployment: + approximately 0.2 percent (≈ four times the response to a monetary tightening).
  - Hours: − about 0.5 percent.
  - Wages in manufacturing: − about 1 percent, depressed for eighteen months to two years.
  - Net core CPI: −0.25 percent.
  - Core CPI: −0.5 percent.
  - PPI: falls by over 0.5 percent.
- Appendix III selected impulse-response rows for Risk-Premium Shock (horizons 0, 3, 6, 12, 24):
  - Industrial Production: 0 -1.95 -1.98 -0.47 1.92.
  - Loans to Non-Financial Corporations: Transactions, FX: 0 -14.40 -2.14 -8.45 1.74.
  - Monetary Policy Interest Rate: 0.68 0.02 -0.23 -0.65 -0.04.
  - JPM EMBI Global Hungary - blended spread: 100 42.33 24.68 -33.65 -28.64.
  - BSE: Market Index (BUX): -4.06 -4.16 -3.79 0.12 0.96.
  - Forint/Euro exchange rate: 1.79 1.83 1.85 0.28 0.51.

### Variance decomposition and relative importance
- Both monetary and external risk-premium shocks explain a relatively small share of variation in most non-financial variables.
- Risk-premium shocks explain a larger share of short-term fluctuations in real activity (industrial production, unemployment, hours, wages) than monetary policy shocks.
- Appendix III variance decomposition snippets (selected, horizons 0, 3, 6, 12, 24):
  - Industrial Production (Monetary Policy decomposition): 0 0.38 0.64 2.46 3.75.
  - Industrial Production (Risk-Premium decomposition): 0 1.88 3.55 5.63 6.54.
  - Monetary Policy Interest Rate (Monetary Policy decomposition): 20.82 20.61 19.72 18.30 15.00.
  - JPM EMBI Global Hungary (Risk-Premium decomposition): 22.85 26.05 27.10 25.17 24.5.

### Forecasting performance (out-of-sample evaluation)
- Forecast experiment: models estimated until 2006:12; forecasts evaluated up to twelve months ahead for IP, CPI, IR, NEER.
- Compared models: small VAR (4 variables), medium VAR (10 variables), large VAR (full/large BVAR).
- Key result: increasing model scale improves forecast accuracy for most variables and horizons; large BVAR performs considerably better than smaller models.
- Representative large BVAR relative mean squared forecast errors for CPI (relative to random walk benchmark): 0.62 (1 month), 0.51 (3 months), 0.67 (6 months), 0.86 (9 months), 0.95 (12 months).
- Representative large BVAR relative MSFEs for IP: 0.78 (1 month), 0.86 (3 months), 0.75 (6 months), 1.05 (9 months), 1.34 (12 months).
- NEER relative MSFEs reported show mixed performance at short horizons but improve at some longer horizons (1.34, 1.27, 1.31, 0.88, 0.65).

### Conclusions and policy implications
- Monetary policy transmission in Hungary during the sample operated fast and efficiently, with rapid pass-through across interest rates, exchange rate, balance sheets, asset prices, and expectations.
- Balance sheet vulnerabilities (high public and external debt, high private sector foreign-currency exposure, foreign-funded banking system) make monetary policy constrained by macro-financial stability considerations.
- Risk-premium shocks induce endogenous procyclical monetary policy responses that, combined with balance sheet effects and depreciation, produce deeper and more persistent contractions than comparable pure monetary policy shocks.
- Policy recommendations (as presented in the source):
  - Prevent the build-up of significant public or private balance sheet vulnerabilities to allow monetary policy to pursue traditional stabilization aims.
  - Consider ex-ante macro-prudential measures to reduce vulnerabilities.
  - Strengthen the financial stability mandate of the central bank to potentially yield both financial stability and macroeconomic benefits over the long run.

*Source: _wp11259 (Appendixes and References; June 2001–September 2010 dataset and analysis).*

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

### _wp11259 - References

### References
- References ................................................................................................................................47

### Tables
- Table 1. Relative Mean Squared Forecast Error ......................................................................32

### Figures
- Figure 1. CE3: Monetary and Financial Developments ...........................................................21
- Figure 2. Hungary: Interest Rate Transmission .......................................................................22
- Figure 3. Hungary: Monetary and Financial Developments ....................................................23
- Figure 4. Hungary: Indebtedness—Total Stock of Liabilities .................................................24
- Figure 5. Hungary: Loans to the Non-Financial Private Sector ...............................................25
- Figure 6.1. Selected Impulse Responses to a Monetary Policy Shock ....................................26
- Figure 6.2. Selected Impulse Responses to a Monetary Policy Shock ....................................27
- Figure 6.3. Selected Impulse Responses to a Monetary Policy Shock ....................................28
- Figure 7.1. Selected Impulse Responses to a Risk-Premium Shock........................................29
- Figure 7.2. Selected Impulse Responses to a Risk-Premium Shock........................................30
- Figure 7.3. Selected Impulse Responses to a Risk-Premium Shock........................................31

### Appendixes
- Appendix I: Methodology ........................................................................................................33
  - A. Prior Distributions ...................................................................................................33
  - B. Model Selection: the Tightness of the Prior ............................................................36
  - C. Forecast Evaluation .................................................................................................37
- Appendix II: Data ....................................................................................................................38
- Appendix III: Detailed Results ................................................................................................41

- 3

*Source: _wp11259 - References .........................................................................................................*

### Appendix Tables

### _wp11259 - Appendix Tables

### Introduction and research objectives
- Focus: transmission mechanism of monetary policy and effects of risk-premium shocks in Hungary during the inflation targeting period.
- Sample period: June 2001 until September 2010.
- Methodology: a large Bayesian Vector Autoregression (large BVAR) using information from "more than 100 macroeconomic and financial series" to extract richer dynamics than small VARs.
- Main aims:
  - "look inside the 'black box' of monetary transmission" by using a comprehensive dataset;
  - analyze how a monetary policy shock and a risk-premium shock feed through the economy;
  - evaluate forecasting performance of the large BVAR compared with smaller models.

### Stylized facts about the Hungarian economy
- Hungary characterized as a "textbook" small open economy; international shocks have strong effects via trade and capital flows.
- Exchange rate history and regime features:
  - Prior to inflation targeting the exchange rate followed a crawling-peg (central bank preannounced a rate of depreciation and allowed some deviations (±2.5 percent) from this path).
  - After 2001, the exchange rate moved within horizontal bands of ±15 percent (removed in February 2008).
  - IMF-supported program introduced in November 2008 allowed for intervention to avoid disruptive exchange rate movements.
- Key macro-financial vulnerabilities (figures refer to 2010Q3):
  - General government debt: close to 90 percent of GDP.
  - Gross external debt: over 160 percent of GDP.
  - Private sector euroization:
    - Deposits in foreign currency: about one fifth of total deposits.
    - Share of foreign currency liabilities of households: exceeded 60 percent.
    - Share of foreign currency liabilities of corporations: around 25 percent.
  - Capital market openness and foreign investor dominance:
    - Portfolio investments held by nonresidents: close to 50 percent of GDP (about half of which was government debt).
    - Banking sector foreign ownership: foreign bank subsidiaries holding about two thirds of total banking assets.
    - Banks funded from wholesale markets: loans exceeded deposits by about 30 percent.
- Policy environment implications:
  - Government bond yields and country risk premium systematically higher than regional peers (CE-3) for much of the last decade.
  - The policy rate has been systematically higher and more responsive to government bond yields than in regional peers.
  - International risk premia correlate with government yields and policy rates, and are passed on to private lending rates.
  - A rise in relative risk premium compared to regional peers can lead to tightening even with large capacity gaps and relatively low inflation.

### Modeling approach and estimation strategy
- Rationale for large BVAR:
  - Small VARs risk omitted variable bias; factor models and large BVARs are data-rich alternatives.
  - Literature evidence: factor models and large BVARs outperform smaller VARs in forecasting; large BVARs can outperform factor models.
- Prior structure and implementation:
  - Priors based on Minnesota prior, augmented with an inverse Wishart prior for residual covariance and a prior on the sum of coefficients (Normal inverted Wishart conjugate prior).
  - Tightness of shrinkage hyperparameter and sum-of-coefficients prior set in relation to model size (following Banbura and others, 2010; detailed implementation in Appendix 1).
  - Procedure: benchmark small VAR (industrial production, net core inflation, nominal effective exchange rate, monetary policy rate) is used to set desired in-sample fit; overall tightness chosen so large model matches this in-sample fit for those four variables.
- Data:
  - Monthly series from June 2001 to September 2010 from MNB and other official sources.
  - Dataset contains 98 domestic real and nominal indicators (aggregate and sectoral) — categories include industrial production, construction, housing starts, retail trade, prices, labor markets (unemployment, hours and wages), monetary and credit aggregates, interest rates (short and long-term, for government, corporations and households), exchange rates (bilateral, effective nominal and real), survey measures of expectations, stock market index, EMBI global sovereign bond yield spread.
  - Additional block of 13 "foreign" variables: international commodity prices, industrial production and imports in Germany and the world, economic sentiment in Germany, stock market indices in Germany and the world (MSCI), and the Euro area policy rate.
  - Note: net core inflation is used (indirect taxes and excises are deducted).

### Identification strategy for structural shocks
- Two shocks analyzed: a monetary policy shock and a risk-premium shock — both set to a one time increase of 100 basis points.
- Identification method: recursive identification via Cholesky decomposition with exclusion restrictions on contemporaneous responses.
- Ordering for monetary policy shock identification:
  - External block (slow-moving) first;
  - Domestic slow-moving variables (real activity: industrial production, construction, sales, hours, trade), nominal variables (prices and wages), confidence measures, monetary and credit aggregates;
  - Policy rate ordered next;
  - Financial (fast-moving) variables last: interest rates, exchange rates, stock market index, country risk-premium.
  - Assumption: slow-moving variables do not respond contemporaneously to monetary policy shocks, while financial variables do.
- Ordering for risk-premium shock identification:
  - Slow-moving variables first;
  - EMBI sovereign spread next (the shock);
  - Fast-moving variables third (including the policy interest rate placed right after the shock in the benchmark calibration).
  - Assumption: monetary policy may respond contemporaneously to a risk-premium shock; risk-premium shock may affect fast-moving variables within the period.
  - Robustness: placing the monetary policy rate prior to the risk-premium has no significant impact on results.

### Main empirical findings (structural analysis and forecasting)
- Transmission of monetary policy:
  - The monetary policy signal transmits "fast and effective" throughout the economy — attributed to high openness and flexibility.
  - Channels identified beyond interest rate and exchange rate: balance sheet channel, asset price channel, expectations channel.
  - Despite high euroization and foreign ownership, a monetary policy tightening induces a sharp decline in economic activity — "more so than in advanced economies."
- Risk-premium shocks and monetary policy interaction:
  - Risk-premium shocks lead to an endogenous procyclical monetary policy response (driven by financial stability concerns) that compounds and deepens the economic downturn.
  - The monetary policy tightening in response to the risk-premium shock does not fully offset the shock to avoid excessive contraction, leading to some depreciation.
  - As a result of depreciation, net worth of foreign currency indebted private agents declines, and negative balance sheet effects reinforce the contractionary effects of the risk-premium shock.
- Forecasting performance:
  - Adding information (the large dataset) significantly improves forecasting performance relative to smaller specifications.
  - The large BVAR is deemed an appropriate tool for the central bank’s toolkit.

*Source: Appendix Tables and surrounding sections of _wp11259 - Appendix Tables (June 2001–September 2010 dataset and analysis).*

### Appendix 3.

### Appendix 3. Impulse Responses to a Monetary Policy Shock

### Overview
- Study of an unexpected monetary policy tightening of 100 basis points in Hungary.
- Key comparisons: small VARs for Hungary (previous literature) and a large BVAR for the US (Banbura and others, 2010).
- Complete results summarized in Appendix 3 Table 1 (not reproduced here); discussion focuses on key results.

### Persistence and Policy Reaction
- The unanticipated 100 basis points policy rate hike has a persistence of about four months.
- Monetary policy becomes accommodative in less than one year, with the policy rate falling about 25 basis points under the equilibrium value after the initial persistence period.
- Policy volatility is relatively higher in Hungary compared with advanced economies.

### Interest Rate Channel
- Monetary policy signal transmits rapidly along the government yield curve and to private sector lending rates.
- Government yields:
  - 3-month treasury bills increase by 75 basis points immediately after the shock.
  - 10-year treasury bonds increase by 10 basis points immediately and decline by 25 basis points after a quarter.
- Corporate lending rates (domestic currency):
  - Short-term corporate borrowing rate increases by 0.6 percent, peaking at about three months and returning to zero in approximately one year.
  - Long-term corporate borrowing rate increases by 0.5 percent, peaking at about three months and returning to zero in approximately one year.
- Household lending rates:
  - Consumption credit rates increase by about 25 basis points.
  - Mortgage loan rates increase by about 15 basis points.
- Interpretation:
  - Fast pass-through to corporates is consistent with shorter loan maturities and links to the 3-month interbank lending rate.
  - Slower pass-through to households may reflect longer mortgage maturities and banks competing to retain individual customers.

### Exchange Rate Channel
- Nominal effective exchange rate (forint) appreciates by about 1 percent after three months.
- The forint appreciation is less persistent than in some earlier studies and is reversed after about six months.
- Quantitative difference from earlier literature (e.g., Vonnák (2005)) may be due largely to methodological and sample differences.

### Balance Sheet and Credit Channels
- Balance sheet effects:
  - Stock of forex loans declines significantly by about 1 percent for the first three months after the shock (valuation effect of a stronger currency).
  - Little change in the stock of forint-denominated debt for corporations and households.
- New bank credit to the private sector declines:
  - Net loans (transactions) to corporates decline significantly immediately after the shock.
  - Loans in forint fall by about 20 percent immediately after a 100 basis points tightening, while foreign currency loans increase initially (substitution effect).
  - Forint lending recovers in less than six months; foreign currency lending takes around nine months to revert to equilibrium.
- Net loans to households decline by about 5 percent for both forint-denominated loans and foreign currency lending.
- Interpretation:
  - Part of the sharp decline in bank lending to firms may reflect substitution to intra-company (parent) financing.
  - Comparable declines in domestic and foreign market credit for domestic agents provide evidence of a shift in credit demand rather than a pure domestic credit supply shock.

### Asset Price and Expectations Channels
- Stock market:
  - Declines by around 4 percent for about one quarter following the tightening.
  - Confidence rebounds once monetary policy reverses stance.
  - Comparison: US decline is 1.3 percent after three months.
- Expectations and risk premia:
  - Economic activity expectations fall temporarily (about two quarters), followed by a medium-term improvement in sentiment.
  - Country risk increases by 20 basis points; EMBI spread remains elevated for about three quarters before returning to equilibrium.

### Overall Effects on Activity, Labor Market and Prices
- Real activity:
  - Aggregate and sectoral indices of industrial production and manufacturing fall between 1 and 2 percent in a hump-shaped response, with a trough after approximately one quarter and reversion to trend after about one year.
  - Consistent with tighter financing conditions and reduced investment.
  - Comparison: Vonnák (2007) finds a 25 basis points rate hike leads to a 0.3 percent decline in industrial production; Banbura and others (2010) find a 0.7 percent drop in US industrial production after six months and a 0.8 percent decline in manufacturing after one year.
- Labor market and wages:
  - Unemployment rises by about 0.05 percent.
  - Hours worked decline by 0.5 percent; hours revert after one year.
  - Wages fall by up to 0.5 percent for approximately one year (result not significant).
- Consumption and retail:
  - Retail sales do not decline significantly, consistent with relatively resilient wages, substitution effects from appreciation, and positive balance sheet effects—implying a moderate effect on private consumption.
- Prices:
  - Core CPI and net core CPI fall by approximately 0.1 percent after six months.
  - (PPI response referenced but not fully specified in the provided text.)

### Key Quantitative Findings (select)
- Policy shock considered: 100 basis points.
- Policy persistence: about four months.
- Policy rate below equilibrium after easing: about 25 basis points.
- 3-month treasury bill response: +75 basis points immediately.
- 10-year treasury bond response: +10 basis points immediately; −25 basis points after a quarter.
- Corporate short-term lending: +0.6 percent (peak at ~3 months; back to zero ~1 year).
- Corporate long-term lending: +0.5 percent (peak at ~3 months; back to zero ~1 year).
- Consumption credit rates: +25 basis points.
- Mortgage loan rates: +15 basis points.
- Forint appreciation: about 1 percent after three months.
- Stock of forex loans: −1 percent for first three months.
- Forint loans to corporates: −20 percent immediately.
- Forint lending recovery: < six months; foreign currency lending recovery: ~ nine months.
- Net loans to households: −5 percent.
- Stock market decline: ~4 percent for about one quarter.
- Country risk (EMBI) increase: 20 basis points; elevated for ~ three quarters.
- Industrial production and manufacturing: −1 to −2 percent; trough ~ one quarter; revert ~ one year.
- Unemployment: +0.05 percent.
- Hours worked: −0.5 percent.
- Wages: − up to 0.5 percent (about one year).
- Core CPI and net core CPI: − approximately 0.1 percent after six months.

_Appendix 3 (Impulse Responses to a Monetary Policy Shock) — _wp11259 - Appendix 3._

### 0.3 percent.

### _wp11259 - 0.3 percent.

### Transmission of monetary policy
- The transmission of monetary policy to prices is fast and occurs simultaneously with the adjustment in output.
- Back-of-the-envelope calculation (monthly model without GDP): implies about 0.3 percent decline in GDP after a quarter. This is based on OLS regression GDP = 1.00(3.83) + 0.30(12.39)*IP, R2 = 0.74.
- The disinflationary effect of monetary tightening is significant; there is no evidence of a “price puzzle.”
- Comparison to prior literature: Vonnák (2007) reported that a 25 basis points tightening leads to a fall in inflation of about 0.1 percent; this study finds a much lower impact on prices.

### Empirical evidence on transmission channels
- Documented channels (including some not previously shown for Hungary): balance sheet channel, asset price channel, expectations channel in addition to standard interest-rate and exchange-rate channels.
- Monetary tightening is contractionary for activity; balance sheet effects operate in the opposite direction to other channels but do not reverse the contractionary conclusion.

### Impulse responses to a 100 basis point risk-premium shock (external tightening)
- Identification: shock = unexpected rise in Hungary’s external sovereign risk-premium as measured by JP Morgan’s EMBI Global bond spread.
- Persistence: risk-premium shocks last about nine months (more persistent than monetary policy shocks).
- Monetary policy response: central bank immediately tightens by about 65 basis points and keeps the rate high for one quarter; policy begins easing about half a year after onset as the shock subsides.
- Yield curve and yields:
  - Short- and long-term government yields rise; term structure flattens or inverts.
  - 3-month rate increases more than long-term yield: by about 45 basis points versus 2 basis points.
- Private lending rates:
  - Corporate lending rates increase by about 35 and 55 basis points for the short-term and long-term rates respectively.
  - Mortgage interest rates increase by about 15 basis points.
  - Consumption lending rates increase by about 50 basis points.
- Exchange rate and balance-sheet effects:
  - Forint depreciates by almost 2 percent on impact; depreciation persists for about nine months.
  - Depreciation increases the forint value of foreign debt; both foreign corporate and household loan stocks expressed in forint terms increase by about 2 percent.
- Credit dynamics:
  - New forint corporate loans: an initial increase of about 10 percent, then fall by about 10 percent; recover after 2–3 quarters.
  - New foreign-currency corporate loans fall on impact by about 30 percent; recover after 2–3 quarters.
  - Net loans to households: domestic currency denominated loans fall by 7.5 percent; foreign currency denominated loans decline by about 12 percent; both do not recover before one year.
  - Household lending recovery is slower than corporate lending, plausibly because households are not hedged against currency depreciation.
- Financial and confidence effects:
  - Economic confidence declines more persistently than after a monetary policy shock (up to twelve months versus six months).
  - Stock market falls by about 4 percent; decline lasts about one year.
- Real activity and labor market:
  - Industrial production drops by about 2 percent, lasting for one year; manufacturing falls by about 2.5 percent.
  - Unemployment rises by approximately 0.2 percent (about four times the response to a monetary tightening).
  - Hours decline by about 0.5 percent.
  - Wages in manufacturing decline by about 1 percent, remaining depressed for eighteen months to two years.
- Inflation:
  - Disinflationary effect is larger than for a monetary policy shock.
  - Net core CPI falls by 0.25 percent; core CPI falls by 0.5 percent; PPI falls by over 0.5 percent.
- Policy interpretation:
  - MNB tried to counter the increase in relative risk premium but not fully, weighing the cost of a sharper economic contraction against financial stability risks.
  - Historical episodes: October 2008 EMBI spread increased by 350 basis points; MNB tightened by 300 basis points while the currency depreciated by about 30 percent. EMBI reached up to 500 basis points by mid-March 2009; monetary policy was loosened by 200 basis points as spreads began coming down in 2010.

### Variance decomposition
- Both monetary and external risk-premium shocks explain a relatively small share of variation in most non-financial variables (consistent with prior literature).
- Risk-premium shocks play a larger role than monetary policy shocks in explaining short-term fluctuations in real activity (industrial production, unemployment, hours, wages).
- In a small open economy like Hungary, external risk-premium volatility contributes more to economic fluctuations than monetary policy shocks.

### Forecasting performance evaluation (out-of-sample)
- Forecast experiment: models estimated until 2006:12; forecasts evaluated up to twelve months ahead for industrial production (IP), inflation (CPI), policy interest rate (IR), and exchange rate (NEER).
- Three specifications compared: small VAR (4 variables), medium VAR (10 variables), large VAR (full specification / large BVAR).
- Key result: increasing model scale improves forecast accuracy for most variables and horizons; the large BVAR performs considerably better than smaller models.
- Representative large BVAR relative mean squared forecast errors (relative to naive random walk benchmark) for CPI: 0.62 (1 month), 0.51 (3 months), 0.67 (6 months), 0.86 (9 months), 0.95 (12 months). Large BVAR also performs well for IP (0.78, 0.86, 0.75, 1.05, 1.34) and NEER at longer horizons (1.34, 1.27, 1.31, 0.88, 0.65).

### Conclusions and policy implications
- Monetary policy transmission in Hungary operated fast and efficiently during the sample period.
- Balance sheet vulnerabilities can make monetary policy hostage to macro-financial stability considerations, potentially outweighing domestic stabilization objectives.
- Risk-premium shocks induce endogenous pro-cyclical monetary policy responses and, combined with balance sheet effects, cause deeper and more persistent economic contractions than comparable pure monetary policy shocks; this highlights the trade-off faced by an inflation-targeting central bank between financial stability and domestic stabilization objectives (the “perverse depreciation effects” dilemma).
- Policy recommendations:
  - Prevent the build-up of significant public or private balance sheet vulnerabilities to allow monetary policy to pursue traditional stabilization aims.
  - Consider ex-ante macro-prudential measures to reduce vulnerabilities.
  - Strengthening the financial stability mandate of the central bank may yield both financial stability and macroeconomic benefits over the long run.

*Source: _wp11259 - 0.3 percent.*

### Appendix I: Methodology

### Appendix I: Methodology

### A. Prior Distributions
- Model form:
  - Reduced form VAR: Yt = c + A1 Yt−1 + ... + Ap Yt−p + ut, where Yt is an n-variable dataset and ut is n-dimensional Gaussian white noise with covariance matrix E[ut ut′].
- Minnesota prior moments for coefficients A_k:
  - E[(A_k)_{ij}] = { 1 if i=j and k=1; 0 otherwise }.
  - Var[(A_k)_{ij}] = { σ_i^2 / k^2 if j=i; σ_i^2 σ_j^2 / k^2 otherwise }.
  - Coefficients A_k assumed independent and normally distributed; intercept prior is diffuse.
- Persistence settings:
  - For highly persistent variables set φ_i = 1 (random walk with drift).
  - For stationary variables set φ_i = 0 (white noise).
- Intuition on prior variance:
  - Decay across lags governed by 1/k^2.
  - Own lags more important than other-variable lags via variance ratios σ_i^2/σ_j^2.
  - Cross-variable lag relative importance governed by ρ ∈ (0,1).
- Overall tightness hyperparameter:
  - λ controls overall tightness around random walk/white noise.
  - Lower λ → stronger prior influence; for λ → 0 posterior equals prior. For λ → ∞ prior has no influence and posterior equals OLS.
  - λ set tightly and in relation to cross-section size: more variables → require more tightness to avoid overfitting.
- Residual covariance prior:
  - Use Normal–Inverted Wishart prior (vec(B)|Σ ∼ N(vec(B0), Σ ⊗ V0) and Σ ∼ iW(S0, ν0)) following Kadiyala and Karlsson (1997) and Robertson and Tallman (1999).
  - Prior parameters B0, V0, S0, ν0 chosen so prior moments of B match Minnesota prior and E[Σ] equals Minnesota fixed residual covariance.
- Dummy-observation implementation:
  - Adding T_d dummy observations Y_d and X_d is equivalent to imposing the Normal–Inverted Wishart prior with:
    - B0 = (X_d′ X_d)^{-1} X_d′ Y_d
    - V0 = (X_d′ X_d)^{-1}
    - S0 = (Y_d − X_d B0)′(Y_d − X_d B0)
    - ν0 = T_d − k
  - Dummy matrices defined to match Minnesota prior moments (J_p = diag(1,2,...,p)).
- Prior on coefficient scales:
  - σ_i^2 set equal to the variance of the residual from a univariate autoregressive model of order p for y_it.
- Augmented regression with dummies:
  - Y* = X* B + U*, where T* = T + T_d, Y* = [Y′, Y_d′]′, X* = [X′, X_d′]′, U* = [U′, U_d′]′.
  - Improper prior on Σ: p(Σ) ∝ |Σ|^{−(n+3)/2} ensures existence of prior expectation of Σ.
  - Posterior remains Normal–Inverted Wishart:
    - vec(B)|Σ,Y ∼ N(vec(B_hat), Σ ⊗ (X*′ X*)^{-1})
    - Σ|Y ∼ iW(S_hat, ν_hat) with ν_hat = T_d + 2(T − k)
    - B_hat = (X*′ X*)^{-1} X*′ Y*
    - S_hat = (Y* − X* B_hat)′(Y* − X* B_hat)
- Sum-of-coefficients prior (long-run neutrality / unit root shrinkage):
  - Error-correction form implies restriction In − A1 − ... − Ap = 0.
  - Prior shrinks (In − A1 − ... − Ap) toward 0 via dummy observations:
    - Y_d = diag(ζ1,...,ζn)/0_{n(p−1)×n} ... (as specified)
    - X_d = [1 2 ... p] ⊗ diag(ζ1,...,ζn)/0_{np×1} ... (as specified)
  - Hyperparameter c controls degree of shrinkage:
    - As c → 0 approach exact differences (strong shrinkage).
    - As c → ∞ return to no shrinkage on sum of coefficients.
  - ζ_i set to sample averages of y_it (Sims and Zha, 1998).
  - τ set relatively loose relative to λ: τ = 10λ (experiments with values from 1 to 100 showed proportionality not critical).

### B. Model Selection: the Tightness of the Prior
- Objective: choose overall tightness λ using a training sample (first half of the sample) to make large BVAR comparable to a small VAR for key variables (IP, IR, CPI, NEER).
- Fit metric in training sample:
  - Fit_{1,4} = (1/4) ∑_i MSFE_i(λ,m) / MSFE_i(λ=0,m=SMALL), where MSFE is one-step-ahead mean squared forecast error evaluated using training samples t = 1,...,T−1.
- Procedure:
  1. Set training sample (first half) and compute average in-sample fit for 4-variable small VAR (IP, IR, CPI, NEER).
  2. Define a large grid for λ. Set prior on sum of coefficients with c = k*λ, experimenting with k ∈ {0.01, 0.1, 1} (experiments showed results not affected).
  3. Estimate the Bayesian VAR for each λ to obtain corresponding fit.
  4. Choose λ_m = argmin_λ |Fit − (1/4) ∑_i MSFE_i(λ,m) / MSFE_i(λ=0)| to match large-model fit to small-model in-sample fit.
- Rationale: penalize larger model for lack of parsimony and ensure comparability across different cross-sectional sizes.

### C. Forecast Evaluation
- Training and evaluation setup:
  - Use training sample from T0 to T1 to estimate model.
  - Construct h-step-ahead forecasts up to horizon H using all available observations at each forecast origin.
  - Denote forecast: ŷ_{i,T+h|T}(λ,m) for variable i, model m, tightness λ, using posterior mean estimates.
- Forecast accuracy metric relative to naive benchmark (random walk where λ = 0):
  - MSFE_{i,h}(λ,m) / MSFE_{i,h}(λ=0,m)
  - MSFE_{i,h}(λ,m) = (1/(T1−T0−H+1)) ∑_{T=T0}^{T1−h} (ŷ_{i,T+h|T}(λ,m) − y_{i,T+h})^2
- Interpretation:
  - If RMSFE_{i,h}(λ,m) < 1 then model m performs better than the benchmark naive prior model.

---

### Appendix II: Data (as provided in the source)
- Data sources: National Statistic Agency and the Hungarian Central Bank.
- Seasonal/working-day adjustments: used where available; adjustments performed by the source.
- Units: series expressed in real-terms where applicable; loan data not seasonally adjusted, reflect end-of-period information, expressed in HUF.
- Lending rates for households: computed as monthly average agreed interest rate of HUF loans and HUF deposits to households weighted by amount of new business.
- Transformations applied: 1 = no transformation; 2 = log multiplied by 100.
- Note: Unlike factor-model papers, data are not transformed to stationarity; trends are preserved.
- Variable groups and transformation codes (selection shown):
  - Real activity (transformation code 2 for Industrial Production, Construction, Retail Sales, Building Permits, etc.).
  - Labor market (Unemployment Rate transformation code 1; many wages and hours series transformation code 2 or 1 as listed).
  - Confidence indicators (IFO climate, EABCI, Economic Sentiment Indicator, etc.; transformation code 1).
  - Prices (CPI and subcomponents, Core CPI, Producer price index; transformation code 1).
  - Monetary and credit aggregates (M1, M2, M3 transformation code 2; loan series transformation code 2; stock and transaction series specified).
  - Interest rates (Monetary Policy Interest Rate, government yields, lending rates, JPM EMBI Global Hungary spreads, BSE BUX, NEER and Forint/Euro exchange rate; transformation codes 1 or 2 as listed).
  - External environment (Germany and World series, ECB Main Refinancing rate, commodity price indices; transformation codes 1 or 2 as listed).
- Selected exact entries (preserving numbering and transformation codes as provided):
  - 1: Industrial Production: Industry Including Construction — 2
  - 22: Retail Sales Volume Index: Adjusted for Calendar Effects — 2
  - 31: Unemployment Rate — 1
  - 60: Consumer Price Index (CPI) — 1
  - 71: M1 — 2
  - 78: Loans to Non-Financial Corporations, Transactions, HUF — 2
  - 86: Monetary Policy Interest Rate — 1
  - 96: Nominal effective exchange rate, all countries — 2
  - 99: Germany: Industrial Production: Total Industry including Construction — 2
  - 102: Euro Area11-16: Main Refinancing Operations: Minimum Bid Rate (EOP, %) — 1
  - 107: All primary commodities price index — 1

*Source: _wp11259 - Appendix I: Methodology (PDF content provided).*

### Appendix III: Detailed Results

### _wp11259 - Appendix III: Detailed Results

### Overview
- Contains Table 1 and Table 2 with full-model impulse responses and variance decompositions for two identified shocks:
  - Monetary Policy Shock (Table 1)
  - Risk-Premium Shock (Table 2)
- Reported horizons (months): 0, 3, 6, 12, 24 for impulse responses and variance decompositions.
- Variables covered include detailed sectoral industrial production series, construction and building permits, retail sales and retail subcomponents, labor market series (unemployment, wages, hours), confidence indicators, CPI and subcomponents, monetary aggregates (M1, M2, M3), MFI loans and loan flows (transactions and stocks, HUF and FX), interest rates and yields, JPM EMBI Global Hungary statistics, BSE: Market Index (BUX), nominal and real effective exchange rates, and Forint/Euro exchange rate.

### Monetary Policy Shock — Key reported impulse responses and variance decompositions (selected rows; values correspond to horizons 0, 3, 6, 12, 24 for impulse responses and 0, 3, 6, 12, 24 for variance decomposition)
- Industrial Production: Industry Including Construction
  - Impulse Response: 0 -1.03 -0.83 0.18 0.30
  - Variance Decomposition: 0 0.38 0.64 2.46 3.75
- Loans to Non-Financial Corporations: Transactions, FX
  - Impulse Response: 0 -9.67 -15.89 9.01 -0.57
  - Variance Decomposition: 0 0.35 1.75 2.45 1.03
- Monetary Policy Interest Rate
  - Impulse Response: 1.00 0.19 -0.18 -0.23 0.03
  - Variance Decomposition: 20.82 20.61 19.72 18.30 15.00
- JPM EMBI Global Hungary - blended spread
  - Impulse Response: 22.54 7.59 5.67 -8.08 1.16
  - Variance Decomposition: 6.42 6.07 5.71 5.40 5.21
- BSE: Market Index: Budapest Stock Index (BUX)
  - Impulse Response: -4.23 -0.37 0.45 4.27 -2.76
  - Variance Decomposition: 4.40 4.14 3.86 3.64 3.45
- Forint/Euro exchange rate
  - Impulse Response: -0.04 -0.97 -0.13 0.14 0.29
  - Variance Decomposition: 0.27 0.53 1.02 1.32 1.38

### Risk-Premium Shock — Key reported impulse responses and variance decompositions (selected rows; same horizon ordering)
- Industrial Production: Industry Including Construction
  - Impulse Response: 0 -1.95 -1.98 -0.47 1.92
  - Variance Decomposition: 0 1.88 3.55 5.63 6.54
- Loans to Non-Financial Corporations: Transactions, FX
  - Impulse Response: 0 -14.40 -2.14 -8.45 1.74
  - Variance Decomposition: 0 2.14 3.74 4.84 5.15
- Monetary Policy Interest Rate
  - Impulse Response: 0.68 0.02 -0.23 -0.65 -0.04
  - Variance Decomposition: 8.57 6.55 4.13 2.73 1.68
- JPM EMBI Global Hungary - blended spread
  - Impulse Response: 100 42.33 24.68 -33.65 -28.64
  - Variance Decomposition: 22.85 26.05 27.10 25.17 24.5
- BSE: Market Index: Budapest Stock Index (BUX)
  - Impulse Response: -4.06 -4.16 -3.79 0.12 0.96
  - Variance Decomposition: 6.18 7.01 5.97 3.72 1.68
- Forint/Euro exchange rate
  - Impulse Response: 1.79 1.83 1.85 0.28 0.51
  - Variance Decomposition: 1.06 2.27 1.26 2.34 3.24

### Notable patterns in the reported tables (based directly on tabulated values)
- Many industrial production and retail series show negative impulse responses at short horizons (3 and 6 months) to both shocks, with varying rebound or further decline by 12 and 24 months as documented in the impulse response columns.
- Several loan flow series (Transactions, FX) display large negative impulse responses under the Monetary Policy Shock and even larger negative responses under the Risk-Premium Shock (e.g., Loans to Non-Financial Corporations: Transactions, FX; Loans to Households, Transactions, FX).
- JPM EMBI Global Hungary - blended spread exhibits very large positive impulse responses to the Risk-Premium Shock (100 at horizon 0) and sizeable responses to the Monetary Policy Shock (22.54 at horizon 0).
- Monetary Policy Interest Rate responds positively at horizon 0 to the Monetary Policy Shock (1.00) and less strongly to the Risk-Premium Shock (0.68); variance decompositions show the monetary policy shock accounts for a substantial share of variance in the policy rate series (20.82 at horizon 0 down to 15.00 at horizon 24).
- Exchange rate series (Forint/Euro, nominal and real effective ER) show differing sign responses across shocks: negative initial impulse responses to the Monetary Policy Shock and positive initial impulse responses to the Risk-Premium Shock (Forint/Euro: -0.04 at horizon 0 for Monetary Policy; 1.79 at horizon 0 for Risk-Premium).

*Source: _wp11259 - Appendix III: Detailed Results; https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11259.pdf*

### REFERENCES

### _wp11259 - REFERENCES

### References

- Allard, C., 2009, Competitiveness in Central Europe: What Has Happened Since EU Accession, IMF Working Paper 09/121.  
- Banbura, M., D. Gianonne, and L. Reichlin, 2010, "Large Bayesian VARs," Journal of Applied Econometrics, Vol. 25(1), pp. 71–92.  
- Benk, S., Z. M. Jakab, M. A. Kovács, B. Párkányi, Z. Reppa, and G. Vadas, 2006, The Hungarian Quarterly Projection Model (NEM), MNB Occasional Papers 2006/60, Magyar Nemzeti Bank.  
- Bloor, C. and T. Matheson, T., 2008, Analysing Shock Transmission in a Data-Rich Environment: A Large BVAR for New Zealand, Reserve Bank of New Zealand, DP2008/09.  
- Bernanke, B., J. Boivin, and P. Eliasz, 2005, "Measuring Monetary Policy: A Factor Augmented Autoregressive (FAVAR) Approach," Quarterly Journal of Economics, 120, pp. 387–422.  
- Calvo, G.A. and C. M. Reinhart, 2002, "Fear Of  Floating," The Quarterly Journal of Economics, MIT Press, Vol. 117(2), May, pp. 379–408.  
- Carriero, A., G. Kapetanios, and M. Marcellino, 2009, "Forecasting Exchange Rates with a Large Bayesian VAR," International Journal of Forecasting, 25(2), pp. 400–417.  
- Chang, R. and A. Velasco, 2001, "Monetary Policy in a Dollarized Economy where Balance Sheets Matter," Journal of Development Economics, Elsevier, Vol. 66(2), December, pp. 445–464.  
- Christiano, L. J., M. Eichenbaum, and C. L. Evans, 1999, Monetary Policy Shocks: What Have We Learned and to What End?, in Handbook of Macroeconomics, ed. by  J. B. Taylor, and M. Woodford, Vol. 1, chap. 2, Elsevier, pp. 65–148.  
- Csermely, Á. and B. Vonnák, 2002, The Role of the Exchange Rate in the Transmission Mechanism in Hungary, presented at the ECB conference "Monetary Policy Transmission in the Euro Area and in Accession Countries" October.  
- De Mol, C, D Gianonne, and L Reichlin, 2008, "Forecasting Using a Large Number of Predictors: Is Bayesian Regression a Valid Alternative to Principal Components?," Journal of Econometrics, Vol. 146, No. 2, October, pp. 318–328.  
- Doan, T. and R. Litterman and C. Sims, 1984, "Forecasting and Conditional Projection Using Realistic Prior Distributions," Econometric Reviews, Taylor and Francis Journals,  Vol. 3(1), pp. 1–100.  
- Forni, M., D. Giannone, M. Lippi, and L. Reichlin, 2009, "Opening the Black Box: Structural Factor Modelswith Large Cross Sections," Econometric Theory, 25 (5),  pp. 1319–1347.  
- Giannone, D., and L. Reichlin, 2006, "Does Information Help Recovering Structural Shocks from Past Observations?," Journal of the European Economic Association, 4(2-3),  pp. 455–465.  
- Jakab, Z. and B. Világi, 2008, An Estimated DSGE Model of the Hungarian Economy, MNB Working Papers No. 2008/9, Magyar Nemzeti Bank.  
- Koop, G., 2010, Forecasting with Medium and Large Bayesian VARs, Manuscript, University of Strathclyde.  
- Leeper, E. M., C. Sims, and T. Zha, 1996, "What Does Monetary Policy Do?," Brookings Papaers on Economic Activity, 1996(2), pp. 1–78.  
- Litterman, R., 1986, "Forecasting with Bayesian Vector Autoregressions - Five Years of Experience," Journal of Business and Economic Statistics, vol 4(1), pp. 25–38.  
- LLosa, G., Tuesta V. and M. Vega, 2005, A BVAR Forecasting Model for Peruvian Inflation, Central Bank of Peru.  
- Kadiyala, K. Rao and S. Karlsson, 1997, "Numerical Methods for Estimation and Inference in Bayesian VAR-models," Journal of Applied Econometrics, Vol. 12(2),  March–April, pp. 99–132.  
- Rezessy, A., 2005, Estimating the Immediate Impact of Monetary Policy Shocks on the Exchange Rate, MNB Occasional Papers No. 38, Magyar Nemzeti Bank.  
- Sims, C. A., 1992, Bayesian Inference for Multivariate Time Series with Trend, Mimeo.  
- Sims, C. A. and T. Zha, 1998, "Bayesian Methods for Dynamic Multivariate Analysis," International Economic Review, 39(4), pp. 949–68.  
- Stock J.H. and M.W. Watson, 2002, "Forecasting Using Principal Components From a Large Number of Predictors," Journal of the American Statistical Association, American Statistical Association, Vol. 97, December, pp. 1167–1179.  
- Stock J.H. and M.W. Watson, 2005, An Empirical Comparison of Methods for Forecasting Using Many Predictors, Manuscript, Princeton University.  
- Vonnák, B., 2005, Estimating the Effect of Hungarian Monetary Policy within a Structural VAR Framework, MNB Working Papers 2005/1, Magyar Nemzeti Bank.  
- Vonnák, B., 2007, The Hungarian Monetary Transmission Mechanism: an Assessment, MNB Working Papers 2007/3, Magyar Nemzeti Bank.  
- Vonnák, B., 2010, Risk Premium Shocks, Monetary Policy and Exchange Rate Pass-Through in the Czech Republic, Hungary and Poland, MNB Working Papers 2010/1, Magyar Nemzeti Bank.

*Source: _wp11259 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11259.pdf_