## wp1737

## Source details

**Canonical URL:** [wp1737](https://www.imf.org/-/media/files/publications/wp/2017/wp1737.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1737.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1737.pdf.json)

---

### I. Introduction and scope
- Postcrisis average growth in CESEE fell from precrisis average growth of 5.4 percent (2000-08) to 1.2 percent in the postcrisis (2009-15), a drop of 4.2 percentage points.  
- The drop exceeded that in advanced economies (1.3 percentage points) and other emerging economies (1 percentage point).  
- Key question: Was the slowdown largely cyclical or structural? The answer affects policy mix and sequencing because hysteresis from prolonged downturns could permanently scar potential output.  
- Challenges: potential output is unobservable; short samples with one boom-bust cycle for many countries; large structural shifts (including financial deepening) and end-point bias of statistical filters.  
- Scope: consistent potential output estimates for 18 CESEE countries using four approaches:
  - Hodrick-Prescott (HP) filter;
  - Multivariate filter (MVF) linking GDP, unemployment, and inflation;
  - Multivariate filter incorporating financial/credit cycles (MVF-FIN);
  - Production function (PF) approach.
- Methodological refinements for CESEE: adjust financial variables (credit and house prices) to separate financial deepening from cyclical credit; include wage inflation, hours worked, and capacity utilization to better capture cyclical pressures in low-inflation and externally disinflationary settings.  
- Caveats: frequent data revisions and large confidence bands due to structural change and short samples.

### II. Methodologies (overview and key features)
- HP filter:
  - Univariate trend extraction applied to GDP.
  - Advantages: simplicity and transparency.
  - Limitations: purely statistical, no direct economic interpretation (inflation-neutral or sustainable), end-of-sample revisions.
- MVF (Multivariate filter):
  - Defines potential output as the level achievable under flexible nominal prices and wages (Okun/Phillips framework).
  - Uses inflation (or wage inflation), inflation expectations, and unemployment gap; imposes Phillips curve and Okun’s law; estimated in a Bayesian simultaneous system.
  - Advantages: limited data requirements; more robust to end-of-sample problems; can incorporate medium-term forecasts.
  - Application note: MVF uses wage inflation for EU members of CESEE (Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, Slovenia).
- MVF-FIN (Multivariate filter with financial frictions):
  - Seeks sustainable potential output that is both inflation-neutral and finance-neutral (Borio, 2013).
  - Adds financial variables (credit growth, house price inflation, capacity utilization) to identify demand vs supply shocks; estimated by maximum likelihood in state-space form.
  - Key CESEE adjustment: use IMF (2015) credit gap estimates to adjust credit series so structural financial deepening is not mistaken for cyclical credit.
  - Ignoring structural credit increases understates potential output in booms and overstates it postcrisis—important for Baltic states and SEE-EU countries.
- Production function (PF) approach:
  - Cobb-Douglas specification: Y_t = A_t (K_t CU_t)^(1-α) (L_t AHW_t)^α
    - K_t from perpetual inventory K_t = (1-ρ)K_{t-1} + I_t; ρ calibrated from historical average depreciation; initial capital from Penn World Tables.
    - CU_t: survey-based capacity utilization.
    - L_t: number of employed persons (national accounts).
    - AHW_t: average hours worked.
    - α: labor share from compensation of employees to gross value added.
  - TFP computed as Solow residual; smoothed trends in unemployment, average hours, and TFP use HP filter (smoothing factor 1600).
  - Potential output uses calibrated labor share, smoothed labor inputs and TFP, and actual capital stock.

### III. Main findings and magnitudes
- All methods indicate a large sustained share of the postcrisis growth slowdown was structural:
  - Potential output growth slowed by about 2 percentage points after the crisis compared with precrisis years—accounting for about half of the total growth slowdown.
- Drivers of postcrisis potential growth slowdown:
  - Stagnant Total Factor Productivity (TFP) is the main drag.
  - Slower capital accumulation contributed to a lesser but material extent.
  - Estimates of TFP are uncertain (TFP is measured as a residual); further analysis is needed to identify drivers.
- External/common vs domestic/idiosyncratic factors:
  - Region-wide slowdown associated with common external factors (trading partners’ potential growth, global trade, global supply chains).
  - For some countries, negative demographics also contributed.
  - For CESEE excluding the CIS and Turkey, common factors strongly co-moved with potential growth before the crisis; CIS and Turkey show more important idiosyncratic factors, especially after the crisis.
- Cross-method differences:
  - Generally small because of overlapping information sets and interlinked real and financial sectors.
  - Larger differences arise when real and financial cycles are poorly synchronized—possibly due to financial frictions or rigidities.
- Output gaps and cyclical recovery:
  - Despite slower potential growth, the postcrisis rebound narrowed negative output gaps; for much of the region gaps appear largely closed in 2015, except the CIS and Turkey.
  - This is consistent with dissipating disinflationary pressures in non-tradable sectors and declining unemployment, while headline inflation remained low due to falling import (notably energy) prices.

### IV. Role of capacity utilization and average hours worked
- Not accounting for cyclical components of production intensities (capacity utilization and average hours worked) may understate booms and busts:
  - Capacity utilization rises in booms and falls in busts; omitting cyclical CU can overstate potential during booms and understate during busts, reducing output gap volatility.
  - Average hours worked cycle notably in the Baltics and SEE-EU countries—incumbent workers work more in booms, fewer hours in busts.
- Quantitative note:
  - Comparison in Appendix A (Figure A.3) shows output gap amplitude is smaller by 2-8 percentage points when PF is estimated without adjusting for capacity utilization and average hours worked (CESEE EU countries).

### V. Reconciling methods and interpreting divergence
- Differences across methods arise mainly from differences in information sets and model richness:
  - Richer specifications (e.g., MVF-FIN and PF with financial and labor market variables) lead to closer agreement across methods.
  - Synchronization of labor and capital market cycles matters: more flexible labor markets and deeper capital markets reduce cross-method differences.
- Example — SEE-EU divergence postcrisis:
  - Precrisis: credit, capacity utilization, hours, and inflation moved together; methods agreed on dynamics.
  - Postcrisis: credit stayed flat while capacity utilization and wage inflation rose.
    - MVF and PF show a closed gap in 2015.
    - MVF-FIN interprets low credit growth as a negative output gap.
  - Two interpretations:
    - Benign: financial cycle lags business cycle and will catch up—no immediate policy action.
    - Less benign: impaired balance sheets (firms, households, intermediaries) prevent growth at full potential—policy action needed to repair balance sheets.
  - Distinguishing requires country-specific balance-sheet information.
- Note: creditless recoveries can feature substantial job creation with no real credit growth.

### VI. Results — potential growth decomposition and country patterns
- Timing and magnitude:
  - Potential growth was broadly stable 2001-07, declined significantly after the global financial crisis, and by 2015 had not recovered to precrisis levels.
  - In more than two thirds of sample countries, potential growth slowdown started before 2007.
- Precrisis (2005-09) average contributions:
  - Capital accumulation contributed more than 2 percentage points to regional potential growth.
  - Bulgaria, Estonia, and Latvia: capital contribution about 4 percentage points.
  - Average estimated TFP contribution during 2005-09 was about 2 percentage points.
  - Latvia, Lithuania, Serbia, Slovakia: TFP contribution close to 3 percentage points.
  - Romania and Moldova: outward migration produced a negative labor contribution.
- Postcrisis:
  - Widespread significant TFP slowdown.
  - Capital accumulation contribution mixed:
    - In 7 out of 18 countries (including Bulgaria, the Baltics, and Slovenia) capital contribution halved after the crisis—precrisis rapid credit growth and external borrowing preceded investment reductions.
    - Czech Republic and Poland (no excessive precrisis borrowing) continued with strong capital contributions.
    - Romania: excessive precrisis credit partly fueled consumption; higher saving after the crisis aided current account adjustment.

### VII. How large were output gaps in 2015? (selected aggregate and country outcomes)
- General: Given lower postcrisis potential growth, the 2015 growth rebound implies narrowing output gaps; output gaps largely closed for CESEE excluding the CIS by 2015.
- In 10 out of 18 countries, at least two of three methods (MVF, MVF-FIN, PF) produce positive but small output gap estimates for 2015—most positive estimates are below one percent (treated as largely closed).
- CIS countries: among 8 countries where at least two methods indicate negative output gaps.
- Inflation and labor market signals 2011–2015:
  - Headline inflation decelerated since end-2011, while estimated output gaps narrowed since 2010 and appear largely closed in CEE, SEE-EU, and the Baltics by 2015.
  - Headline inflation was largely driven by falling import prices (notably oil).
  - Inflation in non-tradable services and falling unemployment indicate dissipating disinflationary pressures.
  - Core and services inflation in most countries remained notably below inflation targets by end-2015.

### VIII. Selected 2015 country-specific output gap and potential growth estimates (exact figures)
- Output gap estimates (percent of potential output, 2015) — method-specific values and sign (based on at least two of MVF, MVF-FIN, PF):
  - Ukraine:
    - HP-filter: -7.5
    - MVF: -0.5
    - MVF FIN: -12.9
    - PF: -6.5
    - Average (excl. HP-filter): -6.6
    - Sign: -
  - Russia:
    - HP-filter: -1.0
    - MVF: -0.1
    - MVF FIN: -3.6
    - PF: -4.1
    - Average (excl. HP-filter): -2.6
    - Sign: -
  - Moldova:
    - HP-filter: -0.9
    - MVF: 0.2
    - MVF FIN: -3.2
    - PF: -4.9
    - Average (excl. HP-filter): -2.6
    - Sign: -
  - Belarus:
    - HP-filter: -1.4
    - MVF: -0.2
    - MVF FIN: -1.5
    - PF: -2.6
    - Average (excl. HP-filter): -1.5
    - Sign: -
  - Poland:
    - HP-filter: -0.2
    - MVF: -0.3
    - MVF FIN: -1.5
    - PF: 3.0
    - Average (excl. HP-filter): 0.4
    - Sign: -
  - Romania:
    - HP-filter: 0.4
    - MVF: 0.1
    - MVF FIN: -3.7
    - PF: 0.2
    - Average (excl. HP-filter): -1.1
    - Sign: +
  - Serbia:
    - HP-filter: -1.2
    - MVF: -0.5
    - MVF FIN: -3.0
    - PF: 0.3
    - Average (excl. HP-filter): -1.1
    - Sign: -
  - Macedonia:
    - HP-filter: -0.2
    - MVF: -0.1
    - MVF FIN: 0.5
    - PF: 1.5
    - Average (excl. HP-filter): 0.6
    - Sign: +
- Potential growth estimates (2015, y-o-y, percent) — selected countries:
  - Ukraine:
    - HP-filter: -4.0
    - MVF: -10.7
    - MVF FIN: -6.6
    - PF: -4.4
    - Average (excl. HP-filter): -7.2
  - Russia:
    - HP-filter: -0.2
    - MVF: -1.4
    - MVF FIN: 0.3
    - PF: 0.7
    - Average (excl. HP-filter): -0.1
  - Poland:
    - HP-filter: 3.1
    - MVF: 3.4
    - MVF FIN: 3.3
    - PF: 1.8
    - Average (excl. HP-filter): 2.9
  - Romania:
    - HP-filter: 2.8
    - MVF: 3.8
    - MVF FIN: 3.1
    - PF: 2.1
    - Average (excl. HP-filter): 3.0
  - Turkey:
    - HP-filter: 3.4
    - MVF: 3.8
    - MVF FIN: 3.2
    - PF: 3.3
    - Average (excl. HP-filter): 3.4
- Methodological notes on these estimates:
  - MVF and HP estimates incorporate data up to 2015Q4; MVF-FIN up to 2015Q2; PF up to 2015Q3.
  - Estimates produced mechanically using same methods for all countries without country-specific judgment.

### IX. Policy implications and recommendations
- Overall prescriptions:
  - Well-targeted structural reforms are needed to raise potential growth—addressing TFP and investment constraints.
  - Expansionary fiscal and monetary policies, subject to availability of policy space, can help mitigate hysteresis by boosting employment and demand and support structural reform to lift potential output (IMF, 2016b).
  - Active labor market policies may be necessary in some countries to increase labor force participation and offset negative impacts of shrinking labor force on potential growth (IMF, 2016a).
- Country-specific diagnostics required:
  - When MVF-FIN and MVF/PF diverge (e.g., SEE-EU in 2015), evaluate private sector balance sheets and nature of financial frictions to decide whether financial repair policies are warranted or whether creditless recovery is benign.

*Source: IMF staff calculations and analyses (wp1737).*

### References  _______________________________________________________________25

### wp1737 - References  _______________________________________________________________25

### I. INTRODUCTION
- Postcrisis average growth in CESEE fell from precrisis average growth of 5.4 percent (2000-08) to 1.2 percent in the postcrisis (2009-15), a drop of 4.2 percentage points.  
- This drop was larger than in advanced economies (1.3 percentage points) and other emerging economies (1 percentage point).  
- Key analytic question: Was the slowdown largely cyclical or structural? Implications affect policy mix and sequencing of macroeconomic support and structural reforms; hysteresis from prolonged downturns could permanently scar potential output.  
- Challenges in estimating potential output for CESEE: potential output is unobservable; measurement errors; short time series with just one boom-bust cycle for many countries; major structural shifts including financial deepening.  
- Scope: consistent potential output estimates for 18 CESEE countries using multiple approaches: (i) Hodrick-Prescott (HP) filter; (ii) multivariate filter (MVF) linking GDP, unemployment, and inflation; (iii) expanded multivariate filter incorporating financial/credit cycles (MVF-FIN); and (iv) production function (PF) approach.  
- Methodological refinements for CESEE: adjusted financial variables (credit and house prices) to better capture cyclical developments; include wage inflation, hours worked, and capacity utilization to better gauge cyclical pressures in low-inflation and externally disinflationary settings.  
- Caveats: frequent data revisions and end-point bias of filters; large confidence bands due to structural changes and short samples.

### Main findings
- All methods indicate that a large share of the postcrisis growth slowdown in CESEE was sustained: potential output growth slowed by about 2 percentage points after the crisis compared with precrisis years, accounting for about half of the total growth slowdown.  
- Postcrisis potential growth was mainly dragged down by stagnant total factor productivity (TFP) and, to a lesser extent, slower capital accumulation. Estimates of TFP are uncertain as TFP is measured as a residual; further analysis is needed to identify drivers of the TFP slowdown.  
- The region-wide slowdown in potential growth has been associated with common external factors, which may reflect trading partners’ potential growth and the evolution of global trade and global supply chains. For some countries, negative demographics also contributed.  
- Differences between potential growth estimates across methods are generally small, reflecting overlapping information sets and interconnected real and financial sectors. Larger differences arise when synchronization between real and financial cycles is low, possibly due to financial frictions and real economy rigidities.  
- Despite the slowdown in potential growth, the postcrisis rebound suggests gradual narrowing of negative output gaps, which appear largely closed in 2015 for much of the region, except the CIS and Turkey. This aligns with dissipating disinflationary pressures in non-tradable sectors and declining unemployment, while headline inflation remained low due to falling import (notably energy) prices.

### II. METHODOLOGY — Overview
- Employed methods reflect different concepts of potential output:
  - HP filter: purely statistical trend extraction from GDP.
  - MVF (multivariate filter): potential output under flexible nominal prices and wages; uses inflation, inflation expectations, and unemployment gap; relevant for monetary policy (Okun/Phillips framework).
  - MVF-FIN (multivariate filter with financial frictions): estimates sustainable (inflation- and finance-neutral) potential output by incorporating credit growth, house prices, and capacity utilization.
  - PF (production function): supply-side measure using capital, labor (including average hours), capacity utilization, and TFP (Solow residual).

### II.A HP Filter
- Tool: Hodrick-Prescott univariate filter applied to GDP trend extraction.  
- Advantages: simplicity and transparency.  
- Limitations: purely statistical (no direct economic interpretation as inflation-neutral or sustainable potential); end-of-sample problem with significant revisions.

### II.B Multivariate Filtering Approach (MVF)
- Concept: potential output as level achievable under flexible nominal prices and wages (Okun, 1962).  
- Model elements:
  - Uses actual inflation, inflation expectations, and unemployment gap to distinguish demand and supply shocks.
  - Two economic relations imposed: Phillips curve linking output gap to inflation; and Okun’s law linking cyclical unemployment to output gap.
  - Bayesian estimation of a simultaneous system for output, inflation, unemployment, and their trends; additional equations govern stochastic trends and shocks.
- Advantages: limited data requirements; more robust to end-of-sample problems; can incorporate medium-term forecasts to improve outer-year estimates.
- Application to CESEE:
  - Headline and core inflation may be contaminated by imported inflation; wage inflation is incorporated to better capture domestic demand pressures.
  - Note: Unless stated otherwise, MVF uses wage inflation for EU members of CESEE comprising Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, and Slovenia.
  - Empirical result: using wage inflation suggests less slack in several CESEE countries in 2014-15.

### II.C Multivariate Filter with Financial Frictions (MVF-FIN)
- Concept: sustainable potential output is inflation-neutral and finance-neutral (Borio, 2013); uses financial variables as additional information to identify demand vs. supply shocks.  
- Method:
  - Decompose observed GDP into trend (sustainable) output y_t* and cycle c_t with constraint on variance ratio λ mirroring HP filter.
  - Allow output gap to depend on lagged values and observable variables x_t (credit growth, consumer price inflation, house price inflation where available, and capacity utilization).
  - Model estimated by maximum likelihood in state-space context.
- Application to CESEE:
  - CESEE experienced strong credit expansion precrisis and contraction postcrisis; disentangling financial cycles from financial deepening is essential.
  - Use IMF (2015) credit gap estimates to adjust credit series so structural increases in credit stocks (financial deepening) are not treated as credit cycles.
  - Ignoring structural credit increases would understate potential output in boom years and overstate it after the crisis—particularly important for Baltic states and SEE-EU countries.
  - Note: authors acknowledge use of provided STATA codes from Mico Mrkaic for MVF-FIN.

### II.D Production Function Approach (PF)
- Framework: Cobb-Douglas production function to decompose output into capital, labor, and TFP (Solow residual).  
- Specification: Y_t = A_t (K_t CU_t)^(1-α) (L_t AHW_t)^α
  - K_t: capital stock from perpetual inventory K_t = (1-ρ)K_{t-1} + I_t; ρ calibrated using historical average depreciation; initial capital stock from Penn World Tables.
  - CU_t: survey-based capacity utilization.
  - L_t: number of employed persons (national accounts concept).
  - AHW_t: average hours worked (national accounts).
  - α: labor share, calculated as ratio of compensation of employees to gross value added.
- TFP:
  - Computed as residual; includes human capital (whose contribution before and after the crisis is small).
  - In some countries TFP also includes capacity utilization and average hours worked due to data constraints.
- Potential output derivation:
  - Uses calibrated labor share, smoothed trends in unemployment and TFP, and actual capital stock.
  - Unemployment rates, average hours worked, and TFP are smoothed using HP filter (smoothing factor 1600).

*Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1737.pdf*

### 0.04 percentage points.

### wp1737 - 0.04 percentage points.

### Application to CESEE
- Not accounting for intensities of production factors may potentially understate booms and busts.  
- Capacity utilization and average hours worked have cyclical components that contribute to the overall business cycle.
- Capacity utilization:
  - Increases during booms when growing output is hitting production limits and falls during busts when capacity becomes excessive for a new lower level of output.
  - Not accounting for cyclical components in capacity utilization may overstate potential output growth during booms and understate it during busts—making output gap less volatile.
- Average hours worked:
  - Exhibits cyclical fluctuations; during booms incumbent workers work more hours, during busts hours are reduced.
  - This is most pronounced in the Baltic states and SEE-EU countries.
- Comparison of output gaps:
  - Figure A.3 in Appendix A compares output gaps derived using production function approach with and without adjusting capacity utilization and average hours worked for the CESEE EU countries.
  - The output gap amplitude appears to be smaller by 2-8 percentage points in the case of production function without adjustments for capacity utilization and average hours worked.

### How can different approaches be reconciled?
- All approaches aim to disentangle cyclical and structural components in real GDP but use different economic concepts, information sets, and estimation methods.
- Differences in trend growth estimates seem to come mainly from differences in information sets.
- Two factors matter most in explaining divergence between potential output estimates:
  - Richness of model specification:
    - The richer the specification of each method, the smaller the differences in potential output estimates across methods.
    - Information inputs overlap to a large degree across methods, more so between MVF-FIN and PF than with the MVF.
    - A richer specification results in better identification of cyclical component of business cycle, increasing overlap between cyclical components captured by different methods.
  - Synchronization of labor and capital markets cycles:
    - More flexible labor markets and deeper capital markets tend to reduce differences in potential output estimates across methods.
    - Structural characteristics and rigidities—especially frictions in financial (related to level of financial deepening) or labor markets—may lead to larger differences across methods.
    - Correlation between output (through capacity utilization) and credit cycles is more pronounced for countries with relatively deeper financial markets (CEE countries).
    - Correlation between average hours worked and wage inflation appears stronger in countries with more flexible labor markets (the Baltics).
- Example: SEE-EU countries
  - Precrisis: credit growth, capacity utilization, hours worked, headline and wage inflation moved in tandem; all methods depict similar output gap dynamics.
  - Postcrisis: credit remained flat while capacity utilization and wage inflation increased robustly.
    - MVF and PF show a closed gap in 2015.
    - MVF-FIN interprets low credit growth as negative output gap.
  - Two possible interpretations of divergent estimates in 2015 (range of negative MVF-FIN and positive MVF and PF):
    - Benign: the financial cycle is lagging the business cycle and would soon catch up; no policy action required.
    - Less benign: balance sheets of firms, households and/or financial intermediaries remain impaired, preventing growth at full potential; further policy action may be needed to address balance sheet weaknesses.
  - Choosing between interpretations requires more country-specific information about private sector balance sheets and nature of financial frictions.
- Note on creditless recovery:
  - Creditless recoveries are characterized by significant job creation and no growth in real credit.

### Results — Potential Growth and its Drivers
- Postcrisis potential growth has been below precrisis levels across CESEE countries.
  - With some variation—1-2 percentage points across methodologies—all methodologies point to a substantial postcrisis decline in potential growth.
  - Estimated potential growth was broadly stable over 2001-07, declined significantly after the global financial crisis, and as of 2015 had not recovered to precrisis level.
  - In more than two thirds of countries in the sample, potential growth slowdown started before 2007.
- Decomposition (production function framework) and role of external/common vs domestic/idiosyncratic factors:
  - Before the crisis:
    - Most countries enjoyed strong TFP growth; some had dynamic capital accumulation; labor played a smaller role.
    - On average, capital accumulation contributed more than 2 percentage points to the region’s potential growth during 2005-09.
    - In Bulgaria, Estonia, and Latvia, capital contribution was about 4 percentage points.
    - Average estimated TFP contribution during 2005-09 was about 2 percentage points.
    - In Latvia, Lithuania, Serbia, and Slovakia, estimated TFP contribution was close to 3 percentage points.
    - Some countries (Romania and Moldova) experienced outward migration, giving rise to a negative contribution of labor to potential growth.
  - After the crisis:
    - Most countries experienced a significant TFP growth slowdown.
    - Contribution of capital accumulation was mixed:
      - In 7 out of 18 countries (including Bulgaria, the Baltics, and Slovenia), contribution of capital accumulation to growth halved after the crisis.
      - These countries had rapid credit growth before the crisis fueled by external borrowing and adjusted current account balances mainly through reduction in investments.
      - Countries like the Czech Republic or Poland, which had not experienced excessive borrowing precrisis, continued to record strong capital contributions.
      - Romania experienced excessive credit boom financed by external borrowing, but much credit financed consumption; increase in saving rate after the crisis contributed to current account adjustment.
- External versus country-specific factors:
  - External/common factors closely correlated with potential growth in CESEE countries excluding the CIS and Turkey; in CIS and Turkey idiosyncratic/country-specific factors were more important, particularly after the crisis.
  - Euro area growth:
    - Above average regional potential growth before the crisis appears associated with strong precrisis potential growth in the euro area; this correlation persisted during the crisis.
  - Other common factors:
    - For CESEE excluding CIS and Turkey, “other common factors” exhibited strong co-movement with region’s potential growth before the crisis.
    - “Other common factors” grew much faster before the crisis, slowed notably after the crisis, then accelerated more recently.
    - Positive co-movement before the crisis may reflect expansion of global supply chains and growing trade links with rest of Europe, on the back of EU accession (8 CESEE countries joined the EU in 2004 and another two countries joined the EU in 2007).
    - Negative contribution after the crisis could be due to slowdown in global trade, structural bottlenecks, and slow resolution of crisis legacies (NPLs, debt overhang) that held back investment.
    - Increased absorption of EU structural funds in more recent years could underlie positive contribution of “other common factors”.

### How Large Were Output Gaps in 2015?
- Given lower postcrisis potential growth, recent growth rebound implies output gaps have been narrowing and appear largely closed for CESEE countries, excluding the CIS, by 2015.
- In 10 out of the 18 countries in the sample, at least two out of three methods (MVF, MVF-FIN, and PF) produce positive, albeit small, output gap estimates for 2015, suggesting recent growth pickup may have been more cyclical than structural.
  - In most cases positive output gap estimates are below one percent, which could be treated as largely closed.
- CIS countries:
  - The CIS countries are among 8 countries where at least two methods indicate negative output gaps.
- Inflation and labor market signals:
  - Headline inflation has been decelerating since end-2011, while estimated output gaps have been narrowing since 2010 and appear largely closed in CEE, SEE-EU, and the Baltics as of 2015.
  - Headline inflation over this period was largely driven by falling import prices, notably amid sustained declines in oil prices.
  - Inflation in non-tradable services—more sensitive to slack—shows disinflationary pressures have been gradually dissipating.
  - Unemployment rates in many countries have declined to precrisis levels.
  - Both core and services inflation in most countries of the region were notably below inflation targets by end-2015.

*Source: IMF staff calculations and analyses in the provided content.*

### 2015. Ukraine exhibits particularly large negative gap, followed by Russia, Moldova, and

### wp1737 - 2015. Ukraine exhibits particularly large negative gap, followed by Russia, Moldova, and

### Output gap estimates (2015) — key country results
- Ukraine (percent of potential output, 2015):
  - HP-filter: -7.5
  - MVF: -0.5
  - MVF FIN: -12.9
  - PF: -6.5
  - Average (excl. HP-filter): -6.6
  - Sign of output gap (based on at least two of 3 methods): -
- Russia (2015):
  - HP-filter: -1.0
  - MVF: -0.1
  - MVF FIN: -3.6
  - PF: -4.1
  - Average (excl. HP-filter): -2.6
  - Sign: -
- Moldova (2015):
  - HP-filter: -0.9
  - MVF: 0.2
  - MVF FIN: -3.2
  - PF: -4.9
  - Average (excl. HP-filter): -2.6
  - Sign: -
- Belarus (2015):
  - HP-filter: -1.4
  - MVF: -0.2
  - MVF FIN: -1.5
  - PF: -2.6
  - Average (excl. HP-filter): -1.5
  - Sign: -
- Poland (2015):
  - HP-filter: -0.2
  - MVF: -0.3
  - MVF FIN: -1.5
  - PF: 3.0
  - Average (excl. HP-filter): 0.4
  - Sign: -
- Romania (2015):
  - HP-filter: 0.4
  - MVF: 0.1
  - MVF FIN: -3.7
  - PF: 0.2
  - Average (excl. HP-filter): -1.1
  - Sign (based on at least two of 3 methods): +
- Serbia (2015):
  - HP-filter: -1.2
  - MVF: -0.5
  - MVF FIN: -3.0
  - PF: 0.3
  - Average (excl. HP-filter): -1.1
  - Sign: -
- Macedonia (2015):
  - HP-filter: -0.2
  - MVF: -0.1
  - MVF FIN: 0.5
  - PF: 1.5
  - Average (excl. HP-filter): 0.6
  - Sign: +

### Potential growth estimates (2015) — selected country figures (y-o-y, percent)
- Ukraine (2015):
  - HP-filter: -4.0
  - MVF: -10.7
  - MVF FIN: -6.6
  - PF: -4.4
  - Average (excl. HP-filter): -7.2
- Russia (2015):
  - HP-filter: -0.2
  - MVF: -1.4
  - MVF FIN: 0.3
  - PF: 0.7
  - Average (excl. HP-filter): -0.1
- Poland (2015):
  - HP-filter: 3.1
  - MVF: 3.4
  - MVF FIN: 3.3
  - PF: 1.8
  - Average (excl. HP-filter): 2.9
- Romania (2015):
  - HP-filter: 2.8
  - MVF: 3.8
  - MVF FIN: 3.1
  - PF: 2.1
  - Average (excl. HP-filter): 3.0
- Turkey (2015):
  - HP-filter: 3.4
  - MVF: 3.8
  - MVF FIN: 3.2
  - PF: 3.3
  - Average (excl. HP-filter): 3.4

### Cross-methodology differences and interpretation
- Several countries show large dispersion across methodologies in 2015 output gap and potential growth estimates (examples below illustrate methodological variance):
  - Romania: PF and MVF indicate slightly positive 2015 output gap, while MVF-FIN shows a significantly negative gap — possibly reflecting low synchronization of business and credit cycles due to financial market frictions.
  - Ukraine: extreme variation across methods (HP-filter -7.5 vs MVF-FIN -12.9 vs MVF -0.5) indicates high model sensitivity and severe negative gap consensus across methods.
- Regional patterns:
  - Among CESEE EU countries, Poland, Slovenia, and Croatia are the only countries with at least two methods producing negative but small output gaps for 2015.
  - Output gap estimates for the Baltic countries and for the rest of the CE-5 countries are positive, albeit relatively small.
  - In SEE-EU, two out of three methods show small positive output gaps for both Romania and Bulgaria.
  - SEE non-EU: two out of three methods produce negative output gap for Serbia in 2015; two out of three methods suggest positive gap for Macedonia; aggregate dominated by Serbia.
- Methodological notes:
  - MVF and HP estimates incorporate data up to 2015Q4, MVF FIN up to 2015Q2, PF up to 2015Q3.
  - Methods used: MVF = Multi-variate filter, MVF FIN = Multi-variate filter with financial frictions, HP = Hodrick-Prescott filter, PF = Production function.
  - Estimates are generated applying same methods to all countries without country-specific judgment.

### Inflation, labor market signals, and cyclical pressure (2011–2015 signals)
- Headline inflation may mask emerging domestic inflationary pressures because of imported low inflation and lower oil prices.
- Wage inflation and non-tradable services inflation indicate dissipating disinflationary pressures, consistent with:
  - falling unemployment rates
  - increasing capacity utilization
- Despite these signs, core inflation in most of the region remains considerably below inflation targets.

### Main conclusions and policy implications
- Overall regional trend:
  - Potential growth in the CESEE region has declined appreciably across approaches.
  - Total Factor Productivity (TFP) played a key role: boosting precrisis trend growth and weighing down growth in recent years.
  - Slow capital accumulation contributed to weaker potential growth in several countries.
  - Labor contribution varied: growth-supportive in CEE, Turkey, and Russia; negative in SEE and other CIS.
- External vs idiosyncratic drivers:
  - For CESEE countries excluding the CIS and Turkey, potential output growth has been largely associated with trading partners’ potential growth and other common factors.
  - In the CIS countries and Turkey, idiosyncratic factors became more important, particularly after the crisis.
- Policy recommendations:
  - Well-targeted structural reforms are needed to raise potential growth.
  - Expansionary fiscal and monetary policies, subject to availability of policy space, could help mitigate hysteresis effects by boosting employment and demand and support structural reform to lift potential output (IMF, 2016b).
  - Some countries may need active labor market policies to increase labor force participation and offset the negative impact of shrinking labor force on potential growth (IMF, 2016a).

### Methodological appendices — data and decomposition approach
- Decomposition of potential growth into common and idiosyncratic components:
  - Regression relates country potential growth to weighted average potential growth of trading partners (weighted by exports) plus average potential growth across countries (other common component) and a country-specific term. Data are demeaned and detrended to control for fixed and time effects.
- Data sources and variables used across methods (examples):
  - Real GDP: World Economic Outlook, April 2016 — used in HP filter, MVF, MVF-FIN, Production Function.
  - Unemployment rate, Headline inflation: World Economic Outlook, April 2016 — used in MVF and MVF-FIN respectively.
  - Wage inflation: Eurostat, Haver Analytics — used in MVF.
  - Credit growth and credit growth adjusted for financial deepening (credit gap): used in MVF-FIN.
  - House price indices, capital stock, average hours worked, capacity utilization, labor force participation rate, number of employed persons: Haver Analytics, Penn World Tables 8.1, IMF sources — used in various approaches as specified for each method.

*Source: IMF staff calculations (wp1737).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1737.pdf_