## _wp10250

## Source details

**Canonical URL:** [_wp10250](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10250.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10250.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10250.pdf.json)

---

### Executive summary and main conclusions
- Between 2000 and 2007 nonfinancial private sector credit expanded rapidly in the Baltic countries; this paper uses an experimental Debt Overhang Index (DOI) to assess whether legacy debt could hold back the economic recovery.
- Key channels linking debt overhang to activity: overleveraged households cutting spending when perceptions of permanent income and wealth change; shareholder debt overhang discouraging new projects that mainly benefit existing debt holders; impaired bank lending capacity from deteriorating credit portfolios; and relatively large output losses from house price busts.
- Principal findings:
  - When individual leverage indicators are ambiguous, jointly considering a range of balance sheet measures and distributional effects is useful. The DOI facilitates a consistent, broad-based assessment across countries and sectors.
  - All Baltic countries appear to be at risk of debt overhang, with differences across sectors and countries:
    - Risk could be more acute in the household than in the corporate sector due to limited household financial assets.
    - Accounting for the distribution of household leverage and shares in consumption increases risk in Estonia and lessens it in Lithuania.
    - The corporate sector shows an aggregate problem of overstretched balance sheets mainly in Latvia.
    - In Estonia, distributional factors mitigate corporate risks, with sectors contributing considerably to investment not exhibiting particularly high leverage.
    - Foreign corporate ownership, although not captured in the DOI, could play a small role in alleviating debt overhang problems.
  - Preliminary empirical indications (partial information up to mid-2010):
    - For consumption, preliminary proxies for debt-overhang–induced losses have averaged around 3 percent of GDP in the Baltic countries since peaking pre-crisis.
    - For fixed investment, the DOI suggests potential losses in Latvia worth about 4 percent of GDP and the absence of major effects in Estonia and Lithuania.
  - The DOI is experimental; confronting it with the full subsequent economic experience of the Baltic countries will be important.

### How indebted is the nonfinancial private sector? — stylized facts and key statistics
- Credit and timing:
  - Between 2004 and 2007–08 the Baltic countries experienced considerable expansions of private sector credit.
  - Bank credit to the private sector as a percent of GDP doubled in Lithuania and rose by 60–80 percent in Estonia and Latvia.
  - The share of household borrowing in bank credit rose in all three countries, accounting for close to half at end 2009 and up from around 30–35 percent five years earlier.
- Measurement caveats:
  - Some models estimate that only in Estonia credit vastly exceeds levels commensurate with fundamentals; Lithuania would be considerably below its fundamental level of credit depth and Latvia roughly in-line — but these estimates ignore nonbank and nondomestic debt.
  - This paper focuses on measures of total indebtedness rather than bank credit only.
- Household balance sheets and distributional facts:
  - Baltic household indebtedness does not yet exceed the euro area average on some measures, but debts are not matched by financial wealth as in the euro area.
  - Net financial asset position relative to disposable income in the Baltics is less than 70 percent versus 170 percent in the euro area.
  - House prices fell between 50-80 percent from their peaks, severely diminishing housing wealth.
  - Distributional facts:
    - In Estonia, 30 percent of the top income quintile households reported being indebted compared with well below 2 percent for the lowest quintile.
    - In Estonia and Latvia, high income households still have below average debt service ratios, in contrast with Lithuania.
- Corporate balance sheets:
  - Gearing levels of the nonfinancial corporate sector are at or below euro area levels, except for Latvia, whose debt-to-equity ratio is more than double that in Estonia and Lithuania.
  - Volatility of Operating Surplus; Standard deviation of quarterly changes; 2002Q2-2008Q3: Estonia 9.01; Lithuania 16.76; Latvia 14.16; Euro area (12) 0.60.
  - Sectoral distribution:
    - Latvia and Lithuania display considerable variation across sectors; Estonia less so.
    - Hotels and restaurant, retail wholesale, real estate and construction recorded the highest leverage ratios in Latvia and Lithuania.
- Debt service and currency composition:
  - Declining private sector incomes and deflationary pressures increased debt burden during the crisis, but extensive euroization of debt contracts helped mitigate the increase in interest service:
    - Lithuania: 70 percent foreign-currency denominated.
    - Latvia and Estonia: over 90 percent foreign-currency denominated.
  - Interest service payments fell during 2009 as monetary policy was loosened in the eurozone.
  - In Lithuania, high local currency rates and falling incomes may have offset the benefit of foreign-currency denomination.
  - Banks have been rescheduling debt service payments into the future, alleviating short-term pressures, but rescheduling is typically offset by more collateral or higher interest rates.

### Debt Overhang Index (DOI): construction, interpretation, and preliminary implications
- Purpose and intuition:
  - The DOI is a composite index of different flow and stock measures of aggregate indebtedness, scaled relative to the euro area average (aggregate DOI), and weighted by each sector’s or household-quintile’s share in demand (weighted DOI). It is an organizing framework for joint comparison of balance sheet situations across countries and sectors, not a precise measure of debt overhang.
- Schematic construction (as presented in the source):
  - Aggregate DOI:
    - Aggregate ijeurozone
      i
      ijjxxDOI/10
      3
      1
      
      
       with j=household or corporate sector; x_i = balance sheet indicator i, covering 3 indicators.
  - Weighted DOI:
    - Weighted )/(/(())
      1
      jjmj
      M
      m
      jmjj
      DDyyDOIDOI
      
       with y_m = debt indicator of sector/quintile m; D_m = demand of sector /quintile m; M=5 for household sector; M=9 for corporate sector.
- Indicators included (conceptual):
  - Aggregate leverage indicators such as debt-to-income, debt-to-equity, debt service ratios, and net financial assets over income are scaled relative to the euro area and combined to generate a single aggregate DOI.
  - The aggregate DOI is applied to corporate sectors and household quintiles using distributional information; sectoral/quintile DOIs are summed weighted by respective shares in investment/consumption to obtain the weighted DOI.
- Preliminary empirical indications:
  - Concentration of nonperforming loans during the crisis in highly leveraged sectors (real estate development, construction, hotels) reveals undue sectoral indebtedness in parts of the region.
  - Balance sheet adjustments can be protracted in the context of deflation (examples cited from post-Asian crisis Hong Kong and 1990s Sweden).
  - Consumption: preliminary proxies suggest debt-overhang losses averaging around 3 percent of GDP in the Baltics since pre-crisis peaks.
  - Investment: DOI suggests potential losses in Latvia worth about 4 percent of GDP; little evidence of major effects in Estonia and Lithuania.
- Limitations and next steps:
  - The DOI is experimental and not yet comprehensively tested; data constraints limit cross-country/over-time assessments using other episodes.
  - Many DOI stock variables are based on data before the crisis; confronting the DOI with the full subsequent economic experience of the Baltic countries will be necessary.

### Policy-relevant implications highlighted in the analysis
- Diagnostic approach:
  - Use a broad, consistent framework (like the DOI) that combines multiple balance sheet indicators and distributional information rather than relying on single indicators.
- Sectoral focus:
  - Household vulnerability suggests policies addressing household balance sheet repair and targeted social safety nets may be important to support consumption recovery.
  - In Latvia, corporate-sector balance-sheet strains and sectoral pockets of leverage (real estate, construction, hotels) suggest measures to address corporate deleveraging and to facilitate reallocation of resources toward less-levered, productive sectors.
- Financial sector and debt resolution:
  - Bank rescheduling alleviates near-term stress but may maintain elevated debt burdens without addressing net present value; policies to improve debt restructuring, collateral provisioning, and bank balance sheet repair could help restore lending capacity.
- Role of external factors:
  - High degree of foreign-currency denomination of debt in Estonia and Latvia (over 90 percent) and somewhat lower in Lithuania (70 percent) influences debt-service dynamics and the transmission of eurozone monetary conditions; currency composition should be considered in policy responses.
- Monitoring and evaluation:
  - Further data collection (including distributional detail on leverage and demand shares) and ongoing testing of the DOI against realized consumption and investment outcomes are recommended to refine policy choices and better forecast debt-overhang effects.

### Annex II — DOI methodology, preserved tables, and country computations
- Methodology and definitions:
  - Notation: D = debt; GDI = gross disposable income; R = debt service; NFA = net financial assets; NVA = net value added; GOS = gross operating surplus.
  - Timing convention:
    - Ratios involving stock variables (e.g. D/GDI or NFA/GDI) refer to 2008 unless indicated otherwise.
    - Ratios based on flow variables (e.g. R/GOS or R/GDI) refer to observations in 2009 unless indicated otherwise.
  - Aggregation procedure (household sector example for Lithuania):
    - Step 1: Three aggregate balance-sheet indicators are scored relative to the euro area (euro area score = 30).
    - Step 2: Country-level indicators are scored and summed to produce the aggregate DOI (example: Lithuania aggregate DOI = 48).
    - Step 3: Aggregate DOI allocated across income quintiles based on debt service ratios.
    - Step 4: Quintile DOIs weighted by consumption shares to produce consumption/quintile weighted DOI (example: Lithuania consumption/quintile weighted DOI = 41).
- Aggregate household DOI — preserved values (stock variables 2008, flow variables 2009 unless flagged):
  - D/GDI: Estonia 9.2, Latvia 8.0, Lithuania 6.0, Euro area 10
  - NFA/GDI: Estonia 28.3, Latvia 73.9, Lithuania 34.0, Euro area 10
  - R/GDI: Estonia 11.9, Latvia 10.3, Lithuania 7.8, Euro area 10
  - Sum (aggregate DOI): Estonia 49.9, Latvia 247.8, Lithuania 30.0, Euro area 30.0
- Example household quintile computations (exact preserved figures):
  - Estonia:
    - Debt service per quintile (2007) — Percentage debt service/income Quintile 1–5 and Average = 10.0, 8.0, 5.0, 6.0, 6.0, 6.0
    - Relative to average: 1.7, 1.3, 0.8, 1.0, 1.0, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.4, 1.0
    - DOI per income quintile Quintile 1–5: 82.3, 65.9, 41.2, 49.4, 49.4
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 8.2, 8.6, 6.6, 11.4, 17.8, 52.5
  - Latvia:
    - Debt service per quintile (2007) — Percentage debt service/income Quintile 1–5 and Average = 10.0, 10.0, 8.0, 7.0, 5.0, 7.0
    - Relative to average: 1.4, 1.4, 1.1, 1.0, 0.7, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.3, 1.0
    - DOI per income quintile Quintile 1–5: 131.8, 131.8, 105.4, 92.2, 65.9
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 14.5, 17.1, 17.9, 21.2, 22.4, 93.1
  - Lithuania:
    - Debt service per quintile (2007) — Percentage debt service/income Quintile 1–5 and Average = 2.0, 2.0, 5.0, 3.0, 4.0, 4.0
    - Relative to average: 0.5, 0.5, 1.3, 0.8, 1.0, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.3, 1.0
    - DOI per income quintile Quintile 1–5: 23.9, 23.9, 59.8, 35.9, 47.8
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 2.6, 3.3, 11.4, 8.2, 15.3, 40.9
- Special treatment:
  - For Latvia, NFA/GDI is capped at the ratio of the next lowest eurozone country (Slovakia) with a ratio of 0.23; capping is justified by diminishing additional demand effects beyond a threshold.
- Corporate sector preserved leverage ratios (cross-country):
  - D/NVA: Estonia 2.7, Latvia 2.4, Lithuania 1.6, Eurozone 3
  - D/E: Estonia 0.96, Latvia 2.5, Lithuania 1.1, Eurozone 1.3
  - R/GOS: Estonia 0.15, Latvia 0.23, Lithuania 0.11, Eurozone 0.17
- Corporate DOI scoring (eurozone = 10 per indicator; scores preserved):
  - D/NVA: Estonia 9.7, Latvia 9.5, Lithuania 3.1, Eurozone 10
  - D/E: Estonia 7.4, Latvia 19.2, Lithuania 8.5, Eurozone 10
  - R/GOS: Estonia 8.8, Latvia 13.5, Lithuania 6.5, Eurozone 10
  - Sum (aggregate corporate DOI): Estonia 25.2, Latvia 40.6, Lithuania 20.3, Eurozone 30.0
- Country-level corporate sector computations (selected preserved figures):
  - Estonia:
    - Average (2008) D/E = 0.96; sectoral DOIs and shares in total corporate investment lead to DOIs weighted by share of investment Total = 22.1 (sector contributions Manufacturing 2.9; Mining 0.2; Electricity, Gas and Water 1.4; Hotels and Restaurants 0.3; Transport 0.3; Real Estate 8.1; Wholesale, Retail etc 2.8; Construction 1.0; Other 5.0).
  - Latvia:
    - Average (2008) D/E = 2.50; sectoral DOIs and shares in total corporate investment lead to DOIs weighted by share of investment Total = 38.3 (sector contributions Manufacturing 4.6; Mining 0.1; Electricity, Gas and Water 1.1; Hotels and Restaurants 0.9; Transport 3.4; Real Estate 8.5; Wholesale, Retail etc 5.8; Construction 4.0; Other 10.0).
  - Lithuania:
    - Average (2008) D/E = 1.10; sectoral DOIs and shares in total corporate investment lead to DOIs weighted by share of investment Total = 20.4 (sector contributions Manufacturing 3.2; Mining 0.0; Electricity, Gas and Water 0.4; Hotels and Restaurants 0.5; Transport 1.7; Real Estate 7.0; Wholesale, Retail etc 3.1; Construction 1.3; Other 3.2).
- Key analytic intuition:
  - DOI combines balance-sheet leverage indicators with distributional information (household quintiles or corporate sectors) and demand weights (consumption shares or investment shares) to obtain a demand-adjusted measure of debt overhang.
  - Capping rules (example: Latvia NFA/GDI capped at 0.23) are applied where extreme values imply negligible additional demand effects beyond a threshold.

### Annex III — OLS regressions of household consumption and corporate investment (selected results and diagnostics)
- General variables and notation:
  - C = real household consumption; GDP = real GDP; UN = EUROSTAT harmonized unemployment rate (percent or D prefix = quarterly change); R = three-month euribor nominal interest rate (or D prefix = quarterly change); INV = real fixed capital formation.
  - Suffixes: EURO, LT, LV, EST denote eurozone, Lithuania, Latvia, Estonia respectively.
  - DUMEST and DUMESTINV = dummies for the Russian crisis sudden sharp collapse in consumption and investment. DUMLV = dummy for erratic decline in 2000:Q3.
- Selected household consumption regression diagnostics (preserved key metrics):
  - EURO regression: R-squared 0.273459; Adjusted R-squared 0.176587; S.E. of regression 0.002907; F-statistic 2.822882; Prob(F-statistic) 0.042329; Durbin-Watson stat 2.067796.
  - Estonia consumption regression: R-squared 0.453083; Adjusted R-squared 0.380160; S.E. of regression 0.020520; F-statistic 6.213225; Prob(F-statistic) 0.000914; Durbin-Watson stat 2.190036.
  - Lithuania consumption regression: R-squared 0.230567; Adjusted R-squared 0.134387; S.E. of regression 0.022501; F-statistic 2.397260; Prob(F-statistic) 0.070766; Durbin-Watson stat 2.066192.
  - Latvia consumption regression: R-squared 0.312063; Adjusted R-squared 0.193453; S.E. of regression 0.026054; F-statistic 2.631004; Prob(F-statistic) 0.044335; Durbin-Watson stat 1.585196.
- Selected investment regression diagnostics (preserved key metrics):
  - EURO investment regression: R-squared 0.341913; Adjusted R-squared 0.278227; S.E. of regression 0.007348; F-statistic 5.368741; Prob(F-statistic) 0.004283; Durbin-Watson stat 2.184276.
  - Estonia investment regression: R-squared 0.257775; Adjusted R-squared 0.209369; S.E. of regression 0.055280; F-statistic 5.325260; Prob(F-statistic) 0.003107; Durbin-Watson stat 2.257652.
  - Lithuania investment regression: R-squared 0.299526; Adjusted R-squared 0.233857; S.E. of regression 0.058959; F-statistic 4.561119; Prob(F-statistic) 0.009043; Durbin-Watson stat 2.123108.
  - Latvia investment regression: R-squared 0.220917; Adjusted R-squared 0.137444; S.E. of regression 0.029346; F-statistic 2.646560; Prob(F-statistic) 0.068466; Durbin-Watson stat 2.047282.
- Empirical interpretation:
  - Out-of-sample (static) projections for 2008–10 indicate persistent overprediction of consumption of about 4 to 8 percent in cumulative terms for the three Baltic countries, in contrast to absence of projection bias for the eurozone.
  - The on average 6 percent estimation gap could be interpreted as a rough proxy for lost household consumption resulting from debt overhang.
  - Given the share of private consumption in GDP, everything equal, the proxy for debt overhang indicates a reduction in GDP in the Baltic countries by around 3 percent.
  - The cumulative 14 percent gap between actual and projected gross capital formation between end-2007 and mid 2010 for Latvia is equivalent to around 4 percent of GDP and could be interpreted as a proxy for lost fixed capital formation from firms’ above average leverage.
  - For Estonia and Lithuania, there is no systematic overprediction, in line with DOI indicators.
  - Regression model fit: goodness of fit (R2) between 0.2 and 0.5; robustness checks conducted across specifications.

*Source: conclusions for domestic demand. (Working Paper content provided in the supplied PDF extract.)*

### conclusions for domestic demand.

### conclusions for domestic demand.

### Executive summary and main conclusions
- Between 2000 and 2007 nonfinancial private sector credit expanded rapidly in the Baltic countries; this paper uses an experimental Debt Overhang Index (DOI) to assess whether legacy debt could hold back the economic recovery.  
- Key empirical/theoretical channels linking debt overhang to activity include: overleveraged households cutting spending when perceptions of permanent income and wealth change; shareholders with debt overhang refraining from new projects whose returns mainly benefit existing debt holders; impaired bank lending capacity from deteriorating credit portfolios; and relatively large output losses from house price busts.  
- Principal findings:
  - When individual leverage indicators are ambiguous, jointly considering a range of balance sheet measures and distributional effects is useful. The proposed DOI does that, facilitating a consistent, broad-based assessment across countries and sectors.
  - All Baltic countries appear to be at risk of debt overhang, with differences across sectors and countries:
    - Risk could be more acute in the household than in the corporate sector due to limited household financial assets.
    - Accounting for the distribution of household leverage and shares in consumption increases risk in Estonia and lessens it in Lithuania.
    - The corporate sector shows an aggregate problem of overstretched balance sheets mainly in Latvia.
    - In Estonia, distributional factors mitigate corporate risks, with sectors contributing considerably to investment not exhibiting particularly high leverage.
    - Foreign corporate ownership, although not captured in the DOI, could play a small role in alleviating debt overhang problems.
  - It is too early for comprehensive testing of the predictive power of the DOI. Preliminary results (partial information up to mid-2010) indicate:
    - For consumption, preliminary proxies for debt-overhang–induced losses have averaged around 3 percent of GDP in the Baltic countries since peaking pre-crisis.
    - For fixed investment, the DOI suggests potential losses in Latvia worth about 4 percent of GDP and the absence of major effects in Estonia and Lithuania.
  - Going forward, confronting the DOI with the full economic experience of the Baltic countries as they settle toward new steady states will be important.

### How indebted is the nonfinancial private sector? — stylized facts and key statistics
- Credit and timing:
  - Between 2004 and 2007–08 the Baltic countries experienced considerable expansions of private sector credit.
  - Bank credit to the private sector as a percent of GDP doubled in Lithuania and rose by 60–80 percent in Estonia and Latvia.  
  - The share of household borrowing in bank credit rose in all three countries, accounting for close to half at end 2009 and up from around 30–35 percent five years earlier.
- Comparisons and measurement caveats:
  - Some models estimate that only in Estonia credit vastly exceeds levels commensurate with fundamentals; Lithuania would be considerably below its fundamental level of credit depth and Latvia roughly in-line — but these estimates ignore nonbank and nondomestic debt.
  - This paper focuses on measures of total indebtedness rather than bank credit only.
- Household balance sheets:
  - Baltic household indebtedness does not yet exceed the euro area average on some measures, but debts are not matched by financial wealth as in the euro area.
  - Net financial asset position relative to disposable income in the Baltics is less than 70 percent versus 170 percent in the euro area.
  - Households borrowed mainly to purchase real estate; house prices fell between 50-80 percent from their peaks, severely diminishing housing wealth.
  - Distributional facts:
    - In Estonia, 30 percent of the top income quintile households reported being indebted compared with well below 2 percent for the lowest quintile.
    - In Estonia and Latvia, high income households still have below average debt service ratios, in contrast with Lithuania.
- Corporate balance sheets:
  - Gearing levels of the nonfinancial corporate sector are at or below euro area levels, except for Latvia, whose debt-to-equity ratio is more than double that in Estonia and Lithuania.
  - Table 1 (Volatility of Operating Surplus; Standard deviation of quarterly changes; 2002Q2-2008Q3) — Estonia 9.01; Lithuania 16.76; Latvia 14.16; Euro area (12) 0.60.
  - The distribution of leverage across corporate sectors differs:
    - Latvia and Lithuania display considerable variation across sectors; Estonia less so.
    - Hotels and restaurant, retail wholesale, real estate and construction recorded the highest leverage ratios in Latvia and Lithuania.
- Debt service and currency composition:
  - Declining private sector incomes and deflationary pressures increased debt burden during the crisis, but extensive euroization of debt contracts helped mitigate the increase in interest service:
    - Lithuania: 70 percent foreign-currency denominated.
    - Latvia and Estonia: over 90 percent foreign-currency denominated.
  - In all countries, it is estimated that interest service payments fell during 2009 as monetary policy was loosened in the eurozone.
  - In Lithuania, high local currency rates and falling incomes may have offset the benefit of foreign-currency denomination.
  - Banks have been rescheduling debt service payments into the future, alleviating short-term pressures, but rescheduling is typically offset by more collateral or higher interest rates.

### Debt Overhang Index (DOI): construction, interpretation, and preliminary implications
- Purpose and intuition:
  - The DOI is a composite index of different flow and stock measures of aggregate indebtedness, scaled relative to the euro area average (aggregate DOI), and weighted by each sector’s or household-quintile’s share in demand (weighted DOI). It is an organizing framework for joint comparison of balance sheet situations across countries and sectors, not a precise measure of debt overhang.
- Schematic construction (as expressed in the source):
  - Aggregate DOI:
    - Aggregate ijeurozone
      i
      ijjxxDOI/10
      3
      1
      
      
       with j=household or corporate sector; x_i = balance sheet indicator i, covering 3 indicators.
  - Weighted DOI:
    - Weighted )/(/(())
      1
      jjmj
      M
      m
      jmjj
      DDyyDOIDOI
      
       with y_m = debt indicator of sector/quintile m; D_m = demand of sector /quintile m; M=5 for household sector; M=9 for corporate sector.
- Indicators included (conceptual):
  - Aggregate leverage indicators such as debt-to-income, debt-to-equity, debt service ratios, and net financial assets over income are scaled relative to the euro area and combined to generate a single aggregate DOI.
  - The aggregate DOI is applied to corporate sectors and household quintiles using distributional information; sectoral/quintile DOIs are summed weighted by respective shares in investment/consumption to obtain the weighted DOI.
- Preliminary empirical indications:
  - The concentration of nonperforming loans during the crisis in highly leveraged sectors (real estate development, construction, hotels) already reveals undue sectoral indebtedness in parts of the region.
  - Balance sheet adjustments can be protracted in the context of deflation (examples from post-Asian crisis Hong Kong and 1990s Sweden are documented), and the DOI’s preliminary results are consistent with observed behavior of consumption and investment up to mid-2010:
    - Consumption: preliminary proxies suggest debt-overhang losses averaging around 3 percent of GDP in the Baltics since pre-crisis peaks.
    - Investment: DOI suggests potential losses in Latvia worth about 4 percent of GDP; little evidence of major effects in Estonia and Lithuania.
- Limitations and next steps:
  - The DOI is experimental and not yet comprehensively tested; data constraints limit cross-country/over-time assessments using other episodes.
  - Many DOI stock variables are based on data before the crisis; confronting the DOI with the full subsequent economic experience of the Baltic countries will be necessary.

### Policy-relevant implications highlighted in the analysis
- Diagnostic approach:
  - Use a broad, consistent framework (like the DOI) that combines multiple balance sheet indicators and distributional information rather than relying on single indicators.
- Sectoral focus:
  - Given the potentially greater household vulnerability (limited financial assets, large housing wealth losses), policies addressing household balance sheet repair and targeted social safety nets may be important to support consumption recovery.
  - In Latvia, corporate-sector balance-sheet strains (high debt-to-equity) and sectoral pockets of leverage (real estate, construction, hotels) suggest a need for measures to address corporate deleveraging and to facilitate reallocation of resources toward less-levered, productive sectors.
- Financial sector and debt resolution:
  - Bank practices of rescheduling alleviate near-term stress but may maintain elevated debt burdens without addressing net present value; policies to improve debt restructuring, collateral provisioning, and bank balance sheet repair could help restore lending capacity.
- Role of external factors:
  - The high degree of foreign-currency denomination of debt in Estonia and Latvia (over 90 percent) and somewhat lower in Lithuania (70 percent) influences debt-service dynamics and the transmission of eurozone monetary conditions; currency composition should be considered in policy responses.
- Monitoring and evaluation:
  - Further data collection (including distributional detail on leverage and demand shares) and ongoing testing of the DOI against realized consumption and investment outcomes are recommended to refine policy choices and to better forecast debt-overhang effects.

*Source: conclusions for domestic demand. (Working Paper content provided in the supplied PDF extract.)*

### 2009. The DOI is subject to a number of limitations. For a start, euro area gearing ratios

### _wp10250 - 2009. The DOI is subject to a number of limitations. For a start, euro area gearing ratios

### Debt Overhang Index (DOI): scope and limitations
- The DOI synthesizes and compares the risk of debt overhang for domestic demand across sectors and countries, taking into account jointly flow and stock indicators of household and corporate sector indebtedness, and their distribution.
- The DOI has limitations: the current measure "imposes arbitrarily equal weighting and proportional scaling of indicators and relies on dated information on consumption shares by income quintiles."
- Euro area gearing ratios should probably be seen "as upper end benchmarks given that the euro area nonfinancial private sector itself is considered relatively highly leveraged."

### Household sector: levels, distribution, and housing equity
- Baltic household sectors score relatively poorly relative to the euro area, reflecting the low level of financial assets and indicating a relatively high risk of debt overhang.
- Latvia scores over three times the level in the euro area; Lithuania and Estonia nearly double the euro area level.
- Distributional effects:
  - In Estonia, distribution of debts further aggravates the DOI: "a 6 percent increase of the DOI."
  - In Lithuania, distributional effects lessen the DOI: "a 15 percent reduction of the DOI."
  - In Lithuania the first, second, and fourth quintile have below average DOIs, but account together for a larger share of consumption than the third income quintile who has an above average DOI.
  - In Estonia’s low income households are leveraged above average and account for a sufficiently large share of consumption to raise the weighted DOI above the aggregate.
- Negative housing equity (not captured in the DOI) could weigh on household consumption despite owner-occupied housing mitigating this risk.
  - Reflecting the sharp decline in house prices in 2008 and 2009, 15 percent of total mortgage contracts (25 percent of exposures) in Estonia were in negative equity in 2009:Q2.
  - At end June 2009, Swedbank reported that 54 percent of its mortgages in Latvia were in negative equity, compared with 37 percent for Lithuanian mortgages and 24 percent for loans granted in Estonia.
- Collateral channel importance: research suggests the collateral channel can be important, albeit time varying.

### Corporate sector: heterogeneity and investment links
- Corporate sector DOIs are more heterogeneous than household DOIs.
  - Latvia, due to the high debt-to-equity ratio, is well above the euro area average.
  - Lithuania and to a lesser extent Estonia are well below the euro area average.
- Weighted DOI (accounting for sectoral differences in leverage and contribution to investment) lowers the score significantly in Estonia because ‘other sectors’, which account for a large share in investment, have a below average DOI.
- Factors not captured in the DOI but relevant for investment:
  - Corporate investments have been partly financed through outside borrowing, not just cash flows. Analysis of corporate accounts from companies listed on local stock exchanges shows cash from operations and outside borrowing both important. Banks may curtail new lending if corporate balance sheets are overstretched, with implications for business investment.
  - Construction and real estate sectors were facing excess supplies in 2009, which typically take a while to be absorbed:
    - Construction activity in 2009 relative to the averages in 2006-2008: Estonia 50.8, Latvia 52.2, Lithuania 80.4 (percent).
    - Office vacancy rates in the capital in Q3 2009: Estonia 20.0, Latvia 27.0 (Q4 2009 for Latvia), Lithuania 18.0 (percent).
    - With office vacancy rates in respective capitals of between 20–30 percent (higher outside) and construction activity still at 50–80 percent of the average in the previous two years, investment in housing is unlikely to resume quickly.
  - Leverage in foreign-owned firms:
    - In Estonia, 10 percent of total corporate debt is estimated to be owed to affiliated foreign firms.
    - The share could be similar for Lithuania given that 16-18 percent of all corporate debt is owed to the rest of the world.
    - Foreign ownership reaches 70–95 percent in the financial sector.
    - Leveraged hotels and restaurant sectors and construction seemingly did not attract much FDI and may therefore be mainly domestic-owned.
    - Higher relative FDI stocks in leveraged real estate and wholesale retail sectors could suggest a possible ‘parental’ safety net.

### Empirical assessment: consumption and investment projections vs. DOI predictions
- Preliminary assessment based on partial information up to mid-2010 indicates behavior of consumption and investment in the Baltic countries is overall consistent with DOI predictions.
- Consumption:
  - Consumption in all three Baltic countries contracted from their respective 2007 or 2008 peaks by more than GDP, in contrast with the eurozone where the contraction was less than GDP.
  - Regression analysis (data starting in the late 1990s to just before the crisis) regressed quarterly changes in household consumption on lagged changes in income, consumption, lagged short-term euribor interest rates, and lagged changes in unemployment.
  - Out-of-sample (static) projections for 2008–10 indicate persistent overprediction of consumption of about 4 to 8 percent in cumulative terms for the three countries, in contrast to absence of projection bias for the eurozone.
  - The on average 6 percent estimation gap could thus be interpreted as a rough proxy for lost household consumption resulting from debt overhang.
  - This compares to overall reductions in personal consumption by one quarter in both Lithuania and Estonia and close to one third in Latvia.
  - Given the share of private consumption in GDP, everything equal, the proxy for debt overhang indicates a reduction in GDP in the Baltic countries by around 3 percent.
- Investment:
  - Quarterly changes in real gross capital formation were regressed on lagged changes in GDP (or gross operating surplus), lagged changes in capital formation and lagged short-term euribor interest rates.
  - The cumulative 14 percent gap between actual and projected gross capital formation between end-2007 and mid 2010 for Latvia is large and consistent with the high DOI.
  - This 14 percent investment gap is equivalent to around 4 percent of GDP and could be interpreted as a proxy for lost fixed capital formation from firms’ above average leverage.
  - This compares with an overall 60 percent decline in investment over that period for Latvia.
  - For Estonia and Lithuania, there is no systematic overprediction, in line with DOI indicators.
- Regression model fit: goodness of fit (R2) between 0.2 and 0.5; caution advised in overinterpreting cross-country differences. Projection errors could reflect missing variables, but debt overhang is "a stickier variable and hence less correlated with variables included in the regression."

### Conclusion and policy challenges
- All Baltic countries appear to be at risk of debt overhang, with differences across sectors and countries.
  - Household risk could be more acute due to limited financial assets.
  - Distribution of household leverage increases risk in Estonia, lessens it in Lithuania.
  - Corporate overstretched balance sheets mainly in Latvia.
  - In Estonia, sectors contributing considerably to investment do not exhibit particularly high levels of leverage.
  - Foreign corporate ownership could play a small role in alleviating debt overhang problems.
- Quantitative illustrative estimates:
  - Preliminary proxies for debt overhang induced consumption losses have averaged around 3 percent of GDP in the Baltic countries since peaking pre-crisis.
  - Rough estimate suggests potential fixed investment losses in Latvia worth about 4 percent of GDP and absence of major effects in Estonia and Lithuania.
- Macro policy tension:
  - Policymakers face tension between deflating the economy to restore competitiveness (promoting debt deflation) and avoiding further weakening of demand and growth.
  - Extensive wage cuts are correcting imbalances but fueling deflationary pressures as the real value of debt rises.
  - Pay-off of restored competitiveness depends on exogenous factors such as demand developments in key export partners.
- Micro policy tension:
  - Trade-off between optimizing bankruptcy legislation for debt relief and orderly processes vs. protecting private contracts and limiting fiscal costs.
  - In Estonia and Lithuania personal bankruptcy lasts for at least five years before debts are discharged, longer than in many other countries.
  - Bankruptcy for both households and corporations typically results in liquidation rather than reorganization.
  - Proposals to optimize legal frameworks may be viewed as borrower-friendly bail-outs, undermining payment discipline and contract credibility.
  - Courts’ lack of familiarity with insolvency procedures may undermine speedy exit of nonviable firms and rehabilitation of viable ones.

*Source: IMF staff calculations and analysis as presented in the supplied content.*

### ANNEX II. CONSTRUCTION OF THE DEBT OVERHANG INDEX (DOI)

### ANNEX II. CONSTRUCTION OF THE DEBT OVERHANG INDEX (DOI)

### Methodology and definitions
- The DOI aggregates indicators of leverage in the household and nonfinancial corporate sector in relation to a benchmark (aggregate DOI), accounts for distributional differences in leverage, and is reweighted in relation to the share of different segments of the population in demand (weighted DOI).
- Notation:
  - D = debt
  - GDI = gross disposable income
  - R = debt service
  - NFA = net financial assets
  - NVA = net value added
  - GOS = gross operating surplus
- Timing convention:
  - Ratios involving stock variables (e.g. D/GDI or NFA/GDI) refer to 2008 unless indicated otherwise.
  - Ratios based on flow variables (e.g. R/GOS or R/GDI) refer to observations in 2009 unless indicated otherwise.
- Aggregation procedure (household sector example for Lithuania):
  - Step 1: Three aggregate balance-sheet indicators are scored relative to a benchmark (the euro area). The euro area receives a score of 10 on each indicator; total euro area score = 30.
  - Step 2: Country-level indicators are scored on the same scale and summed to produce the aggregate DOI (example: Lithuania aggregate DOI = 48).
  - Step 3: The aggregate DOI is allocated across income quintiles based on debt service ratios; quintiles with higher-than-average debt service receive proportionally higher DOI (example: a quintile with debt service ratio 30 percent above average sees its DOI rise to 60).
  - Step 4: Each quintile DOI is weighted by the share of that quintile in consumption and summed to produce the consumption/quintile weighted DOI (example: Lithuania consumption/quintile weighted DOI = 41).
- Corporate-sector reweighting mirrors the household approach but replaces quintiles and consumption with sectoral leverage and shares in corporate investment.

### Aggregate household DOI — cross-country leverage indicators (preserved values)
- Note: Unless flagged, ratios refer to stock variables in 2008 and flow variables in 2009.
- Shares / Indicators (Estonia, Latvia, Lithuania, Euro area):
  - D/GDI: 9.2, 8.0, 6.0, 10
  - NFA/GDI: 28.3, 73.9, 34.0, 10
  - R/GDI: 11.9, 10.3, 7.8, 10
  - Sum (aggregate DOI): 49.9, 247.8, 30.0, 30.0
- Example country-specific household computations (selected exact figures preserved):

  - Estonia
    - Debt service per quintile (2007) — Percentage debt service/income: Quintile 1–5 and Average = 10.0, 8.0, 5.0, 6.0, 6.0, 6.0
    - Relative to average: 1.7, 1.3, 0.8, 1.0, 1.0, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.4, 1.0
    - DOI per income quintile Quintile 1–5: 82.3, 65.9, 41.2, 49.4, 49.4
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 8.2, 8.6, 6.6, 11.4, 17.8, 52.5

  - Latvia
    - Debt service per quintile (2007) — Percentage debt service/income: Quintile 1–5 and Average = 10.0, 10.0, 8.0, 7.0, 5.0, 7.0
    - Relative to average: 1.4, 1.4, 1.1, 1.0, 0.7, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.3, 1.0
    - DOI per income quintile Quintile 1–5: 131.8, 131.8, 105.4, 92.2, 65.9
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 14.5, 17.1, 17.9, 21.2, 22.4, 93.1

  - Lithuania
    - Debt service per quintile (2007) — Percentage debt service/income: Quintile 1–5 and Average = 2.0, 2.0, 5.0, 3.0, 4.0, 4.0
    - Relative to average: 0.5, 0.5, 1.3, 0.8, 1.0, 1.0
    - Consumption per quintile (2005) — Share Quintile 1–5 Total: 0.1, 0.1, 0.2, 0.2, 0.3, 1.0
    - DOI per income quintile Quintile 1–5: 23.9, 23.9, 59.8, 35.9, 47.8
    - Weighted DOI by share in consumption Quintile 1–5 and Total: 2.6, 3.3, 11.4, 8.2, 15.3, 40.9

- Special treatment:
  - For Latvia, the ratio of net financial assets to gross disposable income (NFA/GDI) is capped at the ratio of the next lowest country in terms of NFA/GDI in the eurozone, Slovakia, with a ratio of 0.23. The capping is justified by the intuition that beyond a certain threshold the effect of a lower NFA/GDI on demand falls to zero; the capping has no effect on the thrust of the results.

### Leverage ratios in the nonfinancial corporate sector (preserved values)
- Cross-country shares / ratios (Estonia, Latvia, Lithuania, Eurozone):
  - D/NVA: 2.7, 2.4, 1.6, 3
  - D/E: 0.96, 2.5, 1.1, 1.3
  - R/GOS: 0.15, 0.23, 0.11, 0.17
- Corporate DOI scoring (scores assigned where eurozone = 10 per indicator):
  - Scores (Estonia, Latvia, Lithuania, Eurozone):
    - D/NVA: 9.7, 9.5, 3.1, 10
    - D/E: 7.4, 19.2, 8.5, 10
    - R/GOS: 8.8, 13.5, 6.5, 10
    - Sum (aggregate corporate DOI): 25.2, 40.6, 20.3, 30.0

### Country-level corporate sector computations (sectoral DOIs and investment-weighted DOIs)
- Estonia (selected exact figures)
  - D/E by sector (2007), Average (2008) D/E = 0.96 with sector breakdowns: Manufacturing 1.01, Mining 0.86, Electricity, Gas and Water 0.82, Hotels and Restaurants 1.21, Transport, Storage and Communication 1.06, Real Estate 0.95, Wholesale, Retail etc 1.27, Construction 1.06, Other 0.82
  - Relative to average: Manufacturing 1.05, Mining 0.89, Electricity, Gas and Water 0.85, Hotels and Restaurants 1.26, Transport 1.11, Real Estate 0.99, Wholesale 1.32, Construction 1.10, Other 0.85, Average 1.00
  - Sectoral DOIs: Manufacturing 26.4, Mining 22.5, Electricity, Gas and Water 21.6, Hotels and Restaurants 31.9, Transport 27.9, Real Estate 25.1, Wholesale, Retail etc 33.2, Construction 27.8, Other 21.5, Average 25.2
  - Shares in Total Corporate Investment (2006-2008): Mining and Quarrying 0.01, Manufacturing 0.11, Electricity, Gas and Water Supply 0.07, Construction 0.04, Wholesale and Retail Trade; Repair of Motor Vehicles etc. 0.08, Hotels and Restaurants 0.01, Transport, Storage and Communication 0.13, Real Estate, Renting and Business Activities 0.32, Other 0.23
  - DOIs weighted by share of investment (Estonia) — sector contributions and Total: Manufacturing 2.9, Mining 0.2, Electricity, Gas and Water 1.4, Hotels and Restaurants 0.3, Transport 0.3, Real Estate 8.1, Wholesale, Retail etc 2.8, Construction 1.0, Other 5.0, Total 22.1

- Latvia (selected exact figures)
  - D/E by sector (2007), Average (2008) D/E = 2.50 with sector breakdowns: Manufacturing 1.83, Mining 1.29, Electricity, Gas and Water 0.98, Hotels and Restaurants 2.64, Transport, Storage and Communication 1.67, Real Estate 2.64, Wholesale, Retail etc 3.73, Construction 3.25, Other 2.38
  - Relative to average: Manufacturing 0.73, Mining 0.52, Electricity, Gas and Water 0.39, Hotels and Restaurants 1.06, Transport 0.67, Real estate 1.06, Wholesale, Retail etc 1.49, Construction 1.30, Other 0.95, Average 1.00
  - Shares in Total Corporate Investment (2008): Mining and Quarrying 0.004, Manufacturing 0.154, Electricity, Gas and Water Supply 0.069, Construction 0.075, Wholesale and Retail Trade; Repair of Motor Vehicles etc. 0.096, Hotels and Restaurants 0.020, Transport, Storage and Communication 0.124, Real Estate, Renting and Business Activities 0.198, Other 0.260
  - Sectoral DOIs: Manufacturing 29.8, Mining 21.0, Electricity, Gas and Water 15.9, Hotels and Restaurants 42.9, Transport, storage and Communication 27.1, Real estate 42.9, Wholesale, Retail etc 60.6, Construction 52.8, Other 38.7
  - DOIs Weighted by Share of Investment (Latvia) — sector contributions and Total: Manufacturing 4.6, Mining 0.1, Electricity, Gas and Water 1.1, Hotels and Restaurants 0.9, Transport 3.4, Real Estate 8.5, Wholesale, Retail etc 5.8, Construction 4.0, Other 10.0, Total 38.3

- Lithuania (selected exact figures)
  - D/E by sector (2007), Average (2008) D/E = 1.10 with sector breakdowns: Manufacturing 1.22, Mining 0.32, Electricity, Gas and Water 0.26, Hotels and Restaurants 2.38, Transport, Storage and Communication 0.78, Real Estate 1.51, Wholesale, Retail etc 1.74, Construction 1.75, Other 0.66
  - Relative to average: Manufacturing 1.11, Mining 0.29, Electricity, Gas and Water 0.24, Hotels and Restaurants 2.16, Transport 0.71, Real Estate 1.37, Wholesale, Retail etc 1.58, Construction 1.59, Other 0.60, Average 1.00
  - Shares in Total Corporate Investment (2006-2008): Mining and Quarrying 0.00, Manufacturing 0.14, Electricity, Gas and Water Supply 0.07, Construction 0.04, Wholesale and Retail Trade; Repair of Motor Vehicles etc. 0.10, Hotels and Restaurants 0.01, Transport, Storage and Communication 0.12, Real Estate, Renting and Business Activities 0.25, Other 0.26
  - Sectoral DOIs: Manufacturing 22.4, Mining 5.9, Electricity, Gas and Water 4.9, Hotels and Restaurants 43.9, Transport 14.5, Real estate 27.8, Wholesale, Retail etc 32.0, Construction 32.3, Other 12.2
  - DOIs Weighted by Share of Investment (Lithuania) — sector contributions and Total: Manufacturing 3.2, Mining 0.0, Electricity, Gas and Water 0.4, Hotels and Restaurants 0.5, Transport 1.7, Real Estate 7.0, Wholesale, Retail etc 3.1, Construction 1.3, Other 3.2, Total 20.4

### Key analytic intuition
- The DOI construction combines balance-sheet leverage indicators with distributional information (household quintiles or corporate sectors) and demand weights (consumption shares or investment shares) to obtain a demand-adjusted measure of debt overhang.
- Capping rules (example: Latvia NFA/GDI capped at 0.23) are applied where extreme values imply negligible additional demand effects beyond a threshold.

### Annex III — OLS regressions of household consumption and corporate investment (selected regression results and diagnostics)
- General notation and variables:
  - C = real household consumption
  - GDP = real GDP
  - UN = EUROSTAT harmonized unemployment rate in percent (or D prefix = quarterly change)
  - R = three-month euribor nominal interest rate (or D prefix = quarterly change)
  - INV = real fixed capital formation
  - Suffixes: EURO, LT, LV, EST denote eurozone, Lithuania, Latvia, Estonia respectively.
  - DUMEST and DUMESTINV = dummies for the Russian crisis sudden sharp collapse in consumption and investment.
  - DUMLV = dummy for erratic decline in 2000:Q3.
  - All data apart from the interest rate are seasonally adjusted.
  - Estimation sample endpoints: up to end-2007 for eurozone, Estonia, Latvia; up to 2008:Q2 for Lithuania. Latvian investment estimations start in 2000 due to late-1990s volatility.

- Selected household consumption regressions (coefficients preserved exactly):

  - Dependent Variable: DLOG(CEURO)
    - Sample (adjusted): 1999Q2 2007Q4
    - Included observations: 35
    - Coefficients:
      - C: 0.004450, Std. Error 0.002108, t-Statistic 2.111226, Prob. 0.0432
      - DLOG(GDPEURO(-1)): 0.351239, Std. Err 0.256644, t-Statistic 1.368587, Prob. 0.1813
      - DLOG(CEURO(-1)): -0.200550, Std. Err 0.201098, t-Statistic -0.997278, Prob. 0.3266
      - DUNEURO: -0.005916, Std. Err 0.005011, t-Statistic -1.180428, Prob. 0.2471
      - R(-1): -0.000408, Std. Err 0.000550, t-Statistic -0.741544, Prob. 0.4641
    - Diagnostics:
      - R-squared 0.273459, Adjusted R-squared 0.176587, S.E. of regression 0.002907
      - F-statistic 2.822882, Prob(F-statistic) 0.042329, Durbin-Watson stat 2.067796

  - Dependent Variable: DLOG(CEST) (Estonia)
    - Sample (adjusted): 1999Q2 2007Q4
    - Included observations: 35
    - Coefficients:
      - C: 0.028079, Std. Err 0.014164, t-Statistic 1.982370, Prob. 0.0567
      - DLOG(GDPEST(-1)): 0.628865, Std. Err 0.306454, t-Statistic 2.052073, Prob. 0.0490
      - DLOG(CEST(-1)): -0.275571, Std. Err 0.139944, t-Statistic -1.969160, Prob. 0.0582
      - DUMEST: -0.051201, Std. Err 0.014948, t-Statistic -3.425304, Prob. 0.0018
      - R(-1): -0.004422, Std. Err 0.003833, t-Statistic -1.153570, Prob. 0.2578
    - Diagnostics:
      - R-squared 0.453083, Adjusted R-squared 0.380160, S.E. of regression 0.020520
      - F-statistic 6.213225, Prob(F-statistic) 0.000914, Durbin-Watson stat 2.190036

  - Dependent Variable: DLOG(CLT) (Lithuania)
    - Sample (adjusted): 1999Q2 2008Q2
    - Included observations: 37
    - Coefficients:
      - C: 0.057236, Std. Err 0.016014, t-Statistic 3.574057, Prob. 0.0011
      - DLOG(GDPLT(-1)): 0.208557, Std. Err 0.239220, t-Statistic 0.871821, Prob. 0.3898
      - DLOG(CLT(-1)): -0.455723, Std. Err 0.167193, t-Statistic -2.725733, Prob. 0.0103
      - UNLT(-1): -0.001390, Std. Err 0.001141, t-Statistic -1.217649, Prob. 0.2323
      - R(-1): -0.005561, Std. Err 0.004202, t-Statistic -1.323539, Prob. 0.1950
    - Diagnostics:
      - R-squared 0.230567, Adjusted R-squared 0.134387, S.E. of regression 0.022501
      - F-statistic 2.397260, Prob(F-statistic) 0.070766, Durbin-Watson stat 2.066192

  - Dependent Variable: DLOG(CLV) (Latvia)
    - Sample (adjusted): 1999Q2 2007Q4
    - Included observations: 35
    - Coefficients:
      - C: 0.077108, Std. Err 0.053788, t-Statistic 1.433567, Prob. 0.1624
      - DLOG(CLV(-1)): -0.235982, Std. Err 0.168996, t-Statistic -1.396378, Prob. 0.1732
      - UNLV(-1): -0.004259, Std. Err 0.004748, t-Statistic -0.896947, Prob. 0.3771
      - DLOG(GDPLV(-1)): 0.243557, Std. Err 0.395984, t-Statistic 0.615067, Prob. 0.5433
      - DUMLV: -0.076397, Std. Err 0.028066, t-Statistic -2.722086, Prob. 0.0109
      - R(-1): -0.004945, Std. Err 0.005752, t-Statistic -0.859739, Prob. 0.3970
    - Diagnostics:
      - R-squared 0.312063, Adjusted R-squared 0.193453, S.E. of regression 0.026054
      - F-statistic 2.631004, Prob(F-statistic) 0.044335, Durbin-Watson stat 1.585196

- Selected investment regressions (coefficients preserved exactly):

  - Dependent Variable: DLOG(INVEURO)
    - Sample (adjusted): 1999Q2 2007Q4
    - Included observations: 35
    - Coefficients:
      - C: 0.010247, Std. Error 0.005058, t-Statistic 2.026048, Prob. 0.0514
      - DLOG(GDPEURO(-1)): 1.356363, Std. Err 0.547350, t-Statistic 2.478055, Prob. 0.0189
      - DLOG(INVEURO(-1)): -0.069286, Std. Err 0.217861, t-Statistic -0.318029, Prob. 0.7526
      - R(-1): -0.003308, Std. Err 0.001470, t-Statistic -2.249711, Prob. 0.0317
    - Diagnostics:
      - R-squared 0.341913, Adjusted R-squared 0.278227, S.E. of regression 0.007348
      - F-statistic 5.368741, Prob(F-statistic) 0.004283, Durbin-Watson stat 2.184276

  - Dependent Variable: DLOG(INVEST) (Estonia)
    - Sample (adjusted): 1995Q3 2007Q4
    - Included observations: 50
    - Coefficients:
      - C: 0.007338, Std. Err 0.016178, t-Statistic 0.453574, Prob. 0.6523
      - DLOG(GDPEST): 1.875057, Std. Err 0.764611, t-Statistic 2.452303, Prob. 0.0180
      - DLOG(INVEST(-1)): -0.253770, Std. Err 0.133300, t-Statistic -1.903758, Prob. 0.0632
      - DUMESTINV: -0.076061, Std. Err 0.045556, t-Statistic -1.669610, Prob. 0.1018
    - Diagnostics:
      - R-squared 0.257775, Adjusted R-squared 0.209369, S.E. of regression 0.055280
      - F-statistic 5.325260, Prob(F-statistic) 0.003107, Durbin-Watson stat 2.257652

  - Dependent Variable: DLOG(INVLT) (Lithuania)
    - Method: Least Squares
    - Sample (adjusted): 1999Q3 2008Q2
    - Included observations: 36
    - Coefficients:
      - C: -0.010563, Std. Error 0.015315, t-Statistic -0.689761, Prob. 0.4953
      - DLOG(GDPLT): 2.131092, Std. Err 0.684112, t-Statistic 3.115123, Prob. 0.0039
      - DLOG(INVLT(-1)): -0.160126, Std. Err 0.156750, t-Statistic -1.021537, Prob. 0.3147
      - DR(-1): 0.002167, Std. Err 0.034933, t-Statistic 0.062039, Prob. 0.9509
    - Diagnostics:
      - R-squared 0.299526, Adjusted R-squared 0.233857, S.E. of regression 0.058959
      - F-statistic 4.561119, Prob(F-statistic) 0.009043, Durbin-Watson stat 2.123108

  - Dependent Variable: DLOG(INVLV) (Latvia)
    - Sample: 2000Q1 2007Q4
    - Included observations: 32
    - Coefficients:
      - C: 0.087546, Std. Err 0.029872, t-Statistic 2.930689, Prob. 0.0067
      - DLOG(INVLV(-1)): -0.255569, Std. Err 0.202867, t-Statistic -1.259786, Prob. 0.2181
      - R(-1): -0.016172, Std. Err 0.006327, t-Statistic -2.555951, Prob. 0.0163
      - DLOG(GDPLV(-1)): 0.195820, Std. Err 0.504723, t-Statistic 0.387976, Prob. 0.7010
    - Diagnostics:
      - R-squared 0.220917, Adjusted R-squared 0.137444, S.E. of regression 0.029346
      - F-statistic 2.646560, Prob(F-statistic) 0.068466, Durbin-Watson stat 2.047282

- Summary methodological note from the regressions:
  - Estimations were tested for robustness to different specifications (changes in sample period and choice of explanatory variables including interest rates, gross disposable income or operating income, GDP, unemployment rate, level or changes). While goodness of fit varies across specifications and countries, the results on the gap between actual and projected variables were robust to these alternative specifications.

*Source: _wp10250 - ANNEX II. CONSTRUCTION OF THE DEBT OVERHANG INDEX (DOI), and Annex III OLS regressions as presented in the supplied PDF content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10250.pdf_