## _wp08176

## Source details

**Canonical URL:** [_wp08176](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08176.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08176.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08176.pdf.json)

---

### I. Introduction — motivations and recent developments
- Six developments increasing attention to central bank financial strength:
  - (i) Increased international financial integration and “Redemption from original sin” raising focus on emerging markets.
  - (ii) Worldwide trend of lower inflation partly eroding traditional central bank financing models; central banks sometimes absorbed large amounts of problem assets.
  - (iii) Large-scale foreign reserve accumulation exposing central banks to revaluation losses and carrying costs.
  - (iv) Recent financial market turbulence showing mature market central bank balance sheets can face severe stress tests.
  - (v) General equilibrium models (Sims, 2004) indicating price level determinacy can depend on public beliefs and fiscal backing of monetary policy.
  - (vi) Adoption of more transparent accounting standards, including fair value accounting, revealing greater volatility in equity.
- Examples and contemporary references cited include Stella (2005), Cargill (2005), Ueda (2004), Bernanke (2003), Lonnberg and Stella (2008), Johnson and Zelmer (2007), and media (Fitch Ratings (2006), Reuters online (2007), The Economist (2005)).

### II. Research gap and paper objectives
- Aggregate empirical evidence linking central bank financial strength (CBFS) and macroeconomic performance remains elusive and controversial.
- Reasons for limited causal inference:
  - Lack of econometric evidence; reliance on case studies and descriptive statistics; insufficient control for reverse causality and heterogeneity.
- Stated aim: provide an econometric evaluation of the relationship between measures of central bank financial strength and policy outcomes using multiple data sources, cross section and panel approaches, and systematic study of long-run profitability trends.

### III. Theoretical considerations and related literature
- Three organizing questions:
  - Causes of weak central bank balance sheets.
  - Theoretical relationship between central bank financial strength and policy outcomes.
  - Empirical implications of leaving a central bank financially precarious.
- Reasons for lack of central bank financial strength (case-study evidence):
  - Aftermath of systemic banking-sector instability (absorption of liquidity, issuance of debt certificates, rollover effects).
  - Assumption of problem assets, expanded mandates, valuation complications.
  - Prolonged deflation and asset slumps (example: Japan).
  - Fiscal abuse and quasi-fiscal losses externalized to the central bank.
- Theories linking CBFS and policy performance:
  - Pragmatic approach: financial health affects policy choices; weak balance sheets can reduce anti-inflation aggressiveness or induce financial repression.
  - Theoretical approach: examines whether undercapitalization matters absent fiscal dominance; distinguishes circular arguments and spurious correlations.
- Alternatives available to a loss-making central bank without treasury recapitalization:
  - Lower cost of monetary operations (economic costs, potential financial repression).
  - Create money (interest-free liabilities) — if unsterilized, inflationary pressure; if sterilized by interest-bearing debt, generates additional interest expenditures and may perpetuate losses.
  - Possibility of return to profitability if demand for currency grows relative to operating costs (Bindseil et al., 2004).
  - Example: Chile — prolonged central bank difficulties mitigated by treasury depositing surpluses (termed “benevolent” fiscal dominance).
- Costs and incentive problems of perpetual losses:
  - Market segmentation, impeded government securities market development, reputational damage, soft budget constraint concerns.
  - Ize (2006): central banks adjust operating expenditures upward as profitability increases, illustrating governance concerns.
- Literature gap: role of financial structure in incentives and corporate governance of central banks underexplored.

### IV. Conceptualizations of CBFS and data sources
- Four conceptualizations employed:
  - CBFS1t = (CBCt + OINt) / TAt
    - CBC = Central Bank Capital; OIN = Other Items Net; TA = Total Assets.
    - Advantage: wide coverage via IMF IFS; disadvantage: imperfect proxy, large OIN/TA may signal low transparency.
  - CBFS2t = ROAAt
    - ROAA = Return on Average Assets from Bankscope; coverage up to 131 countries in recent years, sparse before 1995.
  - CBFS3t = CBINCOMEt / GDPt
    - Central bank profits/losses as percent of GDP for Latin America and the Caribbean, data period 1987–2005 (built from Leone (1994) and IMF staff/country sources).
  - CBFS4 = accumulated losses (moving windows of 2, 3 and 4 years) for Latin America and the Caribbean.
- Data challenges:
  - Variation in accounting practices, non-standard valuation of gains/losses, implicit liabilities not captured in accounting; accounting capital imperfect proxy for net worth.
- Recommendation: employ multiple data sources and conceptualizations.

### V. Recent trends in central bank profitability and key statistics
- General descriptive finding: all three datasets show a trend decline in central bank profitability starting at least in the late 1990s.
- Specific observed trends:
  - Median return on assets five-year average declined from 1.66 percent in the second half of the 1990s to 1.25 in the first half of the current decade.
  - Median return on assets (Bankscope) declined from 2.5 percent in 1995 to substantially less than 1 percent in 2005 (figure-referenced).
  - Share of central banks with very high net worth declined from more than 25 percent in 1992 to around 15 percent in 2005 (Figure 3).
  - Share of central banks reporting a negative capital account in IFS increased from less than two percent to more than eight percent over the same period (Figure 4).
  - In Latin America and the Caribbean, recuperation during the 1990s was reversed thereafter (Figure 5).
  - Number of countries with extreme negative CBFS declined over time (Figure 6).
  - Decreasing share of countries with very high OIN interpreted as improved transparency/statistical reporting (Figures 6 and 7).
- Changes in income and cost components:
  - Net interest margins have declined markedly since the mid-1990s (Figure 8, upper left panel).
  - Other operating income declined substantially (Figure 8, lower left panel).
  - Cost-to-income ratio increased substantially despite attempts to limit non-interest expenditure growth (Figure 8, upper right panel).
- Two major factors for decline in interest revenues:
  - Worldwide decline in inflation: in Ize (2006) sample of 100 countries, average inflation fell from around 14 percent in 1985–1995 to around 7 percent in the following decade; currency holdings as percent of GDP declined by around 1.5 percentage points.
  - Reserve accumulation in countries with large capital inflows: reserve build-up can lead to revaluation losses and higher carrying costs because sterilization yields often exceeded yields on reserve holdings.
- Implications: continuation of trends could threaten improved inflation performance; re-thinking central bank financing models may be necessary though empirically assessing such proposals is limited.

### VI. Econometric analysis — Data, approach, and key results
- Data and samples:
  - Panel: Latin America and the Caribbean, yearly data 1987–2005, including 15 countries (originally 25; 10 excluded for missing information).
  - Cross-section: up to 176 countries (CPI inflation in IFS), effective sample less than 100 depending on specification.
- Dependent variables:
  - CPI inflation rates (IFS) 1987–2005.
  - Rescaled measure d = πt / (1 + πt) ranging from 0 to 1; robustness: log(1+πt) also used.
- Main explanatory variables: CBFS1–CBFS4 measures as described; lagging CBFS measures by one period to address endogeneity; CBFS4 used to capture sustained weakening.
- Controls (panel): central bank independence (CBI), trade openness, GDP per capita, institutional quality, exchange rate regime, banking crisis dummy, average G7 inflation.
- Panel results — Latin America and the Caribbean:
  - Stable negative relationship between lagged CBFS3 and inflation measure d.
  - Examples of pooled OLS (Table 1) CBFS3,t-1 coefficients: -.019**, -.032***, -.029***, -.034***, -.020**, -.036***, -.027*** across specifications; world inflation coefficients positive and significant (e.g., .054***, .040***, .046***, .037***, .058***, .062***, .030***).
  - Fixed effects (Table 2) CBFS3,t-1 coefficients: -.038***, -.033***, -.031***, -.029***, -.018**, -.036***, -.021**; within R-squared values reported as .61, .63, .64, .67, .70, .59, .50 in examples.
  - Accumulated losses (CBFS4) coefficients remain negative and significant at the one percent level in many specifications (Table 4); estimates more stable than CBFS3.
  - Robustness: including central government deficit (Table 3) does not markedly change results; using log(1+πt) yields similar results.
  - Heteroscedasticity and serial correlation detected; FGLS (Table 5) accounting for heteroscedasticity and AR(1) across panels yields CBFS coefficients that remain significant in many specifications though sensitivity increases with exclusions and trending variables.
  - Arellano-Bond GMM estimates are consistent with main findings.
- Cross-section results — large sample:
  - CBFS1 (IFS stock measure) and CBFS2 (Bankscope ROAA) used.
  - Baseline OLS cross-section (Table 6) examples:
    - CBFS1 coefficient examples: -.0010** (.00038), -.0008** (.00034), -.0008* (.00041); constants .1737*** (.02214), .1562*** (.02654), .1889*** (.02829); observations 97, 97, 56; R-squared .14, .16, .11.
  - Sensitivity to sample composition:
    - Recursive inclusion by maximum inflation (Figure 10) shows CBFS1 estimates sensitive when adding observations with inflation between 5 and 15 percent; parameter estimates converge and remain negative thereafter.
    - Restricting CBFS1 ratio to +/- 50 percent (Table 7) strengthens relationship: CBFS1 coefficients examples -.0018*** (.0005), -.0016*** (.0003), -.0020** (.0008); observations 93, 93, 52; R-squared .17, .18, .16.
  - Nonlinearity/thresholds:
    - Evidence consistent with Stella (2008): severe impairment of central bank balance sheet produces significant macro effects; negative CBFS1 coefficient becomes significant only after including observations with CBFS1 below threshold (example threshold around 5 percent in Figure A2).
  - CBFS2 (ROAA) results:
    - Coefficient signs as expected but highly sensitive to sample composition; CBFS2 becomes statistically significant (p-value < 10 percent) only when sample restricted to average inflation below 75 percent.
- Interpretation caveats:
  - OLS likely biased; fixed effects reduces bias but endogeneity remains a concern.
  - Small sample sizes and potential selection bias caution interpretation; need for better instruments and expanded coverage emphasized.

### VII. Policy conclusions and research agenda
- Policy implications:
  - Econometric results, consistent with case studies, suggest weak central bank finances may hinder ability to respond effectively to inflationary pressures.
  - For countries with sustained central bank losses, recapitalization and efforts to restore central bank finances to a safe footing are likely worth fiscal costs, especially given potential additional benefits (e.g., domestic debt market development).
  - Re-thinking central bank financing models to ensure stable income streams could be necessary, though such proposals face opposition and limited empirical assessment.
- Open empirical research priorities:
  - Address endogeneity more directly (find better instruments for CBFS).
  - Expand coverage across time and geography.
  - Study effects of CBFS on measures of financial repression and domestic debt market development.
  - Complement cross-sectional/panel approaches with time-series techniques (e.g., examine effects of central bank recapitalization; use dependent variables capturing central bank activism more directly).
- Theoretical and modeling priorities:
  - Develop theory of central bank corporate governance modeling bureaucratic incentives and implications of the central bank’s ability to create means to pay its own liabilities.
  - Clarify channels linking CBFS and inflation: trust and expectations, price-level indeterminacy without fiscal backing, and soft budget constraints shaping incentives.

### VIII. Supporting empirical material and data notes
- Conceptualizations summarized:
  - CBFS1: Sum of central bank capital (CBC) and other items net (OIN), divided by average total central bank assets (TA).
  - CBFS2: Central bank Return on Average Assets (ROAA).
  - CBFS3: Central bank profits/losses as percent of GDP.
  - CBFS4: Sum of CBFS3 over preceding two, three, or four years.
- Variables and sources listed (examples):
  - Inflation: Annual Change in Consumer Price Index, IMF IFS.
  - GDP per capita (PPP): IMF WEO.
  - Institutional Quality, Bureaucratic Quality, Democratic Accountability, Government Stability, Corruption, Law and Order: ICRG scales.
  - Openness: (Exports + Imports)/GDP, IMF WEO.
  - Banking Crisis dummy: Caprio and Klingebiel (2003).
  - Fixed Regime dummy: Ghosh, Gulde, and Wolf (2002) and IMF Annual Report REAER.
  - CB TOR: Turnover rate for central bank governors, De Haan and Sturm (2001).
  - Deficit: General Government Balance in Percent of GDP, IMF IFS and WEO.
- Appendix provides Latin America panel raw data and pair-wise correlations; selected panel means/medians reported (examples include sequences of values such as -2.7, -2.7, -3.0, -2.0, -1.4, -1.2, -0.7, -0.6, -0.9, -0.5, -0.5, -0.5, -0.5, -0.3, -0.5, -0.7, -0.8, -0.8, -1.0).
- Graphical diagnostics include recursive regressions for relationship between d and CBFS1 (covariates: bureaucratic quality and openness) and country/year plots of d vs. cbfs4.

*Source: _wp08176 - Section V concludes.*

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

### _wp08176 - References..............................................................................................................

### I. Introduction — motivations and recent developments
- The capitalization and profitability of central banks has long been a relatively neglected topic; recent research into the nature, determinants and consequences of “weak” central bank balance sheets has grown considerably.
- Six distinct developments have increased attention to central bank financial strength:
  - (i) Increased international financial integration has led to greater attention being paid to emerging markets, where problems of weak central bank balance sheets have been much more prevalent. “Redemption from original sin”, i.e., the increasing ability of emerging market governments to issue debt in local currency, has reinforced these tendencies.
  - (ii) The worldwide trend of lower inflation rates has partly eroded traditional central bank financing models. Increased price stability has sometimes been accompanied by major disruptions in financial sectors, with central banks assuming large amounts of problem assets, further weakening their balance sheet structure.
  - (iii) Large-scale foreign reserve accumulation by emerging markets has brought to the fore central banks’ exposure to a major re-balancing of exchange rates. In addition to revaluation losses, the carrying cost implied by large-scale reserve accumulation currently represents a major financial burden to many central banks, further depressing income margins.
  - (iv) Recent financial market turbulence has highlighted that even mature market central bank balance sheets can be subjected to potentially serious stress tests: tail events related to financial crises can materially change the risk profile of a central bank’s asset-liability mix. This has been evident in debates surrounding the Bank of Japan’s perceived cautiousness to address deflation dynamics through the large-scale acquisition of long-term debt or foreign exchange.
  - (v) Research into general equilibrium models of monetary policy has shown that price level determinacy can crucially depend on the public’s beliefs about the dynamic properties of an economy following extreme disturbances (Sims, 2004). Avoiding explosive equilibria often requires assuming that fiscal behavior backs monetary policy actions in specific ways; the experience of countries that have faced decisions about whether and how to “back” the central bank financially may provide valuable insights.
  - (vi) The adoption of more transparent accounting standards by central banks worldwide, including fair value accounting for assets, has revealed far more volatility in equity than had hitherto been the case.
- Examples and citations referenced in the text include Stella (2005), Cargill (2005), Ueda (2004), Bernanke (2003), Lonnberg and Stella (2008), Johnson and Zelmer (2007), and contemporary media (Fitch Ratings (2006), Reuters online (2007), The Economist (2005)) as noted in the source material.

### II. Research gap and paper objectives
- Aggregate empirical evidence linking central bank financial strength and macroeconomic performance remains elusive and controversial.
- Reasons for limited causal inference include a lack of econometric evidence, reliance on case studies and descriptive statistics, and insufficient control for reverse causality and observed and unobserved heterogeneity.
- The paper’s stated aim: provide an econometric evaluation of the relationship between measures of central bank financial strength and policy outcomes.
- Methodological approaches described:
  - Use of a variety of data sources and techniques.
  - Employment of both cross section and panel data approaches.
  - Systematic study of long-run trends in central bank profitability.

### III. Organization of the paper (as described)
- Section II: Reviews related contributions on the causes and consequences of weak central bank balance sheets, and relates the approach to literature on institutional and organizational determinants of macroeconomic policy performance, with emphasis on central bank independence.
- Section III: Provides a detailed account of available indicators of central bank financial conditions and describes major trends in central bank profitability.
- Section IV: Presents econometric results for both a panel-based approach and a larger cross-section of countries.

### IV. Tables and Figures (listed in source)
- Tables:
  - 1. Baseline Results with Alternative Control Variables and Excluding Outliers
  - 2. Fixed Effects Estimates with Alternative Control Variables
  - 3. Fiscal Variables and Alternative Dependent Variable
  - 4. Results for Central Bank Financial Strength
  - 5. Feasible Generalized Least Squares Estimation
  - 6. Results for the Cross-Section
  - 7. Accounting for Extreme Cases of Balance Sheet Distortions
- Figures:
  - 1. Return on Average Assets in a Large Sample of Central Banks, 1995–2005
  - 2. Median Return on Average Assets in Bankscope Sample
  - 3. Share of Central Banks with High Central Bank Financial Strength
  - 4. Share of Central Banks with Negative IFS Capital
  - 5. Central Bank Financial Strength in a Group of Western Hemisphere Countries
  - 6. Distribution of Reported Sum of Capital and OIN in Percent of Total Assets
  - 7. Distribution of Reported OIN in Percent of Total Assets
  - 8. Measures of Central Bank Revenues and Costs
  - 9. Price Stability and Central Bank Financial Strength
  - 10. Relationship Between Central Bank Financial Strength and d
  - 11. Relationship Between Central Bank Financial Strength and d

*Source: _wp08176 - References..........................................................................................................*

### Section V concludes.

### _wp08176 - Section V concludes.

### II. THEORETICAL CONSIDERATIONS AND RELATED LITERATURE
- Three organizing questions in the literature:
  - What are the causes that usually lead to weak central bank balance sheets?
  - What is, in theory, the relationship between central bank financial strength and policy outcomes?
  - What are, empirically, the implications of leaving the central bank in a financially precarious situation?
- Recent surge in academic and applied contributions on central bank quasi-fiscal losses and monetary authorities’ net worth (references in text include Stella (2005, 2008 forthcoming), Dalton and Dziobek (2005), Leone (1993), Markiewicz (2001)).

### A. Reasons for a Lack of Central Bank Financial Strength
- Typical causes identified from case study evidence:
  - Aftermath of systemic banking-sector instability: need to absorb liquidity extended to ailing banks; reliance on costly issuance of debt certificates; rollover may continue for years; credibility and money demand effects.
  - Assumption of problem assets mirroring increased liabilities: expanded mandates, higher operational cost, valuation complications for central bank equity.
  - Prolonged deflation and asset slumps: purchase of large amounts of low-yielding instruments whose value may fall when economy recovers and rates rise; revaluation might reduce or wipe out net worth; example cited: Japan (see Cargill (2005); perception effects discussed).
  - Fiscal abuse: attempts by other public institutions to externalize losses or costs using the central bank’s balance sheet; quasi-fiscal losses that could in principle be duplicated by budgetary measures (see Fry (1993); Mackenzie and Stella (1996), pp. 3 and 17).
- Three recent drivers highlighted: build-up of large stocks of foreign reserves; worldwide decline in inflation; financial innovation and development.

### B. Theories Linking Central Bank Financial Strength and Policy Performance
- Two approaches to “irrelevance” assertions:
  - Pragmatic approach: central bankers factor financial health into policy choices; financial constraints can reduce aggressiveness of anti-inflationary policies or push toward financial repression (e.g., excessive reserve requirements).
    - Motives include self-interested behavior (reputation, future employment, bureaucratic rents) and public-welfare concerns (confidence, relationship with government).
    - Weak balance sheet can impair political/economic independence; approaching government for funding risks interference; can be a by-product of fiscal dominance (see Hawkins (2003), Fry (1998)).
  - Theoretical approach: examines economic function of capital and whether undercapitalization matters absent fiscal dominance; distinguishes between circular arguments and spurious correlations where weak CBFS merely signals broader fiscal unsustainability.
- Alternatives available to a loss-making central bank when treasury does not inject capital:
  - Lower cost of monetary operations (e.g., adjust minimum reserve requirements) — involves economic costs and financial repression.
  - Create money (use interest-free liabilities); if not sterilized, lowers money market rates and induces inflationary pressure; if sterilized by issuing debt certificates at interest rates consistent with price stability, incurs additional interest expenditures and may perpetuate losses (Ernhagen et al., 2002).
    - Under plausible assumptions, losses need not go on indefinitely; return to profitability possible if demand for currency grows sufficiently relative to operating costs (Bindseil et al., 2004).
  - Examples: Chile faced prolonged central bank financial difficulties without direct treasury transfers; treasury deposited large surpluses at the central bank — termed “benevolent” fiscal dominance (Stella (2005)).
- Feasibility and cost considerations for perpetuation:
  - Perpetual loss sustainability depends on credibility of separation/consolidation with government and common knowledge of debt accumulation mechanics (Bindseil et al., 2004).
  - Costs to society from perpetuation include market segmentation, impeded development of a unified and liquid government securities market, conflicts when monetary and debt management are handled within same organization, reputational damage (Vaez-Zadeh, 1991).
  - Incentive problems: self-interested bureaucrats may respond to weak finances by adjusting behavior (Rogoff (1985) quote included); Ize (2006) finds central banks adjust operating expenditures upward as profitability increases — implying “soft budget constraint” concerns and governance problems in financially weak central banks.
- Literature gaps: role of financial structure in incentives is underexplored; many models assume central bank loss function identical to social one (Walsh (1995)), limiting analysis of CBFS-related incentive issues.

### C. Related Empirical Evidence
- Existing empirical work is limited:
  - Ize (2006): regresses central bank expenditure residuals on macro/financial stability measures; finds positive relationship between excess expenditures and inflation at very low significance.
  - Stella (2008): t-statistics comparing inflation performance of financially weak vs strong central banks using a stock measure of CBFS and IFS data; reports pronounced and statistically highly significant differences.
- Limitations of prior empirical work: largely descriptive, often do not control for confounders or endogeneity; potentially spurious correlations.
- This paper’s aim: provide first econometric evidence quantifying macroeconomic consequences of weak central bank finances, accounting for joint determination with inflation and other factors.
- Related cross-country and regional literature:
  - Difficulty establishing causal link between legal/de facto central bank independence and inflation; sensitivity to outliers, measures used, control variables, sample (Arnone, Laurens and Segalotto, 2006a,b).
  - General consensus: independence, combined with reforms, contributes to credibility and low inflation, but interactions with openness, exchange rate regime, institutional quality, and economic development remain debated.
  - Latin America findings:
    - Jácome (2001): decomposed independence into economic, political, financial sub-indices; higher “economic” and “financial” independence associated with lower inflation; general legal independence index yields counter-intuitive results.
    - Gutiérrez (2003), Jácome and Vázquez (2005): mixed findings; after controlling for endogeneity and structural reforms, causal link between legal independence and inflation may disappear, suggesting independence must be viewed in broader policy context.

### III. EVALUATING CROSS-COUNTRY INDICATORS OF CENTRAL BANK FINANCIAL STRENGTH

#### A. Conceptualization of Central Bank Financial Strength and Data Sources
- Data challenges:
  - Central bank accounting practices vary across countries and time; non-standard valuation approaches for gains/losses; implicit liabilities and assets not captured in typical accounting.
  - Accounting capital may be an imperfect proxy for net worth; mark-to-market trends and worldwide decline in inflation may bias parameter estimates.
- Recommendation: employ multiple data sources and multiple conceptualizations of CBFS.
- Four conceptualizations used (drawn from text):
  - Stock measure of CBFS using IMF IFS and Monetary and Banking Statistics:
    - CBFS1t = (CBCt + OINt) / TAt
    - CBC = Central Bank Capital; OIN = Other Items Net; TA = Total Assets.
    - Advantage: wide coverage and some standardization via inclusion in IFS.
    - Disadvantage: imperfect proxy; large OIN/TA may signal low transparency or data quality.
  - Flow/profitability measure from Bankscope:
    - CBFS2t = ROAAt
    - ROAA = Return on Average Assets from Bankscope.
    - Coverage variation: up to 131 countries in recent years, sparse before 1995; authors present whole sample and an adjusted sample excluding countries without a minimum number of observations (three per five-year period).
  - Central bank profits/losses for Latin America and the Caribbean expressed as percent of GDP:
    - CBFS3t = CBINCOMEt / GDPt
    - Data period: 1987-2005 (starting from Leone (1994) for 1987-1992, extended using IMF staff reports and country-provided information).
    - Advantage: scrutinized data (IMF country teams or country authorities); amenable to panel methods and lagging to address endogeneity.
    - Disadvantage: limited coverage; potential bias toward countries where CBFS was policy-relevant.
  - Accumulated losses measure (moving windows of 2, 3 and 4 years) for Latin America and the Caribbean:
    - t-n 4ttt t-1 CBFS  =CBINCOME /GDP ∑
    - Used to cross-check econometric results (not discussed further in the next section).
- Note: Table A1 in Appendix summarizes the four conceptualizations.

#### B. Recent Trends in Central Bank Profitability
- Secular trends flagged as exerting pressure on central bank profitability: build-up of foreign reserves, worldwide decline in inflation, financial innovation.
- Figures and sample descriptions (as presented in text):
  - Figures 1–5 summarize global trends in central bank financial conditions using the three data sources.
  - Figures 1 and 2: central bank ROAA in a sample of more than 130 countries between 1995 and 2005.
    - Median and average plotted for whole sample and median for countries with data for at least 9 of 11 years (91 countries).
  - Figures 3 and 4: IFS data on capital-to-asset ratio and the authors’ measure of financial strength.
  - Figure 5: central bank results as a percentage of GDP for Latin America and the Caribbean since 1990.
- Key descriptive finding:
  - All three datasets show a similar pattern: a trend decline in central bank profitability, with the decline starting at least in the late 1990s.
  - Using Bankscope data, the median return on assets declined from around (figure referenced but exact starting value not provided in supplied excerpt).
- Definition note used in figures:
  - Countries in which the ratio of capital and other items net to total assets exceeds 20 are defined as those with high CBFS.

*Italicized source attribution: _wp08176 - Section V concludes.*

### 2.5 percent in 1995 to substantially less than 1 percent in 2005. For the adjusted sample, the

### _wp08176 - 2.5 percent in 1995 to substantially less than 1 percent in 2005. For the adjusted sample, the

### Observed trends in central bank financial strength (CBFS)
- Median return on assets five-year average declined from 1.66 percent in the second half of the 1990s to 1.25 in the first half of the current decade.
- Share of central banks with very high net worth declined from more than 25 percent in 1992 to around 15 percent in 2005 (Figure 3).
- Share of central banks reporting a negative capital account in IFS increased from less than two percent to more than eight percent over the same period (Figure 4).
- In Latin America and the Caribbean, recuperation during the 1990s was reversed since then (Figure 5).
- Distributional evidence:
  - Number of countries displaying extreme negative values for CBFS declined over time (Figure 6).
  - Decreasing share of countries with very high OIN (Figures 6 and 7) — interpreted as accounting becoming more reflective of economic reality and improved transparency/statistical reporting (including IMF safeguards assessments).

### Changes in income and cost components
- Net interest margins have been experiencing a marked decline since the mid-1990s (Figure 8, upper left panel).
- Other operating income declined in importance substantially (Figure 8, lower left panel).
- Non-interest expenditures attempts to limit growth were only partly successful; cost-to-income ratio increased substantially (Figure 8, upper right panel).
- Two major factors attributed to overall decline in interest revenues:
  - Worldwide decline in inflation: in Ize (2006) sample of 100 countries, average inflation fell from around 14 percent in 1985-1995 to around 7 percent in the following decade; currency holdings as a percentage of GDP declined by around 1.5 percentage points.
  - Reserve accumulation in countries with large-scale capital inflows:
    - Reserve build-up can lead to revaluation losses if global exchange rates re-balance.
    - Carrying cost increase because yields on instruments used to sterilize reserve growth often exceeded yield on foreign exchange reserve holdings.

### Implications of declining CBFS
- Continuation of current trends could threaten the improved inflation performance in many countries because central banks operating at low or negative profit margins may face greater obstacles to ensure price stability.
- Potential policy response: re-think financing models for central banks to ensure a stable income stream to finance monetary policy operations — though such proposals would face substantial opposition and empirical assessment is limited.

### Econometric analysis — data and approach
- Panel: Latin America and the Caribbean, yearly data 1987 to 2005, including 15 countries (originally 25; 10 excluded for missing information).
- Cross-section: up to 176 countries (all with CPI inflation in IFS), but availability of CBFS and covariates reduces effective sample to less than 100 countries depending on specification.
- Dependent variables:
  - CPI inflation rates from IFS for 1987–2005.
  - Rescaled measure d = πt / (1 + πt), which ranges from 0 to 1 and can be interpreted as rate of depreciation of purchasing power.
  - Robustness: log(1+πt) also used.
- Main explanatory variables: measures of CBFS (CBFS1–CBFS4 as described in section III).
- Endogeneity concerns addressed by lagging CBFS measures by one period and using accumulated measures (CBFS4) for robustness.
- Controls in panel estimations: central bank independence (CBI), trade openness, GDP per capita, institutional quality, exchange rate regime, banking crisis dummy, and average G7 inflation to capture world trend.
- Cross-section controls limited to openness, institutional quality, GDP per capita, and CBI (where available); interaction terms included to capture reinforcement effects between CBFS and CBI.
- Expectation: estimates may be downward biased because financially weak central banks might neutralize effects on inflation via other distortions (e.g., financial repression).

### Panel results — Latin America and the Caribbean
- Pooled OLS and fixed effects estimations (Tables 1–4) show a stable negative relationship between lagged CBFS (CBFS3) and inflation measure d:
  - Table 1 pooled OLS: CBFS3,t-1 coefficients range (examples) -.019**, -.032***, -.029***, -.034***, -.020**, -.036***, -.027*** across specifications; world inflation coefficients positive and significant (examples .054***, .040***, .046***, .037***, .058***, .062***, .030***).
  - Excluding hyperinflationary observations (inflation > 1000 percent) does not materially change significance (column 7 of Table 1).
- Fixed effects (Table 2) confirm negative relationship:
  - CBFS3,t-1 coefficients (examples) -.038***, -.033***, -.031***, -.029***, -.018**, -.036***, -.021**.
  - Within R-squared values high (examples .61, .63, .64, .67, .70, .59, .50).
- Accumulated losses (CBFS4) used to capture sustained weakening:
  - CBFS4,t-1 coefficients remain negative and significant at the one percent level in many specifications (Table 4), with estimates more stable than CBFS3.
- Robustness and specification checks:
  - Including central government deficit (Specification B, Table 3) does not markedly change results.
  - Using log(1+πt) as dependent variable yields similar results (Specification C, Table 3).
  - Heteroscedasticity across panels and serial correlation detected; FGLS (Table 5) accounting for heteroscedasticity and AR(1) across panels yields CBFS coefficients that remain significant in many specifications, though sensitivity increases when excluding hyperinflationary episodes and including trending variables.
  - Generalized Method of Moments (Arellano-Bond) estimates are consistent with main findings.
- Interpretation caveats:
  - OLS likely biased upwards; fixed effects reduces bias but does not eliminate endogeneity risk.
  - Finding improved instruments for CBFS remains an important next step.
  - Small sample size and potential selection bias caution interpretation; expanding coverage desirable.

### Cross-section results — large sample of countries
- Two alternative cross-sectional CBFS measures:
  - CBFS1: stock variables from IFS (sum of capital and OIN as percent of assets).
  - CBFS2: return on average assets from Bankscope.
- Baseline OLS cross-section (Table 6):
  - CBFS1 coefficient examples: -.0010** (.00038), -.0008** (.00034), -.0008* (.00041) across three specifications; constant terms .1737*** (.02214), .1562*** (.02654), .1889*** (.02829); observations 97, 97, 56.
  - R-squared modest (.14, .16, .11).
- Sensitivity to sample composition:
  - Recursive inclusion by maximum inflation in sample (Figure 10) shows CBFS1 estimates sensitive when adding observations with inflation between 5 and 15 percent; parameter estimates converge and remain in negative territory thereafter. Inclusion of turnover rate widens confidence bands but does not fundamentally change sign.
  - Restricting CBFS1 ratio (sum of capital and OIN to total assets) to +/- 50 percent (Table 7) strengthens the relationship: CBFS1 coefficients examples -.0018*** (.0005), -.0016*** (.0003), -.0020** (.0008); observations 93, 93, 52; R-squared .17, .18, .16.
- Nonlinearity / threshold effects:
  - Evidence consistent with Stella (2008) hypothesis that only severe impairment of CB balance sheet produces significant macroeconomic effects: negative CBFS1 coefficient becomes significant only after including observations with CBFS1 below a threshold (example threshold around 5 percent in Figure A2).
- CBFS2 (return on assets) results (Figure 11):
  - Coefficient signs as expected but substantially more sensitive to sample composition.
  - CBFS2 coefficient becomes statistically significant (p-value < 10 percent) only when sample is restricted to average inflation below 75 percent.

### Policy conclusions and implications
- Econometric results, consistent with case study evidence, suggest weak central bank finances may hinder policymakers’ ability to respond effectively to inflationary pressures.
- For countries experiencing sustained central bank losses, recapitalization and efforts to return central bank finances to a safe footing are likely to be worth the fiscal costs, especially when potential additional benefits (e.g., domestic debt market development) are considered.
- Further research priorities identified:
  - Address endogeneity more directly (better instruments for CBFS).
  - Expand coverage over time and geography.
  - Study effects of CBFS on measures of financial repression and debt market development.

*Source: International Monetary Fund working paper content (excerpt provided).*

### section III, it becomes clear that there is a need for central banks and their stakeholders to

### _wp08176 - section III, it becomes clear that there is a need for central banks and their stakeholders to

### Key findings on central bank funding modes and profitability
- The study documents a decline in central bank profitability using time series evidence; many observers had suspected such a decline prior to this study.
- It is "probably too early to declare an erosion of traditional models of central bank financing," in part because the "(still incomplete) data for 2005 and 2006 could be viewed as an indication of a slowing trend or even a trend reversal."
- A review of central bank funding modes is identified as a "promising area of further investigation."

### Open empirical research issues
- Need to examine other measures of central bank performance, "in particular measures of financial repression."
- Important to disentangle causes and effects:
  - Determine whether "a lack of central bank independence, potentially combined with fiscal dominance" leads to low CBFS, or whether "CBFS itself is one of the major determinants of political and economic independence."
- Endogeneity concerns must be assessed more thoroughly, "taking into account the circumstances under which central banks end up in a financially vulnerable situation."
- Complement cross-sectional and panel approaches with time series techniques, for example:
  - "looking at the effects of central bank recapitalization efforts."
  - Using dependent variables that "capture central bank activism more directly" in a time series context.

### Data expansion and measurement priorities
- A pre-condition for additional empirical research is "to expand and improve available data on CBFS."
- For panel analysis, first step: "extend the dataset on Latin America and the Caribbean to a broader set of countries in the region, and eventually to other regions in the world."
- Improve conceptualizations of CBFS by "developing measures of CBFS that come closer to existing theoretical concepts, such as the central bank net worth."
- To achieve better measures, it may be necessary to "go beyond accounting conventions to incorporate elements such as franchise value or implicit liabilities from future financial stability events."

### Theory and modeling agenda
- Develop a theory of central bank "corporate" governance, modeling:
  - "monetary authorities’ bureaucratic incentives and their relation to CBFS."
  - Implications of the central bank's unique ability "to create the means for paying its own liabilities" for "control right allocation and optimal intervention schemes."
- Clarify how issues of CBFS complement macroeconomic theories emphasizing the fiscal–monetary nexus for price level determination, referencing Sims, 2004.
- Future studies should focus on the exact channel linking CBFS and inflation, differentiating:
  - the role of trust and inflation expectations,
  - "the potential indeterminacy of the price level without fiscal backing,"
  - and "the role of soft budget constraints in shaping central bankers’ incentives."

### Supporting empirical material (appendix highlights)
- Conceptualizations of Central Bank Financial Strength include:
  - CBFS1: "Sum of central bank capital (CBC) and other items net (OIN), divided by average total central bank assets (TA)."
  - CBFS2: "Central bank Return on Average Assets (ROAA), defined as the ratio of net income per annum divided by average total assets."
  - CBFS3: "Central bank profits/losses as a percentage of GDP."
  - CBFS4: "Sum of CBFS3 over the preceding two, three, or four years respectively."
- Data sources and variables listed include: Inflation (Annual Change in Consumer Price Index, IMF IFS), GDP per capita (PPP, IMF WEO), Institutional Quality (0 to 100, ICRG), Bureaucratic Quality (0 to 4, ICRG), Democratic Accountability (0 to 6, ICRG), Government Stability (0 to 12, ICRG), Corruption (0 to 6, ICRG), Law and Order (0 to 6, ICRG), Openness ((Exports + Imports)/GDP, IMF WEO), Banking Crisis (dummy, Caprio and Klingebiel (2003)), Fixed Regime (dummy, Ghosh, Gulde, and Wolf (2002) and IMF Annual Report REAER), CB TOR (Turnover rate for central bank governors, De Haan and Sturm (2001)), Deficit (General Government Balance in Percent of GDP, IMF IFS and WEO).
- Latin America panel raw data table and pair-wise correlation tables are provided for CBFS measures and covariates; selected reported statistics include panel means and medians by year (e.g., Mean series showing values such as -2.7, -2.7, -3.0, -2.0, -1.4, -1.2, -0.7, -0.6, -0.9, -0.5, -0.5, -0.5, -0.5, -0.3, -0.5, -0.7, -0.8, -0.8, -1.0).
- Robustness and graphical diagnostics include "Recursive Regression for Relationship between d and CBFS1 (Covariates: Bureaucratic Quality and Openness)" and country/year graphs of d vs. cbfs4.

*Italic: Source — _wp08176 - section III, it becomes clear that there is a need for central banks and their stakeholders to*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08176.pdf_