## _wp10210 - References (excerpt)

## Source details

**Canonical URL:** [_wp10210 - References (excerpt)](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10210.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10210.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10210.pdf.json)

---

### Main research question and approach
- Investigates whether financial development leads, follows, or is unrelated to economic growth and what policies or environments are conducive to financial deepening.
- Defines "financial accelerations" as episodes measured by the growth rate of the credit-to-GDP ratio.
- Compares short-lived financial deepening episodes with long-lasting ones to identify conditions prevailing around the take-off that increase the likelihood of sustained financial development.
- Analysis covers events "since the early 1960s" and "the past 50 years" across a broad country sample.

### Key empirical findings and stylized facts
- Episodes and prevalence
  - Approximately 210 such episodes in a sample of about 160 countries over the past 50 years.
  - Identified 209 periods of financial accelerations for 1960–2005 in the sample of about 160 countries.
  - Of these, 161 episodes lasted between 5 and 10 years; 48 episodes lasted longer than 10 years.
  - Only about one-quarter of all financial accelerations in the past 50 years have led to long-term financial deepening.
- Episode outcomes and patterns
  - Average length of short periods is 6.9 years (5–10 years); average length of sustained deepening episodes is 14.7 years (10 years+).
  - Countries with one or more short-term accelerations have, on average, doubled their credit-to-GDP ratio by 2005.
  - Countries with at least one long episode have, on average, more than tripled their credit-to-GDP ratio by 2005.
  - Post-acceleration outcomes for 10-year+ episodes (2005):
    - Higher than at the end of episode: 23 countries (48 percent)
    - Same as at the end of episode: 1 country (9 percent) [note: "1/ Mainly countries whose episodes ended close to 2005"]
    - Lower than at the end of episode: 16 countries (33 percent)
- Decadal and regional patterns
  - 1990s account for 44 percent of the total 5–10 year episodes; the 1960s are the least populated decade.
  - Of the 70 episodes of the 1990s and the 9 that started after 2000, 30 are unfinished; of unfinished periods, one third are in CIS or CEE countries and one fourth in Sub-Saharan Africa.
  - Europe, Asia/Pacific and the Western Hemisphere together account for 80 percent of all long events; Europe hosted 50 percent of long episodes occurring in the 1990s.
  - Sub-Saharan Africa has the highest incidence of short episodes, mainly in the 1970s and 1990s.
- Cross-country dispersion in credit-to-GDP (2005)
  - average = 0.50
  - median = 0.35
  - st. dev = 0.46

### Classification of financial acceleration types
- Type 1: cyclical upturn — credit expands faster than output (conventional accelerator).
- Type 2: “credit booms” — sometimes after financial liberalization; can lead to soft landing or banking/real-sector crisis.
- Type 3: longer-term “financial deepening” — slower expansion but sustainable; leads to more sophisticated financial system.
- At onset it is difficult to distinguish among the three types.

### Measurement and episode-identification methodology
- Yardstick: ratio of private sector credit to GDP.
- Notation: uses ∆
஼
ೖ
௒
ೖ
 to denote country k’s three-year moving average of its credit-to-GDP ratio’s annual growth rate (notation reproduced from source).
- Takeoff criteria:
  - i. Country k is experiencing a financial takeoff if ∆
஼
ೖ
௒
ೖ
൒2 %;
  - ii. The episode of financial acceleration lasts at least 5 years; labeled “sustained financial deepening” if it lasts at least 10 years.
- Rationale:
  - 2-percent threshold chosen to capture longer periods of sustained deepening; a centered three-year moving average avoids one-year flukes.
  - Minimum length of 5 years excludes incidental short-lived accelerations.
  - 10-year cut-off informed by lending-boom literature (Gourinchas et al. average 6.7 years, st. dev 3.6; Hilbers et al. 6.8 years for crises vs 9.6 years without crisis).

### Data, variables, and sources (Appendix II)
- Key variables and definitions (exact labels preserved):
  - y_t — Real Gross Domestic Product (GDP ) — GDP in constant prices
  - π_t — Rate of inflation — annual rate of inflation as measured by consumer p ric e   ind e x
  - credit/GDP — Ratio of Bank credit to private sector over GDP
  - GDP per capita — Real GDP per capita
  - Govt fiscal balance — C entral government fiscal balance/GDP
  - Real effect exch rate — Real effective exchange rate
  - Openness — Openness of the economic measured as (exports+imports of goods and services)/GDP)
  - Fin lib index — Index of financial liberalization
  - Bank sup index — Index of financial supervisory reform
  - Polity — P o lity 2 / P O LITY  IV
  - Qual pol inst — Quality of political institutions
  - Duration democracy and autocracy — Duration of democracy and autocracy
  - Constr on Exec — C onstraints on Executives
  - ∆ Positive regime — Positive regime change (Dummy that takes value 1, if a positive change took place in four years before start of financial acceleration.)
  - ∆ Negative regime — N egative regime change (Dummy that takes value 1, if a negative change took place in four years before start of financial acceleration.)
  - Legal origin — English, French, German, Scandinavian (Legal origin (UK, France, Scandinavia and Germany) (dummy))
- Data sources (exact attributions preserved):
  - y_t: World Development Indicators (WDI)
  - π_t: World Economic Outlook (WEO)
  - credit/GDP: WDI
  - GDP per capita: WDI
  - Govt fiscal balance: WEO
  - Real effect exch rate: International Financial Statistics (IFS)
  - Openness: own calculations based on IFS
  - Fin lib index: Tressel-Detragiache (2008) and own calculations based on Tressel-Detragiache (2008). We took bank supervision out of the index.
  - Bank sup index: Extracted from the index of financial liberalization (TD, 2008 and own calculations)
  - Polity / Qual pol inst / Duration democracy and autocracy / Constr on Exec / ∆ Positive regime / ∆ Negative regime: Own calculations based on POLITY 2 index, P O LITY IV
  - Legal origin: La Porta et al. 1999

### Econometric specification and identification
- Dependent variable: dummy equal to 1 in years associated with a financial acceleration (take-off) and 0 otherwise; for each episode starting at t the dummy =1 at t-1, t, and t+1 to reduce mistiming.
- Baseline estimated with a probit where the dependent dummy equals F(x'β), with explanatory variables lagged to reduce endogeneity risk.
- Explanatory variables include lagged averages of real GDP growth, GDP per capita, private credit-to-GDP, inflation, changes in the Fin lib index, changes in the Bank sup index, regime-change dummies, polity measures (Polity, durability, constraints on executive), and year-fixed effects.

### Empirical results — selected estimates and predictive performance
- General correlates
  - Real GDP growth in previous years increases the probability of a financial acceleration by 0.0217*** (coefficient ∆y t-1, t-2 = 0.0217***; standard error (0.00347)).
  - Private-sector credit-to-GDP at t-1: ln credit/GDP t-1, t-2 = -0.169*** (standard error (0.0174)).
  - GDP per capita t-1: ln GDP per capita t-1 = 0.0871*** (standard error (0.0125)).
- Financial liberalization and supervision
  - ∆ Fin. Lib index t = 0.689*** (standard error (0.196))
  - ∆ Fin. Lib index t-1 = 0.575*** (standard error (0.201))
  - ∆ Fin. Lib index t-2 = 0.568*** (standard error (0.195))
  - ∆ bank sup index t = 0.183 (standard error (0.114))
  - ∆ bank sup index t-1 = 0.315*** (standard error (0.117))
  - ∆ bank sup index t-2 = 0.437*** (standard error (0.112))
- Political institutions
  - polity t-1,...,t-5 = -0.00512** (standard error (0.00222)) in one specification (negative effect on probability of take-off lasting less than ten years).
  - Quality of political institutions (quality pol. Inst t,...,t-5) = -0.00522** (standard error (0.00211)) in one robustness specification.
  - constr on exec t,...,t-5 = -0.0117* (standard error (0.00684)) in Table 6.
  - Durability of democratic regimes greatly increases the probability of sustained financial development (duration variables enter with mixed signs across specifications; selected coefficients include duration autocracy = 0.00242* (standard error (0.00137)) in one column).
- Predictive performance
  - Observations: 1433 (all episodes specification)
  - R2: 0.1970 (all episodes specification)
  - Percentage of correctly classified observations: 81.1 (all episodes)
  - p-value: 0
  - Significance notation: *** p<0.01, ** p<0.05, * p<0.1

### Robustness checks — summary of outcomes
- Institutional measures
  - Results robust to alternative measures of institutional quality (quality of political institutions, constraints on the executive).
  - Political-institutions variables remain positive and significant only for long-lasting financial episodes in many specifications.
- Legal origin
  - Controlling for legal origin does not materially change main results: political institutions and durability of democratic regimes remain positive and significant only for long-lasting financial episodes.
  - Example coefficients: legal origin English: -0.0879* (standard error (0.0496)); legal origin French: -0.131** (standard error (0.0565)); legal origin Scandinavian: -0.110*** (standard error (0.0377)) in selected specifications.
- Macro controls
  - Including fiscal balance, changes in the real effective exchange rate, and openness generally does not change main findings.
  - Real effect exch rate t-2 enters positively in some specifications (e.g., 0.000240* (standard error (0.000139))).
  - Sample size falls when including the real exchange rate due to data availability; such estimates are less reliable.

### Interpretation, theory, and relation to prior literature
- Political-institutions view emphasized:
  - Financial development depends on macroeconomic environment, policy design, and contracting institutions, but even more on the quality of political systems that protect these institutions.
  - Effective protection of property rights ultimately depends on political institutions that limit government power via checks and balances.
- Mechanisms by which weak political constraints impede financial development:
  - Banks' fear of expropriation.
  - Government limiting the number of banks, reducing competition.
  - High costs for banks to enforce borrower contracts, favoring lending to well-known customers.
  - Depositor fears of imprudent bank behavior and government expropriation.
- Relation to existing literature
  - Complements work on financial liberalization, legal origin, property-rights protection, and political institutions (references to Keefer (2008), Roe and Siegel (2008), Haber and North, Bordo and Rousseau, and others).
  - Links to literature on lending booms and crises that finds many liberalizations since the 1980s led to short-term lending booms followed by crises (e.g., Gourinchas et al. 2001; Mendoza and Terrones, 2008; Barajas et al. 2008).

### Conclusions and policy implications
- Main conclusions
  - Financial liberalization is a very strong driver of short-term financial accelerations (lending booms) but is not sufficient for sustained deepening of the financial sector.
  - Quality and stability of political institutions strongly increase the probability of long-term financial deepening; political competition and checks and balances provide guarantees that property rights will be effectively respected.
  - Short-term accelerations are more likely in weaker political environments and can overheat the system, sometimes ending in financial crises; their long-term impact on financial development is often negligible.
  - Long-term accelerations tend to slowly push the financial system to higher levels of activity and sophistication and are associated with democratic institutional quality and regime durability.
- Policy recommendations
  - Policies aimed at jumpstarting financial development should address political-institutional quality, not only financial-sector liberalization or technical reforms.
  - Strengthening checks and balances and the durability of political constraints increases the likelihood that financial accelerations will translate into long-term financial deepening.
  - Financial liberalization without concomitant improvements in political-institutional quality risks producing short-lived accelerations or crises rather than sustained development.
- Open questions for future research
  - As economies develop and credit growth slows while financial development shifts toward non-bank finance and capital markets, do political institutions play the same role in maintaining effective financial intermediation? The paper suggests this is an open question for future research.

*Source — content unit: _wp10210 - References (excerpt of IMF working paper).*

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

### _wp10210 - References .............................................................................................................

### Main research question and approach
- Investigates whether financial development leads, follows, or is unrelated to economic growth and what policies or environments are conducive to financial deepening.
- Defines "financial accelerations" as episodes measured by the growth rate of the credit-to-GDP ratio.
- Compares short-lived financial deepening episodes with long-lasting ones to identify conditions prevailing around the take-off that increase the likelihood of sustained financial development.
- Timeframe and sample described in the paper: analysis covers events "since the early 1960s" and "the past 50 years" across a broad country sample.

### Key empirical findings
- The past 50 years witnessed "approximately 210 such episodes in a sample of about 160 countries."
- Only about "one-quarter of all financial accelerations in the past 50 years have led to long-term financial deepening."
- Most financial take-offs stalled after a number of years; some resulted in financial crises; a minority led to sustained financial development.
- Short-term accelerations are:
  - Intimately associated with financial liberalization.
  - Generally negatively associated with the quality of prevailing political institutions.
- Long-term financial deepening episodes are:
  - Positively and closely linked with the quality of political institutions before and at the time of the start of the acceleration.
  - More likely when political systems exhibit checks and balances and durability.

### Interpretation and theoretical contribution
- Argues that financial development depends not only on macroeconomic environment, policy design, and institutions (property rights, contract enforcement) but even more on the quality of political systems that protect these institutions.
- Political-institutions view: effective protection of property rights ultimately depends on political institutions that limit government power via checks and balances (preventing expropriation and ensuring enforcement).
- Mechanisms by which weak political constraints impede financial development include:
  - Banks' fear of expropriation.
  - Government limiting the number of banks, reducing competition.
  - High costs for banks to enforce borrower contracts, favoring lending to well-known customers.
  - Depositor fears of imprudent bank behavior and government expropriation.
- Constraining government power through political competition, free elections, competitive parties, and separation of powers builds confidence that laws will be applied and enforced, thereby promoting financial deepening.

### Relationship to prior literature
- Situates contribution within debates and empirical work on:
  - Financial liberalization (mixed success; disappointing results in many countries despite reforms).
  - Role of legal origin and contracting institutions (mixed and non-robust findings; time-invariance issues).
  - Importance of property-rights protection for financial development (cites evidence favoring property-rights institutions over contracting institutions).
  - Political institutions school (authors cited: Haber and North; Bordo and Rousseau; Keefer; Roe and Siegel; Tressel and Detragiache).
- Notes that earlier findings provided partial explanations but that quality and durability of political institutions provide stronger explanatory power for which accelerations lead to sustained deepening.

### Policy implications and recommendations
- Policies aimed at jumpstarting financial development should address political-institutional quality, not only financial-sector liberalization or technical reforms.
- Strengthening checks and balances and the durability of political constraints increases the likelihood that financial accelerations will translate into long-term financial deepening.
- Financial liberalization without concomitant improvements in political-institutional quality risks producing short-lived accelerations or crises rather than sustained development.

*Italic: Source — content unit: _wp10210 - References (excerpt of IMF working paper).

### conclusion that effective protection of property rights is an important determinant of

### _wp10210 - conclusion that effective protection of property rights is an important determinant of

### Relation to existing literature and main contribution
- Confirms the conclusion that effective protection of property rights is an important determinant of sustained financial development.
- Situates the contribution relative to specific studies:
  - Keefer (2008): instruments effective protection of property rights on political institutions variables; finds the component of secure property rights explained by political institutions is a significant determinant of financial sector development.
  - Roe and Siegel (2008): use indices of political instability to show a consistent and significant link between political instability and financial backwardness.
  - The present work is complementary: while others use measures of political instability, this paper shows that higher democratic content and greater stability increase the likelihood of episodes of financial deepening.
- Historical and regional narratives referenced:
  - Malmendier (2009): documents that during the Roman Empire financial deepening occurred when political stability prevailed, not when the legal system was strongest.
  - Campos and Coricelli (2009): highlight interactions between early stages of democracy and financial development in Eastern Europe; early democracy often lacks political stability and thus slows financial development. The paper’s empirical results support this view across a larger cross-country sample.

### Link to literature on lending booms and crises
- Connects to a literature that finds financial liberalization since the 1980s has often led to short-term lending booms followed by crises (examples cited: Gourinchas et al. 2001, Mendoza and Terrones, 2008, Barajas et al. 2008; Cottarelli et al. 2003; Hilbers et al. 2005).
- The paper aims to:
  - Shed light on conditions under which lending-boom episodes occur.
  - Contrast short-term accelerations that may precede crises with long-term accelerations leading to sustained financial deepening.

### Yardstick and scope for measuring financial development
- Uses the ratio of private sector credit to GDP as the primary yardstick for financial development, justified because:
  - It indicates banks’ role as intermediaries of financial resources.
  - It is widely available across countries and captures broad developments in bank-dominated financial systems.
- Acknowledges limitations:
  - Less complete than measures incorporating quality of financial services or stock market development (see Campos and Coricelli, 2009).

### Stylized facts and examples of financial accelerations
- Observes a large cross-country disparity in credit-to-GDP ratios.
- Notes that well-developed financial systems often experienced prolonged and accelerated financial deepening episodes; such accelerations are highly concentrated among well-developed systems.
- Empirical exemplars (rebased to start in the same year “t”):
  - Australia: financial growth accelerated in 1983 and lasted over 20 years by the paper’s criteria; credit-to-GDP was close to 25 percent at the beginning and close to 100 percent at the end; classified as a type-3 period of sustained financial deepening.
  - Egypt: shorter episode starting in 1980 and dying out after 7 years; level of credit-to-GDP was close to 20 percent at onset and returned to that level after the episode — illustrating a soft landing with no lasting effect.
  - Sweden: rapid type-2 credit boom following financial liberalization, ending in a banking crisis (1992-93); credit-to-GDP rose from just under 40 to 55 percent and then fell below the take-off level.

### Classification of financial acceleration types
- Draws on Hilbers et al. (2005) to distinguish three types:
  - Type 1: cyclical upturn — credit expands faster than output due to financing investment/working capital (conventional accelerator).
  - Type 2: “credit booms” — inappropriate market responses to changing risk perceptions, sometimes after financial liberalization; can lead to soft landing or banking/real-sector crisis.
  - Type 3: longer-term “financial deepening” — slower expansion rate than types 1 and 2 but sustainable for longer, leading to a more sophisticated financial system that accompanies or contributes to economic growth.
- Emphasizes that at onset it is difficult to distinguish between these three types.

### Methodology for identifying episodes of financial acceleration
- Sample and period:
  - Identifies takeoff episodes between 1960 and 2005 for a sample of about 160 countries.
- Definition and criteria:
  - Uses ∆
஼
ೖ
௒
ೖ
 to denote country k’s three-year moving average of its credit-to-GDP ratio’s annual growth rate (notation reproduced from source).
  - i. Country k is experiencing a financial takeoff if ∆
஼
ೖ
௒
ೖ
൒2 %;
  - ii. The episode of financial acceleration lasts at least 5 years; labeled “sustained financial deepening” if it lasts at least 10 years.
- Rationale for methodological choices:
  - The 2-percent threshold is acknowledged as somewhat arbitrary but considered reasonable for capturing longer periods of sustained deepening (noting that credit booms can see annual growth rates of 30 – 40 percent).
  - A centered three-year moving average is used to avoid one-year “accidents” or fluke changes.
  - Minimum length of 5 years eliminates incidental short-lived accelerations.
  - A 10-year cut-off for sustained deepening is informed by the lending-boom literature:
    - Gourinchas et al. (2001) estimate average lifetime of a lending boom as 6.7 years with a standard deviation of 3.6.
    - Hilbers et al. (2005) find that credit booms ending in a crisis last on average (text cuts off here in source).

### Objective and analytic aim
- Objective: examine economic and institutional conditions under which different types of acceleration episodes are likely to take off, distinguishing short-term lending booms from long-term, sustainable financial deepening.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10210.pdf*

### 6.8 years, while those ending without a crisis have a lifetime of, on average 9.6 years.  Thus,

### _wp10210 - 6.8 years, while those ending without a crisis have a lifetime of, on average 9.6 years.  Thus,

### Data and descriptive statistics
- Sample and episodes
  - Identified 209 periods of financial accelerations in a sample of about 160 countries for the period 1960 – 2005.
  - Episodes by duration: 161 episodes lasting between 5 and 10 years; 48 episodes longer than 10 years.
  - Appendix I provides the detailed list of the country episodes (not reproduced here).
- Decadal and regional patterns
  - For 5–10 year periods: the 1960s are the least populated decade; the 1990s account for 44 percent of the total; the 1970s and 1980s each witnessed about 30 episodes.
  - Of the 70 episodes of the 1990s and the 9 that started after 2000, a total of 30 are still unfinished; of these unfinished periods, a third are in CIS or CEE countries and another fourth in Sub-Saharan Africa.
  - Sub-Saharan Africa has the highest incidence of short episodes, mainly in the 1970s and 1990s, followed by the Western Hemisphere and the Asia/Pacific region.
  - Europe, Asia/Pacific and the Western Hemisphere together account for 80 percent of all long events; Europe hosted 50 percent of the long episodes occurring in the 1990s.
  - Asia/Pacific and the Western Hemisphere each account for about 20 percent of the short periods and 25 percent of the long ones.
  - The Middle East and North African region is largely absent among the long events and ranks low in short events.
- Income-group breakdown
  - Middle-income countries (MICs) account for approximately 50 percent of all short episodes and close to 50 percent of long episodes.
  - High-income countries (HICs) account for the remaining roughly 50 percent of long episodes.
  - Low-income countries (LICs) account for less than 20 percent of the short episodes and only 6 percent of the long episodes.
- Episode-level statistics (Tables 2a–2c summarized)
  - Average length of short periods is just under 7 years.
  - Periods of sustained deepening are on average twice as long as short periods.
  - The average level of the credit-to-GDP ratio at the beginning of short and sustained episodes is very similar; take-offs can start at ratio levels as low as 1 percent and as high as 120 to 130 percent.
  - Average growth rate in short episodes is slightly higher than in sustained episodes; extremes for both can exceed 100 percent per annum.
  - Initial credit-to-GDP in countries with no accelerations and those with episodes lasting 5–10 years or over 10 years is around 19 to 22 percent.
  - Countries with one or more short-term accelerations have, on average, doubled their credit-to-GDP ratio by 2005.
  - Countries with at least one long episode have, on average, more than tripled their credit-to-GDP ratio by 2005.
  - Post-acceleration outcomes (sustained episodes): the financial system continued to grow in almost 50 percent of cases; in another 19 percent it is too early to judge (growth period finished shortly before 2005); in 33 percent of cases gains were partly or completely reversed (reversals often correlated with political instability, war, or financial crises).

### Methodology for explaining financial accelerations
- Dependent variable and timing adjustments
  - Dependent variable: dummy equal to 1 in years associated with a financial acceleration (take-off) and 0 otherwise.
  - For each identified episode starting at time t, the dummy equals 1 at t-1, t, and t+1 to minimize mistiming.
  - For ongoing episodes, data for years t+2 until the end of the episode are dropped so that 0 corresponds to non-take-off years rather than mid-episode years.
- Determinant groups and data sources
  - Macroeconomic and structural variables: real GDP growth, inflation, government fiscal position, real exchange rates, GDP per capita, initial credit-to-GDP, trade openness.
  - Financial liberalization: index of financial liberalization developed in TD (2008), 21 indicators, available for 85 countries for 1975–2006 and expanded into the 1960s for this study; bank supervision singled out as a separate index.
  - Institutional variables: Polity IV dataset (Marshall and Jaggers, 2008) used for Polity score (–10 to +10), durability of democratic and autocratic regimes, and a composite quality-of-political-institutions variable (sum of constraints on executives, competition/openness in access to mandates, competitiveness of parties/election process; individual elements range 0–10).
  - Additional institutional dummies: positive and negative regime change dummies (significant change = at least 3 points in the Polity variable over the preceding five years).
  - Dummies for legal origin (English, French, German, Scandinavian) are included in robustness checks.
- Econometric specification
  - Baseline estimated with a probit where the dependent dummy equals F(x'β), with explanatory variables lagged to reduce endogeneity risk.
  - Explanatory variables include average real GDP growth between t-1 and t-2, GDP per capita at t-1, private credit-to-GDP at t-1, inflation at t-j, changes in TD’s financial liberalization index at t-j, changes in bank supervision index at t-j, regime-change dummies, political institutions measures (Polity t or average Polity over preceding five years, or durability), and year-fixed effects.

### Empirical results and key findings
- General effects
  - Real GDP growth in previous years increases the probability of a financial acceleration by 2 percentage points.
  - Private-sector credit-to-GDP at t-1 has a negative coefficient: countries with high levels of financial development are less likely to experience take-offs.
  - GDP per capita t-1 has a positive and significant coefficient: the likelihood of a take-off increases with higher GDP per capita; effect robust to episode length.
- Financial liberalization and supervision
  - Financial liberalization has a significant and large impact on take-off probability, with effects varying by episode duration:
    - Likelihood of a short episode increases significantly following successive efforts to liberalize the financial system (changes in the liberalization index).
    - Contemporaneous financial liberalization matters more for episodes lasting more than 10 years.
  - Improved bank supervision has a weak effect overall; it increases the likelihood of take-offs with a lag, with more robust effects for sustained accelerations.
- Political institutions
  - Polity score:
    - Significant and negative effect on the probability of a take-off lasting less than ten years.
    - Significant and positive effect on the probability of sustained episodes of financial development (long episodes).
    - Interpretation: high institutional quality reduces likelihood of short-lived accelerations but increases likelihood of sustained deepening.
  - Regime-change dummies (recent positive or negative changes) do not increase the likelihood of short or long acceleration episodes in general.
  - Durability of democratic regimes greatly increases the probability of sustained financial development; durability of autocratic regimes has opposite signs but is generally not significant.
- Predictive performance
  - Using a 50 percent predicted-probability cutoff, the model correctly classifies at least 80 percent of observations (bottom line of Table 5).

### Robustness checks
- Institutions
  - Results are robust to alternative measures of institutional quality (quality of political institutions, constraints on the executive).
  - Countries with more checks and balances (constraints on the executive and political competition) are less likely to experience accelerations lasting less than 10 years; for episodes lasting more than 10 years, the coefficient becomes positive and significant at the 10 percent level.
- Legal origins
  - Controlling for legal origin (English, French, German, Scandinavian) does not materially change the main results: political institutions and durability of democratic regimes remain positive and significant only for long-lasting financial episodes.
- Domestic and external macro factors
  - Including fiscal balance, changes in the real effective exchange rate, and economic openness does not change main findings except that the real exchange rate weakens the impact of financial reform and polity.
  - Sample size is halved when including the real exchange rate due to lack of long-term series, making those estimates less reliable.

### Conclusions and implications
- Main conclusions
  - Financial liberalization is a very strong driver of short-term financial accelerations (lending booms) but is not sufficient for sustained deepening of the financial sector.
  - Quality and stability of political institutions strongly increase the probability of long-term financial deepening; political competition and checks and balances provide guarantees that property rights will be effectively respected, enabling sustained financial development.
  - Short-term accelerations are more likely to emerge in weaker political environments and can overheat the system, sometimes ending in financial crises; their long-term impact on financial development is often negligible.
  - Long-term accelerations tend to slowly push the financial system to higher levels of activity and sophistication and are associated with democratic institutional quality and regime durability.
- Open questions for future research
  - As countries develop and credit growth slows, and financial development shifts toward non-bank finance and capital markets, do political institutions play the same role in maintaining effective financial intermediation?
  - The paper is largely silent on these questions and suggests they be addressed by future research.

*Source: _wp10210 - 6.8 years, while those ending without a crisis have a lifetime of, on average 9.6 years.  Thus,*

### REFERENCES

### _wp10210 - REFERENCES

### Bibliographic inventory
- Full list of cited works (authors, titles, journals/working papers/books) as provided in the References section. Key authors and works include Abiad; Acemoglu and Johnson; Beck and Levine; Bordo and Rousseau; Boyd, Levine and Smith; Claessens and Laeven; Demirgüç-Kunt and Detragiache; Detragiache and Giang Ho; Djankov, McLiesh and Shleifer; Gourinchas, Valdés and Landerretche; Haber, North and Weingast; Hausmann, Pritchett and Rodrik; Kaminsky and Schmukler; King, Plosser and Rebelo; Levine; McKinnon; Mendoza and Terrones; Pagano and Volpin; Rajan and Zingales; Shaw; Tressel and Detragiache; Williamson and Mahar; and many others as listed.

### Key figures and descriptive statistics
- Figure 1 — Cross-country disparities in credit to private sector/GDP (2005):
  - average = 0.50
  - median = 0.35
  - st. dev = 0.46
  - Note: black bars represent countries that had at least one acceleration episode of 10 years or more.

- Figure 2 — Private credit/GDP time profile around acceleration episodes: (plot of private credit/GDP across years t-4 ... t+22 for sample episodes, including Australia 1983-2005; Egypt 1980-1987; Sweden 1986-1990).

### Episodes and counts (Tables 1, 2a–2c)
- Table 1 — Financial deepening episodes by decade and duration (excerpted counts preserved in original tabular structure).
- Table 2b — Selected averages (5–10 years vs 10 years+):
  - Average duration of episode (years): 6.9 (5–10 years), 14.7 (10 years+)
  - Average Credit/GDP at beginning of episode (5–10 years): 0.224; at end (2005) for no acceleration: 0.214
  - At start of period under review / At end (2005):
    - No financial acceleration: 0.224 -> 0.214
    - 5–10 year accelerations: 0.215 -> 0.402
    - 10-year+ accelerations: 0.186 -> 0.683

- Table 2c — Post-acceleration Credit/GDP of countries experiencing 10-year+ episodes (2005):
  - Higher than at the end of episode: 23 countries (48 percent)
  - Same as at the end of episode: 1 country (9 percent) [note: "1/ Mainly countries whose episodes ended close to 2005"]
  - Lower than at the end of episode: 16 countries (33 percent)
  - Gain / loss (percent):
    - less than 49 percent: 12 countries
    - 50–99 percent: 3 countries
    - more than 100: 1 country

### Predictability and correlates of acceleration episodes (Table 4)
- Episodes accompanied by (percent of episodes):
  - Financial liberalization (rise in index ≥ 0.13 basis points in 4 years before takeoff): 5–10 years = 39.1; 10 years+ = 36.0
  - Improvements in supervision (rise in index ≥ 0.13 basis points in 4 years before takeoff): 5–10 years = 32.2; 10 years+ = 24.0
  - Quality political institutions (Polity > 6 for at least 5 years before takeoff): 5–10 years = 34.1; 10 years+ = 56.0
  - Democratic regime of at least 10 years before takeoff: 5–10 years = 25.2; 10 years+ = 46.0
  - Autocratic regime of at least 10 years before takeoff: 5–10 years = 23.8; 10 years+ = 15.0
  - Positive regime change in last 5 years: 5–10 years = 26.4; 10 years+ = 0.5
  - Negative regime change in last 5 years: 5–10 years = 14.0; 10 years+ = 1.0
  - Financial crisis within 5 years following takeoff: 5–10 years = 8.7; 10 years+ = 4.2
  - Financial crisis within 5 years before takeoff: 5–10 years = 5.0; 10 years+ = 0.0

### Baseline regression results (Table 5) — selected coefficients and statistics
- Dependent variable: indicator of financial acceleration episodes (various durations)
- Key estimated coefficients (all numbers preserved exactly as in source):
  - ∆y t-1, t-2: 0.0217*** (all episodes)
    - Standard error: (0.00347)
  - π t-1: 4.77e-05 (all episodes)
    - Standard error: (3.76e-05)
  - ∆ Fin. Lib index t: 0.689*** (all episodes)
    - Standard error: (0.196)
  - ∆ Fin. Lib index t-1: 0.575*** (all episodes)
    - Standard error: (0.201)
  - ∆ Fin. Lib index t-2: 0.568*** (all episodes)
    - Standard error: (0.195)
  - ∆ bank sup index t: 0.183 (all episodes)
    - Standard error: (0.114)
  - ∆ bank sup index t-1: 0.315*** (all episodes)
    - Standard error: (0.117)
  - ∆ bank sup index t-2: 0.437*** (all episodes)
    - Standard error: (0.112)
  - ln credit/GDP t-1, t-2: -0.169*** (all episodes)
    - Standard error: (0.0174)
  - ln GDP per capita t-1: 0.0871*** (all episodes)
    - Standard error: (0.0125)
  - polity t-1,...,t-5: -0.00512** (controlling for polity)
    - Standard error: (0.00222)
  - Observations: 1433 (all episodes)
  - R2: 0.1970 (all episodes)
  - p-value: 0
  - Percentage of correctly classified observations: 81.1 (all episodes)
  - Significance notation: *** p<0.01, ** p<0.05, * p<0.1

### Robustness and extensions (selected outcomes from Tables 6–9)
- Quality of institutions and executive constraints (Table 6):
  - ∆ Fin. Lib index t coefficients remain statistically significant and positive (e.g., 0.694*** for all episodes; standard error (0.196))
  - quality pol. Inst t,...,t-5: -0.00522** (standard error (0.00211))
  - constr on exec t,...,t-5: -0.0117* (standard error (0.00684))

- Legal origin robustness (Table 7):
  - ∆ Fin. Lib index t: 0.710*** (all episodes) with standard error (0.196)
  - Legal origin indicators show negative coefficients for English, French, Scandinavian origins in some specifications (e.g., legal origin English: -0.0879*; standard error (0.0496); legal origin French: -0.131**; standard error (0.0565); legal origin Scandinavian: -0.110***; standard error (0.0377))

- Macroeconomic controls and polity (Table 8):
  - Coefficients on ∆ Fin. Lib index t remain positive and in many specifications significant (examples: 0.538***; (0.204); and 0.696***; (0.197) in different columns)
  - real effect exch rate t-2: 0.000240* (standard error (0.000139)) in one specification
  - ∆ negative regime t,t-5: 0.159** (standard error (0.0761)) in one specification
  - Observations and R2 vary by specification (examples: Observations 1292, 1168, 1244, 853, 664, 593, 638, 475, 1400, 1261, 1342, 920; R2 examples: 0.203, 0.175, 0.201, 0.236, 0.190, 0.144, 0.175, 0.310, 0.202, 0.169, 0.198, 0.235)

- Duration of polity and macro robustness (Table 9):
  - Duration democracy and duration autocracy enter with mixed signs across specifications; selected coefficients include duration autocracy = 0.00242* (standard error (0.00137)) in one column
  - ∆ negative regime t,t-5: 0.225*** (standard error (0.0795)) in one column
  - real effect exch rate t-2: 0.000234* (standard error (0.000134)) in one column
  - Observations and R2 examples: Observations 1293, 1169, 1245, 854, 664, 593, 638, 475, 1400, 1261, 1342, 920; R2 examples: 0.202, 0.160, 0.193, 0.257, 0.201, 0.149, 0.179, 0.349, 0.201, 0.154, 0.190, 0.257

### Appendix I — Episodes of financial deepening (1960–2005)
- Country-level episode listing (excerpt):
  - Examples of episodes with durations recorded exactly as in the appendix (e.g., Australia: 1964–1968 length 5; 1983–2005 length 23; Austria: 1962–1981 length 20; China, P.R.: 1996–2003 length 8; Cyprus: 1986–2002 length 17; Brazil, Chile, Finland, France, UK, USA, etc.; many country episodes listed with start/end years and lengths)
  - Note: Asterisks mark unfinished episodes ("* unfinished episodes (30 short ones)").

*Italic: Source content unit _wp10210 - REFERENCES (PDF chapter/section) as provided.*

### APPENDIX II – DATA SOURCES

### APPENDIX II – DATA SOURCES

### Variables and definitions
- y_t — Real Gross Domestic Product (GDP ) — GDP in constant prices
- π_t — Rate of inflation — annual rate of inflation as measured by consumer p ric e   ind e x
- credit/GDP — Ratio of Bank credit to private sector over GDP
- GDP per capita — Real GDP per capita
- Govt fiscal balance — C entral government fiscal balance/GDP
- Real effect exch rate — Real effective exchange rate
- Openness — Openness of the economic measured as (exports+imports of goods and services)/GDP)
- Fin lib index — Index of financial liberalization
- Bank sup index — Index of financial supervisory reform
- Polity — P o lity 2 / P O LITY  IV
- Qual pol inst — Quality of political institutions
- Duration democracy and autocracy — Duration of democracy and autocracy
- Constr on Exec — C onstraints on Executives
- ∆ Positive regime — Positive regime change (Dummy that takes value 1, if a positive change took place in four years before start of financial acceleration.)
- ∆ Negative regime — N egative regime change (Dummy that takes value 1, if a negative change took place in four years before start of financial acceleration.)
- Legal origin — English, French, German, Scandinavian (Legal origin (UK, France, Scandinavia and Germany) (dummy))

### Data sources
- y_t: World Development Indicators (WDI)
- π_t: World Economic Outlook (WEO)
- credit/GDP: WDI
- GDP per capita: WDI
- Govt fiscal balance: WEO
- Real effect exch rate: International Financial Statistics (IFS)
- Openness: own calculations based on IFS
- Fin lib index: Tressel-Detragiache (2008) and own calculations based on Tressel-Detragiache (2008). We took bank supervision out of the index.
- Bank sup index: Extracted from the index of financial liberalization (TD, 2008 and own calculations) (see above).
- Polity / Qual pol inst / Duration democracy and autocracy / Constr on Exec / ∆ Positive regime / ∆ Negative regime: Own calculations based on POLITY 2 index, P O LITY IV
- Legal origin: La Porta et al. 1999

### Index construction and calculation notes
- Openness is measured as (exports+imports of goods and services)/GDP and was computed from IFS data (own calculations).
- Fin lib index is based on Tressel-Detragiache (2008) with bank supervision removed from the index.
- Bank sup index is extracted from the index of financial liberalization (TD, 2008 and own calculations).
- ∆ Positive regime and ∆ Negative regime are dummy variables constructed from POLITY IV; each dummy takes value 1 if a positive/negative change took place in four years before start of financial acceleration.
- Legal origin is coded as dummies for UK, France, Scandinavia and Germany following La Porta et al. 1999.

*Source: APPENDIX II – DATA SOURCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10210.pdf_