## The Dark Side of the Moon? Fintech and Financial Stability — Section 1

## Source details

**Canonical URL:** [The Dark Side of the Moon? Fintech and Financial Stability — Section 1](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023253-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023253-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023253-print-pdf.pdf.json)

---

### Abstract and Key Questions
- Rapid advances in digital technology are transforming financial services; fintech has potential to improve efficiency, competition, and financial inclusion, but also poses systemic risks through greater technological complexity and cybersecurity exposure.
- Research question: How does fintech (excluding cryptocurrencies) correlate with financial stability across 198 countries over 2012–2020?

### Data and Measurement
- Sample: panel of 198 countries over the period 2012–2020.
- Main dependent variable:
  - Bank z-score defined as z-score_it = (ROA_it + CAR_it) / σ(ROA_it), where ROA_it = return on assets, CAR_it = capital-asset ratio, and σ(ROA_it) = standard deviation of return on assets.
- Key explanatory variables (from CCAF):
  - Digital lending: volume of lending instruments through digital platforms (balance sheet lending, peer-to-peer and marketplace lending, debt-based lending, invoice trading), scaled by GDP.
  - Digital capital raising: volume of capital raising instruments through digital platforms (investment-based and non-investment-based crowdfunding), scaled by GDP.
  - Total fintech: combination of digital lending, digital capital raising, and other fintech types (such as micro finance and pension-led funding), scaled by GDP.
- Control variables: real GDP per capita, real GDP growth, inflation, trade openness, domestic credit to the private sector (share of GDP), government stability, bureaucratic quality.
- Dataset sources: CCAF; GFD; WDI; ICRG.

### Descriptive Statistics (selected, exact values)
- Bank z-score: Observations 1,427; Mean 16.8; Std. dev. 9.5; Minimum -0.3; Maximum 62.4.
- Fintech measures:
  - Digital lending: Observations 594; Mean 0.1; Std. dev. 0.3; Minimum 0.0; Maximum 3.4.
  - Digital capital raising: Observations 1,093; Mean 0.0; Std. dev. 0.0; Minimum 0.0; Maximum 0.5.
  - Total (all fintech): Observations 1,118; Mean 0.1; Std. dev. 0.2; Minimum 0.0; Maximum 3.4.
- Other controls (selected):
  - Real GDP per capita: Observations 1,738; Mean 13,706; Std. dev. 18,765; Minimum 263; Maximum 167,809.
  - Real GDP growth: Observations 1,738; Mean 2.2; Std. dev. 5.9; Minimum -54.2; Maximum 86.8.
  - Inflation: Observations 1,620; Mean 5.3; Std. dev. 21.1; Minimum -4.3; Maximum 557.2.
  - Domestic credit to the private sector: Observations 1,528; Mean 55.0; Std. dev. 43.5; Minimum 1.1; Maximum 258.9.
  - Government stability: Observations 1,242; Mean 7.1; Std. dev. 1.1; Minimum 4.0; Maximum 11.0.
  - Bureaucratic quality: Observations 1,242; Mean 2.2; Std. dev. 1.1; Minimum 0.0; Maximum 4.0.

### Empirical Strategy
- Baseline panel specification:
  - z_it = β1 + β2 fintech_it + β3 X_it + η_i + μ_t + ε_it
  - z_it = log of bank z-score; fintech_it = one of (digital lending share of GDP, digital capital raising share of GDP, total fintech share of GDP); X_it = control vector.
  - Country fixed effects (η_i) and time fixed effects (μ_t) included.
  - Driscoll-Kraay (1998) standard errors used to account for heteroskedasticity, autocorrelation and cross-sectional dependence.

### Main Empirical Results (all countries; coefficients and standard errors preserved)
- Column [1] (Digital lending):
  - Digital lending coefficient: -0.071
  - Standard error: [0.041]
- Column [2] (Digital capital raising):
  - Digital capital raising coefficient: 1.433***
  - Standard error: [0.436]
- Column [3] (Total fintech):
  - Total fintech coefficient: -0.052
  - Standard error: [0.073]
- Selected control variable coefficients (columns pooled as reported):
  - Real GDP per capita: -0.116***, -0.080***, -0.074***; standard errors [0.026], [0.016], [0.018].
  - Real GDP growth: 0.010***, 0.007, 0.007; standard errors [0.004], [0.006], [0.005].
  - Inflation: -0.012***, -0.001***, -0.001***; standard errors [0.002], [0.000], [0.000].
  - Domestic credit to the private sector: 0.004***, 0.003***, 0.002***; standard errors [0.000], [0.000], [0.000].
  - Government stability: 0.025***, 0.025***, 0.025***; standard errors [0.009], [0.010], [0.009].
  - Bureaucratic quality: 0.147***, 0.092***, 0.087***; standard errors [0.052], [0.013], [0.013].
- Sample sizes and fit (columns [1][2][3]):
  - Number of observations: 496, 778, 796.
  - Number of countries: 98, 116, 116.
  - Country FE: Yes; Time FE: Yes.
  - R2: 0.15, 0.19, 0.18.

### Robustness and Heterogeneity
- Subsample analysis (advanced economies vs. developing countries):
  - Impact of fintech on the bank z-score becomes statistically insignificant across specifications when estimated for country subsamples (smaller sample sizes).
  - Digital lending impact: negative in both advanced and developing countries.
  - Digital capital raising impact:
    - Positive in advanced economies.
    - Negative in developing countries.
  - Overall effect of fintech:
    - Positive in advanced economies (when aggregating instruments).
    - Negative in developing countries (though statistically insignificant).
- Additional robustness: pre-pandemic period estimates yield similar results (available upon request).

### Interpretation of Findings
- Instrument-specific effects:
  - Digital lending is associated with a negative effect on financial stability (coefficient negative, though statistically insignificant in the full-sample specification for lending alone).
  - Digital capital raising is associated with a large and statistically significant positive effect on financial stability in the pooled sample (coefficient 1.433*** [0.436]).
- Aggregate effect:
  - Total fintech effect appears negative (coefficient -0.052 [0.073]) in pooled regressions because digital lending dominates the fintech composition, but this aggregate effect is statistically insignificant at conventional levels.
- Cross-country heterogeneity:
  - Benefits of digital capital raising for decentralization and diversification appear to materialize in advanced economies but not in developing countries.
- Scale and future risk:
  - Average fintech volume during 2012–2020 is 0.1 percent of GDP, compared to 55 percent of GDP in domestic credit to the private sector; fintech remains small but rapidly expanding, particularly in riskier segments (lending).
  - Fintech growth and adoption by large incumbents and big-tech firms could magnify effects on financial stability going forward.

### Policy Implications and Recommendations
- Need for appropriate regulatory frameworks to harness fintech benefits while mitigating systemic risks.
- Key policy priorities:
  - Strong regulatory institutions.
  - Better use of technology in regulation (RegTech/suptech).
  - Extensive cross-border coordination.
  - Appropriately calibrated prudential regulations to ensure a level playing field and effective monitoring and supervision of traditional and emerging financial institutions.
- Rationale and specific considerations:
  - Fintech firms may take on more risks themselves and can exert pressure on traditional institutions by degrading profitability, loosening lending standards improperly, and increasing risk-taking in operations and transactions.
  - Systemic risks can arise from institutions that are not individually systemically important; monitoring of complex network structures and stress testing in downturns is important.
  - Entry of big-tech firms into the financial sector, especially at a global scale, may require more entity-based regulation.

### Summary
- The paper uses a novel cross-country dataset (2012–2020) to examine fintech (excluding cryptocurrencies) and financial stability.
- Main empirical insights:
  - Digital lending: negative association with bank z-score (greater risk), coefficient -0.071 [0.041] in pooled estimate.
  - Digital capital raising: positive association with bank z-score (strengthening stability) and statistically highly significant, coefficient 1.433*** [0.436] in pooled estimate.
  - Aggregate fintech: negative but statistically insignificant effect, coefficient -0.052 [0.073], driven by the predominance of digital lending in total fintech.
- Policy action is warranted to monitor and regulate fintech developments proactively, with attention to heterogeneity across advanced and developing economies.

*Source: Serhan Cevik, "The Dark Side of the Moon? Fintech and Financial Stability", IMF Working Paper WP/23/253, December 2023 — Section 1.*

### Section 1

### The Dark Side of the Moon? Fintech and Financial Stability — Section 1

### Abstract and Key Questions
- Rapid advances in digital technology are transforming financial services; fintech has potential to improve efficiency, competition, and financial inclusion, but also poses systemic risks through greater technological complexity and cybersecurity exposure.
- Research question: How does fintech (excluding cryptocurrencies) correlate with financial stability across 198 countries over 2012–2020?

### Data and Measurement
- Sample: panel of 198 countries over the period 2012–2020.
- Main dependent variable: bank z-score, defined as
  - z-score_it = (ROA_it + CAR_it) / σ(ROA_it)
  - where ROA_it = return on assets, CAR_it = capital-asset ratio, and σ(ROA_it) = standard deviation of return on assets.
- Key explanatory variables (from CCAF):
  - Digital lending: volume of lending instruments through digital platforms (balance sheet lending, peer-to-peer and marketplace lending, debt-based lending, invoice trading), scaled by GDP.
  - Digital capital raising: volume of capital raising instruments through digital platforms (investment-based and non-investment-based crowdfunding), scaled by GDP.
  - Total fintech: combination of digital lending, digital capital raising, and other fintech types (such as micro finance and pension-led funding), scaled by GDP.
- Control variables: real GDP per capita, real GDP growth, inflation, trade openness, domestic credit to the private sector (share of GDP), government stability, bureaucratic quality.
- Dataset sources: CCAF; GFD; WDI; ICRG.

### Descriptive Statistics (selected, exact values)
- Bank z-score: Observations 1,427; Mean 16.8; Std. dev. 9.5; Minimum -0.3; Maximum 62.4.
- Fintech measures:
  - Digital lending: Observations 594; Mean 0.1; Std. dev. 0.3; Minimum 0.0; Maximum 3.4.
  - Digital capital raising: Observations 1,093; Mean 0.0; Std. dev. 0.0; Minimum 0.0; Maximum 0.5.
  - Total (all fintech): Observations 1,118; Mean 0.1; Std. dev. 0.2; Minimum 0.0; Maximum 3.4.
- Other controls (selected):
  - Real GDP per capita: Observations 1,738; Mean 13,706; Std. dev. 18,765; Minimum 263; Maximum 167,809.
  - Real GDP growth: Observations 1,738; Mean 2.2; Std. dev. 5.9; Minimum -54.2; Maximum 86.8.
  - Inflation: Observations 1,620; Mean 5.3; Std. dev. 21.1; Minimum -4.3; Maximum 557.2.
  - Domestic credit to the private sector: Observations 1,528; Mean 55.0; Std. dev. 43.5; Minimum 1.1; Maximum 258.9.
  - Government stability: Observations 1,242; Mean 7.1; Std. dev. 1.1; Minimum 4.0; Maximum 11.0.
  - Bureaucratic quality: Observations 1,242; Mean 2.2; Std. dev. 1.1; Minimum 0.0; Maximum 4.0.

### Empirical Strategy
- Baseline panel specification:
  - z_it = β1 + β2 fintech_it + β3 X_it + η_i + μ_t + ε_it
  - z_it = log of bank z-score; fintech_it = one of (digital lending share of GDP, digital capital raising share of GDP, total fintech share of GDP); X_it = control vector.
  - Country fixed effects (η_i) and time fixed effects (μ_t) included.
  - Driscoll-Kraay (1998) standard errors used to account for heteroskedasticity, autocorrelation and cross-sectional dependence.

### Main Empirical Results (all countries; coefficients and standard errors preserved)
- Column [1] (Digital lending):
  - Digital lending coefficient: -0.071
  - Standard error: [0.041]
- Column [2] (Digital capital raising):
  - Digital capital raising coefficient: 1.433***
  - Standard error: [0.436]
- Column [3] (Total fintech):
  - Total fintech coefficient: -0.052
  - Standard error: [0.073]
- Selected control variable coefficients (columns pooled as reported):
  - Real GDP per capita: -0.116***, -0.080***, -0.074***; standard errors [0.026], [0.016], [0.018].
  - Real GDP growth: 0.010***, 0.007, 0.007; standard errors [0.004], [0.006], [0.005].
  - Inflation: -0.012***, -0.001***, -0.001***; standard errors [0.002], [0.000], [0.000].
  - Domestic credit to the private sector: 0.004***, 0.003***, 0.002***; standard errors [0.000], [0.000], [0.000].
  - Government stability: 0.025***, 0.025***, 0.025***; standard errors [0.009], [0.010], [0.009].
  - Bureaucratic quality: 0.147***, 0.092***, 0.087***; standard errors [0.052], [0.013], [0.013].
- Sample sizes and fit (columns [1][2][3]):
  - Number of observations: 496, 778, 796.
  - Number of countries: 98, 116, 116.
  - Country FE: Yes; Time FE: Yes.
  - R2: 0.15, 0.19, 0.18.

### Robustness and Heterogeneity
- Subsample analysis (advanced economies vs. developing countries):
  - Impact of fintech on the bank z-score becomes statistically insignificant across specifications when estimated for country subsamples (smaller sample sizes).
  - Digital lending impact: negative in both advanced and developing countries.
  - Digital capital raising impact:
    - Positive in advanced economies.
    - Negative in developing countries.
  - Overall effect of fintech:
    - Positive in advanced economies (when aggregating instruments).
    - Negative in developing countries (though statistically insignificant).
- Additional robustness: pre-pandemic period estimates yield similar results (available upon request).

### Interpretation of Findings
- Instrument-specific effects:
  - Digital lending is associated with a negative effect on financial stability (coefficient negative, though statistically insignificant in the full-sample specification for lending alone).
  - Digital capital raising is associated with a large and statistically significant positive effect on financial stability in the pooled sample (coefficient 1.433*** [0.436]).
- Aggregate effect:
  - Total fintech effect appears negative (coefficient -0.052 [0.073]) in pooled regressions because digital lending dominates the fintech composition, but this aggregate effect is statistically insignificant at conventional levels.
- Cross-country heterogeneity:
  - Benefits of digital capital raising for decentralization and diversification appear to materialize in advanced economies but not in developing countries.
- Scale and future risk:
  - Average fintech volume during 2012–2020 is 0.1 percent of GDP, compared to 55 percent of GDP in domestic credit to the private sector; fintech remains small but rapidly expanding, particularly in riskier segments (lending).
  - Fintech growth and adoption by large incumbents and big-tech firms could magnify effects on financial stability going forward.

### Policy Implications and Recommendations (as discussed)
- Need for appropriate regulatory frameworks to harness fintech benefits while mitigating systemic risks.
- Key policy priorities:
  - Strong regulatory institutions.
  - Better use of technology in regulation (RegTech/suptech).
  - Extensive cross-border coordination.
  - Appropriately calibrated prudential regulations to ensure a level playing field and effective monitoring and supervision of traditional and emerging financial institutions.
- Rationale:
  - Fintech firms may take on more risks themselves and can exert pressure on traditional institutions by degrading profitability, loosening lending standards improperly, and increasing risk-taking in operations and transactions.
  - Systemic risks can arise from institutions that are not individually systemically important; monitoring of complex network structures and stress testing in downturns is important.

### Summary
- The paper uses a novel cross-country dataset (2012–2020) to examine fintech (excluding cryptocurrencies) and financial stability.
- Main empirical insights:
  - Digital lending: negative association with bank z-score (greater risk), coefficient -0.071 [0.041] in pooled estimate.
  - Digital capital raising: positive association with bank z-score (strengthening stability) and statistically highly significant, coefficient 1.433*** [0.436] in pooled estimate.
  - Aggregate fintech: negative but statistically insignificant effect, coefficient -0.052 [0.073], driven by the predominance of digital lending in total fintech.
- Policy action is warranted to monitor and regulate fintech developments proactively, with attention to heterogeneity across advanced and developing economies.

*Source: Serhan Cevik, "The Dark Side of the Moon? Fintech and Financial Stability", IMF Working Paper WP/23/253, December 2023 — Section 1.*

### Section 2

### Section 2

### Regression results and control variables
- Dependent variable: financial stability measured by the country-level bank z-score.
- Estimation method: Driscoll-Kraay standard errors reported in brackets. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.
- Key patterns from control variables:
  - Real GDP per capita: overall inverse correlation with the bank z-score (financial stability tends to be lower in higher income countries). Disaggregated estimations:
    - Coefficient on real GDP per capita is positive in advanced economies, but negative in developing countries.
  - Real GDP growth: stabilizing effect across all countries.
  - Inflation: significant negative impact on financial stability.
  - Trade openness: generally not statistically significant, except marginal positive impact in advanced economies.
  - Financial development (domestic credit to the private sector as a share of GDP): strong and statistically significant positive relationship with the z-score across all countries.
  - Institutional variables:
    - Government stability and bureaucratic quality strengthen financial stability, with greater statistical significance in developing countries.

### Table 3 — Fintech and Financial Stability: Advanced Economies (coefficients and standard errors)
- Columns labeled [1], [2], [3].
- Digital lending:
  - [1]: -0.347 [0.132]
- Digital capital raising:
  - [2]: 3.483 [2.232]
- Total fintech:
  - [3]: 0.211 [0.203]
- Real GDP per capita:
  - [1]: 0.203 [0.148]
  - [2]: 0.318*** [0.074]
  - [3]: 0.296** [0.102]
- Real GDP growth:
  - [1]: 0.007 [0.003]
  - [2]: 0.008 [0.007]
  - [3]: 0.002 [0.008]
- Inflation:
  - [1]: -0.038 [0.036]
  - [2]: -0.122*** [0.038]
  - [3]: -0.100*** [0.021]
- Trade openness:
  - [1]: 0.000* [0.000]
  - [2]: 0.001* [0.000]
  - [3]: 0.001* [0.001]
- Domestic credit to the private sector:
  - [1]: 0.003*** [0.001]
  - [2]: 0.003*** [0.001]
  - [3]: 0.002** [0.001]
- Government stability:
  - [1]: 0.021 [0.019]
  - [2]: 0.019 [0.028]
  - [3]: 0.002 [0.020]
- Bureacratic quality:
  - [1]: 0.154 [0.062]
  - [2]: 0.020 [0.044]
  - [3]: 0.032 [0.028]
- Number of observations: 216, 258, 268 (for columns [1], [2], [3] respectively)
- Number of countries: 33, 34, 34
- Country FE: Yes (all columns)
- Time FE: Yes (all columns)
- R2:
  - [1]: 0.28
  - [2]: 0.18
  - [3]: 0.14

### Table 4 — Fintech and Financial Stability: Developing Countries (coefficients and standard errors)
- Columns labeled [1], [2], [3].
- Digital lending:
  - [1]: -0.029 [0.049]
- Digital capital raising:
  - [2]: -5.468 [5.424]
- Total fintech:
  - [3]: -0.063 [0.042]
- Real GDP per capita:
  - [1]: -0.124* [0.045]
  - [2]: -0.048*** [0.014]
  - [3]: -0.043 [0.022]
- Real GDP growth:
  - [1]: 0.006 [0.004]
  - [2]: 0.006 [0.007]
  - [3]: 0.008 [0.007]
- Inflation:
  - [1]: -0.015*** [0.004]
  - [2]: -0.001*** [0.000]
  - [3]: -0.001*** [0.000]
- Trade openness:
  - [1]: 0.001 [0.001]
  - [2]: 0.001 [0.001]
  - [3]: 0.001 [0.001]
- Domestic credit to the private sector:
  - [1]: 0.003*** [0.000]
  - [2]: 0.003*** [0.001]
  - [3]: 0.003*** [0.001]
- Government stability:
  - [1]: 0.047*** [0.015]
  - [2]: 0.027 [0.014]
  - [3]: 0.021 [0.013]
- Bureacratic quality:
  - [1]: 0.103*** [0.028]
  - [2]: 0.127*** [0.014]
  - [3]: 0.120 [0.026]
- Number of observations: 280, 520, 528 (for columns [1], [2], [3] respectively)
- Number of countries: 65, 82, 82
- Country FE: Yes (all columns)
- Time FE: Yes (all columns)
- R2:
  - [1]: 0.09
  - [2]: 0.07
  - [3]: 0.07

### Main empirical findings and interpretation
- Dataset and scope:
  - Cross-country dataset covering 198 countries over the period 2012–2020 (fintech excluding cryptocurrencies).
- Principal empirical results:
  - Impact of fintech depends on instrument type:
    - Digital lending (share of GDP): statistically insignificant negative effect on the z-score.
    - Digital capital raising (share of GDP): large and statistically significant positive effect on financial stability (in pooled and advanced-economy estimations).
  - Aggregate effect of all fintech instruments together:
    - Negative but statistically insignificant because digital lending overwhelmingly dominates total fintech.
  - Heterogeneity by country income group:
    - Digital capital raising: positive effect in advanced economies; negative effect in developing countries.
  - Interpretation:
    - Fintech-facilitated lending may involve greater financial risk due to concentration and over-reliance on data-driven algorithms.
    - Fintech-enabled capital raising may decentralize and diversify risk in the financial system, at least in advanced economies.
  - Magnitude context:
    - Average volume of fintech instruments amounts to 0.1 percent of GDP during the period 2012–2020.
    - Domestic credit to the private sector averages 55 percent of GDP over the same period.

### Broader context, trends, and risks
- Fintech investment growth:
  - Total value of investments into fintech worldwide increased from US$1 billion in 2008 to US$247 billion in the first half of 2023.
- Potential benefits of fintech:
  - Make financial systems more efficient and competitive.
  - Broaden financial inclusion for under-served populations.
  - Enhance decentralization, diversification, market depth, efficiency, and transparency.
- Potential risks and vulnerabilities:
  - Exposure to cybersecurity threats.
  - Potential amplification of market volatility.
  - Increased risk-taking and contagious behavior among consumers and financial institutions.
  - Systemic financial risks can arise from institutions that individually are not systemically important.
  - New financial technologies with complex network structures, especially on the lending front, are yet to be tested in economic downturns.
- Additional dynamic:
  - Fintech firms tend to take on more risks themselves and exert pressure on traditional financial institutions to engage in riskier operations.
  - Entry and growth of big-tech firms into finance may require more entity-based regulation.

### Policy implications and recommendations
- Regulatory and supervisory priorities:
  - Proactively assess the adequacy of regulatory frameworks for fintech to harness benefits while mitigating risks to financial stability.
  - Modernize legal principles and macroprudential policies; consider expanding the scope of existing regulations to prevent buildup of systemic risks from fast-growing fintech.
  - Ensure a level playing field and effective monitoring and supervision of traditional and emerging financial institutions.
  - Strengthen regulatory institutions and better use technology in regulation.
  - Extensive cross-border coordination and collaboration to develop common standards and regulatory principles given the international nature of fintech.
- Specific consideration:
  - Entry of big-tech firms into the financial sector, especially at a global scale, may require more entity-based regulation.

*Source: Author's estimations.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023253-print-pdf.pdf_