## wpiea2019085

## Source details

**Canonical URL:** [wpiea2019085](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019085.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019085.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019085.pdf.json)

---

### I. Introduction — context and motivation
- Past decades show large swings in international capital flows, notably linked to the global financial crisis (GFC).
- Global financial factors and risk conditions are established drivers of cross-border flows (Forbes and Warnock 2012; Passari and Rey 2015; Bruno and Shin 2017; IMF 2013, 2016a, 2017).
- Cross-country variation in capital flow volatility driven by global factors is less well understood and has important macroeconomic and financial stability implications (Cerutti et al. 2014).
- Focus of the paper:
  - Role of financial institutions’ balance sheets and portfolio and risk management practices in driving the response of capital flows to global risk conditions.
  - Emphasis on nonbank financial institutions (pension and insurance funds, investment funds and asset managers) as potential drivers of capital flow sensitivity to global risk.
  - Use of IMF Standardized Report Form (SRF) dataset decomposing resident banks’ and nonbank financial institutions’ balance sheets into domestic and external positions.

### Key empirical findings
- Statistically significant association between nonbank financial institutions’ foreign exposures (degree to which nonbanks are net long in foreign assets) and a country’s sensitivity of capital flows to global risk (VIX).
- Advanced-economy subset:
  - Association remains significant when considering only bank-related capital flows.
- Emerging-market sample:
  - Association between bank-related flows and nonbank foreign exposures is not significant.
- No significant association found between cross-border capital flows’ response to global risk factors and on-balance sheet exposures of banks.

### Data and measurement (SRF and flows)
- SRF provides aggregate monetary and financial balance sheet (stock) positions for central bank, other depository corporations (ODCs — banks), and other financial institutions (OFCs — nonbanks).
- Net external position = external assets − external liabilities; positions available by instrument; specific currency denomination of external positions not available; off-balance-sheet positions not observed.
- Flow measures (exact definitions preserved):
  - Net capital inflows = 2* NCAPFLP / (private external assets + private external liabilities).
  - Gross non-resident inflows = ICAPFLP / private external liabilities.
  - Gross resident inflows = (-1) * OCAPFLP / private external assets.
  - Net bank inflows = 2*NOTHFB / (private external assets + private external liabilities).
- VIX: CBOE Volatility index, end of period monthly.
- Sample and frequency:
  - Quarterly focus with a sample of 24 countries (3 Advanced Economies: US, Japan, Sweden; 21 Emerging Markets).
  - Subsample of eight countries with full coverage 2001Q4–2017Q4 for selected statistics.

### Stylized facts from SRF (Section IV)
- Most countries have bank-dominated financial systems:
  - Bank balance sheet share exceeds 0.6 in most countries, with large cross-country variation.
  - United States and South Africa have bank shares below 50%; some emerging markets have banks > 80% of total balance sheet.
- Countries with large financial sectors have larger nonbank sectors:
  - Financial depth = financial institutions’ balance sheet / GDP.
  - Negative association: greater financial depth → smaller bank share (larger nonbank share).
- Countries with large nonbank sectors tend to be net long in foreign exposure:
  - Net foreign exposure = (external assets − external liabilities) / total assets for each sector.
  - Less variation in banks’ foreign exposures than in nonbanks’.
  - Negative relationship between bank share and nonbanks’ net long position.

### Empirical strategy
- Two complementary approaches:
  1. Cross-country patterns during the GFC: compare average quarterly private capital flows in 2008Q4–2009Q1 to 2006Q4–2007Q1 (Bolivia uses three years earlier).
  2. Fixed effects panel regressions (2002Q1–2017Q4) relating capital flows (share of total foreign positions) to dlog(VIX) lagged and interaction dlog(VIX) × NETEXP_j,t-1 for j = total, banks, nonbanks.
- Dependent variables: net capital inflows, net bank inflows, gross non-resident inflows, gross resident inflows (resident gross outflows × −1), and EMP.
- Controls include d(i), d(i*SR), and country fixed effects; robustness checks include financial depth, pegged exchange rate, capital controls (Chinn-Ito), exclusion of oil exporters, exclusion of Advanced Economies, exclusion of the GFC, and time fixed effects.

### Selected regression coefficients and statistics (preserving reported values)
- Regression sample period: 2002Q1 to 2017Q4.
- Typical panel observation counts (Table 2): Nobs reported as 988, 984, 988, 988, 993 across dependent variables; No.cross sections typically 24 (one column shows 23 for EMP).
- From Table 2, panel (c) Net exposure for nonbanks (selected estimates):
  - dlog(VIX): net capital inflows −0.003 (0.003); net bank inflows 0.001 (0.002); gross non-resident inflows −0.009*** (0.003); gross resident inflows 0.011** (0.004); EMP −0.040*** (0.014).
  - dlog(VIX)_NETEXP_NB,t-1 = 0.050** (0.021) in the net capital inflows regression.
  - NETEXP_NB,t-1 = −0.002 (0.037).
- Robustness (Table 3) — baseline interaction remains significant across specifications:
  - Example: Table 3, Column I: dlog(VIX)_NETEXP_NB,t-1 = 0.050** (0.021).
  - Table 3, Column IX (with time effects): dlog(VIX)_NETEXP_NB,t-1 = 0.046** (0.022); R_squared 0.11; Nobs 988; No.cross sections 24.
- Additional selected coefficients from robustness and appendix tables (exact values preserved):
  - Table A3 (Total financial institutions, 2002Q1-2017Q4): dlog(VIX) coefficients include -0.003, -0.003, -0.000, -0.004, -0.007**, -0.004*, -0.003, -0.005*; dlog(VIX)_NETEXP_FC,t-1 = 0.068*, 0.056, 0.079**, 0.077, 0.040, 0.065*, 0.089**, 0.071** across columns; Nobs include 987, 919, 987, 986, 967, 987, 805, 959.
  - Table A4 (Bank exposures, 2002Q1-2017Q4): dlog(VIX) = -0.002, -0.001, -0.001, -0.001, -0.006*, -0.003, -0.002, -0.004; dlog(VIX)_NETEXP_B,t-1 = 0.063, 0.048, 0.066, 0.061, 0.019, 0.065, 0.069, 0.060; Nobs include 988, 919, 988, 987, 968, 988, 806, 960.
  - Table A5 (Nonbank exposures, 2006Q4-2017Q4): dlog(VIX) = -0.003, -0.002, 0.002, -0.003, -0.007**, -0.003, -0.002, -0.004; dlog(VIX)_NETEXP_NB,t-1 = 0.050**, 0.048*, 0.060**, 0.052*, 0.045**, 0.048**, 0.079***, 0.046*; Nobs include 825, 757, 825, 825, 805, 825, 699, 797.
  - Table A6 (Nonbank exposures, 2009Q3-2017Q4): dlog(VIX) = -0.006*, -0.007*, -0.003, -0.006, -0.007*, -0.006*, -0.005, -0.006*; dlog(VIX)_NETEXP_NB,t-1 = 0.054**, 0.055**, 0.062***, 0.056**, 0.053**, 0.048**, 0.071***, 0.054**; Nobs include 688, 621, 688, 688, 675, 688, 595, 688.
  - EM sample (Table A7, panel c nonbanks): dlog(VIX) = -0.002, 0.001, -0.011***, 0.013***, -0.043**; dlog(VIX)_NB,t-1 = 0.078***, 0.024*, 0.028**, 0.041*, 0.217***; Nobs include 806, 802, 806, 806, 811.
  - Excluding the United States (Table A8, nonbanks): dlog(VIX)_NB,t-1 = 0.051**, 0.021**, 0.022*, 0.019, 0.176***; Nobs include 927, 923, 927, 927, 932.

### Interpretation and mechanism (Section VI)
- Nonbank risk management practices as a transmission channel:
  - Nonbanks often hold net long foreign asset positions and dynamically manage risk (e.g., value-at-risk driven hedging).
  - When global or currency risk rises, nonbanks increase hedging (sell foreign assets or enter derivatives like forwards).
  - Nonbanks typically hedge with domestic banks (forwards off-balance-sheet for nonbanks).
  - Banks, as market makers, take the other side of derivatives and hedge by borrowing foreign currency or adjusting foreign asset holdings — these hedges are recorded on-bank balance sheets and as cross-border flows.
  - Result: observed bank cross-border flows (often gross non-resident inflows or bank retrenchment) can reflect nonbank-driven hedging demand rather than banks’ own portfolio rebalancing.
- Stylized transaction example (Figure 6 summary):
  - Resident nonbank sells foreign currency forward to resident bank (off-balance-sheet for nonbank).
  - Resident bank hedges by borrowing foreign currency in the interbank market (on-balance-sheet recorded as gross non-resident bank inflow).
  - At maturity, bank repays borrowings using foreign currency received from the nonbank or reduces foreign assets.
- Case studies:
  - Denmark (early 2015): institutional investors increased hedging via kroner forward purchases from Danish banks after Swiss franc shock; banks hedged by selling euros spot against kroner; transactions recorded as bank-related capital inflows; Danmarks Nationalbank intervened.
  - Korea: exporters, importers, AMCs, insurance funds, and national pension fund generated forward market dynamics; banks acted as market makers and sourced USD in interbank markets to hedge forwards; maturity mismatches during the GFC produced USD squeezes.
- Indicators of bank–nonbank interconnectedness:
  - FSB Global Shadow Banking Monitoring Report 2017 shows banks’ exposure to nonbanks (sum of assets and liabilities) around 20 percent of total bank assets in Brazil and South Africa; interconnectedness high in Indonesia, Turkey, South Africa.

### Sector heterogeneity and patterns
- Nonbank foreign exposure matters more than bank foreign exposure in conditioning capital flow sensitivity to the VIX.
- Interaction effects indicate:
  - When VIX increases, larger nonbank foreign exposure significantly increases total net capital inflow response and the response of net bank-related inflows.
  - Interaction term significant in gross non-resident inflow regressions but not in gross resident inflows — implying adjustments primarily via changes in foreign liabilities (gross non-resident inflows) rather than on-balance-sheet foreign asset repatriations.
- Cross-sample differences:
  - Association between nonbank exposures and sensitivity of bank-specific flows to the VIX is driven by advanced economies, especially the US.
  - Nonbank foreign exposures matter more than bank exposures for sensitivity of capital flows even when advanced economies are excluded.

### Policy implications and recommendations (Section VII)
- Institutional structure shifts (growing nonbank sector) can alter capital flow volatility and sensitivity to global risk.
- Regulatory and macroprudential implications:
  - Nonbanks’ foreign exposures and risk management practices can be significant drivers of capital flows, even when flows are observed primarily in bank credit flows.
  - Policy perimeter could be extended to include:
    - Foreign exposures of nonbank financial institutions.
    - Risk management practices of nonbank financial institutions, particularly where nonbanks have greater exposures to foreign investments or funding.
  - Measures should account for cross-country heterogeneity in bank–nonbank institutional relations and be designed within country-specific institutional contexts.
- Data needs:
  - More detailed and complete data on financial system structure and balance sheets (including off-balance-sheet positions) are needed.
  - SRF provides unique perspective on nonbank balance sheets but the sample is small and off-balance-sheet positions are not recorded.
- Suggested further work:
  - Identify country-specific links between capital flows and bank and nonbank foreign exposures to better inform policy design.

*Source: wpiea2019085 - References / IMF working paper section text as provided.*

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

### wpiea2019085 - References .............................................................................................................

### I. INTRODUCTION — context and motivation
- The past decades have witnessed large swings in international capital flows, notably in connection with the global financial crisis (GFC).
- Global financial factors and risk conditions are established drivers of cross-border flows (Forbes and Warnock 2012; Passari and Rey 2015; Bruno and Shin 2017; IMF 2013, 2016a, 2017).
- Cross-country variation in capital flow volatility driven by global factors is less well understood and has important macroeconomic and financial stability implications (Cerutti et al. 2014).

### Focus of the paper
- Investigates the role of financial institutions’ balance sheets and portfolio and risk management practices in driving the response of capital flows to global risk conditions.
- Emphasizes nonbank financial institutions (pension and insurance funds, investment funds and asset managers) as potential drivers of capital flow sensitivity to global risk.
- Uses a previously unexplored dataset: the IMF’s Standardized Report Form (SRF), which decomposes resident banks’ and nonbank financial institutions’ balance sheets into domestic and external positions.

### Key empirical findings
- There is a statistically significant association between nonbank financial institutions’ foreign exposures (defined as the degree to which nonbanks are net long in foreign assets) and a country’s sensitivity of capital flows to global risk factors, as captured by the VIX.
- For advanced economies in the dataset:
  - The association remains significant when considering only bank-related capital flows.
- For the emerging market sample:
  - The association between bank-related flows and nonbank foreign exposures is not significant.
- No significant association is found between the response of cross-border capital flows to global risk factors and on-balance sheet exposures of banks.

### Mechanisms and interpretation
- Nonbanks are likely to adjust portfolio risk exposures dynamically in response to increasing perceived foreign risks, generating capital flows.
- Typical hedging practices by nonbanks through derivatives with correspondent banks can generate bank-recorded capital flows driven by nonbank portfolio decisions:
  - Nonbanks adjust foreign currency risks off-balance sheet through derivatives contracts with banks.
  - Banks provide OTC foreign currency forwards and swaps to nonbank clients at market rates, hedge these positions in cross-border interbank markets, and record on-balance-sheet adjustments (bank credit-related capital flows).
  - Thus, observed bank flows can be driven by nonbank portfolio risk management rather than banks’ own portfolio management.
- The bank–nonbank nexus as a driver of capital flows is less well understood due to data limitations.

### Supporting evidence and case studies
- The study presents two case studies (Denmark and Korea) showing instances where bank flows were driven by nonbank portfolio adjustments through derivatives positions, rather than by banks’ own portfolio management behavior.
- Prior literature and examples cited:
  - Denmark’s experience during currency pressure in early 2015 suggests banks may respond to and hedge customer FX derivatives demands by changing cross-border positions without changing their own total exposures (Danmarks Nationalbank 2015).
  - Shin (2012) and FSB (2018), Borio et al. (2017) point to broader interactions between banks and nonbanks.

### Policy implications and recommendations
- Nonbanks’ foreign exposures and risk management practices can be significant drivers of capital flows, even when the resulting flows are observed primarily in bank credit flows.
- In countries where capital flow volatility causes disruptive foreign exchange market conditions and where capital flow management measures are considered, the policy perimeter could be extended to include:
  - Foreign exposures of nonbank financial institutions.
  - Risk management practices of nonbank financial institutions, particularly when nonbanks have greater exposures to foreign investments or funding.
- Policy responses should account for significant cross-country heterogeneity in bank–nonbank institutional relations; measures should be designed within country-specific institutional contexts.
- Additional work is desirable to identify country-specific links between capital flows and bank and nonbank foreign exposures to better inform policy design.

*Source: wpiea2019085 - References*

### Section III presents the SRF dataset, and Section IV presents some key stylized facts on the

### wpiea2019085 - Section III presents the SRF dataset, and Section IV presents some key stylized facts on the

### Previous literature
- Financial intermediaries optimize expected returns and risks subject to constraints; the mix of intermediaries can have real effects (Bernanke and Gertler 1989; Kiyotaki and Moore 1997; Adrian and Shin 2010; Adrian and Boyarchenko 2013).
- Prior empirical work has focused mainly on banks (Levchenko and Mauro 2007; Milesi-Ferretti and Tille 2011; Cetorelli and Goldberg 2011, 2012; Bruno and Shin 2015b; Krogstrup and Tille 2018).
- Findings from the literature:
  - Bank-intermediated capital flows are more prone to sudden stops and were main drivers of retrenchment during the global financial crisis.
  - Bank pre-existing balance sheet exposure to global risks conditions the responsiveness of cross-border bank funding flows to global risk.
  - Leverage constraints and other behaviors/constraints affect balance sheet responses to global risk (Miranda-Agrippino and Rey 2018; Devereux and Yetman 2010).
- Gaps: previous work does not consider bank–nonbank interactions and the fact many bank balance sheet operations happen on behalf of nonbank clients.
- Recent contributions examine investor-type mix and sensitivity to global factors (Cerutti, Claessens and Puy 2015; IMF 2014), finding that larger pension funds reduce the response of domestic asset prices and the exchange rate to global factors.

### The SRF dataset (Section III)
- Data source: IMF Standardized Report Form (SRF).
- SRF provides aggregate monetary and financial balance sheet (stock) positions for:
  - central bank,
  - other depository corporations (ODCs — banks),
  - other financial institutions (OFCs — nonbanks).
- Bank category includes commercial banks, credit unions, and other deposit-taking corporations.
- Nonbanks include institutional investors such as pension and insurance funds, mutual funds and other financial institutions.
- Key dataset features:
  - Assets and liabilities divided into domestic and external positions (external = positions with foreign counterparties).
  - Net external position = claims on nonresidents (external assets) minus liabilities to nonresidents (external liabilities).
  - All positions available by instrument; specific currency denomination of external positions is not available.
  - ODC and OFC data are consolidated within the sector.
- Frequency and coverage:
  - Data available monthly or quarterly; banks data from December 2001 for most countries; nonbank data becomes available later in many countries.
  - Focused sample: quarterly frequency, countries and sample periods with good nonbank coverage, yielding a sample of 24 countries (3 Advanced Economies: US, Japan, Sweden; 21 Emerging Markets).
  - Subsample of eight countries with full coverage 2001Q4–2017Q4 used for selected summary statistics.
- Notes on measurement:
  - All bank balance sheet data from SRF are consolidated within the sector.
  - Off-balance sheet positions and currency denomination of external positions are not observed.

### Stylized facts (Section IV)
A. Most countries have bank-dominated financial systems
- Bank balance sheet share exceeds 0.6 in most countries, with large cross-country variation.
- Advanced economies generally have less bank-dominated systems.
- Examples: United States and South Africa have bank shares below 50%; some emerging markets have banks comprising more than 80% of financial institutions’ total balance sheet.
- Banks tend to hold a higher share of financial institutions’ external balance sheet; exceptions include Chile, Colombia and Thailand where bank shares are lower in external than domestic positions.

B. Countries with large financial sectors have larger nonbank sectors
- Financial depth defined as financial institutions’ balance sheet as a share of GDP.
- Clear negative association: the greater the financial depth, the smaller the bank share (i.e., larger nonbank share).
- Relationship is steeper and equally significant within the EM sample alone.

C. Countries with large nonbank sectors tend to be net long in foreign exposure
- Net foreign exposure measured as (external assets − external liabilities) / total assets for each sector.
- Less variation in banks’ foreign exposures than in nonbank foreign exposures across countries.
- Negative relationship between bank share and nonbanks’ net long position: more nonbank-dominated systems tend to have nonbanks with larger net long foreign exposures.
- Implication: well-developed financial systems with large nonbank sectors tend to have nonbanks holding more foreign assets than foreign funding.

### The link between capital flows and net foreign exposures (Sections V.A–V.D)

Data and measurement
- Balance sheet exposures: SRF dataset.
- Capital flows: IMF Financial Flows Analytics database; focus on non-official (private) flows.
- Flow scaling approach:
  - Gross outflows (active resident adjustments of foreign assets) scaled by gross private foreign assets.
  - Gross private inflows (foreign residents’ adjustments of foreign liabilities) scaled by gross private foreign liabilities.
  - Net capital inflows scaled by the average of gross private inflows and outflows.
- Exchange rate regime considerations:
  - Use of an exchange market pressure index (EMP) constructed in Goldberg and Krogstrup (2018) for robustness.
  - Results robust to including exchange rate regime in regressions.

Empirical strategy
- Two complementary approaches:
  1. Cross-country patterns during the Global Financial Crisis (GFC): compare average quarterly private capital flows in 2008Q4–2009Q1 to the same quarters two years earlier (2006Q4–2007Q1); for Bolivia used three years earlier due to anomalies.
  2. Fixed effects panel regressions relating capital flows (share of total foreign positions) to VIX and interaction of VIX with net foreign exposure of financial institutions.
- Regression specification (abbreviated):
  - Dependent variables: five capital flow measures — net capital inflows, net bank inflows, gross non-resident inflows, gross resident inflows (resident gross outflows multiplied by minus one), and EMP.
  - Main regressors: dlog(VIX) lagged; interaction dlog(VIX) × NETEXP_j,t-1 for j = total, banks, nonbanks; NETEXP_j,t-1; controls including d(i), d(i*SR), and other country fixed effects.
- Regression sample period: 2002Q1 to 2017Q4.

Key numerical information from regressions (Table 2 / Table 3)
- Regression sample period: 2002Q1 to 2017Q4.
- Typical panel regression observation counts (Table 2):
  - Nobs reported as 988, 984, 988, 988, 993 across dependent variables in one table panel.
  - No.cross sections typically 24 (one column shows 23 for EMP).
- Selected coefficient estimates (from Table 2, panel (c) Net exposure for nonbanks):
  - dlog(VIX) coefficients: net capital inflows −0.003 (standard error 0.003); net bank inflows 0.001 (0.002); gross non-resident inflows −0.009*** (0.003); gross resident inflows 0.011** (0.004); EMP −0.040*** (0.014).
  - dlog(VIX)_NETEXP_NB,t-1 = 0.050** (standard error 0.021) in the net capital inflows regression (significant).
  - NETEXP_NB,t-1 = −0.002 (0.037) (not significant in that column).
- Robustness (Table 3):
  - Baseline interaction term dlog(VIX)_NETEXP_NB,t-1 remains significant across specifications including:
    - Excluding oil-exporting countries (Column II).
    - Controlling for financial depth (Column III).
    - Controlling for net total foreign asset position in percent of total assets and liabilities (Column IV).
    - Controlling for pegged exchange rate (Column V) and capital controls (Chinn-Ito index) (Column VI).
    - Excluding Advanced Economies (Column VII) — interaction estimate slightly larger and more significant.
    - Excluding the GFC observations (Column VIII) — interaction survives.
    - Including time fixed effects (Column IX) — interaction survives.
  - Example coefficient from Table 3, Column I: dlog(VIX)_NETEXP_NB,t-1 = 0.050** (0.021).
  - Example from Table 3, Column IX (with time effects): dlog(VIX)_NETEXP_NB,t-1 = 0.046** (0.022); R_squared 0.11; Nobs 988; No.cross sections 24.

Main empirical findings
- VIX alone: significant for gross flows but not for net flows (consistent with offsetting repatriation and retrenchment).
- Interaction effect:
  - Net exposure to foreign assets increases the response of net capital flows to the VIX and makes the coefficient significant.
  - Interpretation: when resident financial institutions are net long in foreign exposure, net capital inflows increase in response to higher global risk factors; when net short, inflows decline.
- Sector heterogeneity:
  - Nonbank foreign exposure matters more than bank foreign exposure in conditioning capital flow sensitivity to the VIX.
  - When VIX increases, larger nonbank foreign exposure significantly increases total net capital inflow response and the response of net bank-related inflows.
  - Interaction term significant in gross non-resident inflow regressions but not in gross resident inflows — implying adjustments primarily via changes in foreign liabilities (gross non-resident inflows) rather than on-balance-sheet foreign asset repatriations.
- Cross-sample differences:
  - Association between nonbank exposures and sensitivity of bank-specific flows to the VIX is driven by advanced economies, especially the US.
  - Nonbank foreign exposures matter more than bank exposures for sensitivity of capital flows even when advanced economies are excluded.

### Interpretation: nonbank risk management practices & bank flows (Section VI)
- Mechanism proposed:
  - Nonbanks (institutional investors) often hold net long foreign asset positions and use dynamic risk management (e.g., value-at-risk driven hedging).
  - When global risk or currency risk rises, nonbanks increase hedging (sell foreign assets or enter derivatives like forwards).
  - Nonbanks frequently hedge with domestic banks (enter forwards), which are off-balance-sheet from the nonbank perspective.
  - Banks, as market makers, take the other side of forwards and hedge by borrowing foreign currency in interbank spot markets or adjusting foreign asset holdings — these hedging transactions are on-bank balance sheets and recorded as capital flows (often gross non-resident inflows or bank retrenchment).
  - Thus observed bank cross-border flows can reflect nonbank-driven hedging demand rather than primary bank portfolio rebalancing.
- Stylized transaction example (Figure 6 summary):
  - Resident nonbank with net long foreign assets sells foreign currency forward to resident bank (off-balance-sheet for nonbank).
  - Resident bank hedges by borrowing foreign currency in the interbank market (on-balance-sheet recorded as gross non-resident bank inflow).
  - When forward matures, bank repays borrowings using foreign currency received from the nonbank, or reduces foreign assets.
- Case studies:
  - Denmark (2015): institutional investors increased hedging via kroner forward purchases from Danish banks following Swiss franc shock; banks hedged by selling euros spot against kroner; transactions recorded as bank-related capital inflows; Danmarks Nationalbank intervened heavily.
  - Korea: exporters, importers, AMCs, insurance funds, and national pension fund create forward market dynamics; banks act as market makers and source USD in interbank markets to hedge forwards, linking banks’ cross-border funding to nonbank hedging demand; maturity mismatches during the GFC produced USD squeezes.
- Indicators: FSB Global Shadow Banking Monitoring Report 2017 shows banks’ exposure to nonbanks (sum of assets and liabilities) around 20 percent of total bank assets in Brazil and South Africa; interconnectedness high in several sample countries (Indonesia, Turkey, South Africa).

### Policy implications and conclusions (Section VII)
- Shifts in financial institutional structure (growing nonbank sector) can change capital flow volatility and sensitivity to global risk.
- Regulation and supervision:
  - Risk management practices of nonbank financial institutions can directly affect capital flow volatility through the described mechanisms.
  - Macroprudential and capital flow management measures could, in some cases, beneficially be extended to include regulation of foreign exposures and risk management practices of nonbank financial institutions.
- Data needs:
  - More detailed and complete data on financial system structure and balance sheets (including off-balance-sheet positions) are needed to analyze drivers of cross-border capital flows.
  - The SRF dataset provides a unique perspective on nonbank balance sheets but sample is small and off-balance-sheet positions are not recorded.
- Summary conclusion:
  - Panel regression evidence shows cross-border capital flow responses to global financial conditions (VIX) are associated with foreign exposures in nonbank balance sheets.
  - These flow responses are reflected mainly in gross non-resident inflows and in bank flows in advanced economies, consistent with nonbank-driven hedging transmitted via banks.

*Source: IMF working paper section text as provided.*

### Chapter 2 in Global Financial Stability Report, October 2016.

### wpiea2019085 - Chapter 2 in Global Financial Stability Report, October 2016

### Data coverage and variable definitions
- Country coverage and data frequency mix: examples include United States (US) with start 2001Q4; Sweden (SE) 2001M12; Japan (JP) 2001M12; Turkey (TR) 2002M12 to 2008Q1; sample contains monthly (M) and quarterly (Q) observations and notes ODC: M; OFC: Q for some countries.
- Key constructed variables (exact definitions preserved):
  - Net capital inflows = 2* NCAPFLP/ (private external assets + private external liabilities). NCAPFLP is total net nonofficial inflows; ICAPFLP total gross nonofficial inflows; OCAPFLP total gross nonofficial outflows; NOTHFB = other inflows to banks minus other outflows to banks. Source: IMF, International Investment Positions (IIP) and Financial Flows Analytics Database (FFA).
  - Gross non-resident inflows = ICAPFLP / private external liabilities.
  - Gross resident inflows = (-1) * OCAPFLP / private external assets.
  - Net bank inflows = 2*NOTHFB/(private external assets + private external liabilities).
  - NETEXP_FC (Net exposure for total financial corporations) = (total financial institutions external assets – total financial institutions external liabilities) / total financial institutions total assets. Frequency: Quarterly. Source: IMF Standardized Report Form (SRF).
  - NETEXP_B (Net exposure for banks) = (Banks external assets – Banks external liabilities) / Banks total assets. Quarterly. Source: SRF.
  - NETEXP_NB (Net exposure for nonbanks) = (Nonbanks external assets – Nonbanks external liabilities) / Nonbanks total assets. Quarterly. Source: SRF.
  - VIX: CBOE Volatility index, end of period monthly; extended backwards using VXO from 1986m1 to 1989m12. Source: Chicago Board Options Exchange.
  - i (Monetary policy or short-term rate): in percentage points, end of period, monthly. Constructed from IFS policy rate line 60, national central bank policy rate, 3-month money market interest rate (IFS line 60b), or short-term treasury bond rate (IFS line 60c) as available. Source: IMF International Financial Statistics or national central banks.
  - i*SR (Shadow policy rate in US, EU, Japan and UK): in percentage points, end of period, monthly. Source: Krippner [2016].
  - Financial depth (total) = (Bank domestic assets + Nonbank domestic assets) / GDP (Quarterly). Source: SRF.
  - NFA (Net financial assets) = 2*(external assets – external liabilities) / (external assets + external liabilities). Source: IMF International Investment Positions (IIP).
  - peg: Standard Shambaugh measure of pegs, 1=peg, 0=nonpeg, yearly. Source: Klein and Shambaugh [2008].
  - kaopen: Chinn-Ito index, yearly. Source: Aizenman, Chinn, Ito [2015].
  - dB_FA, dNB_FA: quarterly first differences of banks’ and nonbanks’ foreign asset positions (B_FA, NB_FA). Source: SRF.
  - de_eur_usd, de_i_usd: quarterly change in exchange rates defined as change between end of last quarter and end of current quarter, divided by the change in the end-of-last quarter exchange rate. Source: IMF Global Assumptions Data (GAS).

### Financial depth and balance sheet structure
- Financial depth measures (calculated as described in source):
  - Financial depth for total = (Bank total assets + Bank total liabilities + Nonbank total assets + Nonbank total liabilities) / GDP.
  - Financial depth for domestic = (Bank domestic assets + Bank domestic liabilities + Nonbank domestic assets + Nonbank domestic liabilities) / GDP.
  - Financial depth for external = (Bank external assets + Bank external liabilities + Nonbank external assets + Nonbank external liabilities) / GDP.
- Empirical observation from charts:
  - "Financial depth based on total balance sheet is almost similar (and identical) to financial depths calculated using domestic positions." This implies domestic developments helped financial deepening.
  - Bank share of total assets is plotted against financial depth (domestic and external) using five-year averages; country labels include US, SE, JP, TR, ZA, BO, BR, CL, CO, DO, EC, SV, HN, ID, TH, AM, BY, AL, GE, KZ, MD, UA, MK, RO.
  - Numerical axis values shown in figures: bank share total spans .2 .4 .6 .8 1; financial depth domestic values include 0 5 10 15 22.5; financial depth external includes values 0.5 1.0 1.5 (as presented in figure annotations).

### Net foreign exposure by financial institutions
- Measurement (exact formulas preserved):
  - Net exposure for banks = (Banks external assets – Banks external liabilities) / Banks total assets.
  - Net exposure for nonbanks = (Nonbanks external assets – Nonbanks external liabilities) / Nonbanks total assets.
  - Net exposure for total financial institutions = (total financial institutions external assets – total financial institutions external liabilities) / total financial institutions total assets.
- Stylized patterns documented:
  - "Banks are more balanced where nonbanks share is larger, and nonbanks have larger external asset share in their balance sheets."
  - Interpretation: banks often fund externally and lend domestically; nonbanks are often domestically funded but partly invest abroad.
  - Pattern reverses where banks dominate the financial system.

### Robustness checks and key regression results (selected coefficients and statistics preserved exactly)
- General note on tables: regressions use robust clustered standard errors; asterisks *, **, *** indicate significance at the 10, 5 percent and 1 percent levels, respectively. Fixed effects included. Sample periods vary by table.
- Table A3 (Total financial institutions exposures on capital flows, 2002Q1-2017Q4) — selected coefficients across columns I–VIII:
  - dlog(VIX) coefficients: -0.003, -0.003, -0.000, -0.004, -0.007**, -0.004*, -0.003, -0.005* (standard errors shown in parentheses in source).
  - dlog(VIX)_NETEXP_FC, t-1 coefficients: 0.068*, 0.056, 0.079**, 0.077, 0.040, 0.065*, 0.089**, 0.071**.
  - NETEXP_FC, t-1 coefficients: 0.069, 0.060, 0.070, 0.065, 0.076, 0.070, 0.075, 0.056.
  - d(i) coefficients: 0.005 across columns (0.005 or 0.004 in last column).
  - d(i*SR) coefficients: -0.006, -0.006, -0.005, -0.006, -0.006, -0.006, -0.007, -0.005.
  - R_squared values: 0.02 for columns I–VII, 0.01 for VIII. Adj. R_squared: 0.01 or 0.02 or 0.01 depending on column.
  - Nobs across columns: 987, 919, 987, 986, 967, 987, 805, 959. No. cross sections: 24, 21, 24, 23, 24, 21, 24, 24.
- Table A4 (Bank foreign exposures on capital flows, 2002Q1-2017Q4) — selected coefficients:
  - dlog(VIX): -0.002, -0.001, -0.001, -0.001, -0.006*, -0.003, -0.002, -0.004.
  - dlog(VIX)_NETEXP_B, t-1: 0.063, 0.048, 0.066, 0.061, 0.019, 0.065, 0.069, 0.060.
  - NETEXP_B, t-1: 0.072, 0.064, 0.073, 0.067, 0.073, 0.076, 0.077, 0.059.
  - d(i): 0.005 across columns (0.004 in last column).
  - d(i*SR): -0.006, -0.006, -0.005, -0.006, -0.006, -0.006, -0.007, -0.005.
  - R_squared: 0.02 for columns I–VII, 0.01 for VIII. Nobs: 988, 919, 988, 987, 968, 988, 806, 960.
- Table A5 (Nonbank foreign exposures on capital flows, 2006Q4-2017Q4) — selected coefficients:
  - dlog(VIX): -0.003, -0.002, 0.002, -0.003, -0.007**, -0.003, -0.002, -0.004.
  - dlog(VIX)_NETEXP_NB, t-1: 0.050**, 0.048*, 0.060**, 0.052*, 0.045**, 0.048**, 0.079***, 0.046*.
  - NETEXP_NB, t-1: -0.023, -0.015, -0.008, -0.023, 0.011, -0.023, -0.022, -0.044.
  - d(i): 0.005 or 0.006 across columns.
  - d(i*SR): -0.006, -0.006, -0.001, -0.006, -0.007, -0.006, -0.008, -0.006.
  - findepth_t-1: -0.025** reported in a column (standard error 0.010).
  - R_squared: 0.02 or 0.04 depending on column. Nobs: 825, 757, 825, 825, 805, 825, 699, 797.
- Table A6 (Nonbank foreign exposures, 2009Q3-2017Q4) — selected coefficients:
  - dlog(VIX): -0.006*, -0.007*, -0.003, -0.006, -0.007*, -0.006*, -0.005, -0.006*.
  - dlog(VIX)_NETEXP_NB, t-1: 0.054**, 0.055**, 0.062***, 0.056**, 0.053**, 0.048**, 0.071***, 0.054**.
  - findepth_t-1: -0.044*** (standard error 0.012) in a reported column.
  - R_squared values: 0.01–0.05 across columns. Nobs: 688, 621, 688, 688, 675, 688, 595, 688.
- Panel regressions for EM country sample (Table A7) — selected coefficients (EM sample, 2002Q1-2017Q4):
  - For total financial institutions:
    - dlog(VIX): -0.003, -0.001, -0.010***, 0.011**, -0.044** (coefficients across columns for dependent variables net capital inflows, net bank inflows, gross non-resident inflows, gross resident inflows, EMP).
    - dlog(VIX)_NETEXP_FC, t-1: 0.088**, -0.020, 0.036, 0.039, 0.296**.
    - NETEXP_FC, t-1: 0.074, 0.037, 0.017, 0.061**, 0.202**.
    - d(i): 0.005, -0.000, 0.005*, -0.002, -0.024**.
    - d(i*SR): -0.007, -0.000, -0.008*, 0.002, -0.019.
    - R_squared: 0.02, 0.01, 0.02, 0.02, 0.04. Nobs: 806, 802, 806, 806, 811.
  - For nonbanks (panel c):
    - dlog(VIX): -0.002, 0.001, -0.011***, 0.013***, -0.043**.
    - dlog(VIX)_NB, t-1: 0.078***, 0.024*, 0.028**, 0.041*, 0.217***.
    - NETEXP_NB, t-1: -0.000, 0.000, 0.024, -0.034, 0.024.
    - d(i): 0.005, -0.000, 0.005*, -0.002, -0.024***.
    - d(i*SR): -0.007, 0.000, -0.008*, 0.003, -0.020.
    - R_squared: 0.02, 0.00, 0.02, 0.02, 0.04. Nobs: 806, 802, 806, 806, 811.
- Panel regressions excluding the United States (Table A8) — selected coefficients:
  - For total financial institutions:
    - dlog(VIX): -0.004*, 0.001, -0.010***, 0.009**, -0.044*** (across dependent variables).
    - dlog(VIX)_NETEXP_FC, t-1: 0.068*, -0.013, 0.041**, 0.011, 0.287**.
    - NETEXP_FC, t-1: 0.070, 0.032, 0.013, 0.059**, 0.178**.
    - d(i): 0.005, -0.000, 0.005*, -0.002, -0.023**.
    - d(i*SR): -0.006, -0.001, -0.007*, 0.002, -0.020.
    - R_squared: 0.02, 0.01, 0.02, 0.02, 0.04. Nobs: 927, 923, 927, 927, 932.
  - For nonbanks (excluding US):
    - dlog(VIX): -0.004, 0.001, -0.010***, 0.010**, -0.043***.
    - dlog(VIX)_NB, t-1: 0.051**, 0.021**, 0.022*, 0.019, 0.176***.
    - NETEXP_NB, t-1: -0.001, -0.001, 0.018, -0.030, 0.013.
    - d(i): 0.005, -0.000, 0.005**, -0.002, -0.024***.
    - d(i*SR): -0.006, -0.001, -0.007**, 0.002, -0.021.
    - R_squared: 0.01, 0.00, 0.02, 0.01, 0.03. Nobs: 927, 923, 927, 927, 932.

### Interpretations and analytical implications preserved from source
- Interaction terms dlog(VIX)_NETEXP (for total, banks, nonbanks) often show positive and in many columns statistically significant coefficients, indicating that higher global risk (VIX) combined with prior net-exposure amplifies capital flow responses in the sample specifications reported.
- The direct dlog(VIX) coefficients are often negative (some statistically significant), suggesting that increases in global volatility are associated with declines in capital inflows (or related flow measures), with heterogeneity across institutions and sample splits.
- Financial depth and the nonbank sector matter: findepth_t-1 shows negative and in some specifications statistically significant coefficients for nonbank regressions (e.g., -0.025**, -0.044***), indicating deeper domestic financial systems moderate certain nonbank exposures or flows in robustness checks.
- Shadow policy rate interactions and short-term rate changes (d(i), d(i*SR)) have small coefficients generally reported as 0.005 or -0.006, and are included as controls across specifications.

*International Monetary Fund. Chapter 2 in Global Financial Stability Report, October 2016 (appendices A–E and robustness tables as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019085.pdf_