## _wp1453

## Source details

**Canonical URL:** [_wp1453](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1453.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1453.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1453.pdf.json)

---

### Overview and key conclusions
- Paper examines how Korea’s capital flows and trade have been affected by unconventional monetary policies (UMPs), focusing on QE (U.S.) and QQME (BOJ).
- Main findings:
  - QE had little direct impact on capital flows to Korea; QE tapering is unlikely to cause capital outflows from Korea, owing partially to safe-haven behavior of capital flows to Korea (Section II).
  - Exchange rate spillovers to Korea from QQME will be limited in the near term, although Korea’s newly gained non-price competitiveness may unwind if the yen’s weakening cycle goes deeper or becomes sufficiently prolonged (Section III).
  - The won has appreciated by 25 percent against the yen since May 2012; so far the impact on Korea’s exports has remained muted.

### II. Spillovers through the financial linkage — capital flow trends and UMP effects
- Historic capital-flow dynamics and crisis episodes:
  - Sudden outflow in 2008Q4: about US$ 33 billion (14 percent of Korea’s gross international reserves).
  - Won depreciation during crisis: maximum of 36 percent.
  - Korea’s gross international reserves decreased by 34 billion dollars that quarter (a decline by 16 percent); maximum decline during the GFC was US$ 65 billion (peak-to-trough fall of 24 percent).
  - Banks’ external debt: from 220 billion dollars (short-term debt, 160 billion dollars) pre-crisis to 178 billion dollars (short-term debt, 77 billion dollars) in September 2013.
- Post-crisis capital flow patterns:
  - QE waves: QE1 (November 2008 to March 2010), QE2 (November 2010 to June 2011), QE3 (September 2012 to now).
  - QE1 average monthly net inflows to Korea: 1.5 billion U.S. dollar; bond and equity flows averaged 3.5 billion U.S. dollars offsetting outflows in FDI and banking.
  - Korea’s monthly gross bond and equity flows: 2.1 billion dollars (QE1) → 2.4 billion dollars (QE2) → 2.1 billion dollars (QE3).
  - Bond and equity fund flows to EMs (EPFR): rose to an average of 14 billion dollars per month during QE3 from 4-5 billion dollars per month during QE1 and QE2.
- Differentiation since May 2013 (taper-tantrum):
  - Despite global EM outflows, Korea attracted significant inflows as investors reallocated from weaker EMs.
  - Drivers: reduced banking vulnerability, diversified investor base (including central banks), large current account surplus, market perception of managed won appreciation.
- Empirical estimation framework:
  - Ordinary least-square models for weekly (Jan 2008–Jun 2013 for portfolio flows) and monthly data for banking flows.
  - UMP proxies: Fed balance sheet changes classified into Liquidity (LQ), Treasuries (TR), and MBS measures; event dummies around QE announcements; BOJ_ASST and BOJ_GS for QQME (monthly).
  - Portfolio flows used Korea Exchange transaction breakdowns; banking flows used monthly Balance of Payments statistics.
- Equity-flow regression key quantified effects:
  - A 1 percent increase in VIX can cause 0.6 billion dollars of outflows a week.
  - A one percentage point acceleration of sequential increases in VIX can reduce foreign equity flow response to PBR by 0.2 billion dollars for every one-tenth increase of PBR.
  - Nonlinear PBR response: threshold PBR* decreases by about 0.36 with a 1 percent increase in VIX and collapses to zero when VIX increases by 2.7 percent in the current week.
  - U.S. yield curve steepening: a 10 basis point acceleration of steepening associated with about a 100 million dollar increase in equity flows during the noncrisis period (larger effect during crisis).
  - Baseline weekly regression: D(BOJ_GS) coefficient -0.329963 (Std. Error 0.122261; t-Statistic -2.698837; Prob. 0.0092); LOG(VIX) coefficient -4098.354 (Std. Error 1132.019; t-Statistic -3.620394; Prob. 0.0006).
- Bond-flow regression highlights:
  - Federal Reserve’s Treasury Purchase Program (TR) has a highly significant and negative correlation with foreign bond flows to Korea.
  - VIX nonlinearities: foreign bond flows respond positively to increases in VIX until threshold VIX*; when TB5YKRW-CD3MKRW = 0, VIX* = 31.2.
  - Yield curve steepening increases VIX*, allowing safe-haven flows to withstand higher risk aversion.
  - Cross-currency swap basis spread dynamics: decreases in basis spread (more negative) lower VIX* except during yield curve inversion.
- Banking-flow regression highlights:
  - Federal Reserve MBS Purchase Program had a very large direct impact on short-term banking capital flows to Korea, significantly slowing net outflows since 2008-09.
  - UMPs eased global liquidity strains (negative coefficient on LIBOR-OIS) but contributed to U.S. yield-curve flattening (positive coefficient), which could negatively affect banking flows.
- Implications for QE exit scenarios:
  - Growth-driven smooth QE exit: unlikely to cause capital outflows from Korea; may cause inflows to equities and banking via positive signaling and orderly yield curve steepening.
  - Disorderly QE exit: can cause capital outflows through rises in VIX (equities and bonds) and through LIBOR-OIS (bonds and banking flows); effects amplified with onshore dollar funding dislocation.
  - Asset-class differentiated VIX effects:
    - Equities: unambiguously negative for flows.
    - Bonds: positive (search-for-safety) up to about 1–1½ standard deviations above historical median; beyond that, sell-offs can occur.
  - Nonlinear dynamics and tail risks imply behavior during QE expansions may differ from during tapering.

### III. Spillovers through the trade linkage — export trends, product decomposition, and value-chain effects
- Korea’s export market-share evolution:
  - World export market share: 2.7 percent in 2008 → 3.1 percent in 2010; unchanged since.
  - Three explanations for postcrisis gain:
    - Favorable exchange rate: won depreciation post-2008 sharper and more protracted; won-yen fluctuations 12–15 won per yen until Q3 2012 versus 8–11 won per yen before the crisis.
    - Moving up the value chain: example—Samsung Electronics moved to second place in global brand ranking in 2013; was ranked 43 in 2008.
    - Offshore production increases: smart phone offshore production 16 percent in 2008 → 80 percent in 2012; automobile offshore production 38 percent in 2008 → 73 percent in 2012.
- Was Korea’s gain Japan’s loss?
  - Korea’s and Japan’s global market share changes similar in magnitude (~0.4 percentage point).
  - Decomposition findings:
    - China and ASEAN account for 73 percent of Korea’s 0.4 percent gain during 2008–12.
    - In China: 63 percent of Korea’s market share gain due to Chinese market was in professional and scientific instruments and electrical machinery—same markets where Japan also slightly gained or maintained share.
    - Japan lost only 0.04 percentage point of its global market share in China—one-tenth of its overall market share loss and one-fourth of Korea’s gain (+0.26 percentage point) in this market.
    - In ASEAN: Korea’s largest gain in petrochemicals (+0.027 percentage point) where Japan lost the most (-0.023 percentage point).
- Product-level market-share decomposition:
  - About one-fifth of Korea’s market share gain was in the global automobile market; Japan’s loss in this market closely mirrored Korea’s gain.
  - About one-fourth of Korea’s market share gain was in electrical machinery and appliances; Japan’s loss in electrical machinery amounts to about 40 percent of Korea’s gain in the sector.
  - Aggregate: about 70 percent of Korea’s market share gains during 2008-2012 appear achieved in geographic subsets (notably China and ASEAN) where Japan’s market share has not changed much.
- Sensitivity of Korea’s exports to the won/yen exchange rate (VAR results):
  - VAR sample and specification: 1998M1-2013M8; variables: Korea’s export value (U.S. dollars, seasonally adjusted), Korea’s import value (U.S. dollars, seasonally adjusted), won/yen exchange rate; four lags; crisis dummy; world trade value index (2005=100) as exogenous control.
  - Peak response of Korea’s exports to a 10 percent depreciation of the won/yen:
    - 1998-2005: 1.5 percent (after four months).
    - 2006-13: 0.7 percent (after five months).
  - Alternative real effective exchange rate model: peak response declined from 1.4 percent (after four months) for 1998-2005 to zero for 2006-13.
  - Overall: sensitivity of Korea’s exports to won/yen has been reduced by about one-half since the late 1990s, though it remains larger than for other Asian competitors.
- Value-chain mitigation potential:
  - Foreign value added (FVA) accounts for 40 percent of Korea’s gross export value; Korea’s FVA/gross export ratio is the fifth highest among OECD countries and substantially higher than Japan (15 percent) and China (33 percent).
  - Share of FVA originating in Japan in Korea’s gross export value: 5.1 percent.
  - Sectoral FVA-to-gross-export shares (percent) — columns correspond to Korea Japan China Indonesia U.S. Germany France:
    - Overall: 5.1 85.2 4.4 1.4 0.9 0.8 1.8
    - Electrical: 6.8 82.2 6.3 4.1 1.6 1.7 1.1
    - Transport: 6.8 85.9 6.9 3.0 2.5 1.2 0.2
    - Chemical: 4.1 78.9 4.0 0.9 0.8 0.7 0.9
    - Logistics: 2.4 93.2 1.5 1.5 0.3 0.6 2.0
    - Metal: 6.1 80.5 3.5 1.6 1.0 0.6 0.8
  - Implications:
    - High FVA share implies less exposure to exchange rate shocks due to offsets from imported input prices.
    - Korea’s dependence on Japan as supplier of intermediate inputs is modest (5.1 percent); Korea–China two-way value chain linkages are growing stronger.
    - Most competitive gains/losses from yen depreciation will fall on Japanese exporters; Korean exporters receive some trickled-down gains.
    - Korea’s link to Japan is larger than third-country competitors in certain sectors (e.g., transport: Korea 6.8 percent vs United States 2.5 percent and Germany 1.2 percent).
  - Net assessment: value-chain links could mitigate negative spillovers from yen depreciation but the mitigation effect is likely relatively small for Korea; yen depreciation would likely strengthen Korea’s competitive edge against third-country competitors.

### Conclusion — policy-relevant findings and recommendations
- Financial-linkage conclusions:
  - QE and QQME operations or announcements had little or no direct influence on capital flows to Korea, except cross-border banking capital flows where sustained outflow was substantially offset by the Federal Reserve’s MBS purchase programs.
  - Indirect effects through financial prices were significant and nonlinear: bond flows showed safe-haven behavior (flows increase with higher VIX) up to threshold levels of risk aversion.
  - A growth-driven smooth QE exit that leaves long-term U.S. rates anchored and preserves investor confidence is unlikely to cause capital outflows from Korea and could cause inflows to Korean equity and bank debt via positive signaling.
  - A disorderly QE exit can cause capital outflows via rises in VIX (equities and bonds) and the LIBOR-OIS spread (bonds and banking flows), especially if onshore dollar funding markets are dislocated.
  - Asset-class differentiation: equities unambiguously negative with rising VIX; bonds can be positive except in tail VIX eruptions.
  - Korea may be graduating from a high-capital-flow beta country, supported by resilience of the won and asset prices to recent QE tapering-related turmoil.
- Trade-linkage conclusions:
  - Bulk of Korea’s global market share increase since the crisis appears not to have been at Japan’s loss, although Japan’s loss was similar in magnitude to Korea’s gain.
  - Korea’s exports are less sensitive to the won/yen exchange rate than before, though sensitivity remains larger than other Asian competitors.
  - Expected positive spillovers from yen depreciation through value chain effects would be small for Korea.
- Cautions and strategic implications:
  - Benign assessment of trade spillovers should be read cautiously: large Japanese firms suffered margin compression since 2008-09 and may decompress margins as the yen weakens; some Japanese firms may re-enter top-tier product streams if yen weakness deepens or is prolonged, potentially unwinding Korea’s product differentiation gains.
  - Recommendation: Korean companies should continue to step up non-price competitiveness to turn the challenge into an opportunity.

### Appendices — selected regression variables and highlights
- Key regression-variable definitions (excerpt):
  - BOJ_GS: Bank of Japan's holding of government securities (in billions of yen)
  - BONDFLOW_ALL: Net foreign bond inflows (in millions of U.S. dollars)
  - CCR1Y_RATE: 1-year cross country swap spreads (in basis points)
  - CD3MKRW: Korea's 3 month CD interest rate (in percent)
  - CRISIS: Dummy variable that captures the existence of a financial crisis
  - EQUITYFLOW: Net foreign equity inflows (in millions of U.S. dollars)
  - LIBOIS_SPREAD_USD: LIBOR-OIS spread for U.S. dollars (in percent)
  - MBS: Outstanding amount of the Federal Reserve's MBS facility (weekly average, in billions of U.S. dollars)
  - VIX: The Chicago Board Options Exchange Market Volatility Index
- Selected regression estimates (highlights):
  - Appendix 2 (Equity regression baseline; Dependent Variable: PINL_EQ; sample 2008M03 - 2013M06; Included observations: 64):
    - D(BOJ_GS): Coefficient -0.329963; Std. Error 0.122261; t-Statistic -2.698837; Prob. 0.0092
    - LOG(VIX): Coefficient -4098.354; Std. Error 1132.019; t-Statistic -3.620394; Prob. 0.0006
    - DLOG(KOSPI(-1)): Coefficient 15252.786; Std. Error 356.279; t-Statistic 2.399640; Prob. 0.0198
    - KOSPI_PBR(-1): Coefficient 56054.852; Std. Error 1596.362; t-Statistic 2.595569; Prob. 0.0121
    - R-squared 0.652723; Adjusted R-squared 0.60221; S.E. of regression 2275.16; Durbin-Watson stat 2.109061
  - Appendix 1 (Equityflow OLS; sample 1/15/2008 - 6/25/2013; Included observations: 285):
    - CRISIS: Coefficient -2313.369; Std. Error 356.6128; t-Statistic -6.487062; Prob. 0
    - LOG(VIX): Coefficient -614.2112; Std. Error 198.5488; t-Statistic -3.093503; Prob. 0.0022
    - DLOG(KOSPI(-1)): Coefficient 9075.035; Std. Error 1674.6815; t-Statistic 5.418964; Prob. 0
    - KOSPI_PBR(-1): Coefficient 6357.298; Std. Error 1673.3613; t-Statistic 3.799120; Prob. 0.0002
    - R-squared 0.437907; Adjusted R-squared 0.417393; S.E. of regression 723.2586; Durbin-Watson stat 1.688911
- Appendix 3 (BONDFLOW_ALL regression; sample 1/08/2008 - 6/18/2013; Included observations: 285):
  - C: 1728.232; Std. Error 168.7346; t-Statistic 10.24231; Prob. 0
  - CRISIS: -2162.402; Std. Error 483.8886; t-Statistic -4.468801; Prob. 0
  - D(TR): -0.017879; Std. Error 0.005791; t-Statistic -3.087533; Prob. 0.0022
  - LIBOIS_SPREAD_USD: -802.0172; Std. Error 235.4032; t-Statistic -3.406994; Prob. 0.0008
  - SOV_CDS_KR: -5.204942; Std. Error 1.304547; t-Statistic -3.989847; Prob. 0.0001
  - CCR1Y_RATE*CRISIS: -9.37297; Std. Error 2.345515; t-Statistic -3.996124; Prob. 0.0001
  - CCR5Y_RATE: 2.409907; Std. Error 1.118084; t-Statistic 2.155391; Prob. 0.032
  - CCR1Y_RATE*VIX: -0.180255; Std. Error 0.064917; t-Statistic -2.776677; Prob. 0.0059
  - CCR1Y_RATE*VIX^2: 0.002886; Std. Error 0.000925; t-Statistic 3.121489; Prob. 0.002
  - 100*(TB5YKRW-CD3MKRW)*VIX: 0.063313; Std. Error 0.025471; t-Statistic 2.485738; Prob. 0.0135
  - R-squared 0.319349; Adjusted R-squared 0.297073; S.E. of regression 782.4322; Durbin-Watson stat 1.421518

*Source: IMF staff estimates, _wp1453 - INTRODUCTION.*

### INTRODUCTION

### _wp1453 - INTRODUCTION

### Overview and key conclusions
- Paper examines how Korea’s capital flows and trade have been affected by unconventional monetary policies (UMPs), focusing on QE (U.S.) and QQME (BOJ).
- Main findings:
  - QE had little direct impact on capital flows to Korea; QE tapering is unlikely to cause capital outflows from Korea, owing partially to safe-haven behavior of capital flows to Korea (Section II).
  - Exchange rate spillovers to Korea from QQME will be limited in the near term, although Korea’s newly gained non-price competitiveness may unwind if the yen’s weakening cycle goes deeper or becomes sufficiently prolonged (Section III).
  - The won has appreciated by 25 percent against the yen since May 2012; so far the impact on Korea’s exports has remained muted.

### Context and motivation
- Historical concern: capital flows to recipient economies can cause exchange rate overshooting, credit excess, sudden stops, and balance of payments crises, prompting a global review of policy responses.
- Since May 2013, discussion has shifted to risks associated with QE exit, triggered capital outflows from a number of emerging markets (EMs); BOJ’s QQME in April 2013 is an additional factor.

### Policy actions noted
- Korean authorities’ responses to past stresses included monetary easing, fiscal stimulus, BOK dollar-swap injection, government guarantees on banks’ external debt, and bilateral currency swap facilities with the United States and Japan.
- Authorities introduced macro-prudential regulations, including a ceiling on banks’ FX derivative position, to curb systemic risks from short-term external debt and foreign-currency domestic lending.

---

### II. Spillovers through the financial linkage

#### A. Korea—Capital Flow Trends
- Early 2000s to 2008-09 crisis:
  - Capital account liberalization introduced new opportunities and risks.
  - Expectations of trend appreciation in the won created large hedging demand by exporters, producing liquidity mismatches in the banking system; banks (particularly foreign branches) relied heavily on external borrowing, particularly short-term, low-rate borrowing.
  - Model estimate: a one percentage point increase in the net forward position (scaled by total asset size) of a foreign bank branch is associated with a 0.3 percentage point increase in short-term external debt (same quarter) and 0.2 percentage point increases in subsequent quarters.
- 2008Q4 crisis impact:
  - Sudden outflow of about US$ 33 billion (14 percent of Korea’s gross international reserves) of banking capital.
  - Sharp depreciation of the won (to a maximum of 36 percent during the crisis).
  - Korea’s gross international reserves decreased by 34 billion dollars that quarter (a decline by 16 percent; maximum decline during the GFC was US$ 65 billion or a peak-to-trough fall of 24 percent).
- Post-crisis recovery through 2013:
  - By 2009Q1, Korea stemmed the sudden stop; current account surplus and portfolio inflows more than offset banks’ external-debt reduction.
  - Three QE waves: QE1 (November 2008 to March 2010), QE2 (November 2010 to June 2011), QE3 (September 2012 to now).
  - QE1: monthly capital flows to Korea were 1.5 billion U.S. dollar net inflows on average (average net flow was negative during QE2 and QE3); strong bond and equity flows (3.5 billion U.S. dollars) offset outflows in FDI, banking flows, and settlement payments related to derivative liabilities.
  - Bond and equity fund flows to EMs (EPFR data): rose to an average of 14 billion dollars per month during QE3 from 4-5 billion dollars per month during QE1 and QE2; Korea’s monthly gross bond and equity flows: 2.1 billion dollars (QE1) → 2.4 billion dollars (QE2) → 2.1 billion dollars (QE3).
  - Banks’ external debt fell from 220 billion dollars (short-term debt, 160 billion dollars) immediately prior to the crisis to 178 billion dollars (short-term debt, 77 billion dollars) in September 2013.
- Since May 2013 (taper-tantrum period):
  - Despite global EM outflows, Korea attracted significant inflows of both equity and bonds as investors redirected funds from weaker EMs (India, Indonesia, Brazil, Turkey).
  - Reasons for differentiation:
    - Reduced vulnerability: banks shed external short-term debt.
    - Modest post-crisis equity rebound compared to other EMs; negative equity drags (North Korea risks, index reclassification by a large EM ETF, yen depreciation) began to fade in summer 2013.
    - Bond flows driven by a more diversified investor base, including central banks, showing "safe-haven" style increases when global markets become more risk averse (up to a critical point).
    - Large current account surplus and market perception that authorities pace the won’s appreciation created one-sided expectations on the won.

#### B. Effects of UMPs on Capital Flows: Estimation
- Model specification:
  - Ordinary least-square models using Korean-specific weekly (Jan 2008–Jun 2013 for portfolio flows) and monthly data (for banking flows), with dependent variables: portfolio bonds, portfolio equity, and bank flows.
  - Push factors (global), pull factors (Korea-specific), and UMP proxy variables included.
- UMP proxies:
  - Weekly change in the U.S. Federal Reserve balance sheet classified into Liquidity (LQ), Treasuries (TR), and MBS measures; event dummy = 1 (-1) in two weeks following QE expansion (reduction) announcements.
  - QQME proxies: BOJ total asset stock (BOJ_ASST) and government securities holdings (BOJ_GS) — monthly.
- Data notes:
  - For portfolio flows, used Korea Exchange transaction breakdowns rather than EPFR, due to EPFR’s limitations for Korea (e.g., about 40 percent of foreign flows are from central banks).
  - For banking flows, used monthly Balance of Payments statistics as highest-frequency public data.
- Equity flow regression (baseline weekly):
  - None of the QE variables was significant in the baseline model.
  - QE could have supported equity flows to the extent measures and announcements alleviated risk aversion.
  - Key quantified effects:
    - A 1 percent increase in VIX can cause 0.6 billion dollars of outflows a week (≈ two-thirds of standard deviation from average net flows).
    - A one percentage point acceleration of sequential increases in VIX can reduce foreign equity flow response to PBR by 0.2 billion dollars for every one-tenth increase of PBR.
    - Nonlinear response to PBR: equity flows respond procyclically to PBR until a threshold (PBR*), beyond which behavior becomes countercyclical; PBR* decreases by about 0.36 with a 1 percent increase in VIX and collapses to zero when VIX increases by 2.7 percent in the current week.
    - U.S. yield curve steepening: a 10 basis point acceleration of steepening associated with about a 100 million dollar increase in equity flows during the noncrisis period (larger effect during crisis).
  - Alternative monthly specification: shows a very significant negative association between QQME and equity flows to Korea.
- Bond flow regression:
  - Federal Reserve’s Treasury Purchase Program (TR) has a highly significant and negative correlation with foreign bond flows to Korea.
  - UMPs could positively affect bond flows by easing global dollar liquidity stress (captured by LIBOR-OIS spread) and by dampening VIX during tail elevations.
  - Nonlinear response to VIX:
    - Foreign bond flows respond positively to increases in VIX until a threshold (VIX*); beyond VIX*, flows turn negative.
    - When term premium (TB5YKRW-CD3MKRW) is zero, VIX* = 31.2.
    - Yield curve steepening increases VIX*, implying safe-haven flows can withstand higher risk aversion.
  - Cross-currency swap basis spread dynamics: a decrease in the basis spread (more negative) lowers the VIX* threshold except during yield curve inversion, reflecting episodes of dollar liquidity difficulties.
- Bank flow regression:
  - Federal Reserve MBS Purchase Program had a very large direct impact on short-term banking capital flows to Korea, significantly slowing net outflows since the 2008-09 crisis.
  - Indirect effects: UMPs eased global liquidity strains (negative coefficient on LIBOR-OIS) but also contributed to U.S. yield-curve flattening (positive coefficient), which could negatively affect banking flows.
- Implications for QE exit risks:
  - A growth-driven smooth QE exit that leaves long-term U.S. rates anchored and investor confidence intact is unlikely to cause capital outflows from Korea; may even cause inflows to equities and banking via positive signaling and orderly yield curve steepening.
  - A disorderly QE exit can cause capital outflows through rises in VIX (equities and bonds) and through LIBOR-OIS (bonds and banking flows); effects amplified if unwinding causes onshore dollar funding dislocation.
  - VIX increases have asset-class–specific effects:
    - Equities: unambiguously negative for flows.
    - Bonds: can be positive (search-for-safety) up to about 1–1½ standard deviations above historical median; beyond that, can trigger sell-offs.
  - Caution: behavior of capital flows during QE expansions may differ from during tapering; nonlinear dynamics and tail risks may arise.

---

### III. Spillovers through the trade linkage

#### A. Trends in Korea’s exports after the Global Financial Crisis
- Korea’s world export market share rose from 2.7 percent in 2008 to 3.1 percent in 2010, and has remained unchanged since.
- Three commonly cited explanations for Korea’s postcrisis market-share gains:
  - Favorable exchange rate: won’s depreciation after 2008 was sharper and more protracted than many competitors; won-yen fluctuations were 12–15 won per yen until Q3 2012 versus 8–11 won per yen before the crisis.
  - Moving up the value chain: Korean exporters shifted to high-value final products and stronger branding (example: Samsung Electronics moved to second place in global brand ranking in 2013; was ranked 43 in 2008).
  - Offshore production: offshore production rose sharply for ICT and autos (smart phone offshore production: 16 percent in 2008 → 80 percent in 2012; automobile offshore production: 38 percent in 2008 → 73 percent in 2012), augmenting global export share and diversification.

#### B. Was Korea’s gain Japan’s loss?
- Korea’s and Japan’s global market share changes were similar in magnitude (~0.4 percentage point); question: did Korea’s gain come at Japan’s loss?
- Decomposition analysis framework: change in Korea’s global market share decomposed by geographic/product market segments to detect demand substitution.
- Findings from decomposition:
  - Korea’s recent global market share gains were led by China and ASEAN, which account for 73 percent of the 0.4 percent gain during 2008–12.
  - Exports to China:
    - 63 percent of Korea’s global market share gain due to the Chinese market was in professional and scientific instruments and electrical machinery—same markets where Japan also slightly gained or maintained share.
    - Korea’s market share remained flat or declined in areas where Japan declined most (e.g., petrochemicals and inorganic chemicals).
    - Overall, Japan lost only 0.04 percentage point of its global market share in China’s market—one-tenth of its overall market share loss and one-fourth of Korea’s gain (+0.26 percentage point) in this market.
  - Exports to ASEAN:
    - Korea’s gain in ASEAN was broader based; largest gain in petrochemicals (+0.027 percentage point) where Japan lost the most (-0.023 percentage point), reflecting Japan’s post-2011 energy supply disruptions.
    - Both Korea and Japan increased market share in road vehicles.

---

*Source: IMF staff estimates, _wp1453 - INTRODUCTION.*

### 0.01 percent of its global market share in the ASEAN market.

### _wp1453 - 0.01 percent of its global market share in the ASEAN market.

### Global market gains by products
- Decomposition of Korea’s export products indicates a substantial portion of Korea’s market share gains during 2008-12 may reflect demand substitution from Japan and thus may be prone to unwinding if the won/yen exchange rate permanently reverts to precrisis levels.
- Key product-level findings:
  - About one-fifth of Korea’s market share gain was made in the global automobile market; Japan’s loss in this market closely mirrored Korea’s gain.
  - About one-fourth of Korea’s market share gain was made in electrical machinery and appliances; Japan’s loss in electrical machinery amounts to about 40 percent of Korea’s gain in the sector.
- Aggregate observation:
  - About 70 percent of Korea’s market share gains during 2008-2012 appear to have been achieved in geographical subsets (notably China and ASEAN) of the global market where Japan’s market share has not changed much.
  - The dissection across products is suggestive of some demand shift from Japan.

### Sensitivity of Korea’s exports to the Japanese yen
- Econometric approach:
  - A Vector Auto Regression (VAR) model spanning 1998M1-2013M8 on three variables: (i) Korea’s export value in U.S. dollars (seasonally adjusted); (ii) Korea’s import value in U.S. dollars (seasonally adjusted); and (iii) the won’s exchange rate against the Japanese yen; in the same Cholesky ordering.
  - Model includes a dummy for financial crises (1997 Asian crisis and 2008-09 global financial crisis) and CPB Netherlands Bureau for Economic Policy Analysis’s world trade value index (2005=100) as exogenous controls.
  - VAR uses four lags following Akaike criterion.
- Main numerical results on sensitivity:
  - The sensitivity of Korea’s exports to the won/yen exchange rate has been reduced by about one-half since the late 1990s.
  - Peak response of Korea’s exports to a 10 percent depreciation of the won/yen:
    - 1998-2005: 1.5 percent (after four months of the shock).
    - 2006-13: 0.7 percent (after five months of the shock).
  - Alternative model using real effective exchange rates:
    - Peak response declined from 1.4 percent (after four months of the shocks) for 1998-2005 to zero for 2006-13.
- Comparative perspective:
  - The model suggests the spillover effect of QQME through the exchange rate channel to Korea is relatively larger for Korea than its Asian competitors.

### Will Korea benefit from a weak yen through value chain effects?
- Conceptual point:
  - Cross-border integration of production can mitigate negative spillovers from a competitor’s currency depreciation to the extent that the affected country sources intermediate inputs from the depreciating-currency country.
- Korea–Japan production link facts:
  - Korean manufacturers have strong upstream links to Japanese suppliers of intermediate inputs (e.g., Japanese high-precision equipment for semiconductors and LCDs).
  - Examples: share of Japan in imported inputs was 30 percent in 2010 compared to 42 percent in 2000 for semiconductors; for LCDs the share of Japan was 81 percent in 2010 compared to 83 percent in 2000.
  - Korea’s reliance on Japan as the supplier of intermediate inputs has been steadily declining; China’s importance in production linkages is rapidly growing.
- OECD value-added-based trade statistics (summary findings):
  - Foreign value added (FVA) accounts for 40 percent of Korea’s gross export value; this makes Korea’s FVA to gross export ratio the fifth highest among the OECD countries and substantially higher than Japan (15 percent) and China (33 percent).
  - Share of FVA originating in Japan in Korea’s gross export value: 5.1 percent.
  - Japan’s domestic value-added ratio of exports: 85 percent.
  - Sectoral comparisons (columns: Korea Japan China Indonesia U.S. Germany France; values are "shares of FVA to gross exports" in percent):
    - Overall: 5.1 85.2 4.4 1.4 0.9 0.8 1.8
    - Electrical: 6.8 82.2 6.3 4.1 1.6 1.7 1.1
    - Transport: 6.8 85.9 6.9 3.0 2.5 1.2 0.2
    - Chemical: 4.1 78.9 4.0 0.9 0.8 0.7 0.9
    - Logistics: 2.4 93.2 1.5 1.5 0.3 0.6 2.0
    - Metal: 6.1 80.5 3.5 1.6 1.0 0.6 0.8
  - Implications:
    - High FVA share implies less exposure to exchange rate shocks due to offsets from imported input prices.
    - Korean exporters’ dependence on Japan is modest (5.1 percent), and Korean–Chinese two-way value chain linkages may already be stronger than Korea–Japan links.
    - Most competitive gains or losses from changes in the yen’s exchange rate will fall on Japanese exporters, though Korean exporters receive some trickled-down gains.
    - Korea’s link to Japan is larger than third-country competitors: e.g., Korean transportation equipment manufacturers’ FVA share owing to Japan is 6.8 percent versus United States 2.5 percent and Germany 1.2 percent.
- Net assessment:
  - Value chain links could mitigate negative spillovers from the depreciation of one country’s currency to another, but the mitigation effect would likely be relatively small in Korea’s case.
  - A depreciation of the Japanese yen would likely strengthen Korea’s competitive edge against third-country competitors.

### Conclusion — policy-relevant findings and implications
- Financial-linkage findings:
  - QE and QQME operations or announcements had little or no direct influence on capital flows to Korea, except cross-border banking capital flows where sustained outflow was substantially offset by the Federal Reserve’s MBS purchase programs.
  - Indirect effects of QE through financial prices were significant, with nonlinear behavior: capital flows to Korean bonds showed safe-haven behavior (flows increase with higher VIX) up to threshold levels of risk aversion.
  - Overall, indirect effects point to a largely positive impact of QE through alleviation of risk aversion (captured by VIX) and global dollar liquidity stress (represented by the LIBOR-OIS spread).
  - A growth-driven smooth QE exit that leaves long-term U.S. rates anchored and preserves investor confidence is unlikely to cause capital outflows from Korea and could cause inflows to Korean equity and bank debt via positive signaling.
  - A disorderly QE exit can cause capital outflows through VIX (equities and bonds) and the LIBOR-OIS spread (bonds and banking flows), with bond flows particularly vulnerable if onshore dollar funding markets are dislocated.
  - An increase in VIX associated with a disorderly QE exit has asset-class-differentiated impacts: equities unambiguously negative; bonds can be positive except in a tail eruption of VIX.
  - These analyses point to Korea’s possible graduation from a high-capital-flow beta country, supported by resilience of the won and asset prices to recent QE tapering-related turmoil.
- Trade-linkage findings (summary):
  - Bulk of the increase in Korea’s global market share since the crisis appears not to have been at Japan’s loss, although Japan’s loss was in a very similar magnitude to Korea’s gain.
  - Korean exports are less sensitive to the won/yen exchange rate than before, although sensitivity remains larger than other Asian competitors.
  - Expected positive spillovers from the yen’s depreciation through value chain effects would be small for Korea.
- Cautions and strategic implications:
  - Benign assessment of spillovers from QQME through trade linkages should be read cautiously:
    - Large Japanese companies suffered margin compression since the 2008-09 crisis and are likely to decompress margins as the yen weakens.
    - Some Japanese firms may strategically foray back into top-tier product streams (e.g., smart phones) if the yen’s weakening cycle deepens or is prolonged, potentially unwinding Korea’s product differentiation gains.
  - Recommendation for Korea:
    - Korean companies should continue to step up non-price competitiveness to turn the challenge into an opportunity.

### Appendices — selected regression results and variables
- Key regression-variable definitions (excerpt):
  - BOJ_GS: Bank of Japan's holding of government securities (in billions of yen)
  - BONDFLOW_ALL: Net foreign bond inflows (in millions of U.S. dollars)
  - CCR1Y_RATE: 1-year cross country swap spreads (in basis points)
  - CD3MKRW: Korea's 3 month CD interest rate (in percent)
  - CRISIS: Dummy variable that captures the existence of a financial crisis
  - EQUITYFLOW: Net foreign equity inflows (in millions of U.S. dollars)
  - LIBOIS_SPREAD_USD: LIBOR-OIS spread for U.S. dollars (in percent)
  - MBS: Outstanding amount of the Federal Reserve's MBS facility (weekly average, in billions of U.S. dollars)
  - VIX: The Chicago Board Options Exchange Market Volatility Index
- Selected regression estimates (highlights):
  - Appendix 2 (Equity regression baseline; Dependent Variable: PINL_EQ; sample 2008M03 - 2013M06; Included observations: 64):
    - D(BOJ_GS): Coefficient -0.329963, Std. Error 0.122261, t-Statistic -2.698837, Prob. 0.0092
    - LOG(VIX): Coefficient -4098.354, Std. Error 1132.019, t-Statistic -3.620394, Prob. 0.0006
    - DLOG(KOSPI(-1)): Coefficient 15252.786, Std. Error 356.279, t-Statistic 2.399640, Prob. 0.0198
    - KOSPI_PBR(-1): Coefficient 56054.852, Std. Error 1596.362, t-Statistic 2.595569, Prob. 0.0121
    - R-squared 0.652723; Adjusted R-squared 0.60221; S.E. of regression 2275.16; Durbin-Watson stat 2.109061
  - Appendix 1 (Equityflow OLS; sample 1/15/2008 - 6/25/2013; Included observations: 285):
    - CRISIS: Coefficient -2313.369, Std. Error 356.6128, t-Statistic -6.487062, Prob. 0
    - LOG(VIX): Coefficient -614.2112, Std. Error 198.5488, t-Statistic -3.093503, Prob. 0.0022
    - DLOG(KOSPI(-1)): Coefficient 9075.035, Std. Error 1674.6815, t-Statistic 5.418964, Prob. 0
    - KOSPI_PBR(-1): Coefficient 6357.298, Std. Error 1673.3613, t-Statistic 3.799120, Prob. 0.0002
    - R-squared 0.437907; Adjusted R-squared 0.417393; S.E. of regression 723.2586; Durbin-Watson stat 1.688911

*Italic: Source: _wp1453 - 0.01 percent of its global market share in the ASEAN market.*

### Appendix 3. Equity Regression (Alternative)

### Appendix 3. Equity Regression (Alternative)

### Regression specification and sample
- Dependent Variable: BONDFLOW_ALL
- Method: Least Squares
- Sample (adjusted): 1/08/2008 - 6/18/2013
- Included observations: 285 after adjustments

### Estimated coefficients (variable — Coefficient; Std. Error; t-Statistic; Prob.)
- C — 1728.232; 168.7346; 10.24231; 0
- CRISIS — -2162.402; 483.8886; -4.468801; 0
- D(TR) — -0.017879; 0.005791; -3.087533; 0.0022
- LIBOIS_SPREAD_USD — -802.0172; 235.4032; -3.406994; 0.0008
- SOV_CDS_KR — -5.204942; 1.304547; -3.989847; 0.0001
- CCR1Y_RATE*CRISIS — -9.37297; 2.345515; -3.996124; 0.0001
- CCR5Y_RATE — 2.409907; 1.118084; 2.155391; 0.032
- CCR1Y_RATE*VIX — -0.180255; 0.064917; -2.776677; 0.0059
- CCR1Y_RATE*VIX^2 — 0.002886; 0.000925; 3.121489; 0.002
- 100*(TB5YKRW-CD3MKRW)*VIX — 0.063313; 0.025471; 2.485738; 0.0135

### Model fit and diagnostics
- R-squared: 0.319349
- Adjusted R-squared: 0.297073
- S.E. of regression: 782.4322
- Sum squared resid: 1.68E+08
- Log likelihood: -2298.094
- F-statistic: 14.33608
- Prob(F-statistic): 0
- Mean dependent var: 869.4192
- S.D. dependent var: 933.2361
- Akaike info criterion: 16.19715
- Schwarz criterion: 16.32531
- Hannan-Quinn criter.: 16.24852
- Durbin-Watson stat: 1.421518

### Key findings and implications (based on coefficient signs and significance)
- The CRISIS indicator is associated with a large negative effect on BONDFLOW_ALL (coefficient -2162.402; Prob. 0).
- Higher LIBOIS_SPREAD_USD is significantly associated with lower BONDFLOW_ALL (coefficient -802.0172; Prob. 0.0008).
- Increases in SOV_CDS_KR are significantly associated with reductions in BONDFLOW_ALL (coefficient -5.204942; Prob. 0.0001).
- Interaction CCR1Y_RATE*CRISIS shows a significant negative relationship with BONDFLOW_ALL (coefficient -9.37297; Prob. 0.0001), indicating sensitivity of bond flows to short-term credit conditions during crisis periods.
- CCR5Y_RATE has a positive and significant coefficient (2.409907; Prob. 0.032), suggesting longer-term credit conditions are associated with higher BONDFLOW_ALL.
- Nonlinear interaction terms with VIX (CCR1Y_RATE*VIX negative; CCR1Y_RATE*VIX^2 positive) are both significant, implying a nonlinear effect of volatility on the relationship between CCR1Y_RATE and BONDFLOW_ALL.
- The term 100*(TB5YKRW-CD3MKRW)*VIX is positive and significant (0.063313; Prob. 0.0135), indicating that the combination of term spread and volatility is associated with higher bond flows.

*Source: Authors' estimates.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1453.pdf_