MANAGING CAPITAL FLOWS: WHAT CAN BE DONE?

* This paper draws on join work with Atish R. Ghosh and Mahvash S. Qureshi. The views expressed in this presentation are those of the presenter and do not necessarily represent those of the IMF or IMF policy.

Jonathan D. OstryResearch DepartmentInternational Monetary Fund

National Bank of Poland Annual ConferenceCentral Europe’s Growth Perspectives in a “New Normal” World

Warsaw, October 14, 2016

RISING FREQUENCY OF SURGES & CRASHES

2

Inflow surges to EMEs have been increasing in frequency and magnitude

Regions experiencing largest surges also tend to subsequently experience the largest drop in net flows—heightening the challenge of managing volatility on the up and downsides

Net Financial Flows to EMEs, 1980-2015 (in USD bln.)

o Latin America prior to the 1980s debt crisiso Asia in the runup to the 1997-98 financial crisiso Emerging Europe prior to the 2008 GFC

Source: IMF’s WEO database. Net financial flows exclude reserve assets. Figures for 2015 are provisional. Net flow in percent of GDP is average across countries.

-5

-3

-1

1

3

5

-800

-600

-400

-200

0

200

400

600

800

1980 1987 1994 2001 2008 2015

Africa & Middle EastLatin America & CaribbeanEuropeAsiaNet flow to EMs (in percent of total GDP)-right axis

0

5

10

15

0

20

40

60

1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013

Avg. net financial flow to GDP in surges (in %)-right axis

Surge observations--threshold approach (in % of total obs.)

Surges of Net Capital Flows (to GDP)

Source: Ghosh et al. (JIE, 2014). Sample=53 EMEs (1980-2013). Surges defined as net capital flow (in % of GDP) observations in the top 30th percentile of a country’s distribution and in the top 30th percentile of the full sample’s distribution.

Surges have become more frequent—the share of surge observations rose from about 10 pct. in the 1980s to over 30 pct. in 2000s

SURGES: DRIVEN BY PUSH & PULL FACTORS

3

Surges are synchronized globally—suggesting that common factors are at play

But even in times of global surges, not all EMEs are affected—so pull factors must also be relevant

Most surges are liability-driven (driven by foreign investors) rather than asset-driven (by sale of residents’ assets abroad and repatriation of proceeds)

Recent research finds

Push factors such as US real interest rate, global risk aversion, commodity prices explain global synchronization

Pull factors such as real GDP growth, external financing need, capital account openness, institutional quality also matter

4

4

WHAT DO SURGES DO?

5

5

Macroeconomic imbalances

Financial vulnerabilities

SURGES LEAD TO MACRO & FIN STAB RISKS

-40

-20

0

20

40

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

REER Overvaluation(In percent of trend REER)

Overvaluation=0.2***Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

RealGDP Growth(In percent)

Growth=0.1**Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Output Gap(In percent of potential output)

Output gap=0.2***Net flows+const.

-30

-10

10

30

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in FX Lending/Total Lending(In ppt.)

Change in FX lending=0.1*Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

3-yr. Change in Domestic Private Credit/GDP(In ppt.)

3-yr.change in credit=0.6***Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in Loan-to-Deposit Ratio(In ppt.)

Change in LTD ratio=0.5***Net flows+const.

Note: Net financial flows (to GDP) are lagged one period.

10ppt. increase in net flows to GDP increases overvaluation by 2 ppt.

6

6

Macroeconomic imbalances

Financial vulnerabilities

SURGES LEAD TO MACRO & FIN STAB RISKS

-40

-20

0

20

40

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

REER Overvaluation(In percent of trend REER)

Overvaluation=0.2***Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

RealGDP Growth(In percent)

Growth=0.1**Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Output Gap(In percent of potential output)

Output gap=0.2***Net flows+const.

-30

-10

10

30

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in FX Lending/Total Lending(In ppt.)

Change in FX lending=0.1*Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

3-yr. Change in Domestic Private Credit/GDP(In ppt.)

3-yr.change in credit=0.6***Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in Loan-to-Deposit Ratio(In ppt.)

Change in LTD ratio=0.5***Net flows+const.

Note: Net financial flows (to GDP) are lagged one period.

10ppt. increase in net flows to GDP increases real GDP growth by 1 ppt.

7

7

Macroeconomic imbalances

Financial vulnerabilities

SURGES LEAD TO MACRO & FIN STAB RISKS

-40

-20

0

20

40

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

REER Overvaluation(In percent of trend REER)

Overvaluation=0.2***Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

RealGDP Growth(In percent)

Growth=0.1**Net flows+const.

-20

-10

0

10

20

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Output Gap(In percent of potential output)

Output gap=0.2***Net flows+const.

-30

-10

10

30

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in FX Lending/Total Lending(In ppt.)

Change in FX lending=0.1*Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

3-yr. Change in Domestic Private Credit/GDP(In ppt.)

3-yr.change in credit=0.6***Net flows+const.

-50

-30

-10

10

30

50

-40 -20 0 20 40 60

Actual

Fitted

Net financial flows/GDP (in percent)

Change in Loan-to-Deposit Ratio(In ppt.)

Change in LTD ratio=0.5***Net flows+const.

Note: Net financial flows (to GDP) are lagged one period.

10ppt. increase in net flows to GDP increases rate of credit expansion by 2 ppt. per yr.

8

8

SURGES DIFFER FROM NORMAL TIMES

-1 1 3 5 7

Change in FX lending/total lending (in ppt.)

Change in LTD ratio (in ppt.)

3-yr. change in domestic credit/GDP (in ppt.)

Output gap (in pct. of potential output)

Real GDP growth rate (in pct.)

REER overvaluation (in pct. of REER trend)

SurgeNon-surge

***

***

***

***

***

Macroeconomic and Financial Vulnerabilities in Surges vs. Non-Surges

Note: *** indicates difference between the means of the two groups (surges, non-surges) is statistically significant at the 1 percent level.

9

9

NOT ALL FLOWS EQUAL IN TERMS OF RISKINESS

o Other investment (banking) and portfolio flows are more prone to creating macroeconomic and financial vulnerabilities than FDI flows

o Within portfolio flows, debt flows are more prone to overvaluation, while equity flows are more expansionary (Blanchard, Ostry, Ghosh & Chamon, AERP&P, 2016)

o Liability flows are more expansionary and prone to creating financial vulnerabilities than asset flows

o For example, 10 percent of GDP ↑ in liability flows ↑ real GDP growth rate by about 1 ppt., while a corresponding increase in asset flows ↑ growth by 0.5 ppt.

o Some types of flows thus seem “safer” than others—cyclical as well as structural tools needed (Ostry et al., JIE, 2012)

10

10

SURGES APPRECIABLY RAISE CRISIS RISKo Increase in net financial flows of 5 percent of GDP raises crisis prob. by about 1 percentage point (unconditional crisis prob. in the sample: 4 pct.)o Main factors contributing to increase in crisis likelihood:

o Change in domestic credit (in percent of GDP) o Currency overvaluation

o Crisis likelihood by type of flow: o Portfolio and other investment flows significantly ↑ crisis prob., but not FDIo Within portfolio flows, debt more likely to cause a financial crisis than equity

o NOT all surge episodes end in crisis• Several EMEs (Brazil, India, South Africa, and Turkey) received large capital

inflows in the runup to GFC, and did not suffer a subsequent banking crisis• While some (e.g., Russia and Ukraine) did experience a banking crisis• Overall, of several (about 30) EMEs that had an inflow surge in the run up to

the GFC, only 5 experienced a financial crisis in 2008–09

11

11

WHAT CAUSES CRASHES?

12

12

LIKELIHOOD OF POST-SURGE CRISIS

Banking and Currency Crisis Probability (In percent)

Note: Post-surge crisis probability is defined as a (banking/currency) crisis within two years after a surge episode ends; see Ghosh, Ostry, Qureshi (AER, 2016).

0

10

20

Banking crisis Currency crisis Banking & currencycrisis

Full sample Post-surge episode

0

10

20

30

Asia Europe Latin America &Caribbean

Middle East &Africa

Post-surge episode banking crisisPost-surge episode currency crisisPost-surge episode banking & currency crisis

13

13

Crash: banking or currency crisis occurring within two years of surge end (Ghosh, Ostry, Qureshi, AERP&P, 2016)

Unconditional crash probability in the sample: 20 percent

Changes in both global and domestic conditions are strongly associated with how surge episodes end

Global: US interest rate, commodity prices, and global risk aversion

For instance, predicted crisis probability increases by 7ppt. if US real interest rate rises by 100 bps (relative to no change in interest rates at all)

Domestic: Crash probability is higher when the surge experiences…

Greater credit expansion, economic overheating, currency overvaluation

Smaller stock of foreign exchange reserves

Smaller share of FDI inflows (predicted crisis prob. is 12ppt. lower if surge predominantly FDI driven)

Higher debt inflows/accumulation of bank foreign liabilities (predicted crisis prob. is 11ppt. higher if surge predominantly debt inflow driven)

SOFT-LANDING OR CRASH?

14

14

WHAT CAN COUNTRIES DO?

15

15

Capital flows bring risks, while growth/risk-sharing benefits of full liberalization are hard to identify in the data (unbundling is important)

Global push factors (which are largely exogenous for recipient countries) are important determinants of surge and crash risk:

Countries therefore need latitude to adopt “insulation” policies

Policies may be countercyclical or structural, especially to improve the composition of inflows

During surge and/or normal times, a range of policies seem salient:

Exchange rate management/FXI, macroeconomic policies, macroprudential measures, and capital controls all have a role to avoid macro and financial-stability imbalances

Structural policies also needed to improve mix of flows in favor of FDI and equity (less debt)

IT’S REALLY IMPORTANT TO MANAGE THE SURGE WELL

16

16

Sometimes macroprudential measures are superior (Jeanne, 2012) When all borrowing is risky (credit boom), no need to discriminate between

domestic and foreign lenders

Optimal policy is tax on domestic borrowing

But other times, controls may be the best response (Ostry et al., 2011) When external borrowing is the source of distortions, authorities should

discriminate against resident-nonresident transactions

Optimal policy is capital controls

When borrowing is channeled through unregulated financial sector, broad restrictions on foreign borrowing work better (Ostry et al., 2012)

There is uncertainty regarding how each tool operates Several macroprudential measures are de facto capital controls

Road-testing both measures still at an early stage

Circumvention risks apply to both sets of measures

Which could argue for simple rules, e.g., limits on leverage or open FX positions

In practice, employ both instruments if use of each has convex costs

NO GENERAL CASE FOR MACROPRU VS CONTROLS

17

EFFECTIVENESS OF RECENT CAPITAL CONTROLS Brazilian residency-based measures and tax on FX derivatives

Shifted composition of inflows (lower portfolio debt flows)—Benelli et al. (2014); Baumann and Gallagher (2015)

Reduced currency appreciation pressures, and increased MP autonomy—Baumann and Gallagher (2015); Chamon and Garcia (2016)

Reduced bond inflows and foreign investors’ bond portfolio allocation—Lambert et al. (2013) and Forbes et al. (2016)

Korean currency-based measures Reduced banks’ short-term foreign borrowing and maturity mismatches—IMF

(2012); Kim (2013) Reduced volatility of cross-border bank flows—Bruno and Shin (2014)

Cross-country evidence: Inflow controls/FX-related measures Reduced net/gross inflows (Ahmed and Zlate, 2015); bond and bank inflows

(Bruno et al., 2015) Reduced financial fragilities (Forbes et al., 2015; Zhang and Zoli, 2016)

Not much impact on macro variables (ER, net capital flows, interest rate differential)—Forbes et al. (2015)

17

18

18

ROLE FOR POLICY COOPERATION?

19

19

POLICIES MAY NEED TO BE COORDINATED GLOBALLY

“Almost every country at one time or another, exercises control over the inflow and outflow of investments, but without the cooperation of other countries such control is difficult, expensive, and subject to considerable evasion.”

Harry Dexter White

“But such control will be more difficult to work…by unilateral action than if movements of capital can be controlled at both ends.”

John Maynard Keynes

20

FOUR IMPLICATIONS FOR POLICY ADVICE Capital controls should not substitute for warranted external adjustment

Inflow controls should invite scrutiny when exchange rate is undervalued (exceptions: learning-by-doing? Financial stability?)

Capital controls should not be used to exploit market power Same as any monopolist/monopsonist Far fetched? Think of other policies (e.g., monetary policy) affecting flows

Capital controls imposed to offset genuine domestic externalities may require source-recipient country coordination (convex costs) Such coordination requires source countries to take into account impact of

their policies on others (Sweden/Latvia; USA) May benefit source countries even if at first blush looks costly for them

Capital controls imposed for genuine domestic externalities may require coordination among borrowers to prevent capital control wars Would involve lowering controls in response to generalized surges Would be in their own interests–right cost-benefit balance of controls

Managing Capital Flows: What Can Be Done? Rising Frequency of Surges & CrashesSurges: Driven by Push & Pull factorsSlide Number 4Surges lead to macro & fin stab risksSurges lead to macro & fin stab risksSurges lead to macro & fin stab risksSurges differ from normal timesNot all flows equal in terms of riskinessSurges appreciably raise crisis riskSlide Number 11Likelihood of post-surge crisisSoft-landing or crash?Slide Number 14�It’s really important to manage the surge well�Slide Number 17Slide Number 18�Slide Number 20