A P j t f th IMF’ R h D t t A Project of the IMF’s Research Department

Presented by Steve PhillipsFebruary 2, 2012 February 2, 2012

“EBA” methodology as a successorEBA methodology as a successor to the well-known CGER Today: an introduction to EBA Ancestry: CA regression literature; CGERAncestry: CA regression literature; CGER EBA is a project of the IMF Research

Department comprising:Department, comprising: a research effort to better understand CAs, and de elopment of an approach to e aluating CA s in development of an approach to evaluating CA s in

a normative sense

EBA is also about real exchange rates…

EBA’s new goals, relative to CGER

Focus on current accounts as much as on exchange ratesrates

Distinguish normative and descriptive analyses Add focus on impact of policiesAdd focus on impact of policies

EBA also aims at… taking explicit account of cyclical (higher-

frequency) influences a more explicit treatment of uncertainties

Recall the CGER approach:Panel regression of current account balances:Panel regression of current account balances:

CA/gdp = B ( X ) + e

X regressors: demographics, per capita income, relative growth rates, oil trade balance, fiscal balance, lagged CAg , , , gg

How to interpret the e residuals? As deviations from fitted l d h ’values interpreted as a CA norm. What’s a norm?

Careful interpretation: “CA level implied by fundamentals” Temptation: “equilibrium” and/or appropriate level An Temptation: equilibrium and/or appropriate level. An

implied normative judgment?

Wh h CGER?Why enhance CGER?Issues of CGER interpretation: pResiduals are sizable… reflect ignorance,

not distortions?not distortions?Unclear mix of descriptive analysis and

i [ l fi l]normative assessment [example: fiscal]Assessment sounds normative, but doesn’t

provide a clear link to role of policies, policy implicationsp

Wh h CGER?Why enhance CGER?

A d i i CGER i ifi i And issues in CGER regression specification, including:

What does lagged CA pick up ? Sustained distortions?

Some plausible stories for S and I behavior are not captured by the CGER regressors (e.g., p y g ( g ,social insurance, shifts in capital markets)

Cycle in Global Risk Aversion, y ,Cycle in EM Current Accounts

VIX and Mean Current Account of Emerging Market Economies

0.35

0.40

1%

2%

g g

Mean Current Account / GDP

VIX Lagged One Year (Right Axis)

0.300%

0.20

0.25

-2%

-1%

0.15

%

-3%

0.10-4%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

What is the EBA approach?EBA panel regression aims to better understand the CA:EBA panel regression aims to better understand the CA:

CA/gdp = B1 ( Structural ) + B2 (Policies) + B3 (Cyclical) + v

Want a tight fit (small v residuals) to reduce uncertainty of analysis.But a small CA residual does not give a clean bill of health !

EBA strategy: identify policy gaps and their contributions to CA

EBA is not easy. Challenges: Need to quantify “policy gaps”... and some policies are difficult to measure in the first place ...and some policies are difficult to measure in the first place Can improve fit but can’t eliminate residuals, uncertainty

EBA’s CA regression specificationEBA’s CA regression specification(preliminary, work by Luis Catao and others) Has traditional regressors: demographics , relative

income, relative growth, net oil exports, NFA But no lagged CA No country dummies But no lagged CA. No country dummies.

Adds new variables, of 3 kinds:li l d t f t d l l b l it l cyclical and exogenous: terms of trade cycle; global capital

market conditions (with differential effects) cyclical and policy-influenced: output gap (proxy for

demand shock), real interest differential policies: fiscal balance; public health expenditure, social

protection index, polity, controls on capital inflowsp , p y, p

Signs as expected, most significant

How well does the CA regression work, asHow well does the CA regression work, as a descriptive tool?

Actual vs Fitted CA/GDP for 2010

Std. error = 1.8% GDP.Improved fit, even with

l d t 0.08

0.10

0.12

/

annual data, no dummies, no lagged CA.

0.02

0.04

0.06

CA

/GD

P

Yet some residuals are very large. For 2010:•Largest negative -0.06

-0.04

-0.02

0.00

Act

ual C

Largest negative residual: Greece•Largest positive residuals: Malaysia, Sweden, Thailand -0.12

-0.10

-0.08

0 08 0 06 0 04 0 02 0 00 0 02 0 04 0 06 0 08 0 10 0 12 0 14

actual = fitted(dashed lines are +/- 2 RMSEs)

-0.08 -0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

Fitted CA/GDP

What do panel regression resultsWhat do panel regression results mean for an individual country?

For countries with small CA residuals, we ask:h h k d i f h CA b l ? what are the key drivers of the CA balance?

is the well-fitted CA driven by inappropriate policies? And is the CA sustainable?policies? And is the CA sustainable? Example: prelim. regression suggests that CA deficit of

Turkey comes in part from a weak fiscal balance. y p For countries with large CA surpluses, what are the

estimated contributions of policies?

India Actual and Fitted CA/GDP for India

6%

8%

10%/

Actual CA/GDP

Fitted CA/GDP

2%

4%

-4%

-2%

0%

-8%

-6%

-10%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

TurkeyActual and Fitted CA/GDP for Turkey

6%

8%

10%/ y

Actual CA/GDP

Fitted CA/GDP

2%

4%

-4%

-2%

0%

-8%

-6%

-10%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

SpainActual and Fitted CA/GDP for Spain

6%

8%

10%

/ p

Actual CA/GDP

Fitted CA/GDP

2%

4%

-4%

-2%

0%

-8%

-6%

-10%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

What do panel regression resultsWhat do panel regression results mean for an individual country?

For countries with fairly large CA residuals, we ask: what do we think is missing from the regression’s what do we think is missing from the regression s

story, and is it a good or bad thing? For these countries as well, need to ask whether fittedo t ese cou t es as we , eed to as w et e fitted

values reflect policy distortions... … and to consider sustainability

ItalyActual and Fitted CA/GDP for Italy

6%

8%

10%

/ y

Actual CA/GDP

Fitted CA/GDP

2%

4%

-4%

-2%

0%

-8%

-6%

-10%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

GermanyActual and Fitted CA/GDP for Germany

6%

8%

10%

/ y

Actual CA/GDP

Fitted CA/GDP

2%

4%

-4%

-2%

0%

-8%

-6%

-10%1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

EBA Analysis has 3 legs:1. Current Account analysis

2. Also: Real Exchange Rate analysis 2. Also: Real Exchange Rate analysis Again, add cyclical and policy drivers Also new: aim to assess the level of the RER (not just j

the level of an RER index relative to its past average)

3. And: External Sustainability analysis3 y y Various enhancements of CGER’s ES approach. Hope to add new approach based on intertemporal

constraint (proposed by Martin Evans)

EBA TakeawaysEBA builds on the CGER base by: Moving the boundary between uncertainty and

what we can explain… ….by adding explicit role for policies, and by

modeling responses to higher-frequency shocks Facilitating normative analysis and policy

di idiscussions... ….but need for informed judgment remains.