On the Measurement and Forecasting of Business Cycles and...

www.conferenceboard.org © 2015 The Conference Board, Inc. | 1

On the Measurement and Forecasting of Business

Cycles and Growth Cycles in the Global Economy Ataman Ozyildirim, [email protected]

November 30, 2015

Prepared for 7th Joint EC-OECD Workshop on recent developments in Business and Consumer Surveys, Paris,

France

http://www.conference-board.org/


Key Questions for a Global Leading Economic Index

Where is global growth trending this quarter?

Is the global cycle more or less synchronized across countries/

regions and between mature and emerging economies?

What is the probability of a slowdown or recession in the next

quarter or two?

What is the current impact and importance of cyclical (short-term)

versus structural (long-term) factors in the global cycle?

How do financial and nonfinancial indicators behave, and how do

they interact?

How does the interconnectivity of regional economies work, and

what is their impact on the global economy?

What does this mean for the global economic outlook for the

remainder of the year and next year



-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

12%

World G20 OECD G7

Increased trade and financial integration have strengthened correlations

across aggregate measures of economic activity – particularly output

Real GDP Growth Year-over-year Percent Change

Source: IMF; OECD; The Conference Board

Correlations World G20 OECD G7

World - 0.8523 0.8171 0.8116

G20 0.8523 - 0.9072 0.8665

OECD 0.8171 0.9072 - 0.9837

G7 0.8116 0.8665 0.9837 -



Business Cycle Reference Dates

The recent update of TCB’s Business Cycle Reference Dates mirrors this

degree of conformity across economies in the BCI program

Source: The Conference Board

Latin

America

U.S. Mexico Euro Area France Spain Germany U.K. Australia Japan Korea China India Brazil

Dates at Business Cycle Peaks

1940s Feb-45

Nov-48

1950s Jul-53

Aug-57

1960's Apr-60 May-66 Dec-60

Dec-69

1970's Nov-73 Aug-74 May-73 Jun-73 Aug-74 Feb-73 Mar-79

Nov-79

1980's Jan-80 Nov-81 Feb-80 Mar-80 May-82 Jul-88

Jul-81 Nov-85 Oct-82

1990's Jul-90 Nov-94 Feb-92 Feb-92 Feb-92 Feb-91 May-90 May-90 Feb-92 Aug-97 Mar-91 Oct-97

Jun-95 Mar-97

2000's Mar-01 Oct-00 Feb-08 Aug-02 Feb-08 Mar-01 May-08 Dec-00 Sep-08 Dec-00

Dec-07 Jul-08 Feb-08 Feb-08 Feb-08 Jan-08 Oct-02

Jul-08

2010s Jul-11 Feb-12 Jun-10 Aug-10 Sep-10

Feb-12

Dates at Business Cycle Troughs

1940s Oct-45

Oct-49

1950s May-54

Apr-58

1960's Feb-61 Jun-67 May-61

1970's Nov-70 May-75 Oct-75 Aug-75 Mar-75 Apr-75

Mar-75

1980's Jul-80 Jun-83 Aug-81 Nov-82 Feb-82 May-83 May-80 Oct-89

Nov-82 Jan-87 Jan-85

1990's Mar-91 Oct-95 Sep-93 Dec-93 Jun-93 Jul-93 Dec-91 Jul-91 Aug-93 Jul-98 Nov-91 Feb-99

Mar-96 Feb-99

2000's Nov-01 Mar-02 Aug-09 May-03 Jun-09 Aug-03 Aug-09 Feb-02 Dec-08 Jan-09 Sep-01

Jun-09 May-09 Aug-09 Jun-09 Mar-09 Jun-03

Jan-09

2010s Feb-13 Mar-13 Apr-13 Dec-11 Apr-11

Dec-12

Source: NBER; IBRE/FGV; The Conference Board

Notes: U.S. and Brazilian dates are obtained from the NBER and IBRE/FGV Business Cycle Dating Committees, respectively; all other business cycle reference dates are determined using a business cycle dating algorithm (see Bry

and Boschan (1971) and Harding and Pagan (2002))

Business Cycle Reference Dates

North America Europe Asia-Pacific

Burns & Mitchell

(1946) established

business cycle dating;

Bry & Boschan (1971)

and Harding & Pagan

(2002) provide

updates via

computerized

algorithms

Examine cyclical

peaks and troughs in

coincident economic

indicators, in

accordance set rules

regarding phases and

cycles.

How are Business Cycles Dated?



-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

Global Advanced Economies Emerging & Developing Economies

Real GDP Growth Year-over-year Percent Change

Some economists contend that global GDP growth falling below a

threshold of 2.8% delineates a global recession

Note: Shaded regions depict hypothetical recession as quarters during which global GDP growth was below 2.8%

Source: IMF; The Conference Board

Global Recession

Threshold: 2.8%



There is an ample literature on the global business cycle, much of which

adopts techniques used in country-level business cycle research

• Regional business cycles have become

increasingly important, especially in areas with

growing levels of international trade and financial

regionalization and integration.

• The strong business cycle synchronization of the 1970s

and 1980s reflected systemic “shocks” (e.g. oil price shocks)

that were uniformly destabilizing.

Hirata, Kose, and Otrok (2013)

• Using a Bayesian dynamic latent factor model, the authors

show that a common world factor is an important source

of volatility across a large sample of economies.

• Consumption and Investment cycles, in

comparison to growth cycles, are more so

determined by country and idiosyncratic

factors across multiple economies.

Kose, Otrok, and Whiteman (2003)

• Global business cycles can be decomposed

into global, group (e.g. emerging), country,

and idiosyncratic portions.

• Variance decomposition shows the amounts

attributable to each individual factor.

• Group factors have become less

important since the era of globalization.

Kose, Otrok, and Prasad (2008)

• The authors identify a strong positive impact

of trade intensity on business cycle

synchronization

• Additionally, they find that bilateral intra-

industry trade and trade specialization

correlation increase co-movement of

business cycle dynamics

Duval et. al. (2014)

Global

Business

Cycle

Literature

Source:



More recently, the notion that economies within certain regions follow

similar cyclical trends has been developed

• Even single economic data series– World Steel Production

– have a salient ability to predict global business cycles with a

requisite lag, and in some cases, outperform existing

indicators or groups of indicators.

Ravazzolo and Vespignani (2015)

• The degree to which business cycles synchronize across

countries might depend on, among other things, physical

distance, the amount of bilateral trade, similarities in

institutions or language, or historical trade routes.

• The business cycles of most African and Asian (developed

and developing) countries do not appear to co-move with

either their regional neighbors or the rest of the world.

Cooke, Kose, Otrok, and Owyang (2015)

Source: Cooke, Kose, Otrok, and Owyang (2015)



0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Global Developed Emerging

Synchronization of Business Cycles Percent of Economies (by Gross Domestic Product) in Business Cycle Recession

Diffusion of business cycle movements across the global economy

creates a need for more stylized research and timely monitoring




Individual economic data series often contain noise and lack substantial

breadth – both in geography and in sector


Difficulty in Defining the Global Business Cycle

Production Trade

Output

Sentiment Expectations

Financial

Indicators

Can one composite

measure capture

different sectors and

regions of the global

economy in a

comprehensive and

meaningful way?



0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Developing a more precise definition of the global business cycle provides

a benchmark against which to measure the Global LEI prototypes

Note: Shaded regions depict global recession dates


Business Cycle Analysis of Recession Share Series

by GDP Share

Duration in months of

Peaks (P ) and Troughs (T ) business cycles and phases

P T P P to T T to P P to P

(1) (2) (3) (4) (5) (6)

Jul-80 Jul-81 12

Jul-81 Feb-82 Jul-90 7 101 108

Jul-90 Aug-93 Feb-08 37 174 211

Feb-08 Aug-09 Jul-11 18 23 41

Jul-11 Dec-11 5

Mean 16.8 77.5 120.0

Median 12.5 62.0 108.0

St. Deviation 14.7 75.6 85.6

35% Threshold

Global Business Cycle Percent of Economies (by Gross Domestic Product) in Business Cycle Recession



Global Business Cycle Indicators (BCI) at The Conference

Board

Modeled after U.S. system of monthly leading economic indexes

Aggregated from CEIs and LEIs for 13 countries/areas:

U.S., Brazil, Mexico, Japan, South Korea, China, India, Australia, U.K.,

Euro Area, Germany, France and Spain

Composite indices:

Bring cycles and turning points into focus

Used to define and anticipate turning points in business cycles

Help to identify growth cycles vs. business cycles

Help in forecasting and economic outlook



“Classical” business cycles vs. growth cycles

BCP

BCT

time

time

GCP

GCT

Le

ve

l o

f e

co

no

mic

activity

De

via

tio

n fro

m lo

ng

te

rm tre

nd

BCP: business cycle peak

BCT: business cycle trough

GCP: growth cycle peak

GCT: growth cycle trough



When developing both CEIs and LEIs, components are

selected based on six criteria of cyclical performance

Consistent Timing The series must exhibit a consistent timing pattern as a

leading, coincident, or lagging indicator 1

Conformity The series must conform well to the business cycle 2

Smoothness Month-to-month movements must not be too erratic

3

Economic Significance Cyclical timing must have economic meaning and be

logical 4

Statistical Adequacy Data must be collected and processed in a statistically

reliable way; no large and frequent revisions 5

Currency or Timeliness Series must be published on a reasonably prompt

schedule – preferably on a monthly basis. 6

Source: The Conference Board, “Business Cycle Indicators Handbook”, December 2000; Zarnowitz (2007)



Global CEI Components

United States Australia Japan Korea Mexico United Kingdom China Brazil Euro Area India

Ret

ail

Manufacturing and

Trade SalesRetail Sales

Retail, Wholesale,

and Manufacturing

Sales

Wholesale and

Retail SalesRetail Sales Retail Sales

Retail Sales of

Consumer Goods

Volume of Sales of

the Retail MarketRetail Trade

Car Sales, Passenger

Vehicles

Non-Agricultural

EmploymentEmployed Persons

Number of

Employed PersonsTotal Employment

Employment: IMSS

BeneficiariesEmployment LFS

Employment:

Manufacturing

Occupied Employed

PopulationEmployment

Employment

Workforce

Passenger Carried

Volume

Inco

me Personal Income

less Transfer

Payments

Household Gross

Disposable Income

Wage and Salary

Income

Monthly Cash

Earnings

Real Household

Disposable Income

Average Real

Income of Workers

Industrial

Production

Industrial

Production

Industrial

Production

Industrial

Production

Industrial

Production

Industrial

Production

Industrial

Production: Value-

Added

Industrial

Production

Industrial

Production

Industrial

Production

Electricity

Production

Industrial Electric

Energy Consumption

Shipments of

Corrugated Paper

Manufacturing

Turnover

Total Imports

Pro

du

ctio

nEm

plo

ymen

t



Global LEI Components

United States Australia Japan Korea Mexico United Kingdom China Brazil Euro Area India

Interests Rate

Spread (10-yr T-Bond

Less FFR)

(cumulated)

Yield Spread, 10-yr

less Policy Rate

(cumulated)

Yield Spread

(cumulated)

Yield of Government

Public BondsFederal Funds Rate

Yield Spread

(cumulated)

Swap Rate, 360 days

%

Yield Spread

(cumulated)

Yield spread (10 Yr

90 Day)

Index of Stock

Prices, 500 Common

Stocks

Stock Price Index,

Ordinary ShareStock Prices Stock Prices Stock Prices

Share Price Index,

Stock

Stock Pirces

(Bovespa Index)Eurostoxx Index

BSE: Index: Monthly:

SENSES: Average

Leading Credit Index

(st. dev)Money Supply, M3

Real Money Supply,

M2

Loan: Financial

Institution

Systemic Stress

Indicator

M3: Bank Credit to

Commercial Sector

Real Exchange RateTerms of Trade

IndexREER: 36 Currencies

Rural Goods Exports Real Exports, FOB Exports FOBExports Volume

IndexMerchandise Exports

Mfrs New Orders,

Capital Goods, Non-

defense excl.

Aircraft

Gross Operating

Surplus

Tankan Business

Conditions Survey

Value of Machinery

OrdersOrder Book Volume

PMI: Mfg New

Export Order

Consumer Durable

Goods Production

Index

New Orders of

Capital GoodsIP: Capital Goods

Mfrs New Orders,

Consumer Goods &

Materials

New Orders for

Machinery and

Construction

5000 Ind Enterprises

Diffusion Index:

General Business

Condition

Manufacturing

Survey: Expectations

Index, %

Markit PMI

Manufacturing New

Orders

ISM New Order

Index

Ratio of Sales to

Inventories,

Nonfarm

New Orders for

Machinery and

Construction

Index of Inventories

to Shipment

Net Insufficient

Inventories

Volume of Expected

Output

PMI: Mfg Supplier

Delivery

Services Sector

Survey: Expectations

Index, %

Markit Business

Expectaitons,

Services

PMI: Services

Business Activity

Real Operating

Profits

Operating Surplus of

CorporationsCargo Handles

Building permits for

new private housing

units

Building Approvals Dwelling UnitesPrivate Construction

Orders

Industrial

Production,

Construction

Floor Space Started Building Permits

Avg. consumer

expectations for

business conditions

Business Failures

U.S. Refiners’

Acquisition cost of

Domestic &

Imported crude oil

Consumer

Confidence

Consumer

Expectation Index

Consumers Survey:

Expectations Index,

%

Consumer

expectation of

general economy in

next 12 months

Average Weekly

Hours,

Manufacturing

Index of Overtime

Worked

Average Weekly

Claims,

Unemployment

Insurance

6-Mouth Growth

Rate of Labor

Productivity

Productivity for

whole economy

Fin

anci

alEx

tern

alB

usi

nes

s A

ctiv

ity

& In

vest

men

tC

on

sum

ers

& E

xpec

tati

on

s



Methodology calculates a composite index: CEI and LEI

Compute month-to-

month changes for

each individual

component.

All BCI Program

Indices are used with

the exception of the

France, Germany,

and Spain Indices,

given that the Euro

Area Index is used.

STEP 1

Compute Monthly Changes in Indexes

Because monthly

changes in the

indexes of each

individual index can

have substantial

differences (see

previous slide), the

monthly changes are

standardized in order

to account for regional

volatilities.

STEP 2

Standardization

Each standardized

change in the index is

weighted by that

country’s GDP share

of the entire BCI

Program GDP, which

encompasses about

66% of total output.

The standardized

change is multiplied

by the GDP weight to

compute the

contribution to the

Index’s change.

Source for GDP: The

Conference Board’s

Total Economy

Database.

STEP 3

GDP Weights x Std. Changes

The contributions to

the percent change in

the index values are

summed across all

ten economies.

These percent

changes are then

applied to an index

level that in 2004 =

100.

STEP 4

Sum Contributions & Index to 2010 = 100

The LEI is adjusted to

the trend of the CEI,

as is the practice with

each individual

economy.

The level index is

computed, along with

monthly, six-month,

and annual percent

changes.

STEP 5

Trend Adjustment (for LEI only)



-40%

-30%

-20%

-10%

0%

10%

20%

30%

40% China TCB/FGV Brazil U.S. Euro Area

6-month percent change (annual rate)

While each country exhibits its own business cycle, major

global cyclical downturns affect them all




Global Business Cycle Chronologies

LatAm

U.S. Mexico Euro Area France Spain Germany U.K. Australia Japan Korea China India Brazil

1940s Feb-45

Nov-48

1950s Jul-53

Aug-57

1960s Apr-60 May-66 Dec-60

Dec-69

1970s Nov-73 Aug-74 May-73 Jun-73 Aug-74 Feb-73 Mar-79

Nov-79

1980s Jan-80 Nov-81 Feb-80 Mar-80 May-82 Jul-88

Jul-81 Nov-85 Oct-82

1990s Jul-90 Nov-94 Feb-92 Feb-92 Feb-92 Feb-91 May-90 May-90 Feb-92 Aug-97 Mar-91 Oct-97

Jun-95 Mar-97

2000s Mar-01 Oct-00 Feb-08 Aug-02 Feb-08 Mar-01 May-08 Dec-00 Sep-08 Dec-00

Dec-07 Jul-08 Feb-08 Feb-08 Feb-08 Jan-08 Oct-02

Jul-08

2010s Jul-11 Feb-12 Jun-10 Aug-10 Sep-10

Feb-12

1940s Oct-45

Oct-49

1950s May-54

Apr-58

1960s Feb-61 Jun-67 May-61

1970s Nov-70 May-75 Oct-75 Aug-75 Mar-75 Apr-75

Mar-75

1980s Jul-80 Jun-83 Aug-81 Nov-82 Feb-82 May-83 May-80 Oct-89

Nov-82 Jan-87 Jan-85

1990s Mar-91 Oct-95 Sep-93 Dec-93 Jun-93 Jul-93 Dec-91 Jul-91 Aug-93 Jul-98 Nov-91 Feb-99

Mar-96 Feb-99

2000s Nov-01 Mar-02 Aug-09 May-03 Jun-09 Aug-03 Aug-09 Feb-02 Dec-08 Jan-09 Sep-01

Jun-09 May-09 Aug-09 Jun-09 Mar-09 Jun-03

Jan-09

2010s Feb-13 Mar-13 Apr-13 Dec-11 Apr-11

Dec-12

Asia-Pacific

Date at Business Cycle Peak

Date at Business Cycle Trough

North America Europe



Accurate weighting economies in the Global LEI

is a crucial consideration – BCI economies currently

comprise 66%* of global output

Share of Global GDP

1. Australia - 1.0%

2. Brazil - 2.9%

3. China - 16.7%

4. Euro Area - 12.5%

5. India - 6.2%

6. Japan - 4.5%

7. Korea - 1.7%

8. Mexico - 2.0%

9. United Kingdom - 2.4%

10.United States - 16.5%

Note: This calculation includes the entire Euro Area and excludes France, Germany, and Spain

Source: The Conference Board, Total Economy Database



Australia, 1.3% Brazil, 1.2%

China, 11.4%

India, 2.1%

Japan, 4.0%

South Korea, 2.9%

Mexico, 2.2%

United Kingdom, 3.0%

United States, 10.7%

Euro Area, 13.0%

-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25% GDP (lhs axis) Trade (rhs axis)

Global GDP and Trade Year-over-year Percent Change

Alternatively, weighting by global trade volume share may better capture

and weight cyclical contributions

Source: IMF International Finance Statistics; CPB World Trade Monitor; The Conference Board

Global Trade Share Percent of Global Trade



Each method has trade-offs when considering its viability as an index

construction method


Technical Sophistication Glo

bal/

Reg

ion

al

High

Ind

ivid

ua

l

Co

un

try

Low

Da

ta G

ran

ula

rity

Pros:

• Higher frequency data

• Fewer data revisions and omissions

• Easy to explain to non-technical audience

• Ease of replicability

Pros:

• Avoids dealing with issues of trade, value added, or

GDP-weighting various components

• Easy to explain to non-technical audience

• Requires fewer “balancing” of aggregated date

Pros:

• Higher frequency data

• Fewer data revisions and omissions

• Highly-technical methods can produce strong

goodness of fit and smooth lead times

Pros:

• Avoids dealing with issues of trade, value added, or

GDP-weighting various components

• Highly-technical methods can produce strong

goodness of fit and lead times

Cons:

• Does not use advanced statistical techniques that

can provide for better fit and average lead times.

• Can overly emphasize countries’ idiosyncrasies

(i.e. policy interventions)

Cons:


• Difficult to explain to a non-technical audience

• Parsimonious models can perform equally as well

Cons:

• Regional data is often not as high frequency and is

updated/published infrequently


(i.e. policy interventions)

Cons:

• Regional data is often not as high frequency and is

updated/published infrequently

• Difficult to explain to a non-technical audience

• Parsimonious models can perform equally as well



-6%

-4%

-2%

0%

2%

4%

6%

-15%

-10%

-5%

0%

5%

10%

15% Global LEI (lhs axis) Global CEI (rhs axis)

Annual percent change

Global Coincident Economic Index (CEI) is a measure of

the global cycle and the Global LEI leads its peaks and

troughs

Note: Shaded areas represent global growth cycle recessions as determined by The Conference Board




-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

Australia Brazil China India Japan Korea Mexico United Kingdom United States Euro Area Total

Decomposition analysis demonstrates the relative

contributions of mature vs. emerging economies 2014-15


Contribution to Global LEI Percent Change Percentage Points



-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Australia Brazil China India Japan Korea Mexico United Kingdom United States Euro Area Total

Decomposition analysis demonstrates the relative

contributions of mature vs. emerging economies 2007-2009


Contribution to Global LEI Percent Change Percentage Points



-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

Global CEI World GDP

Global CEI and World GDP have a high contemporaneous

correlation

Annual Percent Change

Correlations

0Q Fwd 0.7511

1Q Fwd 0.6275

2Q Fwd 0.4119

3Q Fwd 0.1730

4Q Fwd -0.0304




-15%

-10%

-5%

0%

5%

10%

15%

-4%

-2%

0%

2%

4%

6%

8%

10%

World GDP Global LEI

Global LEI can be used as an early warning signal for

downturns in global output




Correlations

0Q Fwd 0.7170

1Q Fwd 0.7290

2Q Fwd 0.6235

3Q Fwd 0.4477

4Q Fwd 0.2575



-8%

-6%

-4%

-2%

0%

2%

4%

6%

-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15% Developed Economies LEI (lhs axis) Developed Economies CEI (rhs axis)



Sub-indexes both for mature economies…



-2%

0%

2%

4%

6%

8%

10%

12%

-2%

0%

2%

4%

6%

8%

10%

12%

14% Emerging Economies LEI (lhs axis) Emerging Economies CEI (rhs axis)



…and emerging market economies can help analyze the

diffusion and prevalence of global cyclical movements



There is a high degree of correlation between the Global

LEI and the CPB World Trade Monitor Index of Global

Trade

-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

Global LEI (lhs axis)

CPB World Trade Monitor (rhs axis)



Correlations

0M Fwd 0.8093

1M Fwd 0.8249

2M Fwd 0.8117

3M Fwd 0.7839

4M Fwd 0.7292

5M Fwd 0.6589

6M Fwd 0.5719



Global LEI and World GDP per capita have a strong

correlation, but the frequency aggregation diminishes its

usefulness

-10%

-8%

-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

-3%

-2%

-1%

0%

1%

2%

3%

4%

5%

World GDP per Capita (lhs axis)

Global LEI (rhs axis)


Source: Maddison Project; The Conference Board

Correlations

0Y Fwd 0.7911

1Y Fwd 0.3347



Global LEI Prototypes (1 of 3) Index Value (lhs axis)

0

20

40

60

80

100

120

140

160

-20%

-15%

-10%

-5%

0%

5%

10%

15%

Global LEI: GDP Weighted

Global LEI: Index Method

Global LEI: GDP x Index Method

Global LEI Prototypes Annual Percent Change (rhs axis)

The three initial Global LEI prototypes produce quite uniform results, with

common trends and cycles


GDP weighting is

the most sensitive

to EM growth



Global LEI Prototypes (2 of 3) Index Value (lhs axis)

-20%

-15%

-10%

-5%

0%

5%

10%

15%

0

20

40

60

80

100

120

140

160

Global LEI: Components: GDP x Index Method

Global LEI: Components: Index Method

Global LEI: Trade Weights

Global LEI: Trade Weights x Index Method

Trade-weighting increases volatility during expansions and downturns,

while the component prototypes appear structurally dissimilar


Component methodologies

don’t show similar trend as

other indices – flat especially

in post Recessions

Component methodologies

don’t show similar trend as

other indexes– weak growth

especially post Great

Recession

Global LEI Prototypes Annual Percent Change (rhs axis)



-6%

-4%

-2%

0%

2%

4%

-12

-10

-8

-6

-4

-2

0

2

4

6

8

Global LEI: Principal Component: 6M Chg. (lhs axis)

Global LEI: Principal Component: 12M Chg. (lhs axis)

Global LEI: Principal Component: Five Series* (rhs axis)

Global LEI Prototypes (3 of 3) Index Value

The principal component prototypes show similar cyclical movements with

varying degrees of volatility based on underlying composition

Note: Principal Component of World Industrial Production, World Trade, Dow Jones Global Index, the Baltic Dry Index, and Brent

Crude Oil Prices


12-Mo. has smoothest

monthly changes

Global LEI Prototypes (3 of 3) Annual Percent Change



Note: See Appendix regarding details of model specification and explanation of statistical tests; forecast proportions may not sum to

100% due to rounding.


Out-of-Sample Recession Probability Modeling

A pseudo real-time recession probability modeling exercise

shows that four methodologies excel far beyond the others

Forecast Proportion (%) Theil’s U

Bias Variance Covariance

Global LEI: GDP x Index Method 0.4 2.2 97.4 0.370

Global LEI: GDP Weighted 0.0 9.2 90.7 0.373

Global LEI: Trade Weighted 0.3 4.1 95.4 0.375

Global LEI: Trade x Index Method 1.4 0.7 97.8 0.390

Global LEI: Index Method 1.5 19.9 78.5 0.411

Global LEI: Component Index Method 12.8 27.9 59.4 0.445

Global LEI: Component GDP x Index Method 7.7 46.5 45.6 0.484

0.0

0.2

0.4

0.6

0.8

1.0

00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

Global LEI: GDP Weighted Index Methodology

Forecast: EQ_LEI_4_FActual: GLOBAL_DUM_35_GDPForecast sample: 2000M01 2014M12Included observations: 180Root Mean Squared Error 0.265317Mean Absolute Error 0.107899Mean Abs. Percent Error 4.554984Theil Inequality Coefficient 0.370319 Bias Proportion 0.004009 Variance Proportion 0.021973 Covariance Proportion 0.974018



Note: See Appendix regarding details of model specification and explanation of statistical tests

Source: McGuckin, Ozyildirim, and Zarnowitz (2004); The Conference Board

Out-of-Sample Growth Forecasting: “Horse Race”

An out-of-sample growth forecasting exercise highlights the LEIs

ability to forecast trends in key global economic data

Forecasting Global Industrial Production Relative

FMSE

DM

Statistic P-Value

Global LEI: GDP Weighted 0.889 2.100 0.018

Global LEI: GDP Weighted Index Methodology 0.915 1.757 0.039

Global LEI: Trade Weighted 0.917 1.699 0.045

Global LEI: Trade Weighted Index Methodology 0.936 1.322 0.093

Benchmark: Global LEI: Index Methodology - - -

Global LEI: Principal Component: Five Series 1.147 -1.451 0.073

Global LEI: Component Index Methodology 1.169 -2.082 0.019

Global LEI: Component GDP Weighted Index Methodology 1.227 -2.095 0.018

Global LEI: Principal Component of 12 Mo. Growth 1.251 -1.880 0.030

Global LEI: Principal Component of 6 Mo. Growth 1.447 -1.387 0.083



Next steps: remaining research issues

Global LEI and CEI (and subindexes) are feasible, but

more research needed around the weighting of the

countries and the correlations

Complete benchmarking studies of the existing

European LEIs and the China LEI

Explore inclusion of other economies

Model global recession risk probabilities, and financial

and nonfinancial indexes

Link more formally with global growth forecasts



Existing measures of the global cycle



-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

OECD GDP OECD LEI

OECD GDP Growth and OECD Leading Economic Indicator Year-over-year Percent Change

The OECD Leading Economic Indicator anticipates peaks and troughs in

OECD GDP with a strong one-quarter-forward correlation

Source: OECD; The Conference Board

Correlations 0Q Fwd 1Q Fwd 2Q Fwd 3Q Fwd

0.8162 0.9188 0.8965 0.7735



-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

World GDP OECD LEI

Global Real GDP Growth and OECD Leading Economic Indicator Year-over-year Percent Change

When compared against global GDP, the OECD LEI performs with

relatively good strength, yet appears more coincident

Source: IMF; OECD; The Conference Board

Correlations 0Q Fwd 1Q Fwd 2Q Fwd 3Q Fwd

0.6499 0.7708 0.7431 0.5926



Underlying Components

(1) Belgian and Netherlands Manufacturing Survey, (2) U.S. Consumer Confidence Aggregate, (3) S&P GSCI Industrial Metals

Index, (4) U.S. Initial Jobless Claims, (5) Baltic Dry Index, (6) Global New Orders Less Inventories, (7) Global PMI, (8) GS

Australian and Canadian Dollar Trade Weighted Index, (9) Korean Exports, and (10) Japan IP Inventory/Sales Ratio

Goldman Sachs Global Leading Indicator (GLI)

Goldman Sachs’ leading indicator, constructed from ten underlying

components, leads turnings points in global industrial production

Note: See Appendix for details regarding the methodology.

Source: Goldman Sachs Global Investment Research



Brookings Tracking Indexes for the Global Economic Recovery (TIGER) Indexes

Brookings’ tracking indexes provide group and activity classifications;

however, infrequent publishing and lack of a target series reduce their utility.

Source: Brookings Institution

Group Classifications

1. Total

2. Advanced Economies

3. Emerging Market Economies

4. Euro Periphery

Activity Classifications

1. Overall Growth Index

2. Real Activity Index

3. Financial Index

4. Confidence Index

Principal Component Analysis

“…enables one to construct indicators of co-

movement across al variables in a dataset or a

subset of them. This procedure is ideal for

creating the TIGER indexes as it allows us to

combine information from different types of

economic variables and multiple countries”


On the Measurement and Forecasting of Business Cycles and...

Documents

Transcript of On the Measurement and Forecasting of Business Cycles and...