International Comparison Program Frederic A. Vogel Global Manager, ICP The World Bank.

International Comparison Program

Frederic A. VogelGlobal Manager, ICPThe World Bank

The World Bank 2

Purpose or goal of this presentation

• Provide an understanding of what is a Purchasing Power Parity

• How they are estimated and their statistical and economic properties

• Provide an overview of the International Comparison Program

• Present uses of PPPs

The World Bank 3

The Problem—Comparing economic measures across countries

• Country Level Statistics are in National Currencies.

• Comparisons across countries require converting data into a common currency i.e.– Gross Domestic Product– Inter country comparisons of Agricultural

output

The World Bank 4

How convert national data into a common currency?

• Exchange rates

• or

• Purchasing Power Parities

The World Bank 5

Statistical basis of PPPs

GDP or TVP A =

155

1i

N

j 1PAij QAij

GDP OR TVP B=

155

1i

N

j 1

PBij QBij

GDP or TVP Ausing B prices =

155

1i

N

j 1

PAij . (PBij / PAij ). QAij

Direct comparisons of country A with country B:

Exchange rates or PPPs ?

The World Bank 6

Why not use exchange rates (X-rates)?

• X-rates will convert the GDPs to a common currency

• X-rates are easily understood being determined by the demand for & the supply of currencies

• X-rates are easily observed, cover all countries, readily available

The World Bank 7

Exchange rates differ from PPPs

They do not measure differences in the relative price

levels between countries

Are fixed by policy in some countries

Currency speculation and

short-term capital movements cause exchange rates to fluctuate

Price levels differ across components

of GDP

Fluctuations can make some

countries appear “richer” or “poorer” almost overnight –

e.g. Euro from $ 0.91 to $1.20 < 3 years

Because…

The World Bank 8

What is a Purchasing Power Parity (PPP)?

• Form of Exchange Rate

• Based on comparison of prices between countries– International Price Index– Compares prices at a point in time

across space vs. CPI comparing prices across time

• (Big Mac Example)

The World Bank 9

The Economist has popularized its international price comparisons based on the Big Mac

Calculating PPPs: The Big Mac Example

By comparing the local currency prices of Big Macs, the Economist measures relative price levels and compares these to the exchange rate

The World Bank 10

Calculating the Big Mac Index

Compare the price of a Big Mac in four countries

Price in national currency

PPP toUS $

Exchange Rate

PPP scale factor

Philippines 68.02 23.5 55.30 2.4

China 10.20 3.5 7.50 2.1

Switzerland 6.17 2.1 1.26 0.6

US 2.90 1.0 1.00 1.0

The World Bank 11

Limitations of the Big Mac Index

• It is based on only one product, and is not representative of the output of the full economy

• For example, Big Macs are sold mostly in urban areas and leave out traditional food such as rice

The Big Mac Index is easy to calculate, but it leaves out a lot

What is needed is an index that compares the prices of a large, representative basket of goods across many countries

The World Bank 12

Overview of the International Comparison Program

The governance structure

Explain the process and analytical methods to compute PPPs to Compare GDPs and similar economic variables between countries

Procedure for 2005 benchmark year

First regional comparisons which then are linked to provide PPPs at international $ for every country.

Present uses of PPPs

The World Bank 13

ICP Governance

Africa(48 countries)

Asia(23 countries)

CIS(10 countries)

LAC(10 countries)

West Asia(11 countries)

Eurostat/OECD

(45 countries)

ICP Executive Board

Technical Advisory Group

Global OfficeWorld Bank

ICP under auspices of the UN Statistical

Commission

The World Bank 14

Historical Global ICP Participation

The World Bank 15

Basic Data Required to compute PPPs for GDP --Expenditures

A breakdown of final expenditure on National GDP into 155 categories using a common classification

National annual prices for products chosen from a common basket of well-defined goods & services

The World Bank 16

The steps

• Start with GDP and Main aggregates– Individual Consumption by Households– Individual Consumption by NPISH– Individual Consumption by Government– Collective Consumption by Government– Capital Formation– Change in Inventories– Trade Balance

Divide into 155Categories

The World Bank 17

Individual Consumption by Households—Categories• Food & non-Alcoholic Beverages-29• Alcohol and Tobacco• Clothing & Footwear• Housing, water, electricity• Health• Education• Transport--etc

The World Bank 18

Food Basic Headings

• Rice• Other Cereals• Pasta• Beef and Veal• Fish & Seafood• Fresh Milk

• Fresh Fruit• -• -• -• Sugar

The World Bank 19

Purchasing Power Parities - The Steps

Start with the full GDPdivide into subgroups155 Basic Headings

Start with the full GDPdivide into subgroups155 Basic Headings

E.g. Food29 Basic Headings

E.g. Food29 Basic Headings

Define productsmatching “like with like”

collect prices

Define productsmatching “like with like”

collect prices

about 10 per Basic

Heading

about 10 per Basic

Heading1,000+

products

1,000+ products

The World Bank 20

The Structured Product DefinitionDefine Products

The World Bank 21

Data Collection

• Utilize CPI framework if possible• supplemented to obtain national

coverage– Product coverage– Annual estimates

National Annual Averages vs time to time change

The World Bank 22

Number of items priced by regionCategory Africa Asia W. Asia LAC CIS Ring

Food and non-alcoholic beverages 356 223 353 147 198 281

Alcoholic beverages and tobacco 41 19 21 8 20 30

Clothing and footwear 128 78 162 136 104 132

Housing and utilities 21 17 12 18 22 35

Furnishing and household equipment 95 85 83 77 91 124

Health 144 112 69 51 75 162

Transportation 55 65 29 33 47 96

Communication 19 19 12 8 16 28

Recreation and Culture 49 70 59 54 79 96

Education 7 7 11 10 7 7

Restaurants 51 25 20 14 45 60

Misc goods and services 34 56 31 22 36 44

Total 1000 776 862 578 740 1095

.

The World Bank 23

Purchasing Power Parities - The Steps

Start with the full GDPdivide into subgroups155 Basic Headings

Start with the full GDPdivide into subgroups155 Basic Headings

E.g. Food29 Basic Headings

E.g. Food29 Basic Headings

Define productsmatching “like with like”

collect prices

Define productsmatching “like with like”

collect prices

about 10 per Basic

Heading

about 10 per Basic

Heading

1,000+ products

1,000+ products

Calculate Basic Heading parities

Calculate Basic Heading parities

Regional PPPs atBasic Heading level

Regional PPPs atBasic Heading level

The World Bank 24

Calculation of PPPs

1. First step-compute BH PPPs

2. Then combine / average Basic Heading

PPPs to aggregates such as food,

clothing, and on the consumption and the

GDP

3. ICP Handbook,

www.worldbank.org/data/ICP

The World Bank 25

Basic Heading Country A Country B Country C Rice Prices in Prices in Prices in Price ratios

Nat'l Nat'l Nat'lCurrency Currency Currency

Product B/A C/A C/BLong grain rice 10 40 100 4 10 2.5Medium grain rice 12 16 1.33Small grain rice 15 15 30 1 2 2imported 25 100 4

Geometric means/ PPPs 1.75 4.31 2.24

PPPs using direct and indirect measures 1.81 4.17 2.31

b/a = (c/a)/(c/b)

Countries have equal weights. Products have equal weights.

Basic Heading PPPs

The World Bank 26

PPPs for Basic Headings

Two approaches

– EKS method• OECD-Eurostat countries and the CIS

Region

– CPD method• Used in all other Regions

The World Bank 27

BC

CA

BA

PPPPPPPPP

• Bilateral PPPs are calculated for each pair of countries in the region.

• They are the geometric averages of the price ratios of the individual products included in each Basic Heading.

• If not all the products were priced by all the countries:• Bilateral PPPs obtained in this way will not be

transitive.• If they are not transitive, then for the three countries

A,B and C.

Elteto-Koves-Sulc (EKS) Method

The World Bank 28

Bilateral PPP made transitive by:– Taking the geometric average of the PPP

calculated directly between A and B.– And all the PPPs for A and B that can be

calculated indirectly.

With 3 countries we have:3

1

1

CB

CA

AB

BA

transitiveB

A

PPP

PPP

PPPPPPPPP

Elteto-Koves-Sulc (EKS) Method

The World Bank 29

The CPD is a regression model which assumes:– - Individual product PPPs within a Basic

Heading are constant between any given pair of countries

– - Each country has overall price level that determines the absolute prices of products in the Basic Heading for that country

CPD Method

The World Bank 30

• When there are c countries and n products, the model can be written (in logarithmic terms) as:

• is the price in country i of product j, ai is the country parameter, bj is the product parameter, and uij is the residual term. Results are transitive

ijjiij ubap lnlnlnln

CDP Method

njci ...,2,1;,...,2,1 ijp

The World Bank 31

Combine Basic Headings

Countries have equal weight. EKS used to put all bilateral pairs into Multilateral comparison with transitive results. Do not have additivity

The World Bank 32

Aggregation—Different methods available

• EKS—per example. Weights based on relative shares, thus countries treated equally regardless of economic size.– Components not additive to total

• G-K—Geary-Khamis. Uses quantity weights. Results are additive, however, can overstate real expenditures in poor countries. (Gershenkron effect)

The World Bank 33

Geary-Kharmis (GK) Method• Based on the simple idea of estimating a unique

set of international prices and these, like constant prices in a time series, are then used to revalue expenditures in each country.

)( Gip

C

jC

j

ji

ji

jG

jiG

i

q

qx

PPP

pp

1

1

.

Aggregating PPPs

The World Bank 34

Geary-Kharmis (GK) Method• There are i to n products and j to C countries; G

refers to the group of countries. The group (international) price is a quantity-weighted average of the prices in the C countries with these prices converted to a common currency. The PPP is defined as:

n

i

ji

Gi

n

i

ji

ji

jG

qp

qpPPP

1

1.)( Gip

Aggregating PPPs

The World Bank 35

Geary-Kharmis (GK) Method - Problem• Gerschenkron effect

– Tends to overstate the volumes of countries whose relative prices are very different from the uniform set of prices.

– Tends to resemble the price structure of the larger and richer countries in the group.

– Volume measures obtained for poor countries appear artificially large.

Aggregating PPPs

The World Bank 36

Purchasing Power Parities - The Steps

Start with the full GDPdivide into subgroups155 Basic Headings

Start with the full GDPdivide into subgroups155 Basic Headings

E.g. Food33 Basic Headings

E.g. Food33 Basic Headings

Define productsmatching “like with like”

collect prices

Define productsmatching “like with like”

collect prices

about 10 per Basic

Heading

about 10 per Basic

Heading

1,000+ products

1,000+ products

Calculate Basic Heading parities

Calculate Basic Heading parities Regional PPPs at

Basic Heading level

Regional PPPs atBasic Heading level

Combine Basic HeadingPPPs to subgroups,

then to GDP

Combine Basic HeadingPPPs to subgroups,

then to GDPFirst Regional, then Global GDP PPPsFirst Regional, then Global GDP PPPs

The World Bank 37

Linking regions for Global Comparison

The previous steps done independently by each of 6 regions

Selected countries from each region collect prices using a product list reflecting the world as a whole

The World Bank 38

Ring to link regions

• Combine regional lists to create new set of global specifications– Updated regional SPDs to global set– Supporting software

• Develop linking methodology—see ICP Handbook www.worldbank.org/data/icp

The World Bank 39

Number of items priced by regionCategory Africa Asia W. Asia LAC CIS Ring

Food and non-alcoholic beverages 356 223 353 147 198 281

Alcoholic beverages and tobacco 41 19 21 8 20 30

Clothing and footwear 128 78 162 136 104 132

Housing and utilities 21 17 12 18 22 35

Furnishing and household equipment 95 85 83 77 91 124

Health 144 112 69 51 75 162

Transportation 55 65 29 33 47 96

Communication 19 19 12 8 16 28

Recreation and Culture 49 70 59 54 79 96

Education 7 7 11 10 7 7

Restaurants 51 25 20 14 45 60

Misc goods and services 34 56 31 22 36 44

Total 1000 776 862 578 740 1095

.

The World Bank 40

Results Published by Three Regions

• LAC—Full GDP—10 countries

• Asia—Full GDP—23 countries

• Africa—Consumption—48 countries

• Remaining to follow Oct/Nov. for regional data.

• Global results--December

The World Bank 41

Examples how PPPs are used

World Bank uses PPPs to establish international poverty lines

IMF measures size of economy and aggregate growth rates in PPP terms

EU allocates Structural Funds to member states

System of National Accounts calls for real comparisons between countries to be done using PPPs

The World Bank 42

Examples how PPPs are used

WHO and UNESCO use PPPs to measure health and education expenditures

Compare per capita expenditures for health, education, etc.

The World Bank 43

Other Applications of PPPs

PPPs differ from sector to sector allowing price levels for components of GDP to be compared across countries

Use comparative ratios of investment to GDP to evaluate economic growth

Evaluate living standards using PPP per capita expenditures

The World Bank 44

Sustainability—www.worldbank.org/data/ICP

• ICP Handbook and Operational Manuals

• Software and supporting documentation

• Numerous research papers

• Newsletter

Thank you

Questions????????????????/

The World Bank 46

Fundamental Principles of the ICP

• Comparability of results between countries

• Common understanding of data sharing and confidentiality

• No other international statistical program requires as much cooperation between international, regional and countries

The World Bank 47

The ICP: New Benchmark Estimates

The International Comparison Program, now underway, is a global initiative to produce new PPPs

Includes over 100 countries plus adding in 40+ Eurostat/OECD — 1000+ items priced

New methods of data collection and estimation will improve quality

New PPPs will allow recalibration of estimates of poverty and sizes of economies

The World Bank 48

Table 3. CO² emissions per unit of GDP(tons per Dollar of GDP)Country Using PPPs Using Exchange Rates

Sweden 0.20 0.20

Zambia 0.21 0.53

France 0.23 0.27

Brazil 0.23 0.49

India 0.35 1.90

Mexico 0.39 0.60

United States 0.54 0.54

China 0.66 2.50

South Africa 0.73 2.28

Ukraine 1.09 6.92

Russia 1.17 4.65

The World Bank 49

World Bank Atlas Method (Exchange rate based)

e = / + / + *t

P t

P t-2

P t

P t-2

S $

S $e t-2

1

3

P t

P t-1

P t

P t-1

S $

S $e t-1 e t

P t = GDP deflator year t

e t = Annual average x-rate

P t = IMF Special Drawing Rights (SDR) deflator in US $

S $

Three year moving of annual average exchange rates adjusted by The ratio of the GDP and IMF SDR deflators

The World Bank 50

Number of products to price in each Basic Heading

• Depends on homogeniety of products with the Basic Heading

• Fish and sea food vs.

• Milk and eggs

• Time and cost also factors

• Need 5 minimum up to 20+

The World Bank 51

Why We Need New Estimates of PPPs

Existing PPPs are based on price surveys carried out during the 1993-96 period and updated by comparing domestic price levels

They include estimates for countries that did not participate. For example, China, India, and many African countries were not included in the previous surveys

Size and structure of economies have changed

The World Bank 52

Multilateral Comparison

The parity between any two countries does not depend upon which country is used as the base for the comparisons

Final PPPs are based on a composite of direct and indirect parities

Prices from one country affect PPPs for all other countries

The World Bank 53

15) ln pij = ln + ln 2 xi2 + ln 3 xi3 + …ln c xic +

  ln 2 y2j + ln 3 y3j + … ln n ynj + ij

Country Product Dummy (CPD)

Uses all data—each price receives same weightProvides estimates of Standard errors of PPPs

pij = j i ij i = 1, 2, … n : j = 1, 2, … c 

(12) 1 = 1 = 1

Model can be extended to use weighting based on Expenditure weights for example, or whether or notPrice is for a representative product.

The World Bank 54

Combine Basic Headings

The World Bank 55

Different methods can be used

• EKS—Elteto, Koves, and Szulc– Provides results that are base country invarient

and transitive• CPD—Country Product Dummy

– Regression method-results are invarient and transitive

– Provides measures of variability– Used for this round

• Both provide same results if there are no missing prices

The World Bank 56

PPPs for non-benchmark countries(ICPY) = b1(Atlas Y) + b2(HSER) + e

ICPY PPP adjusted GNI per capita

Atlas Y Atlas GNI per capita

HSER Gross High School enrollment

e Residual term

PPPs for non-benchmark countries in ICP

The World Bank 57

Statistical basis of the ICP

155

1i

GDP country A =

155

1i

N

j 1PAij QAij

GDP country B =

155

1i

N

j 1

PBij QBij

GDP country AB

using B prices =

155

1i

N

j 1

PAij . (PBij / PAij ). QAij

Direct comparisons of country A with country B:

Fundamental Principles

Products must beComparable

Using same pricingConcepts

Same time periodNeed to share data

The World Bank 58

Statistical basis of the ICP

155

1i

GDP country A =

155

1i

N

j 1PAij QAij

GDP country B =

155

1i

N

j 1

PBij QBij

GDP country AB

using B prices =

155

1i

N

j 1

PAij . (PBij / PAij ). QAij

Direct comparisons of country A with country B:

Problem—PPPs Owner occupied Housing---Estimate quantities

The World Bank 59

What are PPPs?

Purchasing Power Parities based on relative price levels between countries

A method to convert national GDPs to a common currency for comparison purposes using national prices instead of exchange rates

The World Bank 60

Statistical basis of the ICP

155

1i

GDP country A =

155

1i

N

j 1PAij QAij

GDP country B =

155

1i

N

j 1

PBij QBij

GDP country AB

using B prices =

155

1i

N

j 1

PAij . (PBij / PAij ). QAij

Direct comparisons of country A with country B:

Problem—PPPs Owner occupied Housing---Estimate quantities

Quantity sq.feet,Number rooms

Quality, water, Electricity, etc

Traditional vs.Modern

The World Bank 61

Real Expenditure Shares Asia =100

010203040506070

PRC IND IRN INO THA PAK

GDP

AFCH

GFCF

The World Bank 62

Relative Per Capita--Asia = 100

0

50

100

150

200

250

300

350

PRC IND IRN INO THA PAK

GDP

AFCH

GFCF

The World Bank 63

Real Expenditure Shares Latin America =100

0102030405060

GDP

AFCH

GFCF

The World Bank 64

Relative Per Capita-LAC = 100

0

50

100

150

200

GDP

AFCH

GFCF

The World Bank 65

Consumption Expenditure Shares Africa =100

0

5

10

15

20

25

SA EGY NIG SUD MOR Tun

EXP

The World Bank 66

Consumption Expenditure per capita relative

Africa =100

0

100

200

300

400

PerCap

The World Bank 67

jkjk nn

i ij

ikjk p

pPPP

1

1

)3(

Purchasing Power Parity betweenTwo countries

The World Bank 68

Statistical basis of PPPs

GDP country A =

155

1i

N

j 1PAij QAij

GDP country B =

155

1i

N

j 1

PBij QBij

GDP country Ausing B prices =

155

1i

N

j 1

PAij . (PBij / PAij ). QAij

Direct comparisons of country A with country B:

Exchange rates or PPPs ?

The World Bank 69

The Steps

Region Consumption Specifications And number of Products priced

Housing, Gov’t comp, Equipment, Construction specs

Within region PPPs

Africa—48 Regional 1000 Global PPPs Africa Asia—23 Regional 778 Global PPPs

Asia LAC—10 Regional 678 Global PPPs LAC W. Asia—11 Regional 862 Global PPPs W. Asia CIS—10 Regional 740 Regional PPP CIS EU/OECD Regional Regional PPP Eurostat/OECD

The World Bank 70

Region Consumption specifications

Housing, Gov’t comp, Equipment, Construction, specs

Within region PPPs

Ring countries

Africa—48 Regional Global PPP af 6 Asia—23 Regional Global PPP as 4 LAC—10 Regional Global PPP

lac 2

W. Asia—11

Regional Global PPP wa

2

CIS—10 Regional Regional PPP cis *(Russia) EU/OECD Regional Regional PPP

eu/oe 2

Ring Program

Global specs, composite of regions--1086

Global specs from above

Between region PPPs

Convert regional PPPs to Global Level

The World Bank 71

Region-Country

Within region PPP

Ring Prices

Ring prices in regional currency

Between Region PPPs

Global PPPs

1-a PPPa = 1 PPP1 = 1 PPPa 1-b PPPb RPb RPb/PPPb PPPb 1-c PPPc PPPc 1-d PPPd PPPd 1-e PPPe RPe RPe/PPPe PPPe 1-f PPPf PPPf 2-g PPPg =1 PPP2 PPP2*PPPg 2-h PPPh PPP2*PPPh 2-i PPPi RPi RPi/PPPi PPP2*PPPi 2-j PPPj PPP2*PPPj 2-k PPPk RPk RPk/PPPk PPP2*PPPk 3-l PPPl =1 PPP3 PPP3*PPPl 3-m PPPm RPm RPm/PPPm PPP3*PPPm 3-n PPPn RPn RPn/PPPn PPP3*PPPn 3-o PPPo PPP3*PPPo

Methodology to Link Regional BH PPPs

The World Bank 72

• Gaps - not all types of rice are sold in all countries A/B rice PPP may be based on a different set of rice products from the A/C and B/C rice PPPs

• Set of PPPs will not be consistent with each other. They are not “transitive.” PPPA/B ≠ PPPA/C * PPPC/B

• Comparison between any two countries should not depend upon choice of base country

A Complication

The World Bank 73

Calculate Basic Heading Parities

Parities are calculated from average prices in national currency for products in each of 155 basic headings of the GDP

Prices by Country

Rice A B C

Long grain 10 40 100

Short grain 12 16

Parboiled 15 15 30

Imported 25 100

The World Bank 74

nn

kjl kl

jljkjk PPP

PPPPPPEKS

1

,

2

Method to make direct parities = indirect parities

The World Bank 75

Moving from Bilateral to Multilateral

The parity between A and B can be made transitive by using the indirect comparisons between other countries:

• Comparison of country A and C (C/A)• Comparison of country B and C (C/B)• The ratio of these two (C/B)/(C/A)=B/A

which is an indirect estimate of the parity between A and B

The World Bank 76

Gross Domestic Product

• Represents size of a country’s economy and structure

• Per capita measures provide evaluation of economic well-being

• Compare sizes and components such as investments, between countries to evaluate growth and development

• Compare income--$1/per day for poverty comparisons.

The World Bank 77

Inter country comparisons of Agricultural measures

• Total value of production

• Per capita

• TVP / GDP comparisons across countries.