Volume 3, No. 3, October 2006 Newsletter 39654 A quarterly e-publication of the International Comparison Program In This Issue Cover Stories ............................................. 1 Building Strategic Partnerships to Sustain the ICP · Building Strategic Partnership to Sustain the Interview with Shaida Badiee, ICP - Interview with Shaida Badiee Director, Development Data Group, World Bank · Purchasing Power Parity Exchange Rates for Counting the Poor How does the International Comparison Program (ICP) contribute to our understanding of development? Notes from the Global Manager .............. 2 The ICP contributes to development work in three important ways. First, the results of the ICP provide a crucial information News in Brief ............................................. 2 base for research and policymaking. Accurate measurement of · Latin America Releases More Detailed real output, the structure of economies, and the relative pur- Consumption Data chasing power of currencies are essential for understanding · Moscow Workshop for Regional Coordinators development. For example, purchasing power parities underpin the widely used · Executive Board Discussion in Paris $1-per-day international poverty threshold. Second, the ICP also encourages · Independent Evaluation of the ICP global data harmonization by demonstrating rigorous compliance with global · Progress Report on Regional and standards for statistical classifications, data collection, processing, and report- Ring Work Online ing. Finally, the program contributes to, and benefits from, national capacity building initiatives, particularly in national accounts and price statistics. >> Feature ....................................................... 3 · The Other Face of ICP ... continued on page 5 Regional Spotlight ..................................... 9 · Relevance of Poverty PPPs in the Asian Development Bank Purchasing Power Parity Exchange Rates for Counting the Poor Angus Deaton, Dwight D. Eisenhower Professor of International Affairs and Methodology .............................................. 10 Professor of Economics, Princeton University · A New Data Collection and Validation Soft- ware Tool ­ For Equipment, Construction Global Poverty Counts and Compensation Items One of the most closely watched statistics in the develop- ment community is the global poverty count: the number of Back Page ................................................... 12 people in the world who live on less than $1 or $2 a day in · ICP in Western Asia Stays on Target 1993 Purchasing Power Parity dollars. When the World Bank despite Crisis in the Region statisticians do the global poverty calculation, the first step is to use Purchasing Power Parity (PPP) exchange rates to convert the $1 and $2 international lines into the local currency equivalent poverty line for each country; note that these will typically be different from the national poverty lines that countries use for their own domestic purposes. Country household surveys are then used to estimate the number of people in each country who live in households whose per capita total expenditure (or per capita income when expenditure is not available) is less than the local PPP value of the international lines. The dollar-a-day line itself was originally chosen because it was close to the PPP equivalent of the national poverty lines of a group of poor countries around the world. >> ... continued on page 7 Published in Arabic, English, French, Russian & Spanish www.worldbank.org/data/icp Volume 3, No. 3 ICPNewsletter ehT Notes from the Global Manager News in Brief Dear Readers, Latin America Releases More Detailed Consumption Data The Latin America region was the first to publish preliminary PPP The highlights of this edition are the contributions by Alan data in June 2006 for household consumption, including health and Heston and Angus Deaton, both eminent members of the education. The preliminary data included PPP estimates and related Technical Advisory Group that is providing unparalleled sup- data such as per capita expenditures. The new release in October port to the International Comparison Program. Special thanks 2006 is restricted to Price Level Indexes and PPP rates, and provides to both for their discussions about the origins of the ICP and detailed information covering 70 selected Basic Headings. Data for methodology to improve PPP-based measures of the number total GDP including Consumption, Government Expenditure, of people living below the poverty line. Bishnu Pant's con- Construction, Machinery and Equipment, and their sub-aggregates tribution adds a user's dimension. Shaida Badiee shares her will be published in early 2007. The new data can be accessed from views on building strategic partnerships to sustain the ICP. the ICP website. The recent publication of preliminary Purchasing Power Moscow Workshop for Regional Coordinators Parities (PPP) and GDP indices from Latin America is rais- TheICPregionalcoordinatorsandGlobalOfficestaff metinMos- ing expectations about what is to come. The Global Office cow from July 24-28, 2006. A primary purpose was for each region has received numerous enquiries about when the remaining to present their average prices and preliminary PPPs by basic head- regional and global data will be available. In addition, some ing to the other regions. This is an essential exercise to ensure con- want advance or informal information, such the likely PPP for sistent methods are followed across regions. This review included Country A. Others are requesting access to average prices. prices and expenditure weights from the regional comparisons, and Regional coordinators have raised questions about how their a first look at prices from the Ring Comparison. The discussion regional PPPs and expenditure shares will be reviewed dur- also included the steps to be followed to calibrate the regional PPPs ing preparations of the global results. Till now, each region's to a common international currency and base. The policy to publish data have been restricted to Global Office access only. The both preliminary and final results was also confirmed. question becomes who sees what, when the regional data are Executive Board Discussion in Paris combined into a global database to prepare global PPPs. The ICP Executive Board met on June 16, 2006 in Paris. Its main As a result, the Global Office has documented policies defin- actions were to direct the Global Office to finalize the data access ing access to the different data sets. The fundamental prin- policy statements, prepare a work plan leading to the publication of ciples followed by national and international organizations the global results for its consideration, initiate plans for an indepen- regarding data confidentiality and the release of official statis- dent evaluation of this round, and to begin addressing the future tics form the backbone of the policy. These policies are being of the ICP program. The Board again meets on October 17-18 in shared with the ICP Executive Board and regional coordina- Washington to take final action on these matters and establish a tors, after which they will be posted on the ICP website. date for the first global data publication. Regions are being reminded of the data quality and time re- Independent Evaluation of the ICP quirements for inclusion in the global results. The Executive This round is unlike any ever completed. A new governance struc- Board will receive a status overview for each region when ture that included the formation of a Global Office (GO) sup- it meets on October 17-18, 2006. It will then set a date for ported by an Executive Board and Technical Advisory Group was when the first global results will be published. Regions and formed. New methodology starting with the development of the countries meeting the data quality and timetables will be in- Structured Product Descriptions, and followed by new or improved cluded in these first results, with the remainder brought into methods for equipment, construction, and housing, was developed the comparisons when final results are published. and implemented. Numerous software systems were developed to The need for quick action regarding the future of the ICP is support many data collection and validation activities. In the light being raised by the regional coordinators and countries be- of these many new developments, the Executive Board has di- cause they have devoted so much energy to the effort. Their rected the GO to prepare Terms of Reference for an independent point is that unless there are plans for the program to con- evaluation. Commissioning a comprehensive evaluation, covering tinue, staff in the national offices will slip away resulting in all aspects of the program at the national, regional and global lev- a huge loss of knowledge and experience. The Global Of- els requires careful planning and a substantial budget. The GO is fice will soon initiate an evaluation of this round to guide a preparing the Terms of Reference and mobilizing the required re- smooth transition into the next one. sources to carry out this mandate. I want to close with special thanks to the 19 countries that Progress Report on Regional and Ring Work Online are collecting prices for Ring List products. This work has Regional validation of GDP weights and results of the Government required special commitment because it is on top of activities Compensation survey for the Eurostat/OECD Ring Countries are still underway to complete the regional comparisons. being finalized. In Asia, expert groups associated with Construc- tion and Equipment price surveys are being formed to assist in Fred Vogel regional data validation. Visit the ICP website for more regional and Ring data collection updates: www.worldbank.org/data/icp. www.worldbank.org/data/icp October 2006 A quarterly print and e-newsletter for the International Comparison Program Feature The Other Face of ICP lister, 1957, for China). The final decision to go with the expen- Alan Heston, Professor Emeritus of Economics, diture side meant that production comparisons can only be made University of Pennsylvania imperfectly at the GDP level, whereas much analytic interest is in sectoral productivity comparisons. When the United Nations International Compar- ison Project (ICP) was established in 1968, it was Use of Direct and Indirect Price Comparisons and Unit Values envisaged that purchasing power and real prod- On both the expenditure and product side, it is often necessary uct comparisons would be carried out among to use indirect, in contrast t.o direct, price comparisons, another member countries. There were several choices decision faced by the ICP. In the two-country case, the price com- to be made. Would the comparison be made for parison for any item in countries A and B is: all of GDP, which was certainly the choice of ICP proponents? More fundamentally, should the comparison be PA/PB = (ExpendituresA/QA})/(ExpendituresB/QB) made from the expenditure side of the national accounts or the production side, since both types of comparisons were envisaged If we make a direct price comparison as is done for consumption in the establishment of the ICP? This note focuses on this ques- items like a dozen eggs, then this can be used to derive QA/QB by tion and how it has played out. dividing the expenditure ratios of A/B by the price ratios. At the time, most countries used specification pricing in their CPI collec- When this decision was made, there were different models for tion, comparing over time the price of the same item in a written both types of comparisons. One widely used academic study at specification. It seemed desirable and a reasonable statistical goal the time was the Conditions of Economic Progress by Colin to apply the same technique to comparing prices across space. Clark. His work emphasized the production side of the accounts and built up comparisons at the level of national income and Returning to the basic identity, it is also clear that there are two expressed them in what he termed International Units for some other indirect ways to obtain the same information. One is to di- 36 countries for the 1925-34 period. International Units had the rectly compare quantity indicators, say hospital bed days per 1,000 purchasing power of the US$ over those years. There are various persons in both A and B, and take this as a direct quantity com- degrees of detail that can be used in production side studies and parison. In this case, the estimation of PA/PB would be indirect, Clark's work covered only the single digit sectors. dividing the expenditure ratios by the directly estimated quantity ratios, QA/QB. The ICP uses these when direct price compari- PPP Comparison of the 1950-65 Period sons are very difficult, like health and education. Milton Gilbert was instrumental in fostering two purchasing power studies, the first involving the United States in binary ex- Still another form of indirect comparison used in the expenditure penditure comparisons with France, Germany, Italy and the UK. approach of the ICP, but even more so, in estimating PPPs from (Gilbert and Kravis, 1954). The second was a study by Paige and the production side, are unit values or indirect measures of unit Bombach (1959) that was a model for subsequent industry-of- values. A unit value estimate of PA is (ExpendituresA/QA}). The origin studies, focusing mainly on manufacturing. Several parallel ratio of two unit values is clearly another type of price ratio. Unit purchasing power studies from the expenditure side were carried values are very commonly available from household expenditure out in the 1960s, including the Economic and Social Commission surveys, which have been used in poverty work by Angus Deaton of Latin America of the United Nations (1963) and the Council and Prasada Rao. These readily produce estimates for household for Mutual Economic Assistance (CMEA). purchases like eggs and grains that are needed for either the ex- penditure or production approaches. These purchasing power studies were landmarks leading to the establishment of the ICP, and CMEA, ECLA and OECD partici- However,forthegovernmentsector,therearefeweasilyidentified pants all took part in the early planning of the 1970 benchmark quantities. As a consequence, there are no direct quantity ratios comparison. The early benchmark comparisons were from the from which indirect prices can be derived, nor can unit values be expenditure side as they appeared easier from the data point of estimated. Typically, price ratios are approximated by comparing view, so long as exports and imports are not treated in detail. But wages and salaries for similar occupational categories in different expenditure comparisons are easier only if you impose standards countries. The underlying assumption is that the government ser- of comparison as high as those in the Paige-Bombach study for vices produced by, say, a mechanic, will be the same across coun- production side studies. There are also very simple production tries so that a comparison of mechanics' wages approximates a side comparisons that can be carried out where most of the price comparison of prices of output of government services. comparisons are indirect, making assumptions about the relation of wage rates to productivity in pairs or groups of countries. This Use of Unit Values from the Production Side technique underlay most of Colin Clark's comparisons. And it was Angus Maddison was at the OEEC at the time of both the Gil- the technique that the US CIA and other defense-related agencies bert-Kravis and the Paige-Bombach studies, and has always had a accepted from consultants for key countries (for example, Hol- ... continued on page 4 www.worldbank.org/data/icp Volume 3, No. 3 ICPNewsletter ehT Feature Alan Heston .... continued from page 3 tors. These examples illustrate why, difficult as they are to carry strong interest in productivity trends. While at the University of out, that production side comparisons are such an important Groningen, Maddiston took the lead in stimulating production aspect of the ICP. side studies utilizing unit values as the basis for PPPs. His stu- dents, most notably Bart van Ark, expanded the number of coun- Fortunately, the work of the Groningen group is widely dis- tries and improved the methodology at The Groningen Growth seminated in terms of productivity growth in a joint effort with and Development Center (GGDC). Their website provides their The Conference Board (www.conference-board.org). An annual methodological papers and estimates of sectoral productivity of GGDC report is published as part of The Conference Board capital and labor on a per person and on an hourly basis. (www. program; it covers output per employee for 97 countries and GGDC.net). output per hour for a smaller number of industrial countries. It offers comparative levels of productivity as well as growth, and Because GGDC builds their estimates up from the production has been well received. rather than the expenditure side, it requires the estimation of in- dustry or sector-specific PPPs. The GGDC, in collaboration with Some Shared Problems and Potential the National Institute of Economic and Social Research (Lon- There are overlapping areas in the production and expenditure ap- don), has undertaken several such studies making extensive use proaches that simplify harmonization. For better or worse, both of producer-based unit values, obtained from manufacturing pro- approaches face similar problems in estimating PPPs for the gov- duction censuses and industry surveys (see Van Ark and Timmer, ernment, education and health sectors. Other overlapping areas 2001, for a review). Only a few attempts have been undertaken to where the production approach tends to accept the expenditure obtain a full reconstruction of industry PPPs which can be ag- side estimates are construction and producer durables. One rea- gregated up to the GDP level, and for a small subset of countries son is that unit values are not readily available for the construction only (Jorgenson 1995; Madison and van Ark, 1989, 2002). sector. And while unit values are available for machinery, the cat- egories are typically very heterogeneous so that diesel engines of Advantages of Production-side Estimates different capacities may be included in the same quantity. There- The ICP has focused on the expenditure-side concept, by collect- fore, use of specification pricing as in the expenditure approach ing price and budget shares for final expenditure categories, and can be used for these sectors in the production approach. An ob- using these to construct PPPs and deflators for nominal GDP. vious example is transport equipment where the quantities are But many users of PWT have in mind the output-side concept, very heterogeneous, while it is relatively straightforward to com- as when they use real GDP to construct and compare country pare prices for a given specified model or estimated prices from a productivities, which is why developing separate measures for the hedonic regression. Because the price concept for the production two concepts of real GDP is a goal of some importance. approach is value-added, the expenditure PPPs for these sectors require some adjustments, but are otherwise suitable. Consider the aggregate GDP measure from the expenditure side. Typically, one finds that for countries like the Netherlands or Other complementarities between the expenditure and produc- Norway, the output per worker is significantly higher than the tion approaches exist, and one example is the distributive trades. US. In 2004, the Netherlands was 103.8 and Norway 130.2% of Although distributive services are not directly priced in the expen- the US labor productivity at the level of GDP, while for GDP diture approach, they are important in explaining relative prices per person it was 73.1 and 94.5%. For the Netherlands, this seem- across space. In their studies of the retail and wholesale trade ing paradox is clearly related to labor force participation rates sectors, the GGDC has found very substantial differences in pro- of women and total hours worked. Not only is the workforce a ductivity across countries. These are highly informative for the higher percent of the population in the United States, they typi- expenditure approach because the productivity differences are cally work more hours per year. frequently a function of scale of outlet. More efficient marketing sectors translate into lower final prices to consumers than would While this is part of the story, one really needs to look at the otherwise be the case. various manufacturing sectors to fully understand what is going on. For example, in Belgium, the productivity per hour is lower Hedonic regressions provide estimates of the weight that vari- than in the United States in all industries, yet average productiv- ous product or service characteristics have in determining price. ity per worker is 16% higher than in the US. The explanation These price-determining characteristics will be partly shared turns on the composition of Belgium production that is heav- acrossbothfinalandintermediategoodsandservices.Thismeans ily into steel, often special steels, where their productivity per that there is potential to use common information about the im- worker is less than in the US, but is more than most industries portance of different characteristics across both approaches. At in either country. So the importance of steel for Belgium drives present, this at best is a research topic, but with the increasing use the average labor productivity of the economy above that of of hedonics in price comparisons, applications may not be far in the US where production is spread out more evenly over all sec- the future. >> .... continued on page 8 www.worldbank.org/data/icp October 2006 A quarterly print and e-newsletter for the International Comparison Program Cover Story Shaida Badiee ... continued from page 1 The current program's success truly hinges on the collective commitment of all partners. The World Bank provides financial, How does this round of the ICP differ from previous rounds? One of the defining elements of the current round is an im- technical, and managerial support to the ICP Global Office, but proved governance structure that emphasizes joint ownership no one institution can undertake a complex global program such and close collaboration between national statistical offices, re- as the ICP alone. In keeping with the principles of promoting gional organizations and international agencies. widespread ownership and shared responsibilities, the Bank sees its primary role as broadening and deepening global partnerships In the past, there was little input from participating countries. In so that a more robust basis for sustaining the work of the ICP contrast, the current round is owned and managed by a consor- can be established. tium of national, regional and international organizations, under One of the issues raised in the Ryten evaluation of the ICP was the auspices of the ICP Executive Board. The Board is com- the classic public goods dilemma when it comes to securing sus- prised of distinguished statistical managers and experts, reflect- tainable funding. How is this being addressed? ing a balanced representation of stakeholders from all regions of the world. As a result, the program has become an example The problem of building a coalition to support a public good of a collaborative effort, moving away from isolated ad-hoc ini- remains the most challenging impediment that the ICP needs to tiatives and promoting instead strategic partnership and a pro- overcome. The Ryten report rightly calls upon the international grammatic approach. development institutions, as primary users of the data, to rise to the challenge and establish a cost-sharing arrangement. The And the current round has addressed data-quality problems by report makes a compelling case that a sustainable partnership means of targeted research. A Technical Advisory Group, com- must be based on the comparative advantages of the different posed of renowned experts, provided essential advice to the partners, in terms of both capacity and resources. Global Office for its methodological work. A Poverty Advisory Group has also been created to focus on the problem of mea- Our strategy has been to develop partnerships on three levels: suring the purchasing power of poor households. The advisory first with the main international organizations using the data; groups have identified a number of outstanding methodological second with the regional organizations involved in capacity problems, which are now being tackled. Two important outcomes building; and third with bilateral partners who recognize the im- will be poverty-specific PPPs and an international comparison of portance of ICP. We have been fairly successful, but more work the cost of capital formation. Notable strides have also been made needs to be done. in data validation procedures and PPP aggregation methods. The current round has come a long way in building the program's A comprehensive handbook, an accompanying operational man- credibility and we hope that this will encourage existing partners ual, and a related suite of software tools also fill critical gaps to increase their stakes and that new partners will join the cost- of previous ICP rounds. These tools have helped to ensure that sharing arrangement. We also need to provide the capabilities for common procedures are followed. Efforts have also been made countries to carry their share of the cost of data collection by in- to maximize the synergy between the ICP and national statistical tegrating the ICP work with their national CPI surveys. Last but programs. The effect of these changes will be increased statistical not least, we look to the ICP Executive Board to play a key role capacity and improved data quality. in building global support and a sustainable financial base. When do you expect the global data to be published? Can you discuss the World Bank's role in the ICP? The World Bank has supported the ICP since its inception in Originally the plan was to release preliminary global PPP esti- 1968, when it was introduced as a pilot project at the University mates in the second quarter of 2007. But some regions expe- of Pennsylvania. The Bank supported the subsequent phases of rienced delays in finalizing their PPP estimates, and because of the program as it progressively moved from a pilot project to a unforeseen problems in deriving national accounts weights, the regular program on a five-year cycle. release schedule has been pushed to the last quarter of 2007. Some participating countries are revising their national accounts In 1999, the UN Statistical Commission requested a new strategic after facing significant problems at basic heading expenditure framework and an operational plan of action, which the Bank levels. This will result in enhanced quality of PPP data and im- developed in close consultation with other international organi- proved national accounts estimates. zations. The new framework and action plan addressed the pro- The Ryten report faulted multilateral development institutions gram's institutional, organizational, operational, and methodolog- for not using PPP data for operational purposes. Do you see the ical shortcomings. Subsequently, the World Bank was mandated World Bank beginning to use PPP results from the current round by the Commission to convene partners of the ICP, mobilize for policy and operational purposes? resources, and house an ICP Global Office responsible for the day-to-day management and coordination of the current round. Recently, the Global Office was requested to prepare a briefing note for the Executive Directors of the World Bank on sources, methods, and use of PPP data for operational decisions, which ... continued on page 6 www.worldbank.org/data/icp Volume 3, No. 3 ICPNewsletter ehT Cover Story Shaida Badiee .... continued from page 5 What is your view on the future strategic direction of the ICP? was presented to a meeting chaired by President Paul Wolfowitz One of the lessons we have learned from the current round is and attended by the Executive Directors and other senior man- that the ICP should be seen as a means to an end, rather than agers. A similar briefing session was organized for the United an end in itself. The critical issue in sustaining the ICP is not Nations in New York. In general, there is a growing demand for focusing on PPPs, but rather focusing on the framework within quality PPP data to support evidence-based policy formulation which data collection, validation and processing occur. The goal and operational decision-making. should be to build a more efficient program that can help build the capacity of countries to improve national accounts and price In principle, the World Bank recognizes that PPP estimates pro- statistics upon which the PPP estimates depend, while fulfill- vide a sound basis for classifying countries by income level, as- ing the narrower objective of generating reliable, relevant and sessing the purchasing power of currencies, and hence determin- timely PPP data. ing people's command over economic resources. In practice, its position has been to use official exchange rates until the quality, We need to identify the key elements that will form the pro- timeliness, continuity, and reliability of PPP data has improved gram's strategic direction in respect of capacity building. In the sufficiently. Before migrating to PPPs, we need to ensure that the next round, our focus should be on those activities that best quality of data passes the "fitness for use" test. Rigorous testing support the harmonization of the ICP with national statistical and analysis must be conducted to ensure the robustness of the programs. The goal should be to establish a cost-effective and data. We also need to ensure the sustainability of the program. well-integrated process. Having said that, I am confident the results of the current round Naturally, future rounds should build on the experience from will meet the expectations of users. The data from the current the current round. Important lessons have been learned from round will offer a fresh opportunity to expand the use of PPP the work of Statistics Canada as regional co-coordinator for the data from research to the policy and operational realms. And Latin American region, bringing a wealth of experience in the upon confirmation of the quality of the new global data, we will management and technical aspects of the program. The Austra- prepare a position paper advising our Board of Directors to con- lian Bureau of Statistics played a key role by providing technical sider using PPP data in operational decision making. assistance for the Asia-Pacific region. The ONS of the United Kingdom and INSEE of France provided technical support to Meanwhile, we need to increase user understanding of PPP sta- ICP-Africa. We should develop similar collaborative links with tistics. The program generates masses of information on prices other experienced statistical agencies. and expenditure categories, but attention has until now focused on PPP-adjusted GDP per capita and $1-per-day poverty PPPs. Another challenge is determining the right mix and the optimal We can demonstrate that the data are useful for a wide variety of number of countries that should be grouped in a regional or policy and operational work at national and international levels. sub-regional setting. From the program management point of This presents both a challenge and an opportunity. view, this will be essential for effective and efficient coordina- tion. From the capacity building perspective, this will allow us to Is there a plan for a new round of the ICP? design a more targeted strategy and action plan. There is strong interest from participating countries and regional implementing agencies to continue the program and to begin Is there anything else you would like to add? planning for the next round. The ICP Executive Board has re- I would like to end with a note of thanks to all our partners. It questedtheGlobal Officeto commission an independent evalua- may be a bit early to celebrate, but we can take pride in the signifi- tion. The purpose of the evaluation is to identify factors that im- cant steps that have been taken to improve the credibility of the peded effective implementation and provide recommendations program, as noted by the UN Statistical Commission at its last essential to preparation for future rounds. session in March 2006. The success achieved thus far would not have been possible without the strong commitment and financial In short, there is strong support for another round. The ques- contributions of our partners. I must also note the ICP's partner- tion is when rather than if. We are advised by regional and na- ship with the OECD and Eurostat comparison program which tional coordinators to begin the next round immediately after the gave us a unique opportunity to coordinate two independently current round concludes in 2007. This is seen as essential if we run programs under a common umbrella. Last, but not least, I are to build on the momentum of the current round, capitalize wish to acknowledge, with grateful thanks, the valuable contribu- on the benefits of the capacity-building steps taken, and main- tions made by the members of the Technical Advisory Group tain the current experienced staff. The rapidly changing world and the Poverty Advisory Group. n economy also necessitates updating PPP estimates without delay. However, it is critical to take stock of the current round, to see what has worked and what needs to be improved. The timing of the next round will be addressed by the ICP Executive Board and ultimately decided by the UN Statistical Commission. www.worldbank.org/data/icp October 2006 A quarterly print and e-newsletter for the International Comparison Program Cover Story Angus Deaton .... continued from page 1 PPP exchange rates clearly play a central role in the global pov- Why is it so hard to collect prices for people near the poverty erty counts, both in setting the international poverty line itself, line? One problem is some disagreement about the facts. It is and in converting it to local currency for each country that con- often found that poor people must buy in small quantities, be- tributes to the count. Measuring the PPPs accurately is also key cause they are poor, so that they cannot benefit from the econo- because the global poverty counts are extremely sensitive to the mies of scale that richer households enjoy. Yet household sur- PPP exchange rates that go into their calculation. In the poor veys often show that the unit-values of staples such as rice or countries that contribute most to the counts, a large fraction of wheat are higher for better-off households, and while some of the population lives near the international poverty line. In such this undoubtedly reflects the higher quality bought by the rich, cases, even small changes in the local value of the line, which it casts doubt on the general importance of economies of scale could come from a small change in the PPP used to convert the in purchasing. For basic food staples such as cereals, the poor $1-a-day, will cause large numbers of people to be reclassified have every incentive to buy in bulk, and there is evidence that as poor or not poor. In 1998, the World Bank revised its global they do so. Another issue is that the poor often purchase in dif- poverty counts and switched from the PPP exchange rates of ferent markets from the rich, for example in small stores or stalls version 5.6 of the Penn World Table to its own recalculated PPP in village markets, rather than in supermarkets in the cities. To estimates. The change raised the poverty rate in sub-Saharan Af- meet this, it is important that ICP enumerators do not only col- rica for 1993 (the same year in both calculations) from 39 percent lect prices in urban areas, as has sometimes been the case in the to 50 percent of the population. past, and the current round is making great efforts in this direc- tion. The poverty group has also instituted a small experimental PPP Exchange Rates and Conversion of study in which country statisticians will try to identify and col- International Poverty Line lect prices typically paid by the poor when they differ from the The Internal Comparison Program (ICP) is essentially an exten- prices collected for the general ICP effort. sion of the international system of national accounts. Its main task is to price the components of each country's national prod- That the poor spend their limited budgets in different ways than uct in common international prices, so as to make possible useful do the rich is clearly correct, and incorporating this fact into the cross-country comparisons in the same units. The PPP exchange PPP rates is where the poverty group has made the most prog- rates that come out of this exercise are price indexes that convert ress. This adjustment is likely to be important because the poor consumption, investment, or GDP from local currencies to the sometimes buy goods that the rich do not. For example, in India, international currency. More specifically, in the Geary-Khamis many poor people eat jowar (sorghum) as their basic staple, while system that is used in the Penn World Table, the PPP exchange the better-off eat rice or wheat. We should also note that the use rate for India or for Kenya, for example, is the Paasche price of national aggregates does not represent all people equally. Be- index obtained by dividing Indian or Kenyan national aggregate cause national aggregates are the totals of all money spent, those consumption in rupees or shillings by its value in international who spend more money are represented in proportion to the PPP dollars. Such price indexes are far from ideal for converting amount that they spend, so that when the expenditure patterns an international poverty line into local prices. For this calculation, of the rich and poor differ, aggregate patterns are much more like we want to compare, not the values of aggregate national con- the former than the latter. When we want to construct interna- sumption bundles, but the values of the bundles of goods and tionally comparable poverty lines between India and Kenya, say, services that are consumed by people at or near the international we do not want the conversion factors to give much weight to the poverty line. The World Bank has always understood this prob- relative prices of automobiles, personal computers, or banking lem, and uses the standard ICP rates because more appropriate and insurance services. Getting this right will make a difference numbers are not available. That lack is something that is current- only if the international relative prices of goods consumed by ly being addressed by the poverty group within the ICP which the poor are systematically different. This would be certainly be is working, in conjunction with the current round, to produce the case if we were comparing poor people in poor countries poverty-relevant PPPs, or "PPPPs," for a range of countries. with poor people in rich countries, and if poor people system- atically consumed fewer traded goods and more non-tradeables. Why are Poverty-Relevant PPPs Different from But it is not obvious in advance that this is true, and the situation PPPs for National Accounts? is even less clear in the currently relevant situation, which is the There are two separate concerns about using the standard con- relative prices for poor people in different poor countries. sumption PPPs for poverty work. The first is that the nation- al prices collected by the ICP may differ from the prices paid How to Calculate Poverty-Relevant PPPs by the poor, or by people near the poverty line. The second The work currently underway will combine the national prices is that, even if the poor pay the same prices as everyone else, collected by the ICP with data on consumption patterns of the their patterns of consumption are different, so that the weights poor obtained from the large number of household surveys for a poverty-relevant PPP index should be different from the available to the World Bank and which have been processed over weights in an index of aggregate consumption. The first prob- the last three years into a uniform framework in a massive data lem is much more intractable than the second, on which good effort in the Bank's data group. Because household surveys are progress is being made. ... continued on page 8 www.worldbank.org/data/icp Volume 3, No. 3 ICPNewsletter ehT Cover Story Angus Deaton .... continued from page 7 its calculation. So we suspect that the sampling errors are likely to not harmonized internationally, the combination of prices and be less important than some non-sampling errors, for example, expenditure patterns can be done only at a relatively high level of in the collection of some of the prices. aggregation, either at or above the basic heading level. In prin- ciple, it is a mechanical exercise to combine expenditure weights In the next few months, the poverty team will be putting together from one source (household surveys) with prices from another the newly collected prices with the data from the household sur- (the ICP) to calculate the price indexes for the PPPs and, once veys, and we shall learn whether our methodological advances the expenditure patterns of the poor are known, there is nothing can be turned into practical estimates, and whether the PPPPs in these calculations that differs from the standard calculations differ in important ways from the PPPs. If so, we shall have of the familiar PPPs. some basis for calculating better poverty counts. And if not, the knowledge will be a useful defense against the argument that the One new issue here is how to select the "internationally poor" current global poverty counts are seriously flawed by the use of households in each survey, since it is from those that we need to inappropriate exchange-rate conversions. n obtain the consumption patterns that are to be used as weights. Identifying these households requires a PPP exchange rate to convert the $1-a-day line, but we can only know the PPP once Alan Heston .... continued from page 4 we have the weights in hand. This simultaneity problem can be solved by trial and error, starting from a provisional set of PPPs, Summing Up identifying the households at the poverty line in each country, averaging their consumption patterns, and using the results with This note has provided some background on why the expenditure the ICP prices to calculate new PPPs. The process is then re- approach to estimation of PPPs and real product was adopted peated until convergence is obtained. Under special assumptions in 1968 as the main research effort of the ICP. Since there is a about household Engel curves, it is possible to short-circuit this large interest in comparable production by sector estimates of process and calculate the PPPs directly, and these "special-case" real product that could not be filled by the ICP, it is not surprising that alternative estimates have been made outside the ICP institu- PPPs can be used as starting values for the iterative calculation. In practice, convergence appears to be rapid, and the "special tional framework, most notably the Groningen Growth and De- case" PPPs are a very good approximation to the final ones, so velopment Center. One research initiative that would clearly serve this aspect of the calculations does not present any great dif- the needs of both the production and expenditure approaches ficulty. Note also that, in any given survey, however large, we would be development of an international price archive that in- are unlikely to find households exactly at the poverty line. This cluded both final product and intermediate goods prices. More difficulty is finessed by looking at the expenditure patterns of meta-studies of hedonic regression studies across space, such as households near the poverty line, and taking a weighted aver- for dwelling rents, would also facilitate both the expenditure and age of expenditures, with weights that are larger the closer the production approaches. household is to the poverty line. If we are prepared to assume a specific functional form for the Engel curves, this averaging near Selected References the line is not necessary, since the functional form can be used to Ark, B. Van, and M.P. Timmer, 2001 "PPPs and International estimate the expenditure pattern at the line itself. Productivity Comparisons: Bottlenecks and New Directions", Joint World Bank­OECD Seminar on PPPs, 30 January ­ 2 Feb- One final issue concerns sample sizes and standard errors. While some of the household surveys contain tens and even hundreds ruary, Washington, D.C. of thousands of households, many are much smaller, and in these cases there is a real concern about the accuracy of the estimates Gilbert, M. and I. Kravis, 1954 An international comparison of na- of the consumption patterns. This concern is exacerbated by the tional products and the purchasing power of currencies, a study of the United fact that we do not use all of the households, but only those that States, the United Kingdom, France, Germany, and Italy, París, OEEC. are close to the international poverty line, and in some middle- income countries, there may not be many households whose ex- Hollister, William Wallace, 1958 China's gross national product and penditures are actually used. In our preliminary work, we have social accounts, 1950-1957, Glenco, Ill., Free Press. devoted a good deal of attention to this issue, using the sampling variability associated with the design of each survey to calculate Maddison, A. and B. Van Ark, 2002 "The International Com- survey-sampling standard errors for the poverty-relevant PPPs. parison of Real Product and Productivity," in A. Maddison, D.S. Although there is no getting around the problem of surveys with Prasada Rao and W.F. Shepherd, eds., The Asian Economies in the very small numbers of poor households, our preliminary calcula- Twentieth Century, Edwar Elgar, Cheltenham, pp. 5-26. tions have suggested that sampling error is not a major source of concern. One reason for this is that the PPP indexes can them- Paige, D. and G. Bombach, 1959 A Comparison of National selves be thought of as averages taken over a substantial number Output and Productivity of the United Kingdom and the United States, of commodity-specificprice ratios, so that the index will be more OEEC, Paris. precisely estimated than the consumption patterns that go into www.worldbank.org/data/icp October 2006 A quarterly print and e-newsletter for the International Comparison Program Regional Spotlight Relevance of Poverty PPPs in the ther. The region is starting a pilot project to conduct sensitivity Asian Development Bank analysis on the Poverty PPPs resulting from the use of alternative Bishnu D. Pant, Assistant Chief Economist, Asian Development Bank sets of price data and from the use of different aggregation pro- cedures. These analyses are designed to shed light on a longstand- The overarching objective of the Asian Develop- ing question: Will PPPs generated from prices of goods and ser- ment Bank (ADB) is to reduce poverty in its de- vices relevant to the poor differ significantly from Poverty PPPs veloping member-countries. To monitor this ob- derived from average prices of goods and services consumed by jective, it is imperative to use accurate measures the general population? This study will help identify areas for im- of poverty incidence that are comparable and can provement in future ICP rounds. be aggregated at the regional level. Such poverty measures form the basis for temporal and spatial A hallmark of Poverty-PPP work in the Asia Pacific is the partici- comparisons of poverty within the countries, across the countries pative approach, which involves countries in all stages of this pro- and through the region. Getting the poverty count right is crucial cess so that countries will take ownership of the enterprise and the to ADB as the Asia-Pacific region remains home to two-thirds of study results. A poverty basket was drawn up for the survey and it the world's poor. Governments and development agencies need contains food and non-food products commonly bought by poor these poverty measures to make informed policy decisions and households. In identifying products to be surveyed, a sub-regional monitor progress in the fight against poverty. approach was adopted to ensure that the different consumption patterns across the region are considered. The 17 countries were There are difficulties in com- divided into three groups, namely, South Asia, paring poverty levels across the Improvement in the accuracy of Poverty PPPs Mekong and Others. Experts from each group countries since economic ag- "will lead to more focused country-specific policies and met to draw up a list of products relevant to gregates such as consumption assistance strategies for its developing member-coun- the consumption basket of the poor. The lists and gross domestic product tries... The sustainability of the ICP is of paramount from the three groups were then consolidated are expressed in national cur- importance as it provides standardized infrastructure to form the Asia-Pacific poverty basket. Out- rencies. Converting data from a for data collection. " lets to be surveyed will be limited to those fre- national currency to a common quently visited by the poor, such as weekly or currency using exchange rates frequently yields results inconsis- open markets, neighborhood stalls, and small tent with actual developments in the countries being compared, as shops. The survey will be conducted nationwide in all 17 partici- exchange rates are influenced by short-term factors and subject to pating countries for the purpose of estimating a national average substantial distortions from speculative movements and, to some price for urban and rural areas separately. extent, government intervention. The Purchasing Power Parity (PPP) concept underlying the International Comparison Program The ADB highly appreciates the countries' unwavering support (ICP) is the most robust method to compare levels of economic for ICP activities, including the special study on Poverty-PPP sen- output as well as the purchasing power of the poor, directly re- sitivity analysis. Improvement in the accuracy of Poverty PPPs flecting differences in prices and volumes of goods and services will enable the ADB and its development partners to better un- in different countries. derstand country conditions, and this will lead to more focused country-specific policies and assistance strategies for its develop- The current ICP round represents a concerted global effort to im- ing member-countries. prove both the quality and the relevance of the PPP data. Twen- ty-three Asia-Pacific countries are participating in this endeavor, The ADB looks forward to the release of PPP-based poverty es- under the general coordination and management of the ADB. timates and will continue to collaborate in, and contribute to, the One very notable milestone in the ongoing revamp of the ICP global effort to improve the quality, relevance and timeliness of is the work to produce poverty-specific PPPs, under the general poverty data. It is also important to ensure that the new data and auspices of the ICP's Poverty Advisory Group (PAG). The ADB the methodology behind it are widely understood and accepted. is adopting the new Poverty PPP aggregation methodology rec- The sustainability of the ICP is of paramount importance as it ommended by the PAG, which involves making use of national provides standardized infrastructure for data collection and pro- PPPs by basic heading and re-weighting these using expenditure cessing for Poverty PPPs. The Poverty PPP series should be fea- patterns of the "poor" households in countries where poverty sible to implement, using data either collected as part of the ICP is prevalent. The ADB welcomes this improvement as a critical work or that is fairly easy to obtain, using the ICP infrastructure first step to address criticisms that general ICP PPPs do not fully with minimal effort and external support for the countries. The reflect consumption patterns of the poor across countries. availability of PPP-based poverty data on a regular basis will go a long way towards tracking progress of the ADB's poverty allevia- The ADB, closely collaborating with 17 member-countries, has tion programs. n ventured to take the development of Poverty PPPs one step fur- www.worldbank.org/data/icp Volume 3, No. 3 ICPNewsletter ehT Methodology A New Data Collection and Validation Software Tool - macro as well as a micro view of the Equipment, Construction For Equipment, Construction and Compensation Items and Compensation prices. It facilitates a very efficient mode of Vilas Mandlekar, Senior Information Officer, World Bank analysis by the regional coordinators. From a technical (software) standpoint, it has innovations in the way this complex data is structured and presented to the users. Why a new software tool? The Price Collection Module of the ICP Tool Pack was used by many countries and regions for From a data collection perspective, the software allows the user their data collection, validation and processing to view detailed descriptions. For example, for equipment goods of household consumption (HHC) items. Re- users are given three options for each specification ­ "Preferred" gions also had the option to use other software product representing tight specification for exact matching; and for HHC data collection, but primarily used the an "Alternate" product that is comparable to the "Preferred" Tool Pack's Data Processing Module for validation and process- product if the "Preferred" product is not available in the market. ing. HHC items are well defined, with short and precise specifi- The third option is "Unspecified Alternate" in the event that nei- cations and data collection requirements that include mainly the ther the "Preferred" nor the "Alternate" product can be priced in item price, unit of measure, and quantity. Countries also have had a country. In this case, countries are expected to provide a detailed vast experience in HHC item price collections through their work description including the make and model of the product closest on the CPI surveys. to the preferred product plus price-determining characteristics. This will allow an expert at the regional or global level to compare However, for data collection and validation on Equipment, Con- such unspecified items across countries and make judgments on struction and Compensation (ECC) items, there is a lacuna, in whether to accept the item as a substitute to the preferred product terms of a well-defined methodology for data collection and as it is, to accept the item with some quality adjustment factor, or analysis, and also in the software tools for this purpose. The ECC to reject the item and its price in the comparison. categories together form a substantial share of the GDP for most For inter-country price and specification validation, the DVM 0 countries and are therefore an important component of the ICP. Unlike HHC items where direct matching is required, for equip- allows all the observed prices (including the detailed price com- ment goods, for example, the approach recommended is to match ponents and characteristics) and product specifications to be like with like where possible. In practice, however, price collection displayed side by side. For equipment goods, the DVM shows templates have been designed to capture major price-determining the components of prices ­ such as installation, taxes and trans- characteristics to provide flexibility for price collection. If exact portation costs. For construction items, it shows the material, matching is not possible, it will still be feasible to make price com- equipment leasing, and labor costs, etc ­ so that the regions parisons, holding constant the main characteristics of the item, can do price comparisons, both intra- country as well as across including performance, reliability, function, technology, etc. Fur- countries. Besides price validation for PPP purposes, the sys- thermore, besides the equipment cost, the installation, transporta- tem allows comparisons of installation, tax, labor or other costs tion and tax costs are also needed to make an effective analysis of across countries that can give countries insights into their price the cost structure and comparison. structure and assess, for example, their tax rate in comparison with other countries. For Construction, in addition to the detailed description, scaled drawings, details of material costs, equipment lease costs, and la- The DVM computes the average prices from the observations bor costs are needed. Compensation items have their own data both in local currency units and in USD, or any other selected collection needs in terms of the detail to be captured to compare currency. The module also produces the variation coefficient to compensations for occupations across countries. indicate variability of average prices across countries. Displaying anomalous prices or price components in red, yellow or black, To meet these challenges, the Global Office (GO) developed two the module gives a visual indication to the user to probe further software tools ­ the ICP Data Capture Forms (DCF) for data into problem data. collection, and the ICP Data Validation Module (DVM) for intra- country and inter-country validations. These software tools pro- To illustrate these validation features of the DVM, the screenshot vide the first efforts toward standardizing and simplifying these shows a sample comparison of a product item for two countries. operations. They would also help to build capacity in this area in Yellow shaded cells show the price components for three prod- countries, particularly since many of them are now establishing ucts ­ two Air Compressors and an Air Conditioner. Regional price indices for these categories of items. data analysts have many options in selecting or discarding individ- ual price observations or choosing en masse only the "Preferred" and/or "Alternate" specified items in a comparison. What is innovative about the software tools and the analyses they support? The user interface is intuitive and simple. The user sees the ag- The DVM software is unique in that it offers simultaneously a gregate information at average level in US$ or Local Currency www.worldbank.org/data/icp October 2006 A quarterly print and e-newsletter for the International Comparison Program Methodology Unit (LCU) (colored yellow) and validates his/her data. There From a process perspective, the data collection forms ­ the DCF is an option to drill down into the details to see if the product module ­ offer an efficient way of price data transfer from coun- was the preferred model, the alternate or the unspecified model tries to the region. Instead of the countries sending the data col- (colored blue) as shown for the Air Compressor (Small). For the lection forms, which can be voluminous (25 MB) since they con- same product, he or she can drill down further to see price details tain product descriptions and images, the software captures only (colored green). In the case of "Unspecified Alternate" the user the data actually entered by the user, in much smaller text files can drill down even further to see product characteristics and val- making data transfer through email easier. The region can then ues (colored gray), where the make, model, other characteristics display the country's detailed price observations and specification and characteristic values entered are displayed. On finalizing the choices using the DCF software. n prices for a category of products, there is an option to create an average prices file in the ICP Tool Pack format that can be seam- For more information on these two software modules, please send Email to lessly integrated for PPP calculation. Vilas Mandlekar at vmandlekar@worldbank.org or to Ram Erabelly at rerabelly@worldbank.org. www.worldbank.org/data/icp Newsletter Volume 3, No. 3 ICP in Western Asia Stays on Target Editor despite Crisis in the Region Yonas Biru Tarik Alami, Regional Coordinator, ICP Western Asia, UN-ESCWA Associate Editor Despite the recent crisis in Lebanon, the ICP activities in Western Asia Amy Siew Hua Lee remain on track. The Economic and Social Commission for Western Asia (ESCWA) headquarters is located in the Lebanese capital of Beirut, Translation Editors from where it manages and coordinates the ICP program for the region. Yuri Dikhanov The program was kept on target in no small part due to the commitment Nada Hamadeh of the countries, the decision of senior ESCWA management to give ICP high priority and support, the dedication of the ICP regional coor- Please send comments, contributions, and dinating team, the assistance from the ICP Global Office in Washington, and financial requests to be included on our distribu- support from the Arab Fund for Economic and Social Development. tion list to Amy Lee slee9@worldbank.org Despite the critical war situation that Lebanon endured in July and August, the ESC- International Comparison Program WA continued implementing its scheduled work program. This was possible through The World Bank constant communication by email and telephone between evacuated staff members, 1818 H Street NW, MC2-209 local staff based in Beirut, and the Washington-based Global Office staff working Washington D.C. 20433 USA on Western Asia. The ICP Regional Office for Western Asia, in collaboration with the Global Office, is now compiling and finalizing consumption data for 2005. Consumption data will be combined with the National Accounts Basic Heading Weights, Consumer Price Indices The International Comparison Program and Regional Spatial Weights which are currently being verified, in order to compute (ICP) is the world's largest statistical initia- National Annual Price Averages and to produce Preliminary Consumption PPPs. These tive, involving 107 countries. It pro- will be published along with a regional report as the first set of deliverables at the begin- duces internationally comparable price ning of 2007. levels, economic aggregates in real terms, and Purchasing Power Parity (PPP) esti- The next phase of the ICP is extremely critical. Along with the vigorous process of cal- mates that inform users about the relative culating National Annual Price Averages and Consumption PPPs for 2005, the Regional sizes of markets, the size and structure Office will validate construction, equipment, and compensation data. Work on the Ring of economies, and the relative purchasing Comparison, which includes Oman and Jordan from the Western Asia region, will be power of currencies. An ICP Global Of- finalized. Preliminary PPP estimates covering total GDP will be produced at the start of fice, housed in the World Bank, manag- the second quarter of 2007. es the global program. National Statistical The Western Asia countries have each shown palpable interest and invested in the ICP, Offices implement the program on the whether through securing financial resources, executing the required duties within dead- ground, under the general guidance and lines, or establishing an ICP department in their National Statistics Offices. The coun- coordination of regional agencies. The tries clearly realize the extent of direct benefits at the national and international levels. Global Office works in close collabora- The countries also know that significant policy implications lie ahead when PPP data tion with the OECD/Eurostat's program shed light on the relative structure of economies, and consequently support evidence- for 43 countries, and publishes global based policymaking and monitoring of progress towards meeting desired objectives, data linking ICP and OECD/Eurostat re- such as the Millennium Development Goals. sults for 150 benchmark countries. An important point to note is that the participating countries have ownership of the program, particularly as a succession of meetings and workshops have highlighted the anticipated outcomes of this crucial program, in terms of producing a "clearer" set of The opinions expressed in The ICP Newsletter are economic indicators and a more reliable economic comparison between countries. n those of the authors, and do not represent the views of the ICP Global Office or the World Bank. www.worldbank.org/data/icp