“Clearly the change will make life more complicated for US companies.
Until now they have had the convenience of using the same currency - Dollars - whether they are paying their workers, importing parts and components, or selling their products to foreign customers....This will all change in the brave new world that is coming.” 

Barry Eichengreen 1st March 2011 - Wall Street Journal 

About The World Price Index

 Analysing international economic data using currency market exchange rates is fraught with problems. One vivid illustration of this issue was published recently:  


“Imagine that over the last four years the US economy has shrunk by 30 per cent and the Chinese economy grown at 2 per cent, not 10 per cent. Imagine UK house prices have dropped 60 per cent and commodity prices have slid back to the levels of the mid-1970s. This frozen wasteland is not some alternate universe, but what reality already looks like when expressed in the world’s remaining hard currency, the Japanese Yen.” [1]

The World Price Index
Download full report (PDF)
Despite the difficulties, cross-country comparisons have to be made on a daily basis for numerous business and political reasons. For example, measurements of the relative size of economies are used to establish the differential contributions required from countries to pay for international organisations such as the United Nations, the World Bank and the International Monetary Fund, and global companies need a reliable method of consolidating income streams measured in different currencies and ways of analysing growth in different markets over time.


The only sensible way to make such comparisons is by using an international yardstick of value based on the calculation of price indices constructed using the Purchasing Power Parity (PPP) method. PPP price indices adjust market exchange rates according to the ratio of relative prices prevailing across countries to make comparisons based on a standard currency usually a construct of ‘international Dollars’.


Unfortunately, a lack of awareness of the importance of PPP estimates means that far too many international comparisons still use market exchange rates, resulting in comparisons that reflect exchange rate changes rather than real changes in the variables under consideration. A former head of statistics at the OECD (along with many others) has pointed out there is simply no justification for using market exchange rates that produce the sort of nonsensical results referred to by Tasker above. See Henderson and Castles (2005)


One major problem facing commercial organisations, which may be responsible for the continuing use of market exchange rates, is that most PPP data is at least a year out of date and even these data are based on relatively crude adjustments of underlying International Comparison Programme ICP PPP indices which are only released every five years.


This paper introduces a new, practical and timely means of comparing prices internationally.  Data is released during the same month that it is collected, making it by far the most current of all PPP systems. The World Price Index (WPI) outlined in this paper, is a work in progress, which will be continually tested and modified. 

WPI data will regularly published at www.worldeconomics.com/wpi. Comments are welcomed and should be sent to Amelia.Myles@worldeconomics.com.


The World Price Index Methodology

 1.      Sampling and Coverage Criteria

The WPI data are derived from tracking the price of a broadly based similar basket of goods (in local currencies) available across different countries. The 60+ items included in the World Price Index basket were chosen based on a number of different criteria. First, the items have been chosen so that a wide variety of categories of private consumer expenditure are represented. Further considerations included ensuring that prices are available from reputable online sources with regular price reviews, along with staple items are included as well as premium brands, and that local brand prices are averaged in cases where no intentionally comparable single brand can be used.


Countries: The WPI headline index has been calculated initially for the largest 15 countries in the world measured by GDP – the US, Japan and the UK, plus three Eurozone nations – Germany, France and Italy and the set of emerging markets known as the BRICs – Brazil, Russia, India and China. This list will be expanded over time but already covers over 60% of world GDP measured using ICP PPP data. The WPI converts six currencies – Sterling, the Euro, the Yen, the Real, the Rupee and the Renminbi into a common standard of international Dollars for comparing economic transactions in these currencies at PPP exchange rates.


Urban Bias: The WPI has an urban/large city bias. The WPI deliberately concentrates on traded consumer items available online that might not be found in the subsistence agricultural segments of large economies such as China and India. The assumption is that the basket of items available in London and Shanghai are related to those that could be bought in Moscow or Paris.  This decision to focus on consumption in large urban areas should minimise the problems faced by broader based PPP measures as a result of the so-called Penn effect (discussed in the Technical Appendix.)


Public Services Excluded: The WPI basket concentrates on readily available consumer items and does not attempt to measure consumption of services often provided by the state such as health and education. This omission is deliberate and it arises from the severe problems and distortions faced in attempting to measure the price of government services by larger surveys such as the ICP. 


2.     Calculation

The World Price Index is simply an index of relative prices, one value for each country that reflects the amount of local currency needed to buy the same representative urban basket of items in each country as can be bought for exactly $1 in the USA. The range of items used in the World Price Index covers a broad base of sectors – food, beverages & tobacco, clothing & footwear, household goods & services, transport, communications, recreation & culture, restaurants & hotels and miscellaneous goods & services (which covers health, personal and non-personal care items). Fifty relative prices are used relating to each country, but in order to calculate a single basket price it is necessary to combine all relative prices to generate a single figure for each country.


Each item price is tracked in its local currency and combined into a three month moving average to smooth individual price anomalies; the resulting figure is then converted into PPP terms relative to the US prices.


All item prices are grouped into categories by taking the arithmetic mean of the available price data to create a category based price index in each country. The geometric average of the category indexes are subsequently taken to combine the categories to form a single monthly World Price Index figure for each country which is relative to the US dollar.



The World Price Index Data

The World Price Index (WPI), a timely monthly index for international economic comparisons. Two examples of how the WPI can be used are show below.


Measuring currency over and under valuation

Table 1 shows WPI data for September 2016 compared with the official exchange rate. An example: At official exchange rates, 101.62¥ was needed in the Japan to buy a $1 basket of items in September 2016; however, the WPI shows that 126.39¥ was actually needed.  Hence, we can draw the conclusion that the Yen is overvalued by 24.4% in comparison to the US Dollar. 


Measuring and comparing gross domestic product across countries

The WPI can be used to compare the relative sizes of gross domestic production in the 16 countries measured.

In general, the WPI measures of GDP and GDP per capita follow a pattern that would be expected given the ‘Penn Effect’ (see Appendix) which suggests that GDP ratios between high and low income countries are systematically exaggerated by GDP conversions at market exchange rates. This is because in emerging markets with large rural populations a subsistence existence is possible at much lower income levels than would be the case in developed country urban environments.

The following tables ( 3 & 4) and charts (4 & 5) show data for GDP and GDP per capita for each of the ten largest countries as measured by nominal exchange rates, the WPI and ICP PPP based exchange rate. As can be seen both WPI and ICP data suggest the size of the emerging economies of China, India and Russia tend to be underestimated by nominal exchange rates and conversely GDP data in most developed economies tend to be overestimated. 



PPP Systems


1.      World Bank International Comparison Program

For purposes such as identifying those countries with the lowest living standards, or creating a global GDP league table ranking countries by the size of their economies or by their economic growth rates PPP conversion rates are typically taken from the World Bank International Comparison Program (ICP).[2] This program assembles data covering over 200 developed and developing countries. The last survey the 2005 benchmark survey was based on the 2003-2007 ICP rounds, the official results of which were published in January 2008, while the next round which started in 2011 is due to deliver statistics by 2013. Prices are collected, or in many cases the value of services are estimated, for the whole range of final items that comprise national GDP: consumer items and services, capital items or government services. The ICP is organised by dividing the world into six regions.[3]  PPP estimates for five of these regions are conducted by the World Bank ICP team directly in cooperation with local co-ordinating institutions such as the Asian Development Bank and the relevant national statistical offices. The remaining region of the world is covered by the Eurostat-OECD PPP Program, which uses a compatible methodology and covers 46 countries but operates on a different timetable. Details of the Eurostat-OECD program are given in subsection 3.


The ICP assembles price data first at a national, annual level.[4] That is, prices are averaged over the whole country and averaged over the year. Prices are the price paid by the purchaser, so they include all wholesale and retail margins, transport costs, insurance charges and non-deductible taxes on products, as well as discounts and rebates that affect the final price paid. The basket of items chosen includes only items whose quality can be expected not to vary across countries. Aside from price and expenditure data, the program collects exchange rate data averaged over the whole year, and mid-year population figures.


Where possible, products are specified by brand and model, but more often products can only be compared according to generic specifications (e.g. 1kg bag of plain rice) rather than brand and model specifications (1kg Tilda Basmati rice).[5] For the main price survey, the part of the ICP that looks only at household consumption, there are over 300 product clusters, within each of which are a number of items for which prices are collected.[6] The ICP uses fairly complex and controversial methodologies for pricing housing, government services, capital equipment, financial services, construction services and others.[7] 


The ICP procedure produces two main datasets: the PPP conversion factor necessary to adjust GDP data and the PPP conversion factor for private consumption expenditure.[8] Figures are expressed in terms of local currency per US Dollar.  While the survey takes place every five years, interim annual figures are published, created by updating the most recent ICP data with nationally produced consumer price indices.


2.   The Penn World Tables (PWT)

The Penn World Tables (PWT) provided by the researchers at the University of Pennsylvania which provide another set of PPP estimates. These are produced by the Center for International Comparisons at the University of Pennsylvania (CICUP) whose present directors Alan Heston and Robert Summers took part in the original ICP work from its inception in 1968 until about 1985 when the 3rd benchmark comparison, covering 34 countries for 1982, was published.


The CICUP turned to expanding the benchmark comparisons to estimates of GDP on a purchasing power basis for non-benchmark countries for one year. The initial exercise was followed by work by Summers and Heston to extend benchmark and non-benchmark estimates over both time and space. Successive updates by the CICUP team have added countries (currently almost 190), additional years (1950-2009), demographic data, and capital stock estimates. The CICUP has cooperated for a number of years with the while the ICOP Center at the University of Groningen which focuses on the production side of national accounts in its PPP comparisons. The CICUP has been primarily concerned with the expenditure side of GDP.


3.      Eurostat-OECD PPP Program

The Eurostat-OECD PPP Program is conducted every three years. Recent surveys were conducted in 2005 and 2008, with the next in line with the ICP in 2011. Surveys take time to analyse, often up to three years. Like the World Bank ICP, PPP estimates for non-EU countries for years in between surveys are calculated using national inflation rates.[9] An excellent summary of PPP measures and the Eurostat-OECD PPP measure in particular is given on the OECD website.[10] Apart from being conducted more frequently than the ICP, and being conducted for only the OECD countries, the World Bank and Eurostat-OECD PPP programs are very similar by virtue of being specifically designed with compatible methodologies. Nonetheless, the Eurostat-OECD PPP program documentation goes in to greater detail about the methodologies used.


First, prices of a specific item (e.g. a 2 litre bottle of Coca Cola) are collected and implicit exchange rates calculated by dividing the price in local currency by the Dollar price in the US. Then within product groups (e.g. soft drinks and concentrates) the average of all implicit exchange rates are taken. Finally, an aggregate measure of PPP is calculated using the average implicit exchange rate in each product group, weighted by the proportion of that product group in total spending.[11] The survey covers approximately 3,000 consumer items and services, 30 occupations in government, 200 types of equipment items and 15 construction projects. The Eurostat-OECD PPP program also divides products into a larger number of product categories than the World Bank ICP.[12]


The Eurostat-OECD PPP program differs in one other way from the World Bank ICP. In addition to giving PPP conversion rates for GDP and private consumption (household final consumption expenditure), it also gives a PPP conversion rate for ‘actual individual consumption’. The difference between household final consumption and actual individual consumption is nuanced. Essentially, household final consumption includes all government final expenditure, while actual individual consumption includes only government final expenditure that is consumed by the individual – i.e. it includes healthcare and education, but does not include defence, police and environmental protection.[13]


Finally the Eurostat OECD data offers a finer level of detail than the ICP data. In particular it details the PPP conversion rate for each of the broad product groups separately, e.g. meat, oils and fats, health, transport and government services.[14]


The Eurostat system is illustrated in Table 4 below, which shows equipment items and construction prices which are updated every two years. However, rents, employee compensation, GDP expenditure weights, exchange rates and populations are updated every year.



A full explanation of the differences between the Eurostat-OECD program and Eurostat comparisons is available on the OECD website.[15] The main reason for the EU’s preoccupation with producing accurate PPP estimates is that EU structural funds, which aim to reduce economic disparity between EU states and make up about a quarter of the total EU budget, are allocated according to GDP per capita based on PPP.


4.   The Economist ‘Big Mac’ Index

The Economist publishes annually a well-publicised non-academic index of a one item basket as a PPP proxy: the McDonalds Big Mac. The Economist describes its index as follows:


“Big Mac index seeks to make exchange-rate theory more digestible. It is arguably the world's most accurate financial indicator to be based on a fast-food item.[16]


The most recent index published for July 2012 calculated this data for 44 currencies. The price of this homogeneous hamburger is published in local currency and in US Dollars (converted at market exchange rates), but an attempt is made to calculate the implied relative price vector of a Big Mac across the world using the US price of a Big Mac as the numeraire or the commodity in terms of which all other items – in this case Big Macs - are valued across the world. Additionally the exchange rate implied by relative prices is calculated to compare against the current market exchange rate in order to estimate the over or undervaluation of the exchange rate against the US Dollar. Table 6 gives figures from the Big Mac Index for the seven largest economies in the world by GDP (based on PPP exchange rates) plus the Euro area.


For example, the price of a Big Mac in China was 16CNY compared to $4.53 in the US, so the bilateral relative price ratio, or the implied PPP of the Dollar, is $4.56/CNY16 or 3.51. At market exchange rates in July the Dollar price of a Renminbi Big Mac of US$4.56 would be, assuming no transaction costs, US$2.61 and therefore a ratio of 6.13. 

Exchanging a Big Mac bought in China back into CNY, then exchanging the CNY for Dollars would not be enough to buy a Big Mac in the US unless an extra was paid. This implies an undervaluation of the Renminbi against the Dollar of -42.76% in terms of a crude one-item basket PPP index. In a similar fashion India’s currency is undervalued against the Dollar by -67.07%.  



The Economist Big Mac index is an interesting tool to illustrate the concepts of purchasing power parity and exchange rates, provided its limited scope are realised.  Further, the exact composition of a Big Mac varies by country. In Islamic countries they use halal beef, kosher beef in Israel, while the Australian Big Mac has 22% fewer calories than the Canadian version. In addition prices of the product vary not only between countries but within countries. For more analysis of the Big Mac Index see footnotes. [17] [18] [19]


A PPP variant of the Big Mac Index was published in 2009 in a report by UBS which calculated the amount of time it took for a worker paid the average net wage to earn enough money to purchase a Big Mac in 73 cities across the world. Reporting on this alternative hamburger index The Economist commented:


Fast-food junkies are best off in Chicago, Toronto and Tokyo, where it takes a mere 12 minutes at work to afford a Big Mac. By contrast, employees must toil for over two hours to earn enough for a burger fix in Mexico City, Jakarta and Nairobi.” [20]


While of the above three PPP measures – World Bank ICP, Eurostat-OECD PPP program, and Eurostat price comparisons – are used for official comparisons of GDP and other statistics, the UBS price and earnings report is a private non-official study of purchasing and earning power across the world. Like the Eurostat-OECD PPP program it is conducted every three years, most recently in 2009. However, unlike any of the previous PPP measures, comparisons are made between cities as opposed to countries. It is also differentiated by the fact that it is published in a more timely manner (for the 2009 report the data was collected in March 2009 and published on 30 June 2009) and it has a long history (the first report was published in 1971, just one year after the first World Bank ICP study). Yet perhaps the most novel aspect of the UBS study is that it compares not just prices across the world, but also earnings. The survey is conducted over the course of a month every three years in 73 cities, reflecting the UBS replication of the Big Mac index, in approximately 50 countries, and the basket is comprised of 122 items. The items are chosen to reflect the typical preferences of a Western European shopper. Sectors are weighted according to the Table 6, and in contrast to official measures of PPP, this weighting does not change according to different consumption habits within cities.



It should be noted that while some government-provided services such as healthcare and education are included in the survey, this is done in only a very rudimentary way. For example, education is represented by prices for continuing-education courses. This is in contrast to the World Bank, Eurostat-OECD, and Eurostat measures which include at the very least government expenditure that is consumed by the individual such as state healthcare and education. The best way to look at the UBS report is to think of it as a comparison of personal disposable spending power, and not a comparison of overall living standards.


In this vein the report also considers income in a different way to the official PPP measures. Rather than using the price data to deflate GDP, the UBS report uses price data to deflate average after-tax individual earnings, collected from a survey of earnings from 14 different professions. This is again consistent with the idea that UBS are not looking at overall living standards, merely personal disposable spending power. It also allows for the provision of statistics such as the number of minutes it takes to earn a Big Mac or an iPod Nano.


6.      Problems with Existing PPP Measures

All PPP measures face shortcomings. The main problems stem from the issues of the timeliness of the data and the frequency of publication, the coverage of products and services, its appropriateness for the purpose it is used for, the underlying data quality and finally a number of technical issues that will be considered in the Appendix. Unfortunately, a solution to one of these problems in the calculation of a price comparison index is often purchased only at the opportunity cost of increasing the seriousness of another issue. There is, of course, a trade-off between timeliness, frequency and the coverage of PPP surveys. The most obvious example is that the publication of data on a regular basis, for example with monthly or quarterly series can only be achieved at cost of a reduction in the coverage of items included in the basket. A comparison of the main existing PPP measures is discussed in Table 7.





7.      Timeliness and Frequency

The World Price Index (WPI) is the most timely of the PPP systems with the most recent price index being published a week after the collection of the underlying inter-country price data. It is also the most frequent of the PPP indices with a monthly data collection and publication cycle.


The ICP PPP system, while the most extensive and comprehensive international of relative prices employed to adjust economic transactions to be measured by a common standard is far from timely.  The most recent data from the 2003-07 ICP round was only released in final form in January 2008 for the 2005 international benchmark while the previous benchmark year before that was 1993. At present the 2009-2013 ICP round is in progress and it will use 2011 as the benchmark year, but the final results are not expected to be published until the end of 2013.


For the developed countries as opposed to the emerging economies, the Eurostat-OECD PPP program produces new PPP estimates on a more timely basis than the ICP programme with PPP rates published every three years with updates based on national inflation in the intervening years. Eurostat, the multinational body responsible for providing data to the European Commission for all Member States also produces PPP estimates for these countries every year. These estimates are not full surveys of PPP in all sectors, but rather they update prices in specific areas that are likely to have changed the most.


The Economist Big Mac Index is more frequent than all these measures being published annually while the UBS survey is a tri-annual survey. However, both of these PPP systems suffer other drawbacks in terms of coverage of countries and the composition of the measured basket of items comprising the index.


Timeliness is important in a rapidly changing world. Whereas between ICP benchmark years, regular PPP figures can be estimated by adjusting the benchmark currency – international Dollars – by actual inflation rates, adjustments due to structural change in the global economy will not be picked up by this method. This makes the ICP programme obviously inappropriate for regular commercial use such as estimating market growth rates or setting international prices, but distortions can also arise that impact on the macroeconomic use of these estimates by international bodies. Based on the result of the release of the 2005 ICP data at the end of 2007, as compared to the previous benchmark, the IMF revised downwards its real PPP adjusted estimates of global growth from 2002 to 2007 to around half a percent per year.  This produced a downward revision of the PPP adjusted GDP growth rates and the share of world output of some emerging economies such as India and China and an upward revision of other economies including some oil-exporting countries. China’s share of global output for 2007 was revised down from 15.8% to 10.9% which is an enormous amount of previously estimated economic activity. [21] Indications that such large step changes which distort global trends had happened could be picked up earlier by more timely PPP measures. This is one of the objectives of the monthly WPI.


8.     Coverage

Existing PPP systems differ significantly in terms of their coverage of countries in the international comparisons made, in the range and number of the basket of items and services covered and in the time period for which comparisons can be made. The combination of the World ICP and the OECD-Eurostat data provides the largest global PPP coverage. The ICP has been steadily increasing its country coverage. One of the reasons for the downward revision of the size of the Chinese economy by about 40% based on the 2005 ICP benchmark was that country’s inclusion in the international price comparison programme for the first time. Prior estimates of China’s economy had been based upon an inaccurate extrapolation of bilateral price comparisons made in 1986 between China and the US.  Furthermore, India participated again in the 2005 ICP price survey for the first time since 1985. Country coverage is expanding further for the 2011 ICP benchmark which has expanded to cover 200 countries including now 47 OECD- Eurostat nations. Many other gaps in international coverage of PPP data have been filled by the Penn World Tables (PWT).


Another coverage problem is to what extent is the basket of items chosen representative of an urban shopper? For example if the index is used to test how much a currency is under or overvalued is the basket representative of the price of traded items? Every PPP measure is unrepresentative and defective in some areas. The ICP makes a downward adjustment for 20% in China for rural prices, but even within the US a 60% range has been found within urban areas in a sample of 25,000 prices by Aten (2006).


The Big Mac and the UBS avoid this problem by concentrating on urban areas and a homogenous set of consumer preferences applied across the world. According to UBS its index is:


“…based on Western European consumer preferences…we weighted our reference basket identically for all of the cities in our survey… The weightings of the individual items in the basket were designed so that all the prices added up to the approximate monthly consumption of a European family of three.” [22]


9.     Quality of Data

There are estimation and methodological problems stemming from the underlying quality and reliability of the data for some of the items included in the basket chosen by the comprehensive ICP PPP survey. The main problem with the larger price indices concerns the estimation of services. The inclusion of poor estimates of the price of service outputs in the ICP survey may render ineffective the higher quality of the data based on surveys of the prices of consumer and producer items. In particular the ICP PPP survey suffers from the treatment of education and health and other government services which put into the perspective some of the theoretical issues about indexing discussed above.  For example, the ICP assumes identical productivity in government services across the world despite vast differences in capital per worker. PPP statistics for China, the second largest economy in the world, are only compiled every five to six years under the World Bank ICP. In the most recent 2005 report China’s price index was revised by 40% and India’s by 36%. This suggests serious issues in the quality of the one existing statistic. [23]


“But the measurement problems for these items are large enough to have major effects on the larger aggregates; indeed we suspect that the largest single factor responsible for the decline in the relative size of the Indian and Chinese economies is a change in the treatment of government services.” [24]


Finally, there is the question of the accuracy and probity of the official government economic data used by the ICP. The problems relating to official data accuracy in Greece, and OECD members have been documented in World Economics in a paper by Sturgess (2010), but the problem is more serious in poorer countries. In 2011 the Ghana Statistical Service revised nominal GDP by over 70% to an estimated US$30.3bn for 2010 from US$17.1bn previously owing to improved measurement of the economy. The data revision was driven by the need to more accurately portray the size and structure of the economy. The base year for GDP measurement was moved to 2006 from 1993, giving a better reflection of the relative contribution to GDP of various economic activities. Moreover, new activities have been included in the GDP measurement, including the oil sector, forest plantations and ICT. Measurement of the wholesale and retail trade industry has been adjusted from a 'commodity flow approach' to a methodology using VAT returns. This has resulted in an upward revision in the nominal value-added of the industry. Consequently, services are now thought to comprise almost 50.0% of GDP, rather than the 35.1% previously estimated.


The implication is that PPP adjustments, although almost always superior to international comparisons involving exchange rates are to be used and improved over time given the current state of economic knowledge. As Dikhanov and Swanson (2010) comment on the China GDP debate:


“The compilation of national accounts and purchasing power parities is a complex and interdependent process. Standards and methods for both are still evolving and are more likely to be more closely adhered to in wealthy countries with well-established statistical traditions.” [25]


10.   Suitability

There is no right or wrong about what type of PPP index to use, although results can differ widely depending on the index chosen. This is why World Bank published ICP figures often differ from those listed in the PWT and why quite large adjustments are made as former non-benchmark countries have entered the ICP programme.  According to Henderson and Castles (2005):


“As to analytical aspects, there is admittedly no single and unique formula for measuring a country’s real GDP, since different sets of price weights can be used to value the respective outputs.” [26]


The researchers involved with the development of PWT data Deaton and Heston et al (2010) have stated about the optimal index choice:


“Which of these is most appropriate depends on the purposes to which the data are put; as is the case with most index number questions there is typically no unique right answer.” [27]


Technical Appendix

An excellent summary of the theory behind PPPs and the pitfalls in measurement is contained in the paper by Deaton and Heston (2010). An earlier version of this paper is available on the Centre for International Comparisons (CI) website along with the Penn World Tables data.[28] This Appendix attempts to throw light on the some of the technical issues raised concerning the construction and use of the WPI and suggest areas where external comments by researchers would be particularly welcome. This Appendix discusses the relevance of the WPI in the light of four technical issues: the Penn Effect; Uniqueness, Transitivity and weighting Methods.


Penn Effect

The Penn effect is an economic finding associated with the PWT that real income ratios between high and low income countries are systematically exaggerated by GDP conversions at market exchange rates. This observed PPP-deviation had been known as the "Balassa-Samuelson effect", but in his review of progress in the area Samuelson acknowledged the debt that his theory owed to the PWT team by coining the term "Penn effect" to describe the "basic fact" they uncovered[29]. In effect the PPP-deviation allows rural Indians to survive on an income below the subsistence level in the developed world. However, the effect implies that the money income level disparity as measured by international exchange rates is an illusion, because these exchange rates only apply to traded items, a small proportion of consumption. The impact of the Penn effect should be minimised by concentrating on an urban basket.



The problems caused by this lack of uniqueness in the choice of index have been known for some time within the country economic accounting literature and are known as the Paasche-Laspeyres spread (LPS) measuring the difference between the two measures depending on the type of index chosen. [30] The main problem comes when relative prices differ across countries since different index number formulae give widely different answers.[31] Heston et al (2009) use the ratio of Laspeyres to Paasche price indexes to measure the size of the LPS – or the bias that needs resolution by further averaging. Using ICP data they show that the LPS is larger for countries further apart in terms of structure and economic development. In other words the Penn effect exacerbates the LPS. The WPI as with UBS assumes an identical basket across countries to get round the uniqueness problem discussed above.



Any set of bilateral price indexes should have two properties.

1.     The price index of a country in terms of the US should be the reciprocal (1/Index) of the US in terms of the country. 

2.     The indexes should be transitive.

E.g. In order to get the price of the basket in China in terms of the basket with Japan as the base, the answer should be the same whether or not it is calculated directly or indirectly via the route China to the UK and then the UK to Japan.


“In practice, it is surely impossible to do without the transitivity assumption: we cannot feasibly work with a matrix of price indexes.” [32]


There are different ways to ensure these properties are satisfied and this is a very controversial area that is beyond the modest aims of the World Price Index described in this paper. For example, there are Gini, Eleto, Koves and Szulc (GEKS) indexes that are all modified Fisher indexes or there is the Geary-Khamis (GK), used by PWT and the Ikle, Dikhanov, Balk (IDB) index which are modified Paasche indexes. Differences in use of index lie behind a recent controversy over the size of the Chinese economy (the Maddison-Wu- ICP debate). In 2008, the late Angus Maddison and Harry Wu dismissed the 2008 ICP GDP estimates for China and some other Asian countries and significantly revised 2003 GDP estimates for China and reduced the official rate of growth from 9.6% to 7.9% per annum. [33]


However, Heston et al (2009) state that although transitivity is an essential theoretical requirement it comes at a price. The price index for any pair of countries depends on prices and budget shares in third countries which violate the property of the independence of irrelevant countries.  Fortunately, they note in reality theoretical differences between indexes in most cases do not appear to have a large effect on calculations. The matrix of more sophisticated indexes, such as the Fisher indexes, are usually close to being transitive without further adjustment using GEKS or other more complex methods. They write:


So that the GEKS step has little effect on the calculations… comparisons between pairs of countries using GEKS price indexes are not very sensitive to prices or budget shares in third countries.” [34]


To give an example of the continuing debate on the use of sophisticated or superlative indexes, the abstract from the following recent paper is instructive:


“This paper shows that just because superlative index numbers approximate each other to the second order does not necessarily imply that they are numerically similar. In fact, the spread between the largest and smallest superlative indexes sometimes even exceeds that between Paasche and Laspeyres indexes. This result has significant implications for the index number literature. It shows that the economic approach does not by itself solve the index number problem, since it does not tell us which superlative index should be used. It may be necessary to combine the economic and axiomatic approaches to arrive at an answer.” [35]


The World Price Index is a simple form of Laspeyres Index which calculates a set of bilateral relative price comparisons which are not transitive. This creates a matrix of different indirect cross country relative prices. Problems are avoided by using the methodology presented in this paper. This is again a pragmatic rather than a theoretical solution.


Weighting Averages

There are a number of means of measuring an average – a statistic which is the most representative number of a distribution. These are the mean, geometric mean, median and mode. There is no categorical right or wrong. The choice depends on the assumed properties of the underlying distribution of the population being estimated and the efficiency and accuracy of the sampling method used. After measuring different calculations against ICP PPP data as a relative test of accuracy the World Price Index settled on the arithmetic mean to group items into categories and then the geometric mean to combine categories. Removing outliers sensibly limits category variations. Although preferable for pragmatic reasons (closeness to ICP calculations) in small samples there is also a strong case for using the arithmetic mean and geometric averages in the algorithm. This addresses issues likely in retail markets as a result of the presence of short term price cutting. In a small sample the calculations will still be different from the true population mean, but the bias will be less than using the sample mean.





Deaton, A. and Heston, A (2010), Understanding PPPs and PPP-based National Accounts, American Economic Journal: Macroeconomics, 2 (4), 1-35.


Dikhanov, Y. and Swanson, E.V. (2010), Maddison and Wu: ‘Measuring China’s Economic Performance’, World Economics, 11(1), January-March, 199-203.


Henderson, D. and Castles, I. (2005) International Comparisons of GDP: Issues of theory and practice, World Economics, 6(1), January-March, 55-84.


Hill, R.J. (2006), Superlative Index Numbers: Not All of them Are Super, Journal of Econometrics 130(1)  January 2006, Pages 25-43


Rogoff, K, 1996, The Purchasing Power Parity Puzzle, Journal of Economic Literature, 34, 647-668.


Samuelson, P. A. (1994) "Facets of Balassa-Samuelson Thirty Years Later," Review of International Economics 2(3), 201-26.


Sarno, L and M P Taylor, (2002), The Economics of Exchange Rates, Cambridge University Press.


Taylor, A.M. and Taylor, M.P. (2004), the Purchasing Power Parity Debate, Journal of Economic Perspectives, 18(4), Fall, 135-158.


UBS (2009), Prices and Earnings: A comparison of purchasing power around the globe, UBS AG, Switzerland.




[1] ‘When a strong currency is not a badge of pride’ Comment by Peter Tasker (a leading expert on Japanese finance at Arcus Research in Tokyo), Financial Times, 29 September, 2011, p.11.

[2] http://siteresources.worldbank.org/ICPEXT/Resources/ICP_2011.html.

[3] http://go.worldbank.org/N5MKS8YYJ0.

[4] Details of the methodology are given at http://go.worldbank.org/N0B1O8ZGW0.

[5] http://go.worldbank.org/975RPR8K50.

[7] Please see http://siteresources.worldbank.org/ICPEXT/Resources/ICP_2011.html and follow links labelled ‘Methodological framework’ and ‘Research agenda’ on the left hand side for details.

[8] http://data.worldbank.org/indicator/PA.NUS.PPP and http://data.worldbank.org/indicator/PA.NUS.PRVT.PP.


[9] One source suggests that is consumer price indices that are used(http:/www.oecd.org/document/47/0,3343,en_2649_34357_36202863_1_1_1_1,00.html) , while another suggests (http://www.oecd.org/std/pricesandpurchasingpowerparitiesppp/purchasingpowerparities-frequentlyaskedquestionsfaqs.htm) implicit deflators are used  (perhaps GDP deflators are used for comparisons of GDP, while consumer price indices are used for comparisons of private consumption)

[10] A full methodology is available on the OECD website.  (www.oecd.org/std/ppp/ )

[11] www.oecd.org/std/ppp/faq.

[12] www.oecd.org/dataoecd/53/15/40210393.pdf.

[13] www.oecd.org/std/ppp/faq.

[14] The World Bank ICP claims to offer the same level of detail, but a long search has found nothing published but the headline GDP and private consumption conversion factors.

[15] www.oecd.org/std/ppp/faq

[16] http://www.economist.com/markets/bigmac/index.cfm

[17] A paper by Clements, Lan and Seah (2010) lists 23 academic papers on the Big Mac Index.

[18] An item summary of the Big Mac Index is given in Pakko and Pollard (2003)

[19] An assessment of the Big Mac Index after 25 years is published in Haidar (2011) http://www.worldeconomics.com/Currency%20Valuation%20and%20Purchasing%20Power%20Parity.details?AID=479

[21] See: http://www.imf.org/external/pubs/ft/survey/so/2008/res018a.htm

[22] UBS (2009) p6

[23]See http://pwt.econ.upenn.edu/papers/deaton%20heston%20complete%20nov10.pdf for details.

[24] Deaton and Heston (2010) p23

[25] Dikhanov and Swanson p202

[26] Henderson and Castles (2005) p59

[27] Deaton and Heston (2010) p3

[28] http://pwt.econ.upenn.edu

[29] Samuelson (1994)

[30] For example, a Laspeyres Index overstates inflation and a Paashe Index understates it. There is an enormous and still inconclusive literature on this area and attempts have been made to make composite indexes from the average or geometric means of the two – or a Fisher Index to take an average of the degree of bias.

[31] For example, Heston et al (2009) find that the US based L index for Zimbabwe was 3.7x the P index, but the US L index for Canada was only 1.06x the P index.

[32] Heston p13

[33] Maddison was dismissive of the use of the EKS index method to compute global PPP estimates since it produces higher PPPs and lower growth estimates of GDP for low-income countries. However, it is argued by Dikhanov and Swanson (2010) that the GK method preferred by Maddison has a price vector reflecting the prices of higher income countries. Consequently, it biases the GDP of poorer countries upwards.

[34] Heston et al p13


[35] Hill (2006) Pages 25-43

Download the full report in PDF format