When I left Merge Global in 2003, I brought the databases I had developed there with me and formed QuERI-International to keep these long time series, consistent, data sets alive. Since then I've improved the models and expanded the coverage until today I want to pass them on to other organizations that can maintain and use these sets of data on industrial and trade for research into the future o the global economy. Global Insight, IHS Markit, has maintained some earlier version of the original data developed at DRI and in a joint venture with WEFA as the World Industry Service without adding much to its structure. After leaving Merge Global, the data was placed for a few years up on the Oxford Economic Forecasting system as a proprietary data set until they decided they wanted to do it all and kicked me off the system but given that the categories are similar if not identical to the original data on the system decided they might peak at the QuERI data. I see they say that there system is the only fully integrated global model of industrial data. So I thought it might be nice to get a good idea of what the QuERI Global Model does cover and how it can provide useful insights into the changing structural patterns of demand in the world economy in the Post-Covid world.
Everything You Wanted to Know About the QuERI Global Databases and Methodologies—One of the Largest, fully integrated, Global Trade and Industry Data sets covering 72 countries for more than 400 commodities over the period 1990 - 2030
IN a few weeks, QuERI-International will make available to governments, research institutes, educational institutes, and private companies with the need and skills to use a global data base covering 72 countries at the ISIC3 level of detail for industrial activity, market demand, international trade, private consumption, investment, employment, prices, and main macroeconomic components in a complete form at a nominal price. Given the complexity of the data and the consistency developed over the nearly forty years of full development, this primer was prepared a few years ago to explain the methodology and approach. For more information, contact me at [email protected].
1. Where does the data come from?
International data on industrial activity and trade comes from many different sources. The problem is that it is often out of date, incomplete, and not detailed enough to allow companies to link operational and performance to market conditions in the countries they are operating in. Solving this problem of getting baseline estimates that are both reasonable, i.e. tied to known information even if not exact, up to date, i.e. close to the current year and with forecasts into the next few years, and in sufficient detail to be linked easily to company products is hard. QuERI solves the problem using an integrated set of global models linking industrial production within a country through an input-output model for the country so that demand for steel is related to growth in demand for automobiles, trucks, construction, etc. Production of steel, for example, is closely tied to demand for steel from all sources of demand -- sales to other companies and final consumption demand (quite limited for steel, but significant for some steel products), it is also tied to foreign demand (exports), and usually is negatively impacted by imports of competing products.
2. Achieving Consistency of Industrial Classifications – Actual & Estimated Data
Accuracy is a more difficult question to answer since even government data is often inaccurate. Actual UNIDO data is generally available with missing data for most countries for manufactures at the ISIC3 detail. We insure as much accuracy in the estimates as possible by using known data to scale more detailed estimates derived from detailed (NAICS 6 level IO models) to available data from UN and country sources. What international data we do have at the NAICS 6 level of product detail (covering more than 400 individual products) is international trade data. At a somewhat lower level of detail we have international data on production of manufactures and some services, at the ISIC 3 level of detail (about 120 commodities) from UNIDO*. Production data on services depends upon UN Standard National Account data. To get down to the NAICS 6 level of service detail we use detailed country-specific IO models.
3. How can you get to NAICS 6 level when the data for 72 countries at that level of detail barely exists for only a few countries and how can you insure consistency across countries?
There is no simple way to get to the detail needed for useful forward projections without using a framework for linking industries together through their relationships to one another and to the economy as a whole. Input-Output models were first developed in the 1930's once sufficient data on the economy became available. Developed by Professor Wasilly Leontieff these relationships were used strategically during the World War II to suggest targets for bombing that would slow or halt German war production. As a result the Allies bombed the German ball bearing plants rather than the truck and tank plants as the Generals had first suggested. Since then IO models have evolved and are available for many countries -- rich and poor. Models, however, are developed over many years and with a frequency of around 5 years between releases. There may, however, be 10 years between the date of the current model and today's date. The 2002 US IO model is the only one now available at the NAICS 6 level of detail. Annual models are estimated at a higher level of detail by the US BEA, but are simply based on limited information and estimates drawn from scaling rows and columns (balancing what is the known production of a commodity with the known demand for the commodity).
QuERI has developed a variant of this approach to estimate models for all countries based on a standardized US detailed IO model (430 rows). Column totals are aggregated to the core model's 63 industry groupings (ISIC2 categories) and to vectors covering investment (63 industry groups), 16 private consumption spending categories, and 4 primary government consumption categories. All of the columns are derived from the core, global econometric forecasting model. Differences between countries in terms of consumption patterns (rows) are based on ratios between the known estimates of apparent consumption and the base model's estimate of apparent consumption. Using a balancing approach then individual differences between countries -- rich or poor, large or small-- are accounted for in the IO model's structure. These adjusted models cover the periods 1998 through 2010 for all 72 countries. This allows for a pure estimate of market demand and production by NAICS 6 categories (410).
4. What are the risks from this approach?
There are risks to assuming that the production and market demand estimates are precise. There's some precision at least in terms of total market demand and production at the ISIC 3 level of detail. This level of industrial detail is only available for manufactures and with a lag that often stretches for several years. Consistency across countries is assumed primarily because of the classification into a standard international definition, but it is likely that many of these sectors are approximations only. Often, as not, for many countries no data is available despite the fact that exports do exist thus suggesting there is production. Production data for the US is through 2011, but for most countries it is through 2009. Forecasting models are then used to extend UNIDO or UN data through the current year. To be absolutely clear we can't guarantee the accuracy of the raw government data which is, like GDP, based on a sample of sales data collected from industries. No government provided data is precise as attested by the fact that Gross Domestic Product is itself just a sample estimate from limited information and is subject to full revisions sometimes years later.
What we can say about the estimates are that they are done in a consistent, and quite reasonable way. They force estimates to line up to known information on production and trade where available. QuERI country IO models are based on a detailed US IO model adjusted for differences in consumption patterns across countries. For example, the changing share of computers in purchases of companies is reflected in the changing percentage of spending within intermediate demand, investment, government, and private consumption vectors. As a result we are re-balancing the IO between rows and columns to insure consistency and it is this structured approach to measuring likely market demand that is the greatest strength of the QuERI model and data sets.
5. Why is this so special? What's new about it?
There are many different levels of data availability. The most detailed data that is easily available is foreign trade data. Export and import data is available at the 6 digit Harmonized Code detail from the United States Commodity Trade Database. It is generally up to date (current data is through 2011), but there are lapses. Some countries simply don't report frequently enough, but if your company is looking for something that describes at least a part of the market in any market in the world you should be able to find it there.
The problem, however, is that it misses what might be a very large market. It doesn't provide any intelligence on which industries are buying the product or the potential of these "customers". The only way to get at that is to do a survey of the market, but as any company doing the survey will tell you without knowing how large the market was at one time and it's relationship to the smaller survey sample, you can't get down to this level of detail with any more accuracy than simply making a guess.
The QuERI data was developed to answer the larger question -- what is the "potential" size of the market. Trade covers only a small portion of the market for most countries. And while production and trade data together might be found, production data is usually out of data and difficult to interpret. QuERI solves the problem by combining data from multiple sources and then processing it through a structured IO model to create pure estimate of likely or potential demand irrespective of the source of supply (domestic or international). By using more detailed IO relationships then estimates can be developed that are more detailed than available from government sources, but which are based on a rational methodology that can be defended because the underlying relationships between buyers and sellers can be understood. Further because it is a detailed build-up from the likely buyers, it splits estimates into four primary areas of demand -- intermediate, investment, government, and private consumption.
6. Is there an underlying model to explain differences across countries?
Countries move through stages of development. This idea was proposed by W.W. Rostow that there are five stages of development:
QuERI models are based on a pooled-cross-sectional econometric model. This allows countries of different stages of development to be included in the model so that the coefficients measuring relative wealth (macro variables like per capita GDP) and market size (measured by urban population size) allow forecasts to reflect changing wealth and country size. Models include other variables that are less generalized and more directly related to the dependent variable such as imports, exports, intermediate demand, final demand and prices. Unlike time series models based on a single countries experience, pooled models take the generalized experiences of many countries -- from poor to rich -- into account when developing coefficients. Additional elements allow for some variables to be more specific to one group of countries while others are allowed to reflect the broader differences across all countries irrespective of their wealth.
Thus the dynamic, iterative, forecasting model is a true reflection of Rostow's theory. It allows longer term projections to be made as well as short-term trends to be integrated. Countries with limited production in more technologically advanced products can develop over time into manufacturing centers of these products as they develop. The non-linear pooled models allow for this kind of transformation making the QuERI models uniquely suited for capturing the changing global dynamics of production, consumption, and international trade.
Everything You Wanted to Know About the QuERI Global Databases and Methodologies—One of the Largest, fully integrated, Global Trade and Industry Data sets covering 72 countries for more than 400 commodities over the period 1990 - 2030
IN a few weeks, QuERI-International will make available to governments, research institutes, educational institutes, and private companies with the need and skills to use a global data base covering 72 countries at the ISIC3 level of detail for industrial activity, market demand, international trade, private consumption, investment, employment, prices, and main macroeconomic components in a complete form at a nominal price. Given the complexity of the data and the consistency developed over the nearly forty years of full development, this primer was prepared a few years ago to explain the methodology and approach. For more information, contact me at [email protected].
1. Where does the data come from?
International data on industrial activity and trade comes from many different sources. The problem is that it is often out of date, incomplete, and not detailed enough to allow companies to link operational and performance to market conditions in the countries they are operating in. Solving this problem of getting baseline estimates that are both reasonable, i.e. tied to known information even if not exact, up to date, i.e. close to the current year and with forecasts into the next few years, and in sufficient detail to be linked easily to company products is hard. QuERI solves the problem using an integrated set of global models linking industrial production within a country through an input-output model for the country so that demand for steel is related to growth in demand for automobiles, trucks, construction, etc. Production of steel, for example, is closely tied to demand for steel from all sources of demand -- sales to other companies and final consumption demand (quite limited for steel, but significant for some steel products), it is also tied to foreign demand (exports), and usually is negatively impacted by imports of competing products.
2. Achieving Consistency of Industrial Classifications – Actual & Estimated Data
Accuracy is a more difficult question to answer since even government data is often inaccurate. Actual UNIDO data is generally available with missing data for most countries for manufactures at the ISIC3 detail. We insure as much accuracy in the estimates as possible by using known data to scale more detailed estimates derived from detailed (NAICS 6 level IO models) to available data from UN and country sources. What international data we do have at the NAICS 6 level of product detail (covering more than 400 individual products) is international trade data. At a somewhat lower level of detail we have international data on production of manufactures and some services, at the ISIC 3 level of detail (about 120 commodities) from UNIDO*. Production data on services depends upon UN Standard National Account data. To get down to the NAICS 6 level of service detail we use detailed country-specific IO models.
3. How can you get to NAICS 6 level when the data for 72 countries at that level of detail barely exists for only a few countries and how can you insure consistency across countries?
There is no simple way to get to the detail needed for useful forward projections without using a framework for linking industries together through their relationships to one another and to the economy as a whole. Input-Output models were first developed in the 1930's once sufficient data on the economy became available. Developed by Professor Wasilly Leontieff these relationships were used strategically during the World War II to suggest targets for bombing that would slow or halt German war production. As a result the Allies bombed the German ball bearing plants rather than the truck and tank plants as the Generals had first suggested. Since then IO models have evolved and are available for many countries -- rich and poor. Models, however, are developed over many years and with a frequency of around 5 years between releases. There may, however, be 10 years between the date of the current model and today's date. The 2002 US IO model is the only one now available at the NAICS 6 level of detail. Annual models are estimated at a higher level of detail by the US BEA, but are simply based on limited information and estimates drawn from scaling rows and columns (balancing what is the known production of a commodity with the known demand for the commodity).
QuERI has developed a variant of this approach to estimate models for all countries based on a standardized US detailed IO model (430 rows). Column totals are aggregated to the core model's 63 industry groupings (ISIC2 categories) and to vectors covering investment (63 industry groups), 16 private consumption spending categories, and 4 primary government consumption categories. All of the columns are derived from the core, global econometric forecasting model. Differences between countries in terms of consumption patterns (rows) are based on ratios between the known estimates of apparent consumption and the base model's estimate of apparent consumption. Using a balancing approach then individual differences between countries -- rich or poor, large or small-- are accounted for in the IO model's structure. These adjusted models cover the periods 1998 through 2010 for all 72 countries. This allows for a pure estimate of market demand and production by NAICS 6 categories (410).
4. What are the risks from this approach?
There are risks to assuming that the production and market demand estimates are precise. There's some precision at least in terms of total market demand and production at the ISIC 3 level of detail. This level of industrial detail is only available for manufactures and with a lag that often stretches for several years. Consistency across countries is assumed primarily because of the classification into a standard international definition, but it is likely that many of these sectors are approximations only. Often, as not, for many countries no data is available despite the fact that exports do exist thus suggesting there is production. Production data for the US is through 2011, but for most countries it is through 2009. Forecasting models are then used to extend UNIDO or UN data through the current year. To be absolutely clear we can't guarantee the accuracy of the raw government data which is, like GDP, based on a sample of sales data collected from industries. No government provided data is precise as attested by the fact that Gross Domestic Product is itself just a sample estimate from limited information and is subject to full revisions sometimes years later.
What we can say about the estimates are that they are done in a consistent, and quite reasonable way. They force estimates to line up to known information on production and trade where available. QuERI country IO models are based on a detailed US IO model adjusted for differences in consumption patterns across countries. For example, the changing share of computers in purchases of companies is reflected in the changing percentage of spending within intermediate demand, investment, government, and private consumption vectors. As a result we are re-balancing the IO between rows and columns to insure consistency and it is this structured approach to measuring likely market demand that is the greatest strength of the QuERI model and data sets.
5. Why is this so special? What's new about it?
There are many different levels of data availability. The most detailed data that is easily available is foreign trade data. Export and import data is available at the 6 digit Harmonized Code detail from the United States Commodity Trade Database. It is generally up to date (current data is through 2011), but there are lapses. Some countries simply don't report frequently enough, but if your company is looking for something that describes at least a part of the market in any market in the world you should be able to find it there.
The problem, however, is that it misses what might be a very large market. It doesn't provide any intelligence on which industries are buying the product or the potential of these "customers". The only way to get at that is to do a survey of the market, but as any company doing the survey will tell you without knowing how large the market was at one time and it's relationship to the smaller survey sample, you can't get down to this level of detail with any more accuracy than simply making a guess.
The QuERI data was developed to answer the larger question -- what is the "potential" size of the market. Trade covers only a small portion of the market for most countries. And while production and trade data together might be found, production data is usually out of data and difficult to interpret. QuERI solves the problem by combining data from multiple sources and then processing it through a structured IO model to create pure estimate of likely or potential demand irrespective of the source of supply (domestic or international). By using more detailed IO relationships then estimates can be developed that are more detailed than available from government sources, but which are based on a rational methodology that can be defended because the underlying relationships between buyers and sellers can be understood. Further because it is a detailed build-up from the likely buyers, it splits estimates into four primary areas of demand -- intermediate, investment, government, and private consumption.
6. Is there an underlying model to explain differences across countries?
Countries move through stages of development. This idea was proposed by W.W. Rostow that there are five stages of development:
- Traditional society -- with limited technology and poorly developed agriculture;
- Pre-conditions to "take-off" - new technologies allow more production, agriculture becomes more efficient allowing shifts of more people to cities and villages, development of market system;
- Take off -- manufacturing starts to allow a rising standard of income, exports and trade expands, new basic industries like textiles and apparel become concentrated in a few countries, but gradually technologies spread leading to new trade opportunities;
- Drive to maturity -- new technologies are developed, new products, consumer spending and life styles improve, new transportation allows longer distance trade and distributed production;
- Age of mass consumption -- rich, socially developed societies, mass consumption allows a higher standard of living for more people, detailed trade networks link countries, technologies are transferred, manufacturing is dispersed, labor costs equalize.
QuERI models are based on a pooled-cross-sectional econometric model. This allows countries of different stages of development to be included in the model so that the coefficients measuring relative wealth (macro variables like per capita GDP) and market size (measured by urban population size) allow forecasts to reflect changing wealth and country size. Models include other variables that are less generalized and more directly related to the dependent variable such as imports, exports, intermediate demand, final demand and prices. Unlike time series models based on a single countries experience, pooled models take the generalized experiences of many countries -- from poor to rich -- into account when developing coefficients. Additional elements allow for some variables to be more specific to one group of countries while others are allowed to reflect the broader differences across all countries irrespective of their wealth.
Thus the dynamic, iterative, forecasting model is a true reflection of Rostow's theory. It allows longer term projections to be made as well as short-term trends to be integrated. Countries with limited production in more technologically advanced products can develop over time into manufacturing centers of these products as they develop. The non-linear pooled models allow for this kind of transformation making the QuERI models uniquely suited for capturing the changing global dynamics of production, consumption, and international trade.
RSS Feed