Methodology for the Freight
Analysis Framework-2:
Forecasts of Inter-regional Commodity Flows
3. Underlying Economic Forecast Drivers
Global Insight's economic forecasting model of the U.S. economy and Global Insight's Regional state economic models provide significant inputs that shape the freight flow forecasts. This section describes the approach used in these models as well as those Global Insight models that directly provide growth rates used in the FAF2 freight flow forecasts.
Global Insight Model of the U.S. Economy
Global Insight's flagship model of the U.S. Economy integrates modern economic theory and behavior in an analytical tool that is widely used in forecasting, assessing derivative risks, and evaluating policy alternatives. The theoretical structure of the Global Insight Model of the U.S. Economy strives to incorporate the best insights of many theoretical approaches to the business cycle: Keynesian, Neoclassical, monetarist, supply-side, and rational expectations. It embodies major properties of the Neoclassical growth models developed by Robert Solow; thus ensuring that short-run cyclical developments will converge to robust long-run equilibrium.
In growth models, the expansion rate of technical progress, the labor force, and the capital stock determine the productive potential of an economy. Both technical progress and the capital stock are governed by investment, which in turn must be in balance with post-tax capital costs, available savings, and the capacity requirements of current spending. As a result, monetary and fiscal policies will influence both the short- and the long-term characteristics of such an economy through their impacts on national saving and investment.
A modern model of output, prices, and financial conditions is melded with the growth model to present the detailed, short-run dynamics of the economy. In specific goods markets, the interactions of a set of supply and demand relations jointly determine spending, production, and price levels. Typically, the level of inflation-adjusted demand is driven by prices, income, wealth, expectations, and financial conditions. The capacity to supply goods and services is keyed to a production function combining the basic inputs of labor hours, energy usage, and the capital stocks of business equipment and structures, and government infrastructure. The "total factor productivity" of this composite of tangible inputs is driven by expenditures on research and development that produce technological progress.
Prices adjust in response to gaps between current production and supply potential and to changes in the cost of inputs. Wages adjust to labor supply-demand gaps (indicated by a demographically-adjusted unemployment rate), current and expected inflation (with a unit long-run elasticity), productivity, tax rates, and minimum wage legislation. The supply of labor positively responds to the perceived availability of jobs, to the after-tax wage level, and to the growth and age/gender mix of the population. Demand for labor is keyed to the level of output in the economy and the productivity of labor, capital, and energy. Tempering the whole process of wage and price determination is the exchange rate; a rise signals prospective losses of jobs and markets unless costs and prices are reduced.
For financial markets, the model predicts interest rates, exchange rates, stock prices, loans, and investments interactively with the preceding GDP and inflation variables. The Federal Reserve sets the supply of reserves in the banking system and the fractional reserve requirements for deposits. In the Global Insight Model, "monetary policy" is defined by a set of targets, instruments, and regular behavioral linkages between targets and instruments. The model user can choose to define unchanged monetary policy as unchanged reserves, or as an unchanged reaction function in which interest rates or reserves are changed in response to changes in such policy concerns as the price level and the unemployment rate.
The Global Insight Model captures the full simultaneity of the U.S. economy, forecasting over 1,200 concepts spanning final demands, aggregate supply, prices, incomes, international trade, industrial detail, interest rates, and financial flows. Chart 1 summarizes the structure of the interactive sectors. The following discussion presents the logic of each sector and the significant interactions with other sectors.
Consumer Spending: The domestic spending, income, and tax policy sectors model the central circular flow of behavior as measured by the national income and product accounts. Consumer spending is divided into eleven durable goods categories, nine nondurable goods categories, and sixteen service categories. Real consumption expenditures are motivated by real income and the user price of a particular category relative to the prices of other consumer goods. Durable and semidurable goods are also especially sensitive to current financing costs, and consumer speculation on whether it is a "good time to buy." The University of Michigan Survey of Consumer Sentiment monitors this last influence, with the index itself modeled as a function of current and lagged values of inflation, unemployment, and the prime rate.
Business Investment: Business spending includes nineteen fixed investment categories. Each equipment and structures spending category is determined by its specific effective post-tax capital costs, capacity utilization, and replacement needs. The cost terms are sophisticated blends of post-tax debt and equity financing costs (offset by expected capital gains) and the purchase price of the investment good (offset by possible tax credits and depreciation-related tax benefits). This Neoclassical structure builds upon the work of Dale Jorgenson, Robert Hall, and Charles Bischoff.
Residential Investment: The residential investment sector of the model includes two housing starts (single and multi-family starts) and three housing sales categories (new and existing single family sales, and new single family units for sale), and five GDP account categories. The housing sector of the Global Insight Model explains new construction as a decision primarily based on the after-tax cost of home ownership relative to disposable income. The equations also include a careful specification of demographic forces.
Government: The government sector is largely exogenous (user-determined) at the federal level and endogenous (equation-determined) at the state and local level. The presence of a large and growing deficit imposes no constraint on federal spending. This contrasts sharply with the state and local sector where legal requirements for balanced budgets mean that declining surpluses or emerging deficits produce both tax increases and reductions in spending growth.
Incomes: Domestic spending, adjusted for trade flows, defines the economy's value-added or gross national product (GNP) and gross domestic product (GDP). Because all value-added must accrue to some sector of the economy, the expenditure measure of GNP also determines the nation's gross income. The distribution of income among households, business, and government is determined in the Tax Policy and Domestic Income sectors of the model. Each pre-tax income category except corporate profits is determined by some combination of wages, prices, interest rates, debt levels, and capacity utilization or unemployment rates. Profits are logically the most volatile component of GNP on the income side. When national spending changes rapidly, the contractual arrangements for labor, borrowed funds, and energy imply that the return to equity holders is a residual that will soar in a boom and collapse in a recession. The model reflects this by subtracting each non-profit income item from national income to solve for profits.
Taxes: Since post-tax rather than pre-tax incomes drive expenditures, each income category must be taxed at an appropriate rate; the model therefore tracks personal, corporate, payroll, and excise taxes separately. Users may set federal tax rates; tax revenues are then simultaneously forecast as the product of the rate and the associated pre-tax income components. However, the model automatically adjusts the effective average personal tax rate for variations in inflation and income per household, and the effective average corporate rate for credits earned on equipment, utility structures, and R&D.
International: The international sector is a critical, fully simultaneous block that can either add or divert strength from the central circular flow of domestic income and spending. Depending on the prices of foreign output, the U.S. exchange rate, and competing domestic prices, imports capture varying shares of domestic demand. Depending on similar variables and the level of world gross domestic product, exports can add to U.S. production. The exchange rate itself responds to international differences in inflation, interest rates, trade deficits, and capital flows between the U.S. and its competitors. Eight aggregate-level categories of goods and two service categories are separately modeled for exports and imports, with one additional category for oil imports.
Investment income flows are also explicitly modeled. The stream of huge current account deficits incurred by the U.S. has important implications for the U.S. investment income balance. As current account deficits accumulate, the U.S. net international investment position and the U.S. investment income balance deteriorate. U.S. foreign assets and liabilities are therefore included in the model, with the current account deficit determining the path of the net investment position.
Chart 1 – Overview of the Global Insight Model of the U.S. Economy
Financial: The use of a detailed financial sector and of interest rate and wealth effects in the spending equations recognizes the importance of credit conditions on the business cycle and on the long-run growth prospects for the economy. Interest rates, the key output of this sector, are modeled as a term structure, pivoting off the federal funds rate. The federal funds rate is determined in response to changes in such policy concerns as inflation and unemployment. Longer-term interest rates are driven by shorter-term rates as well as factors affecting the slope of the yield curve. In the Global Insight Model, such factors include inflation expectations, government borrowing requirements, and corporate financing needs.
Inflation: Inflation is modeled as a carefully controlled, interactive process involving wages, prices, and market conditions. The principal domestic cost influences are labor compensation, nonfarm productivity (output per hour), and foreign input costs; the latter are driven by the exchange rate, the price of oil, and foreign wholesale price inflation. Excise taxes paid by the producer are an additional cost fully fed into the pricing decision. This set of cost influences drives each of the industry-specific producer price indexes, in combination with a demand pressure indicator and appropriately weighted composites of the other producer price indexes. In other words, the inflation rate of each industry price index is the reliably weighted sum of the inflation rates of labor, energy, imported goods, and domestic intermediate goods, plus a variable markup reflecting the intensity of capacity utilization or the presence of bottlenecks.
Supply: The first principle of the market economy is that prices and output are determined simultaneously by the factors underlying both demand and supply. In the Global Insight Model, aggregate supply, or potential GDP, is estimated by a Cobb-Douglas production function that combines factor input growth and improvements in total factor productivity. Factor input equals a weighted average of labor, business fixed capital, public infrastructure, and energy provided by the energy sector. Total factor productivity depends upon the stock of research and development capital and trend technological change. Taxation and other government policies influence labor supply and all investment decisions, and thus potential supply. The growth of aggregate supply is the fundamental constraint on the long-term growth of demand. Inflation, created by demand that exceeds potential GDP, raises credit costs and weakens consumer sentiment, thus putting the brakes on aggregate demand.
Expectations: Expectations influence several expenditure categories in the Global Insight Model, but the principal nuance relates to the entire spectrum of interest rates. Shifts in price expectations or the expected capital needs of the government are captured through price expectations and budget deficit terms, with the former affecting the level of rates throughout the maturity spectrum, and the latter affecting intermediate and long-term rates, and hence affecting the shape of the yield curve. On the expenditure side, inflationary expectations affect consumption via consumer sentiment, while growth expectations affect business investment.
Global Insight U.S. Regional Economic Forecasting Models
The Global Insight approach to regional modeling at the state level represents a departure from many earlier multi-regional modeling and forecasting efforts. Most other regional models are constructed as proportions of the U.S. national economy. In the Global Insight regional forecasting system each area is modeled individually and then linked into the national system. Thus, our models do not forecast regional growth as simple proportions of U.S. totals, but focus on internal growth dynamics and state specific business cycle response. This approach is referred to as "top-down bottom-up." It contrasts with pure share (top-down) models, and models which are not linked to a national macroeconomic model (bottom-up), and contains the best of both approaches. A primary objective is to project how regional activity varies, given an economic environment as laid out by our macroeconomic and industry forecasts. Important regional issues are addressed using information about detailed industrial mix, inter-industry and interregional relationships, productivity and relative costs, and migration trends. Global Insight maintains separate models for 50 states and for Washington DC, as well as for 318 metropolitan areas. The state models have two fundamental characteristics: (1) Each state is modeled individually, with different model structures specified according to the characteristics of the state; and (2) national policy is explicitly captured.
These models were converted from an SIC industry classification basis to a NAICS basis to reflect changes in the industrial classification system used by the U.S. government in reporting state and local industry activity. The individual state models are econometrically estimated and contain about 250 or more equations each. Employment by sector and wage rates and income by type of activity, and Gross State Product (GSP) by sector are modeled in detail. Other coverage includes housing starts, retail sales, consumer price indexes, population by 10-year age groups, the labor force and household employment. The models have the ability to forecast income, wages and GSP in nominal as well as real dollars. The state models have a quarterly periodicity, so they are able to capture the business cycle behavior of the economy, including the timing and amplitude of turning points. Another model characteristic is that they are policy sensitive — they respond to changes in tax rates, military spending, utility costs, etc. The policy simulation capability can be classified into: (1) how a state economy responds to changes in the national economy resulting from national or international events; and (2) how a state responds to a change in government policy.
3A. Global Insight's Business Demographics Model
Global Insight's business demographics forecast contains a consistent set of historical statistical estimates and forecasts by industry sector, by geographic region. The statistics include the number of business establishments, employees, and sales by industry. Industry aggregation levels include the sub-sectors and the 4-, 5-, and 6-digit classifications in the NAICs codes. The model specifically forecasts variables at the county level. Other geographic levels are created by combining, aggregating, or splitting data from this level. All business demographics modeled databases are designed to meet two key criteria. First, they must reflect economic activity that is consistent with actual information available at this level of geography. Second, they must also agree with published values for national and state employment, establishment and sales data.
The table below lists the business demographic concepts included in the BDM.
Business Demographics Model Coverage
|
* Non-standard, and not used in the FAF2 forecasts
The following discussion describes the data and estimation techniques utilized in the Business Demographics Model.
Data
Every BDM forecast starts with at least one observation of activity at the level of geography of interest. This observation, generally collected by a government agency, is treated as an "actual" measurement of the economic activity within a given geographic area. In fact, this observation is actually an estimate of activity. The government surveys a percentage of employers within the region and then imputes the value for the region as a whole from this sample. As with any estimate, these "actual" observations may deviate from the "truth." However, as the size of the geographic area increases, so too does the accuracy of the estimate. This occurs due to the law of averages. It is for this reason that the sum of our county level forecasts will always add up to a measurement or an estimate of state and national level activity.
The following data sources were used as a basis for the first round model of county employment and establishments. U.S. County Business Patterns (CBP) data provides a series of county level employment and establishments from 1980 to 2002 at the four-digit SIC code and six-digit NAICs level of detail. This data serves as our starting observation of "actual" activity for most sectors of the economy. The CBP does not contain data for the government or agriculture sectors. Government data is obtained from the Bureau of Labor Statistics, and the agriculture data is obtained from the Census of Agriculture. Data from the U.S. Bureau of Labor Statistics (BLS) is the basis of Global Insight's national and state level macroeconomic forecasting services. These forecasts are available at the two-digit NAICS and SIC code level of detail for counties, and at the one-digit level of detail for MSAs. Forecasts provided by these services serve as the national and state level constraints on the county level forecasts. The counties add up to the state, and the states sum to the nation. In this way the BDM is always consistent with widely accepted levels of economic activity while also ensuring that county estimates are a valid measure of local activity.
Estimation Techniques
a. Employment and the Number of Establishments
The description of modeling methodology is broken into two sections. First, the modeling of employment and the number of establishments are discussed, followed by a description of the estimation of output.
Like many of the Global Insight models, the underlying technique of county level estimation is the "Top-Down Bottom-Up" model. "Top-Down Bottom-Up" methodology relies on using all of the information available to us at any given time. First, county level data is employed to determine the trend of data in a particular county. Both trending and sharing techniques are used here to create an independent forecast of employment and the number of establishments.
To begin, a first round forecast is calculated using CBP county level data. Employment and the number of establishments for each industry as defined by government four-digit SIC and six-digit NAICs codes are estimated by use of a five-year moving average of historical growth rates (from this point any description of procedures to estimate employment also applies to establishments). This forecast is independent of any information at the state, MSA, or national levels, and returns a unique growth path for each of the nation's 3,141 counties.
Next, a second level forecast is calculated using estimates provided in the first round. Over the period 2002 to 2030, employment in each county for every NAICS code is recalculated as a percentage of the first round estimated total for that industry sector. The resulting series represents the relative movement of employment within the county relative to that at the state level, and to employment in other counties within the state. In other words, is employment in industry X in county Y growing faster, slower, or in step with its counterpart at the state level or in the next county. Next, an estimate of employment levels is made by apportioning the forecast state level employment for that industry to each county based on its share of first round estimated employment.
At this point data for 318 Metropolitan Statistical Areas (MSAs) in the United States are introduced. In an iterative procedure, the county level forecasts are adjusted until the estimates solve for both the state and MSA. A brief description of this procedure follows. Estimates calculated by allocating state level data to the counties are summed to either the MSA to which the county belongs or to a "rest of state" variable. Those counties that comprise each MSA are aggregated into a summed MSA variable. From this, each county's share of MSA employment is calculated, and this share is used to allocate MSA employment to the counties. All of the MSAs in a state are then summed, and subtracted from the sum of the counties for each state. This value, the remainder of employment within each state but not in an MSA, is then allocated to the "rest of state" counties based on their share of the "rest of state" variable calculated above. This process continues iteratively until the selected criteria are met.
b. Output
Output by industry on national level is obtained from Global Insight's Industry Analysis Service. Industry output (as value of sales) is measured in current dollars and is available for all the four-digit NAICS code categories. The Global Insight Industry Analysis Service includes forecasts of constant dollar output and the corresponding price indexes for each of the industry sectors. Nominal dollar output is obtained as identities.
Constant dollar output is estimated as a function of total demand from the input/output block, cyclical variables, and a time trend. The functional form used imposes a unitary elasticity on the demand term, which embodies most of the explanatory power in the relationship. Additional non-demand terms are included in the equations to explain the pattern not well accounted for by the input/output model and its demand indicators – cyclicality and technological change.
National output by industry is transformed to regional measures by using region specific productivity measures from Global Insight's regional models. In addition, the share of employment by industry is used to allocate output to sub-regional geographies.
Data sources include the following: Economic Census, Department of Agriculture, Census of Mining, Annual Survey of Manufactures, Census of Transportation, FCC Statistics of Common Carrier, and Census of Services.
3B. Global Insight's Business Transaction Matrix
Information on inter-industry purchases is provided from Global Insight's Business Transactions Matrix. The primary data source for the Business Transaction Matrix is the latest U.S. Bureau of Economic Analysis (BEA) input/output tables. This data is released every five years as the benchmark input-output accounts of the U.S. The industrial breakdown generally follows a standard six-digit NAICs detail for the manufacturing sectors, and four-digit or three-digit NAICs detail for the non-manufacturing sectors.
Global Insight employs a modified RAS algorithm to forecast changes in the input-output coefficients over time. The chief merits of this method are twofold: its minimal data requirements, and the support of studies that have found the accuracy of the RAS method to be superior to other non-survey coefficient adjustment techniques.
The modified RAS method requires two sets of data: the direct coefficient matrix of an input-output table for an initial year t and a column vector of sectoral gross outputs in year t+1. Given these sets of data, an iterative adjustment procedure is applied to the direct coefficient matrix, which yields an adjusted coefficient matrix for year t+1 that is consistent with the ratio of intermediate input to output and the gross output measures of that year.
Once the input-output matrix forecast estimation is complete, purchases by industry and county can be determined. National use factors (defined as purchases by industry j from industry i per employee in industry j) are calculated, and then multiplied by the number of employees in industry j by county from the BDM, resulting in an estimation of purchases by industry j from industry i in each county.
3C. Global Insight's World Trade Service World Trade Model
The Global Insight world trade forecasting system provides detailed forecasts of international commodity trade to assist decision makers involved with international commodity transportation. The world trade forecasts include all commodities that have physical volume, but not trade in services or commodities without physical volume, such as electricity. The trade forecasts are produced with a system of linked world trade commodity models collectively called the World Trade Model (WTM). The commodities forecast are grouped into Global Insight's own categories derived from the International Standard Industrial Classification (ISIC) and cover 77 ISIC categories. For all trade partners in the world, the WTM has 54 major countries individually and groups the rest of the world into 16 regions according to their geographic location. Therefore, Global Insight forecasts 77 commodities traded among 70 country/regions. This is a framework of 77´70´(70-1), or 371,910 potential trade flows. Because not every country trades every commodity with every other country, there are presently about 270,000 non-zero trade flows in the forecasts. The forecasts of world trade are in nominal and real commodity value and are converted to physical volume by transportation mode. Primary modes of transportation include air, overland and maritime transport, all measured in metric tons as well as in value. The table below shows the aggregate level concepts of world trade in the forecast.
Concept |
|---|
Nominal Value |
Real Value |
Airborne Nominal Value |
Seaborne Nominal Value |
Airborne Real Value |
Seaborne Real Value |
Airborne Metric Tons |
Seaborne Metric Tons |
Over Land / Other Transportation Nominal Value |
Over Land / Other Transportation Metric Tons |
All Transportation Mode Metric Tons |
Trade Data Sources
The primary international trade history data come from the United Nations as processed and published by Statistics Canada. These commodity trade statistics are collected from member countries' customs agencies. Customs departments have records of both the export side and import side of trade flows. Statistics Canada produces export data in f.o.b. (free on board) terms, which are better to use in estimating the real value of commodity trade. These data cover all UN member countries and non-member economies, such as Taiwan. Global Insight also uses OECD international trade by commodity statistics for more current data from developed countries. Because international trade statistics collected by different countries usually have discrepancies and because no one source has complete data, Global Insight also uses U.S. Customs data and IMF Direction-of-Trade data to calibrate and supplement historical commodity trade data. Data from different sources are recorded in different classification systems and units of measurement. Global Insight converts data into thousands of current U.S. dollars and into 1997 real commodity value.
The Global Insight world trade forecasting models also rely on Global Insight's world macroeconomic history and forecast databases. Among the data used are population, GDP, GDP deflators, industrial output, foreign exchange rates, and export prices by country. These data are exogenous variables in the trade forecast models. For international commodity prices, data from the U.S. Bureau of Labor Statistics on international import and export prices are used. Global Insight also uses other data, such as foreign direct investment and import tariffs as determinants of a country's export capacity and import costs.
Modeling International Trade
The basic structure of the model for the trade flow of a commodity is that a country's imports from another country are driven by the importing country's demand forces, enabled by the exporting country's capacity of exporting (supplying) the commodity, and affected by the exporting country's export prices and importing country's import costs for the commodity. A country will import more of a commodity if its demand for this commodity increases. At the same time, the country will import more of this commodity from a particular exporting country if that exporter's capacity to export this commodity is larger and its export price for this commodity is lower than in other exporting countries. Importers will ultimately purchase based on the delivered cost, importing more when the import cost decreases. The distance between two countries is also an important factor in determining the scale of trade between two countries. Our models are constructed to capture the dynamics of international trade so that geographic distance as a constant is embedded in determining the scale of the base. Demand forces are commodity specific. Presently, Global Insight groups 77 commodities into two types: (1) those where major demand forces are the importing country's population and income growth; and (2) those where major demand forces are the importing country's production and technology development.
Export capacity for a commodity is estimated based on the country's capacity to produce this commodity and its ability to export it. Infrastructure, the establishments and resources needed for production determine production capacity. For export capabilities, the models estimate the production capacity that exceeds that needed to meet a country's domestic demand. Export capability is also determined by quality and cost of products facing competition in world markets. Import costs are determined by export prices, import tariffs, and each importing country's foreign exchange rates. The 77 commodity groups are categorized on the basis of the demand response to import costs as price inelastic, low price elastic, and price elastic.
The models are constructed in real value terms. That is, value type variables are in terms of value minus the effect of price inflation. For example, the trade flow of a commodity is measured in the 1997 value of this commodity, and GDP of a country is measured in its 1990 value of GDP. Global Insight uses data in real value terms, because only in real terms do the levels of imports and exports show clear respective responses to changes in demand, supply, and prices. Global Insight does not simply forecast a country's aggregate imports and exports, but forecasts each country's imports and exports with each of its trade partners. Trade between each pair of trading partners can be quite volatile, with importing behavior exhibiting switching of suppliers on an ongoing basis. To capture trade pattern switching, Global Insight use multi-stage switch modeling in the trade forecasting.
The multi-stage switching model approach represents an important improvement on earlier trade model methodologies and better captures the longer term characteristics of individual commodity trade. At the same time this approach is consistent with Global Insight's World Macro and World Industry forecasts.
previous | next
