Constructing the IMPLAN SAM

OVERVIEW

Constructing IMPLAN's annual databases requires gathering data from a large variety of sources, converting them to a consistent sectoring scheme and year, estimating the missing components, and controlling the newly formatted data against other known data sources to maintain accuracy.

Raw data availability differs with each level of regional resolution. At the national level, nearly all database components are available, while at the state, county, and zip code levels increasingly fewer raw data are available. County-level information is typically available for Employment, Employee Compensation, Proprietor Income, Population, Federal and State Government Finances, and selected wealth data, leaving the remaining county data to be estimated. At the zip code level, only County Business Patterns (wage and salary employment) and demographic data from the Bureau of Census are available. Zip-code data are controlled to county totals, county-level data are controlled to the state totals, and State-level data are controlled to the national totals.

This article describes the data and data development processes employed to construct the IMPLAN SAM at both the national and sub-national level. This is described in two sections: the Industry-related I-O matrices (inter-Industry flows) and the Institution-related matrices (inter-institutional flows). Prior to that, however, it is helpful to understand one the techniques IMPLAN employs to balance matrices and estimate missing values. This is known as the RAS Method.

RAS IN IMPLAN

The RAS Method is a widely used technique for balancing matrices and estimating missing values. A matrix is balanced when the sum of the values in both rows and columns sum to the control total for each. At IMPLAN, the RAS Method is used in both the production of annual datasets and in the creation of IMPLAN models (i.e., the construction of balanced Social Accounting Matrices (SAMs) by the model build process).

Within the context of producing the annual IMPLAN datasets, the RAS method is used to ensure that the economic data estimates balance across geographical and industrial hierarchies simultaneously, and across geographical and other hierarchies, such as household income groups.

Read more about RAS in The RAS Method.

NATIONAL I-O TABLES

Each year, IMPLAN gathers current data at the national level, puts it into the IMPLAN data format, and derives new national I-O matrices (use, make, by-products, absorption, and market shares), as well as national tables for deflators, margins, and Regional Purchase Coefficients (RPCs).

IMPLAN uses the RAS process to adjust the annually-released data values to the BEA Benchmark Tables (Use and Byproducts matrices). Because the BEA only releases these tables every five years, and are always lagged relative to the annual IMPLAN data, a rebalancing of the matrix is required annually.

The structural matrices are the basis of the Inter-Industry flows (the flow of dollars between Industries). There are two structural matrices: the Use Matrix and the Make Matrix. The Make Matrix shows the production of Commodities made by each Industry. The Make matrix has two coefficient forms: Byproducts Matrix and Market Shares Matrix. The Use Matrix shows the use of Commodities by each Industry. The Use Matrix, in coefficient form, is the Absorption Matrix. The Commodities an Industry buys, per dollar of Output, is also known as the Intermediate Inputs.

MAKE/BYPRODUCTS MATRICES

Make/Byproducts Matrices: These matrices show the value of all the Commodities each Industry produces. These matrices are thus IxC (rows are industries, columns are commodities).

• Make Matrix = production values
• Byproducts Matrix = coefficients, calculated by dividing each value by the row total (rows sum to 1). The Byproducts Matrix shows the Commodity make-up of each Industry's production.
• Market Share Matrix = coefficients, calculated by dividing each value by the column total (columns sum to 1). The Market Shares Matrix tells us the proportional share each Industry has of the region's production of a Commodity.

The Make Matrix represents the make, or production, of Commodities by a given Industry. The Make Table from the latest Bureau of Economic Analysis Benchmark Tables is price-updated to the current year and forms the basis for the IMPLAN model. Rearranging the U.S. Make Matrix into IMPLAN format allows us to divide each row element by the row total to create a Byproducts Matrix.

Since the Absorption Matrix are coefficients and we do not have Total Commodity Output (TCO) controls, it is not necessary to RAS the Make Table. Accepting the Byproducts Matrix now makes it possible to calculate TCO as the sum of each column of Total Industry Output (TIO), distributed across the matrix.

USE/ABSORPTION MATRICES

Use/Absorption Matrices: These matrices show the Commodity purchases each Industry makes in order to produce its Output. They are CxI matrices (rows are Commodities, columns are Industries).

• Use Matrix = Industry outlays for intermediate goods and services.
• Absorption Matrix (also known as the Direct Requirements table) = coefficients, calculated by dividing each value in the Use matrix by the column total (columns sum to 1). The absorption matrix is also known as the production function.

While IMPLAN's production functions are based on the BEA Benchmark I-O Tables and national data (augmented for greater Industry detail as described in Industry Scheme), the Gross Absorption Coefficient values for all regions change year to year as they are forced to conform to new Industry Output and Value Added totals and new Commodity supply, demand, and trade values each year. This process is described in more detail in Intermediate Input Data.

FROM I-O TO SAM

Once we have a preliminary set of national I-O tables, the absorption table can be adjusted to conform with updated values for Value Added, Final Demands, TIO and TCO. The Intermediate Industry and Commodity Output values are calculated and used as new row and column control totals for the Use Matrix. The Use Matrix is then RASed {bi-proportionally adjusted} to match the new row and column totals. Some values, for which values are known, are held constant and are not included in the RAS process (i.e., they are not adjusted).

After the adjustments are made, the national model balances with total Value Added equaling total Final Demand. TIO equals TCO, making Intermediate Industry and Commodity Output equivalent.

The column and row controls for nearly all elements of the U.S. SAM come from the BEA's National Income and Product Accounts (NIPA). Interior cell estimates come from our annual data development processes, with the exception of some elements of the institutional trade and capital accounts, which are calculated as part of the balancing routine. Balancing refers to the act of forcing each row sum to equal its corresponding column sum; balancing is necessary due to varying levels of precision and resulting rounding discrepancies in the data used to construct the SAM.

The graphic above shows a summary SAM. The full-detail SAM, which contains every economic transaction between all actors in the economy that occurred that year, is too large to be displayed within the IMPLAN application, but it can be exported from Region Details. When viewing the full-detail SAM, it is helpful to have the following list of economic actors and transaction types, and their associated codes, which appear in the SAM.

INSTITUTION CODES

TRANSFER CODES

HOUSEHOLD INCOME AND EXPENDITURES

At IMPLAN, we base Household income on the Bureau of Economic Analysis (BEA)'s "Personal Income" numbers controlled to current BEA National Income and Product Accounts (NIPA) for the nation.

Household personal consumption expenditures are derived from the NIPA Personal Consumption Expenditure data, with household income category detail obtained from the Bureau of Labor Statistics (BLS) Consumer Expenditure Survey (CES) data. Beginning with the 2017 IMPLAN data, we now also incorporate state-level PCE data from the BEA (which are lagged one year but controlled to the current NIPA totals). Read Estimating Household Income and Expenditures for more details on the data sources and estimation process.

GOVERNMENT INCOME AND EXPENDITURES

The BEA NIPA datasets provide the control totals for government transfers. These control totals are allocated to states and counties based on IMPLAN Industry data as well as data from the Annual Survey of State and Local Government Finances and the Census Bureau’s Annual State Government Tax Collections series. Read more about the data sources and estimations processes in Estimating Administrative Government Revenues & Expenditures.

TRADE

Payments by Trade to other Institutions represent a flow of money into the region from outside. Payments by other Institutions to Trade represent a flow of money from inside the region to other regions. Imports and exports of commodities are specified as described in Estimating Trade Flows. Trade flows of Labor Income (i.e., commuter flows) are described in Estimating Commuter Employee Compensation. The remainder of the trade entries are used for balancing. For the purposes of balancing the SAM, balancing is accomplished by making adjustments in the foreign trade and capital accounts. This is done by the software during the model building process.

ENTERPRISE

The Enterprise account holds corporate profits with inventory valuation and capital consumption adjustments (transferred from OPI). The U.S. control total value comes from NIPA table 1.7.5.

CAPITAL

Payments by Capital to other Institutions represent net borrowing of money by that Institution. Payments by other Institutions to Capital represent net savings by that Institution. The Capital accounts are a balancing item that is allowed to float. If the other elements are specified correctly, the capital accounts will be accurate. Balancing in the Capital accounts is done by the software during the model building process.

INVENTORY

Depending on the commodity, net inventory change is based either on data from the Annual Survey of Manufactures or from the latest BEA Benchmark I-O tables. The U.S. control total for net inventory changes comes from NIPA table 5.2.5.

Read more about Private Enterprise, Capital and Inventory data sources and estimation procedures in Estimating Other Final Demand: Capital, Inventory, and Private Enterprise.

REGIONALIZING THE SAM

THE STRUCTURAL MATRIX TABLES

Building a regional SAM model involves combining a wide range of regional-level data with a set of national-level “structural matrix” tables. The national structural matrix tables are used in every regional model, and in the case of the Absorption Matrix, the U.S. values are adjusted in accordance with the regional-level data to yield a balanced regional SAM. The assumption is that the local Output and VA data are correct, and the national coefficients need to adjust to fit the local situation. Read more about this process in Intermediate Input Data. In the case of the By-product Matrix, Deflators, and Margins, U.S. values are applied to regional-level data without adjustments. These national structural matrix tables include:

Absorption Matrix: Each Industry's purchase of intermediate goods and services (Commodities) in coefficient form. When a sub-national model is constructed, these national coefficients are adjusted to match local area Value Added data.

The national Gross Absorption Coefficients (percentages) are adjusted to each region's Value Added per Output ratio for each Industry. For example, if an Industry in a particular region has higher Value Added per dollar of Output, then that Industry must have correspondingly lower Intermediate Inputs per dollar of Output, so the national Gross Absorption Coefficients get adjusted downward proportionately so that the sum of absorption coefficients + VA/Output = 1. To get the Regional Absorption Coefficients (net of imports), each Gross Absorption Coefficient is multiplied by the RPC for that Commodity, which varies across regions.

By-products Matrix: Each Industry's production of Commodities in coefficient form. Using the region's Industry Output, this matrix defines the total value of Commodity production by each Industry.

Deflators: Output deflators indexes of inflation for each Commodity (with the current data year index set at 1.00). GDP deflators are similar indexes which are applied to the Value Added factors. These deflators are not used to create the Social Accounts or Multipliers but can be necessary for impact analysis.

Margins: A set of ratios applied to purchaser value that will convert sales to producer values. This is not used to create the Social Accounts or Multipliers but can be necessary for impact analysis.

Sub-national SAMs are not estimated as part of the data development process; rather, the IMPLAN software creates region-specific SAMs during the model creation stage. With the exception of some inter-institutional transfers, the majority of the state, county, and zip code SAM data come directly from the regional IMPLAN Industry data estimated as described by Introduction to IMPLAN Data and Data Sources. The software allocates the remaining national SAM data to states, counties, and zip codes based on our IMPLAN Industry data and other regional data. The software then combines these data with the Industry data and balances the institutional trade and capital accounts to form a balanced regional SAM. This process allows for custom or combined regions.

TRADE AND COMMUTING

In addition to adjusting national-level absorption coefficients to regional-level data on Industry Output and Value Added values, regional-level data on commuting rates and Commodity trade are sources of additional regional specification in IMPLAN models.

Regional-level foreign trade of goods and services are based on national-level foreign trade rates. IMPLAN’s Gravity Model provides region-specific estimates of the domestic trade of goods and services. Read more about this in Estimating Trade Flows. Region-specific commuting rates are obtained from BEA’s regional data on the net flows of Employee Compensation (net of payroll taxes) and from the Census Bureau’s Journey-To-Work data on county-to-county worker flows. Read more about this in Estimating Commuter Employee Compensation.

 

Written April 15, 2024

Updated July 30, 2024