The purpose of this document is to outline the main data sources and procedures used to compile IMPLAN datasets. IMPLAN data come from many sources, in a variety of sectoring schemes, sometimes with differing definitional frameworks, and most with non-disclosures. Constructing a complete database therefore entails gathering data from the various sources, estimating non-disclosed values, and putting them all into a consistent sectoring scheme and definitional framework, all the while controlling estimated values to known totals or other more accurate or recent data to maintain accuracy. IMPLAN adds value to the available data by:
- Providing estimates for non-disclosed data
- Providing estimates for non-census years
- Providing estimates at a finer geographic scale (i.e., at the local level)
- Providing inter-county trade flow data, which allows for Multi-Regional I/O analysis
- Reconciliation of multiple data sources
- Bringing it all together in a consistent format
Although the U.S. production functions are based on the Benchmark I/O Tables (which are released every 5 years), the U.S. absorption coefficients are forced to sum to new output and value-added totals each year:
- We start with the latest BEA Benchmark I/O tables.
- We derive current industry output, value-added (VA), and final demands.
- Using the byproducts data and the current industry output, we derive current commodity output.
- Multiplying the current industry output through the absorption matrix gives us a first approximation of the current USE matrix (i.e., current production functions).
- The columns of the current USE matrix are forced to sum to Industry Output - Value Added (columns).
- Then the rows of the current USE matrix are force to sum to the Commodity Output - Final Demand control totals (rows). This will be the first time the proportions of the columns change.
- This is repeated until no further adjustment is needed.
Absorption coefficients for a particular industry will also vary across regions because the ratio of Value-Added to Output varies from region to region, which forces the national gross absorption coefficients to adjust (so that Total Absorption coefficient + Value-Added coefficient = 1.0). The assumption is that the local data is correct and the national coefficients need to adjust to fit the local situation. Applying the trade flow assumptions (RPCs) will then further modify the regional absorption coefficients by pulling out the imports (which vary by region).
In general, BLS Quarterly Census of Employment and Wages (QCEW) data provide the county-level industry structure for the IMPLAN database. The Census Bureau's County Business Patterns (CBP) data are used to estimate non-disclosed values, while the BEA Regional Economic Accounts (REA) data are used for control totals to incorporate proprietors and non-covered sectors1.
1. Since these data capture only covered employees, the data set does not include self-employed persons, railway employment, religious organizations, military, elected officials, or any other establishments that have their own social insurance program and/or do not pay into the Unemployment Insurance program. Since much farm employment is self-employment, CEW data has especially sparse coverage of farms. More information about estimating employment for these sectors can be found here.
Value Added Data
The calculation of value-added data starts with calculating earnings. The sources of data for earnings are the same as for employment. However, CBP provides employment only, so if a county does not have income disclosed, then state-level income-per-worker ratios are used with the employment estimates to calculate a first estimate. Next, the income estimates are used to disclose the CEW data and the CEW data are used to non-disclosure adjust the BEA's REA data. The REA data are expanded to separate wage and salary Income from proprietor's Income. The REA data are then used as final control totals, with the CEW data providing the 6-digit NAICS industry structure.
Other Property Income (OPI) is mostly corporate income and is one of the most difficult items to estimate. The data sources we use are state-level 3-digit NAICs which explains why OPI phenomena tend to occur in 'clusters':
- We use BLS QCEW data to derive wage and salary income, which is at the 6-digit NAICS level. We use that along with the compensation to wage and salary income ratio (which is available by county at 3-digit NAICS rom REA) to derive Employee Compensation (EC).
- REA's total income (which is analogous to our Labor Income) includes proprietor income but not property income, so Proprietor Income (PI) is found by subtracting EC from total income by 3-digit NAICs industry.
- Another set of tables in the BEA includes 3-digit NAICS state GDP data. GDP includes proprietor income and property income - together they are Gross Operating Surplus (GOS) - so OPI for 3-digit NAICS is derived by subtracting proprietor income from GOS. These 3-digit control values are distributed to the detailed industries based on the Benchmark I/O characteristics for property income. From one year to the next, it is possible for the sign to flip. This is also true for proprietor income.
For IMPLAN, total industry output (TIO) is the value of annual calendar year production. It can be measured as the total value of purchases by intermediate and final consumers or as intermediate outlay plus value added. Most output data is from the BEA's Annual Industry Accounts and the Annual Survey of Manufacturers. Retail data come from the U.S. Census Bureau's Annual Census of Retail Trade. Other sectors use information from other various surveys and censuses.
National household Personal Consumption Expenditures (PCE) are estimated using the BEA Benchmark I/O-to-PCE bridge tables and current NIPA PCE data. National PCE are distributed to states and counties based on the number of households and household income for each of the nine income categories. The spending patterns for each of the nine household income categories are based on the BLS Consumer Expenditure Survey (CES).
Federal Sales and expenditures data are estimated using NIPA control totals and the Benchmark I/O distribution, with the exception of the timber sales data, which are from the U.S. Forrest Service. Data for State and Local Government sales are obtained from the current Annual Survey of Governments, while State and Local Government expenditures are estimated using NIPA control totals and the Benchmark I-O distribution.
For the manufacturing sectors, the Annual Survey of Manufacturers provides the inventory data. Other sectors are derived from Benchmark I/O ratios.
For the U.S., data for the foreign trade of commodities come from the Department of Commerce import and export trade data, which includes a concordance that maps the data to NAICS. Service trade is based on the BEA Benchmark I/O tables and controlled to current NIPA values. For sub-national regions, foreign imports are assumed to make up the same proportion of the region's demand as for the U.S. Similarly, foreign exports are assumed to make up the same proportion of the region's supply as for the U.S. For example, if the U.S. satisfies 80% of its sugar demands with foreign imports, then each state and county will also satisfy 80% of their sugar demands with foreign imports.
After foreign trade has been removed, the trade flow or econometric model is used to determine domestic trade. Domestic export is a residual - after multiplying a region's RPC for a commodity by the region's total demand for that commodity, we subtract this number from the region's total supply of the good. Any residual is presumed to have been shipped to the rest of the U.S.
We use current-year NIPA investment data by aggregated industry making the investment goods and allocate that to more detailed industries according to the latest Benchmark I/O tables.
Personal Consumption Expenditures
NIPA PCE Data
- Annual and current
- National level
- Only one spending pattern, (i.e., not separated by income class)
- The NIPA table has 100 or so expenditure categories. The BEA benchmark I/O tables are used to distribute these expenditure categories among the IMPLAN sectors.
- The PCE data are in purchaser prices, so margining and re-sectoring are necessary to obtain producer prices for use in IMPLAN.
Census Bureau CES Data
- Annual but lagged
- National level
- Gives us the expenditures by income class; we control these to the NIPA PCE totals.
Data accessible for editing (IMPLAN Pro)
- All study area industry variables for each industry (e.g., VA, output, employment)
- All study area final demands by commodity purchased
- All study area institution-to-institution transfer payments
- All trade flow assumptions (regional purchase coefficients)
- All industry production functions (absorption matrix)
- All industry production (by-products matrix)
- Built-in price indices and margins
- All trade flow assumptions (regional purchase coefficients)
- lndustry production functions (absorption matrix)
- Built-in margins