The Washington State Input-Output Model

Table of Contents Previous Next

An economic model for Washington State originally developed in the 1960s by UW researchers has become the workhorse for economic forecasting studies in the region. Famous the world over for its accuracy and quality, this regional economic model is used as the "gold standard" in many academic studies, and has been used in countless applications ranging from an analysis of the economic impact of the Mariners baseball team to the budgeting process in Olympia.

The state economic model stems from the Nobel-Prize winning work of Harvard University professor Wassily W. Leontief, who published the first such model of the U.S. economy in 1936 for the years 1919 and 1929. Called an input-output model, the approach is "used all over the planet in almost every country for a variety of planning and impact analyses," says UW geography professor William Beyers.

As a graduate student, Beyers worked with a group of UW and WSU faculty who pioneered the development of a survey-based input-output model for the State of Washington. The team in 1965 included UW professor Charles Tiebout, who was a pioneer in the development of regional economic models; Philip Bourque, a professor in the UW School of Business; and UW geography professor Morgan Thomas. There were about ten faculty in all who were involved from the UW and from WSU, which handled the agriculture-related component, notes Beyers.

The model is constructed by surveying all of the industrial sectors in the state about their total sales and purchases. That information goes into a matrix, or table of numbers, that relates the outputs of all industrial sectors (sales to consumers, to industry, and exports) to the inputs to those industries (all the goods, services, labor, and so forth that the industries must purchase in order to generate the outputs). In a balanced system, the total outputs must equal the total inputs.

Survey data have been painstakingly collected over the years from all the industrial sectors of the regional economy, including everything from livestock, fisheries, and forestry, to construction, freight, postal services, industrial machines, aluminum, and so on. Data are collected and entered into the matrix.

The beauty of the approach is that once these figures have been accurately obtained, techniques of matrix mathematics can be used to predict, for any new perturbations or new scenario, what the changes in the entire system would be. The process is called impact analysis, and it can predict the change in outputs, employment, and income in all regional industries as a result of a change in one particular industry. In mathematical terms, it boils down to solving a system of linear equations that encode the interrelations that exist among industries through their purchases and sales.

The Washington State input-output model was first published in 1967, for the year 1963; then, various updates took place with new surveys for the years 1967, 1972, 1982, and most recently, 1987. "In every one of these updates, there's been new survey work done, with an ongoing, changing cast of characters," says Beyers. "Professor Bourque has been a key player in every one of these versions of the model."

Washington was one of the first regions to actually develop this kind of model, notes Beyers. The model has been further refined and applied by Dick Conway, "one of most highly respect economists in the area," says Beyers, into what are called "Input-Output Econometric Models. "

Who uses the model? "There have been two distinct uses," says Beyers. "First, because we were one of those who pioneered the development of these models by survey processesin most regions, they'd been estimated by non-survey techniquesour model has been viewed as having very good quality, as having a level of accuracy which is a cut above all but a few other regional models, because it's based on actually going out and interviewing companies and collecting huge clouds of numbers."

As a result, one use of the state model involves the development of methods that simulate, by non-survey techniques, what is measured in the survey. "To estimate the model by survey techniques may cost on the order of $250,000; but models can be constructed by non-survey techniques for a couple thousand dollars. Surveys take some three years to do; non-survey approaches can be done in a few seconds with a spreadsheet," he explains. But are the non-survey models accurate? A large academic literature has grown up around comparing non-survey approaches with the results obtained in the Washington model, treating it as the "truth" for comparison purposes. "Those studies have made the Washington model globally famous," says Beyers.

Secondly, there have been "myriad" applications of the Washington State model, notes Beyers. For example, two years ago he carried out an impact analysis for the Corporate Council for the Arts to estimate the economic impact of arts and cultural organizations on the regional economy. Later, he assessed the impact of building and of operating a new downtown hall for the Seattle Symphony--how many jobs would be created, what the effects on income and employment would be. Beyers did a similar impact analysis for the governor and the mayor of Seattle and King County executives relating to the Mariners Baseball Team. He also performed an impact analysis for the Fred Hutchinson Cancer Research Center. "It's kind of the workhorse" says Beyers, noting that it's part of the budgeting process used in Olympia to predict what the revenue streams are going to be. "There must be thousands of applications," he adds. "Hardly a day goes by that someone isn't using this model."

Table of Contents Previous Next