Multi-Regional Input-Output (MRIO) analysis is widely used to track additional economic effects stemming from an impact in your Study Area on another Region or Regions. It is frequently used at a county level to show the impacts on the rest of the state; however, MRIO can be used at any geographic level.
Before the advent of MRIO capabilities, there was an approximation method known as "mock MRIO" which analyzed the same event occurring in a smaller region and in a larger region, like a county and a state and then distributed the difference between the two as the effects that would occur in the rest of the state (i.e., the state less the county in which the direct effect actually took place). At best, this was a shaky methodology. However, since the development of domestic trade data and MRIO capabilities in 2008, this is no longer necessary and therefore is not an acceptable method!
The correct method for analyzing the indirect and induced effects that occur in the rest of the state that stem from a change in one of the state’s counties, is to create two regions, one Region for the county and one Combined Region for the remaining counties in the state, and link them via MRIO. This article will demonstrate the inaccuracy of the mock MRIO method and identify the proper approach to model an MRIO analysis.
In an MRIO analysis, the Direct Effect in one region can trigger Indirect and Induced Effects in linked regions, thereby accounting for some of what would have been a leakage in a single-region analysis.
Why Isn't The Mock Methodology Ideal?
First off, there is Aggregation Bias from different multipliers. This means that the results will be dramatically over or underestimated. The beauty of a true MRIO Model is that each region in the analysis keeps its unique economic identity. Thus, Output per Worker relationships, Labor Income per Worker relationships, Value Added to Intermediate Inputs ratios, and other unique identifiers that contribute to the Industry Multipliers are all maintained for each specified region. With a mock MRIO, the ability to keep these unique identities is lost, so rather than summing the effect on the host county’s economy and residual effects on the rest of the state, the larger Study Area provides results that average all of the counties’ unique identities for that Industry. While this may seem innocuous, it can dramatically affect results when the impacted Industry has a competitive advantage in the smaller region over the larger region, and vice versa.
Second, there is a loss of Inter-regional commuting and trade flows. IMPLAN’s Commodity trade and commuting data provide estimates of the inter-regional flows of goods and services and Labor Income between all state, county, and zip codes in the United States. Linking regions through an MRIO analysis makes use of these Commodity and Labor Income flows cycling back and forth between the linked regions, all the while keeping each region’s local identities distinct. A mock MRIO does not make use of the estimates of trade between the smaller geography and the larger geography. Instead, the larger Study Area provides results that average all of the total supply and demand of Commodities within the area, essentially ignoring the trade flows that occur between the smaller and larger geography. A smaller geography may also have a fairly high in-commuting rate, where Employee Compensation (net of payroll taxes) flows out as a result of a net in-commuting. It is not unheard of for a metropolitan county with a large city to have in-commuting rates upwards of 50%. In contrast, a state model typically has a much lower net in-commuting rate and therefore a lower leakage rate of EC.
Let’s consider an example.
Phillips Steel is one of the mills operating in Pittsburgh, PA that manufactures steel. The Governor has asked Phillips Steel to quantify the economic impact of their operations on Allegheny County and the rest of the state.
An analyst at Phillips has determined they will run the same impact in two separate analyses - one at the county level and another at the state level. To set this up, they created one Event for the $1B in Output that the company generates in Industry 215 - Iron and steel mills and ferroalloy manufacturing. They used the Advanced Fields menu to enter their 800 Employees, $95M in Employment Compensation, and $0 for Proprietor Income. Then, the Event was added to both Groups, Allegheny County and State of Pennsylvania to run the impacts.
Exploring the Mock MRIO Results
Using the Region filter on the Results screen, the Analyst was able to export the effects on the county and the state. The analyst presented the impact results on the local county and rest of the state by subtracting the total results on Allegheny County from the total results on Pennsylvania. As shown below, running the same Event through both Regions resulted in different Direct, Indirect, and Induced effects. But why?
Total Economic Impact Results - Mock MRIO ($MM)
Looking at Industry 215’s Commodity Demand in each of the two Regions, we see that while the Industry in both Regions requires the same goods and services, it requires them at different rates relative to their value of production (Output); that is, the Gross Absorption percentages are different in each region. Just comparing the top 10 Commodities by Gross Absorption, we can see that Industry 215 in Allegheny County requires 1.5% more of each Commodity than the Industry does at the state level.
Pennsylvania - Industry 215 Commodity Demand
Allegheny County, PA - Industry 215 Commodity Demand
Unless the analyst modifies the Event at the state level to have the exact same values for ALL components of the Leontief Production Function (Output, Intermediate Inputs, Employment Compensation, Proprietor Income, Other Property Income, and Taxes on Production and Imports) as the county level, the Event will have different purchase amounts for each of these goods and services (Local Purchase Percentage). Even if the analyst does make those adjustments when setting up the Event, that adjustment can only be made for the direct effects. All of the Indirect and Induced effects in the state-level model will be based on the state’s LPFs for each of the Indirectly- and Induced-affected Industries, which will be a weighted average of each individual county’s LPFs for those industries, and thus will lose some of the regional specificity.
Furthermore, note the differences in RPCs between the two regions. While each region’s RPCs are based on the trade flow estimates mentioned above, looking at each region separately shows two problems: the state model includes the data for the county, so the state RPCs do not represent “the rest of the state” but rather the rest of the state PLUS Allegheny County. If Allegheny County happens to be the only county that produces a given good or service, then the state model will have a positive RPC for that commodity (since a state model includes Allegheny County), whereas a “rest of the state” model would have an RPC of 0 for that commodity since such a model would not (and should not, in this case) include Allegheny County!
Knowing that the Direct effect occurs in Allegheny County, rather than being spread out across the state in a weighted average manner (i.e., according to each county’s Output for Industry 215), we know that a larger proportion of the Indirect effects should occur within Allegheny County rather than spread across the state according to the affected Industries’ Output values, which is what essentially occurs in a state-level model.
The correct way to set up this project would have been to create a Combined Region of all counties in Pennsylvania, except Allegheny County. Let’s set up the same example using the proper MRIO method and compare the results to what the Analyst at Phillips presented.
Setting up the MRIO
The first step to setting up an MRIO is to build the Combined Region for the rest of the state. For this project we are combining all Pennsylvania counties excluding Allegheny. Combining a Region is easily accomplished using the Regions List on the Regions page. Once the combined Region is built, we will also select Allegheny County, PA from the map or through the Search Bar for the second Group.
To set up the impact, create one Industry Output Event for the $1B in Output that the company generates in Industry 215 - Iron and steel mills and ferroalloy manufacturing. Then use the Advanced Fields menu to enter their 800 Employees, $95M in Employment Compensation, and $0 for Proprietor Income.
To perform a MRIO, the Event is only added to the Allegheny County Group and the MRIO box at the top of the Group panel is checked. No event will be added to the second Group consisting of the other 66 counties in Pennsylvania as the steel mill operation only occurs in Allegheny County.
Using the Region filter on the Results screen, the Economic Indicator Results are separated by effects on the county, the rest of the state, and the combined total. As shown below, running the Event in Allegheny County resulted in Direct, Indirect, and Induced effects, while the rest of the state only experiences spillover effects from the Indirect and Induced purchases.
Total Economic Impact Results - Correct MRIO ($MM)
Now, let’s compare the MRIO results to the mock MRIO performed by the analyst at Phillips Steel.
Total Economic Impact Results Difference - Mock MRIO to Correct MRIO ($MM)
There are a few key differences between the results from using mock MRIO and the correct MRIO:
- As previously explained, by running the Event at the county and state level, IMPLAN used different LPFs to estimate the remaining portions of Value Added, including the OPI and TOPI for the Impacted Industry in that Region. At the state level, Industry 215 allocates approximately 0.4% more of Output to Other Property Income and .01% less of Output to TOPI, totaling $4.3 million. Due to this, the results erroneously attribute $4.3 million of Value Added as Direct Effects to the rest of Pennsylvania.
- Due to the loss of inter-regional trade and commuting flows between the host county and the rest of state, the overall impact on Allegheny County is slightly lower using the Mock MRIO method. The analyst at Phillips Steel underestimated the actual impact on the county by more than 21 jobs, $1.7 million in Labor Income, $3 million in Value Added, and nearly $5.8 million in Total Output.
- The overall impact on the rest of the state is significantly lower when using the Mock MRIO method. Using the correct MRIO method, the trade and commuting flows indicate that the actual Indirect and Induced Effects on the rest of state result in 475 more jobs supported, $47.6 million more in Labor Income, $74.3 million more in Value Added, and $133.9 million more in Output than when compared to the Mock MRIO results. These variances dramatically underestimate the actual impact of Phillips Steel on the state economy.
While the Mock MRIO method was once the only way to estimate the residual effects of impacts on other regions, IMPLAN economists no longer acknowledge it as an appropriate methodology. If your project requires an impact report on both a local and larger region’s economy, MRIO is the only appropriate method to accomplish this.
Introduction to Multi-Regional Input-Output Analysis
Local Purchase Percentage (LPP) & Regional Purchase Coefficients (RPC)
Industry Leontief Production Functions in IMPLAN
Written March 30, 2022
Article is closed for comments.