Larger study areas tend to yield larger total impacts because larger geographies can typically capture more production as 'local' and are subject to less in-commuting. Stated differently, larger study areas tend to have less leakage to imports and to commuting.
However, this is not always the case. Analysts are occasionally surprised to find that the economy of a smaller subset region, such as a county, reflects a greater Indirect and Induced impact than that of the larger aggregate region (i.e., the state).
Furthermore, larger study areas are, in effect, weighted averages of all their smaller geographic components and therefore will not accurately reflect any one of its smaller component geographies.
This article will discuss 1) the most common reasons for a smaller geography yielding larger total impacts than its larger parent geography and 2) the aggregation bias that occurs when using a larger study area as a proxy for a smaller geography contained within it.
When a Smaller Geography Yields Larger Impacts than its Larger Parent Geography
Why does this occur? How can a smaller region have greater Indirect and Induced Effects than it has when you include surrounding geographies?
The key thing to remember is that regional purchase coefficients (RPCs) represent local use of local supply relative to local demand. A state may have a higher total supply of a given commodity, but it will also have higher total demand for that commodity. If the production of a commodity is concentrated in one or a few counties, then these counties will likely have higher RPCs than the state as a whole.
Typically larger Indirect and Induced impacts in a smaller subset region are the result of areas of high production surrounded by more rural regions. This creates a situation where we see only a small bump in production between the smaller geography and the larger one, but a significant increase in demand. This change can be economy-wide, or it could be related to a specific commodity as a result of regional specialization or clustering. In these areas, the supply relative to demand is much higher in the smaller region than in the larger region (i.e., the RPCs for what is regionally available in the smaller region exceeds that of the larger region). Therefore, the larger geography sees a much larger increase in demand for the products produced in the smaller geography but does not substantially increase the supply available to meet that demand. Wyoming is a classic example of this type of activity because there are few regions of supply and a vast state of demand.
These same principles can apply in regards to labor income and value-added because the regions of greater production often pay higher wages per worker and may pay higher taxes or be subject to additional taxes (such as city taxes not collected in other counties in the state). Since value-added = labor income + other property income + taxes on production and imports net of subsidies, if labor income and/or taxes and/or profits are higher in the core region, "upside-down" effects, where the results are higher in the smaller region (county) than in the larger region (state), may be generated.
Likewise, using employment to set up the impact of an industry can result in highly variant impacts at the state vs. county level if output per worker ratios vary significantly between counties in the state. Production areas with a greater output per worker than the larger surrounding area may reflect a larger impact than the aggregated region as a whole.
Before IMPLAN had multi-regional input-output (MRIO) capability, analysts were forced to:
- choose the small region where the actual direct impact occurs but lose much of the indirect and induced impact to leakage, or
- choose a larger region to capture those leaked impacts, but now the impact location is less precisely defined.
This is no longer necessary with the ability to use MRIO. Now the smaller region can be chosen for the Direct impact while still affording analysts the ability to see the impact on the neighboring regions (and those regions' feedback effects back on the smaller region). MRIO also allows for each region to keep its unique identity and for you to be able to see how the impacts in the core sub-region and the larger aggregate region occur.
A related topic is the aggregation bias that occurs when a larger geography is used as a proxy for one of its smaller component geographies. As noted in the introduction, larger study areas are, in effect, weighted averages of all their smaller geographic components and therefore will not accurately reflect any one of its smaller component geographies.
Let's consider the state of Wyoming again. In 2019, the average Employee Compensation per Wage & Salary Worker in the Offices of Physicians industry for the state as a whole was $114,759, but the county-level rate ranged from a high of $220,136 in Goshen County to a low of $52,277 in Carbon County. Clearly, a state-level model would not be a good representative of this industry in either of these counties! Similar patterns exist in most industries and states, as well as for additional economic factors that affect impact results, such as output per employee, proprietor income per proprietor, etc.
If a user would like to report the state-wide ripple effects that stem from activity in one or more of the counties in that state, it is best to use MRIO analysis rather than using a state-level model. In this way, there will be no aggregation bias and the results will be more precise. In this way, any purchases made from elsewhere in the state are still captured (not leaked) while at the same time avoiding aggregation bias and maintaining the unique characteristics of each industry in each county.
Updated June 2, 2022