A. Descriptive and Predictive Uses
The practical uses of input-output are of two kinds—descriptive and predictive.
In descriptive uses, the model simply informs users about the current regional economy:
- Which industries have how many jobs and how much in earnings
- What portion of an industry’s supply needs are likely purchased inside and outside the region
- Whether any industry clusters are present, and how well they are developed
- What are the main sources of residents’ income
- What is the composition of the region’s economic base (industries that tend to bring money into the region rather than recycling money already there)
Using a model descriptively can reveal new knowledge about an area’s economy that would be impossible to get otherwise, or it may confirm anecdotal evidence. In either case, when correctly interpreted it provides a great starting point for community discussion.
Predictive uses are just as important, if not more so. The model can estimate the total impact of a hypothetical or actual looming event in one industry. The event and its impact are described in terms of changes in jobs, earnings, or sales. For example, entering into the model a loss of 100 jobs in a local industry might reveal a loss of 75 additional jobs based on multiplier effects in several other local industries. Without the model, planners would not have been as well prepared for these indirect effects on the local economy.
B. Steps for Approaching IO Modeling & Analysis
The general steps involved in using a regional input-output model for policy analysis are:
Step 1: Define the region to be analyzed
The preparation and results of IO models depend greatly on how you choose to define the “region” to be analyzed. Do we analyze the “core” of a city, or should we include suburbs and bedroom communities too? In the case of “twin cities” do we isolate one, or treat them together as a single region? These issues are addressed in detail in Section VI, “Defining Appropriate Regions.” The basic principle, however, is that you should not artificially split up closely neighboring areas that have strong economic ties with each other—especially if there is a lot of commuting among them. Metropolitan Statistical Areas (MSAs) are generally good boundaries for analyzing city economies, while smaller communities are often better defined using one or more ZIP codes.
Step 2a: Inform researchers and community stakeholders
Regional input-output models convey a tremendous amount of baseline information right out of the box, before you’ve even started to run hypothetical scenarios (this is their “de-scriptive” function). As a result, it is always wise for policy makers to take stock of the current industry/occupation make-up of the region’s workforce, the relative magnitude of residents’ outside versus inside incomes, the importance of central functions (e.g., being a retail or transport hub), the leading industry clusters, and so on.
- Your region has a big enough industry base to have $50 million in demand for en-gineering services, but the local engineering services industry only has an output of $5 million. Using this as a basis for on-the-ground research, you could go on to discover which specific types of services are most needed, and then find out how your areas might attract more engineers.
- You discover that wineries are an emerging regional industry. However, the model shows that few or none of the key industries it is likely to purchase from—such as glass bottle manufacturers, fertilizer manufacturers, wholesalers, freight trucking, commercial printers, and cardboard packaging manufacturers—are located re-gionally. You decide it is feasible to recruit businesses in some of these categories into the area, and the result is that the region captures more of the wealth being generated by the wineries.
- You have anecdotal evidence that your town is becoming a “bedroom commu-nity” for a nearby city. But what is the economic effect of this? A look at the model shows that, on the positive side, outcommuters are earning an estimated $250 million, which they spend on services and retailers in your town. On the negative side, your town as a result has a disproportionate number of jobs in low-paying sectors like Services and Retail, bringing average earnings per job well below the state average. Your community then decides to create more good-paying job opportunities in the town itself, while also seeking to preserve the quality of life that has drawn commuters to live there.
This informational step can also generate good feedback from stakeholders with addi-tional local knowledge that can be used to customize the model (see next step).
Step 2b (Optional): Customize the model with local knowledge
Recall that most IO models are calibrated for a specific region with estimates for important variables such as the percentage of each industry’s requirements that come from in-region firms. This calibration is entirely mathematical and uses secondary (that is, non-survey) data sources, which may be incomplete. For detailed industry categories in smaller regions, even data such as the total jobs or earnings in some industries have been estimated in the model, since these numbers are often “suppressed” by government agencies to prevent their statistics from breaching the privacy of a particular business.
So, it is a good idea to spend some time with on-the-ground research (to whatever extent you can) in order to customize your model, especially if:
- You are analyzing relatively small region (town or small city), especially a ZIP-code based region. The accuracy of any model’s estimates will decrease as the size of the region decreases.
- Your region has a handful of important businesses that employ a significant portion of the area’s workforce (especially if your analysis involves the industries or related industries that these businesses are categorized under).
- You intend to study the economic impact of a specific business’s expansion or contraction. IO models are intended to represent industries, or whole categories of similar businesses, not specific businesses. A particular business may have a supply chain and production processes that differ from the average and that will affect the model’s results.
Some important variables to nail down with local knowledge and research include:
- Jobs and earnings by industry—you may need to refine exact figures for total jobs and labor earnings in the region’s most important industries (and industries you specifically want to study), especially if there are only one or two big employers in each of those industries.
- Exports/imports—what percentage of each industry’s sales are to out-of-region vs. in-region customers? This also helps the model determine how much each industry purchases from in-region vs. out-of-region suppliers. Industries that both (a) produce a lot of exports, and (b) purchase a lot of their requirements in the region, will have greater multiplier effects and thus be more beneficial to the region.
- Profits—the percentage of key industries’ profits that remain in the region (extent of local ownership). Locally-owned firms tend to keep all profits in the region, boosting their multiplier effect, while firms owned out-of-region will “leak” this income elsewhere.
Step 3: Translate policy issues/alternatives or looming events into direct effects
Most people want to use IO models not just for their descriptive features—they want to predict the effects of a looming event or policy alternative. To do this, we need to translate the situation facing us into a “direct effect,” which is a change caused by factors outside the model, and which serves as the input to the model. The most common form of a direct effect is a positive or negative change in the jobs, sales, or earnings of one or more specific industries.
Here are some examples of specifying direct effects:
- Translating policy issues to direct effects – It is not always immediately obvious how a certain policy alternative can be translated into direct effects, but it is essential that the modeler do it correctly. For instance, as the US government redefines building standards to make them more energy efficient and environmentally friendly, there will be a direct impact on industry—particularly construction, building retrofitting, architecture, and building materials manufacturing. Policy makers can use IO models to simulate the impact of these projects at the local level and visualize how this could impact the region. In this case, local construction and retrofitting companies would likely have more sales, and thus hire more workers.
- A construction/infrastructure project – A construction project will generate sales to the appropriate regional construction industry(ies), but it will also generate purchases of building materials like concrete, gravel, asphalt, steel, and so on. The researcher’s job is to delineate these dollar figures and find out how many will go to in-region vs. out-of-region industries. The result will be a set of direct effects that can all be entered into the model. However, it is also important to remember that jobs created by such projects, and predicted by the model, will only last as long as the project itself. More advanced and customized analysis is needed to figure out the more far-reaching effects of an infrastructure project. For example, a road improvement project may allow more people to commute into the city center, which could cause residential construction growth in the suburbs and matching decline in the city center. Or a new fiber optic project may allow your town to attract a data center that will create 250 new jobs—but the model wouldn’t be able to predict those jobs simply based on the construction effort involved in laying the cables. Similarly, “quality-of-life” projects like a new park, aquarium, stadium, downtown revitalization, etc. are too complex to be judged with an IO model alone—it cannot predict, for instance, that more mobile young entrepreneurs and the jobs they create will move to your area (or stay there instead of moving elsewhere), just because of your downtown is now more attractive.
- Industry expansion or contraction – When your area is faced with the arrival/departure or contraction/expansion of a specific employer, the task of estimating direct effects is a lot easier. If a company announces that it will be laying off 100 workers or adding 100 workers, you can run this number through the model, in that company’s primary industry, to predict the direct and indirect impacts of a change in that industry on the region. Or if a prospective new employer wants $5 million in incentives from local governments, an IO model (with optional fiscal impact components) can help you determine the cost per job likely to be created, and the possible return on investment of providing those incentives.
- An itemized purchases approach – If you have specific knowledge about an event or project, it may be more accurate to apportion out the total spending to multiple specific industries, rather than merely entering the lump sum as new sales in one industry. For example, if a textile factory is closing down, you could simply enter a loss of 100 jobs in the textile industry into the model. Or, you could manually apportion the value of the factory’s regional input purchases and workers’ earnings to the specific regional industries where they actually go—putting in a long list of smaller direct effects rather than one big one. This is especially useful if the factory’s production processes and inputs are quite different from the national average. This approach is also often useful when evaluating construction projects.
Step 4: Enter direct effects into the model
Now let’s talk about how to run the analysis. First, we add in the estimated direct effects in terms of jobs, earnings, or sales gained or lost. There may be multiple changes in multiple industries. The model runs millions of calculations in a few seconds and displays total impact of the chosen direct effect. The model also breaks down the change (positive or negative) in regional jobs and earnings by industry, and provides other information such as the different multipliers associated with the event.
Let’s look at an example that involves both construction and long-term job impacts. Suppose our area has a successful IT company that is considering moving some of its operations outside the region. After some negotiation with local governments, the company decides to expand locally instead. The expansion will involve the construction of a new $20 million building, and is projected to save and create a total of 500 jobs in the region.
Construction Phase Impacts
Suppose the new building will take two years to complete, thus pumping $10 million a year for each of the two years into the regional construction sector. Using the IO model we first discover that the regional construction sector currently has about $225 million in total sales. Running the model, we find that the addition of $10 million in sales in this sector should produce about 80 jobs in that sector, as well as about 65 additional jobs in other sectors. The jobs multiplier for commercial construction in our region is thus (80+65)/80 = 1.81. So for every one job from the initial change, another .81 jobs are created. The other industry areas most impacted by the addition of these jobs are local government, restaurants, health care, and retail. There will also be an impact in construction materials manufacturing industries, like concrete products and fabricated metal products. The model also shows that the building phase will mean an influx of $6.3 million of earnings (labor income) into the area (note that earnings are very different from sales—not all of the original $10 million in sales will result in local earnings).
In-Place or Recurring Impacts
Once the construction phase in our example is complete, these jobs and the resultant spending will most likely disappear, move elsewhere, or possibly be sustained by effects from other future projects. Remember—construction projects tend to have short-term impacts, unless there is solid evidence that they will create or save jobs in other industries that aren’t directly in the construction supply chain (e.g., new infrastructure needed to attract a major new employer to the area).
Next, let’s take a look at some of the potential long-term impacts of the event. In this case, we are creating/saving 500 jobs in a particular industry—e.g., “Computer facilities management services.” This needs no translation in order to become a “direct effect” we can enter into the model. When we do, we find that the addition of these jobs (with a jobs multiplier of about 1.57) will produce an additional 285 jobs throughout the region. The total annual impact of these new jobs could be $31 million worth of new earnings per year—much higher than the construction phase and also having the potential of staying in the economy for a much longer period of time.
Again, the IO model makes this a very straightforward exercise. The model will show the change, positive or negative, in regional jobs and earnings by industry (with detail down to the 6-digit NAICS code level if you are using EMSI’s EI model).