Institution spending pattern in ABP



  • Official comment


    All spending patterns in IMPLAN are margined. If you are creating a spending pattern from scratch, then you do need to account for margins (see the link below).


    Manually Margining Spend Patterns for Analysis by Parts


    Additionally, when the agency states the local purchase percentage of an item, it is important to know if the percentage is in reference to the purchase from a wholesaler, retailer, or from the producer. It's possible that the purchase was entirely local (from a retailer), but the production was entirely outside of the region.


    Client spending information is often lumped into larger categories (such as those you mention) that don't match IMPLAN sectors and commodities. A few different approaches for resolving the dilemma are:

    1. Leave the value of the vague purchases out of the analysis (though the value should still be included as part of the Direct Effect Output). This is the conservative approach.
    2. Make no assumption as to what exactly was purchased and simply apply the value to the appropriate wholesale or retail category (assuming the items were purchased through a wholesaler or retailer - also note the above comment regarding margining).
    3. Make assumptions as to the makeup of the vague purchase value.
      1. Your approach of using a modified spending pattern for 456 is an interesting one, but do know that, even modified as described, the spending pattern would still include purchases of commodities likely accounted for in your other spending information (e.g., purchases of gasoline for agency vehicles and purchases of utilities).


    Also, how is equipment defined? Intermediate expenditures are current-account expenditures. If the equipment category includes investment in equipment, the investment should be accounted for separately.


    Intermediate Inputs (expenditures)


    The difference between industry spending patterns and institution spending patterns is in how the model treats them. Commodity purchases listed in an industry spending pattern are treated as intermediate expenditures and the results will only show Indirect Effects and Induced Effects. Commodity purchases listed in an institution spending pattern are treated as final demands and the results will show Direct, Indirect, and Induced Effects. If you make the adjustment mentioned previously (adding the Direct Effect to the Indirect Effect to determine the Indirect Effect) then the result is equivalent to having run the exact same commodity purchases as an industry spending pattern.


  • Avatar
    IMPLAN Support


    Thank you for your post. 

    Taxes and license fees paid to the government are classified as part of Taxes On Production and Imports Net of Subsidies (TOPI), not Intermediate Expenditures (IE). As such, they would be part of the federal agencies Value Added (VA). Traditionally, Other Property Income (OPI) and TOPI are not internalized in the multiplier (see the article below for more). However, if you feel confident that the taxes and fees collected by the government will also be will be spent locally, then the use of a general institution spending pattern (such as the State and Local Government Non-Ed) is a reasonable route to take, but all results should be treated as Induced. 

     Explaining the Type SAM Multiplier 

    The government enterprise sectors represent government establishments that operate similar to private establishments (i.e., they charge fees for service). Government agencies, however, operate based on government funding from tax revenue. You do not want to use a government enterprise sector to determine Direct Effects for a government agency. 

    The classification for Direct Effect is somewhat confusing when discussing government agencies. As government agencies are classified as a government institution, government agencies are technically a final demander (like households). Therefore, their spending is treated like a final demand, hence why your institution spending pattern is reporting Direct Effects that include purchases of both commodities such as utilities and government employment sectors. This is confusing if the intent is to describe the results similar to how industry impact results are normally described as the institution spending pattern results mix direct labor with what would be considered a commodity input purchase (indirect effect) in a standard analysis.

     The article below discusses the use of an institutional spending patterns and also describes a method to reorganize results so that they match the more typical description if such is desired. The example in the article is public education, but the method works just as well for government agencies. Note that the spending patterns are completely customizable, so you can apply your detailed budget information to an institution spending pattern.

     Working with Institution Spending Patterns

      For general reference:

    Output = Intermediate Expenditures (IE) + Value Added (VA)

     Value Added (VA) = Employee Compensation (EC) + Proprietor Income (PI) + Other Property Income (OPI) + Taxes On Production and Imports Net of Subsidies (TOPI)

     *Note: you state that you use wage information for Labor Income (LI), but know that Employee Compensation is a fully loaded payroll (wage & salary plus employee paid payroll taxes plus employer paid payroll taxes plus benefits). If you only have wage & salary, you can use our converter spreadsheet to estimate an EC value (found in our downloads section).

    • Technically, labor income also includes proprietor income, but there is no proprietor income for government agencies.

     If your total annual budget value includes all the above components, then it is equivalent to Output. You can determine Value Added by simply subtracting the sum of IE  (input purchases of goods and services) from Output. As there is no proprietor income for government agencies, Labor Income would be equivalent to known EC.

    Hope this helps and please let us know if you have any additional questions.

    Thank you! 


    IMPLAN Staff

  • Avatar

    I got an email alert with your reply although, strangely, I can’t see the comment when I look at my post on the IMPLAN community site.

    I think I’ve got the gist of your recommendation, but I do have a few remaining questions.

    First, I have pretty detailed spending information from this agency, which I should probably mention is a local branch of the EPA, including local purchase percentages for most items. Previously, I had been setting the purchases up in an industry spending pattern. Now, based on your suggestion, I am creating a new institution spending pattern and setting up all known purchases using Analysis by Parts (following the instructions in the link below).

    However, there are some categories (Equipment, Supplies, and Chemicals) which they do not describe in detail. My original plan was to use the 456 – scientific research industry spending pattern, delete all “service” sectors (after wholesale trade), normalize, and then set the level to the value for the three categories combined. My thinking was that the EPA’s purchases (for equipment, supplies, and chemicals) would be more similar to a scientific research firm than to the broad category of “Federal Government Non-Defense”. However, I now wonder about combining an institution spending pattern for the known expenses with an industry spending pattern for the unknown expenses. Is this a valid technique? And, if I do combine them, how can I align my results with the instructions in your article?

    As for your other points, my employee compensation numbers do include wages, benefits, and payroll taxes, so it sounds like value added will equal employee compensation. And regarding the licenses and fees category, the agency does state that they are all spent locally, so I will continue with my original method but move results to induced effects as you suggest.

  • Avatar

    Hi again,

    I have finished the modeling using the methods you suggest, and I just have one final question. Turns out, the agency has 60 federal employees whose wages + benefits total roughly $8 million. However, they also have 68 employees that are contracted through other companies. Some are post-doc students, some work for private companies, etc. These employees are all housed at the physical location, they all live in the community, and the agency provided me with their wages and benefits, excluding all non-local admin fees. I intend to model that amount as another labor income activity but I'm not sure which sector to put the direct effects from these wages. Most of the companies for which these individuals work are not located in the region so the impact is only coming from their wages.

    Thanks again for your help.

  • Avatar

    Sorry, one more question:

    When reporting direct effects (in the case of analysis by parts), should I report the agency's total spending or only the amount spent locally? And is the same rule of thumb true for operational spending as for wages? For example, the agency has an operating budget of about $14.9 million annually. $12.9 million of that goes to wages and about $2.0 million goes to operational spending (supplies, utilities, etc). For both of those categories I have a very detailed picture of what portion is spent within the study area. Should I include only that portion in my direct effects or should I report the full amount? Is there a "best practice" that I should follow?

    Thanks again!

  • Avatar


    Typically (new construction sectors being an exception), contract labor is considered an intermediate expenditure and is captured as part of an industry's consumption (industry spending pattern). Therefore, accounting for contract labor using a spending pattern ideally requires applying the contract amount (which would be the fee paid to the contracted firm, not just the amount the contract employee keeps as compensation) as a payment to the appropriate commodity sector (the commodity provided by the contractor's firm; e.g., 3449 Architectural, engineering, and related services for a contracted architect). 


    That being said, since all of the contract companies are located outside the region, just capturing the induced effects (via a Labor Income Change) and adding the direct payroll tax impact is fine.  In that case, it’s not necessary to select an industry for them – you can simply describe the situation in your report. If you feel that the contract workers fall within particular household income categories, you could also consider a household income change. If you do choose to use a household income change, you will need to reduce your employee compensation amount used to set up the Event by removing payroll taxes (both employee and employer paid).  You would still report the direct payroll taxes as part of the direct tax impact.


    Note: you mention that the contract workers are all housed at a physical location. Is this housing paid for by the agency? If so, you may want to even consider a more advanced approach of utilizing a household spending pattern. Such an approach would allow you to select the particular household income category in which the contract employees would fall and also allow you customize their spending (e.g., reduce or remove spending on commodity 3441 imputed rental services of owner-occupied dwellings). The household spending pattern activity takes spending (not compensation) as its activity level, so you need to remove from your employee compensation value the following items: payroll taxes (both employee and employer paid), income taxes, and savings.  Additional note: while a household spending pattern will provide you with direct, indirect, and induced effects, the effects need to be treated as induced.


    Please see this article for more information:  Comparing Labor Income, Household and Industry Change Activity Types Related to Employee Compensation


    As to what to include in your direct effects, the answer depends on how you intend to describe your results. For a standard industry analysis, direct effects represent the operational value of the industry. X amount of regional output production requires y amount employment and z amount of labor income. Not all of the commodity inputs required to meet the regional production are sourced locally, but the direct effect output still represents total production.


    Government institutions, on the other hand, are typically considered final demanders, such that their list of expenditures are classified as direct effects in those “supplier” industries, rather than indirect. So – with an Industry Change in a private industry, those input purchases would show up in the directly-impacted industry as part of its direct output and would also show up as indirect effects in each of the respective supplier industries, whereas for government they would only show up once (as direct effects in each of the respective supplier industries). With institution impacts (such as government in this case), we often recommend focusing less on classifying as direct vs. indirect and more on describing the situation along the lines of “The EPA spends $X on goods and services, which generates $Y in various industries throughout the regional economy.” In that sense, those input expenditures are kind of direct (they are included in X) and indirect (they are included in Y, which are the results from IMPLAN), which is akin to the standard Industry Change activity. The attached article on Output vs. GDP may prove useful.

  • Avatar


    Thanks again for the quick reply. The only part of your response that I'm unclear on is when you advise "adding the direct payroll tax impact" in addition to the labor income change. I have one big number for labor income, the client didn't break it down into wages, benefits, payroll tax, etc... What are you recommending I do when you suggest adding the payroll tax impact and can I do that if I don't have the breakdown of the labor income components?


  • Avatar

    The reference is to the tax impact report generated by IMPLAN. A labor income change activity does not generate direct effects; therefore, the tax impact report does not include the payroll tax revenue associated to the compensation applied to the labor income change. If you do not intend to use the tax impact report results, then you can ignore the suggestion. If you do intend to use the tax impact report, then you do want to include the payroll tax revenue. The steps below will guide you through a process to generate these values.


    1. Using the same model, create a new Industry Change Activity and label it EC Payroll Tax
    2. Select industry 517, and set the event year to the year you used for the Labor Income Change.
    3. Enter the same value you used for the labor income change in the employee compensation box.
    4. Create a new scenario called EC Payroll Tax and only include your EC Payroll Tax activity.
    5. Run the Scenario.
    6. Set your dollar year for view (monetary year) to the same as the rest of your results.
    7. Navigate to the tax impact report, filter to only DIRECT.
    8. Pull the payroll taxes from the Employee Compensation column.


    *It is important that you filter to only the direct taxes and do not pull any values other than those from the Employee Compensation column. All other values produced by this approach are irrelevant.

Please sign in to leave a comment.