Double counting from combining government Impacts and industry impacts



  • Official comment

    Hello Dale,

    Please clarify your statement that we would "call it a contribution impact." Are you looking at the contribution of the entire industry to the region, or are you looking at change in the industry in the region? If the contribution of the entire industry, we recommend the use of contribution analysis methodology to prevent overestimating the industry's contribution.

    *Note, the link describes multi-industry contribution, but the same process works for a single industry.

    Once built, the only way to affect a model is to use the customize features. The model of the economy is not affected by the events, activities, and scenarios you apply to the model. Additionally, scenarios do not affect each other. That is, you can run an analysis for the contribution of the industry as scenario 1 and another analysis for the government activity as scenario 2 and the results of one scenario will not affect the results of another scenario. So long as you did not combine the results of the two scenarios, double counting would not be an issue.

    If the goal is to combine the two scenarios, then, as you suggest, something will need to be adjusted to account for the double counting. If you are manually entering your government activities as a series of commodity change events or if you are importing government institution spending patterns, you could set the Local Purchase Percentage (LPP) for the purchases of commodities produced by the industry to 0. This will allow the government spending to be accounted for, but it will not generate a first-round impact to the industry of interest. However, there could still be later-round impacts to the industry of interest. If you've customized the model for contribution analysis (as described in the link above), then the industry will be unaffected by any round of indirect and induced effects. In short, you can combine the results from both the industry contribution and the government activities and the Total Industry Output (TIO) for the contribution industry sector will be the same as your original TIO for the industry.

    Multiplier Specification Question:
    In user preferences, you're setting preferences for how you would like the model to be constructed. In Social Accounts, the trade flow method selected determines the methodology used to estimate the local availability of commodities, affecting the transactions that appear in the social accounts. In Multipliers, you're directing the model to calculate the multipliers you prefer (I.e., which social accounts would you like to include in the multiplier calculation).

    The Multiplier Specification dropdown on the Model Overview screen is not a driver of any part of the model; instead, the dropdown is simply a display list to inform the user as to what accounts were included in the multiplier for the model. Once a model is built, this is the only visual representation to the user as to what accounts were included. In short, the dropdown list should match the list of accounts with which you built the model. As to why it's a dropdown (which has the unfortunate effect of suggesting some larger function), I suspect that's simply due to a desire to save space.

    Please let me know if you have any additional questions!


  • Avatar

    Hi Dale,


    I have a couple of follow-up questions before I can get to the meat of your questions.

    1. What study area and data year are you using?

    2. How would you build a model without the sectors that you are referring to (i.e the gov't sectors)? Do you mean customizing the model by putting zeros for that industry?

    I will work with my team to clarify the other things while you let me know the answers to those questions.



  • Avatar

    Hi and Thanks for the response Lynette.

    It is a statewide study area for Connecticut. For this work I am not using any MRIO features.

    As for the gov sector question, I believe I'm suggesting that they not be included in the model build SAM feature. I've not tested what that would do yet though. It is on my list for the day.

    However, as you seem to suggest in your query, yes I could drop them from the production function for the industry. Is that about the best idea of what we have to address this issue before the impacts are run?

    Also in my first exchange, I also suggested that I drop them indirectly by not including (or rather deleting) the indirect business taxes from the events and reducing the total output estimated by the amount I would be removing.

    The personal taxes that would still be generated by the industry change activity also would need to be addressed. I believe they would show up in the fiscal impact reports though so maybe looking at the results after the impacts are run would be best.



  • Avatar
    Let's not worry about the "contribution impact" nomenclature at this point unless there is something specific that you need to point out.  I shouldn't have used and you seem to have given me the information I needed by making some assumptions regarding what I meant by the term.
    I do understand that the results of one scenario do not affect the results of another scenario.  Although, I'm sure it is good to be reminded of these things.
    Yes, that is it, I do want to add the two (gov and industry) impacts together to get one overall effect on the state's economy and, I think, I see exactly what is accomplished by setting the LPP to zero. I'll test that but yes, conceptually that seems like it would work to address the government side of the double counting issue. 
    There is however still a issue of double counting the tax dollars coming from the industry change impact that would be double counted as part of the values in an activity level change in the government spending patterns. 

    I have tried to also express this dimension in my previous post but I don't think it was addressed.  I might be wrong though and is also likely that I wasn't very clear.
    Let me try again. 
    There are also dollars that would be double counted from the taxes associated with the industry change that feed over into the total output measure being using for government activity.  This double counting of dollars, would, in my mind also occur in the induced effects.
    That is, just to repeat it, the industry change scenario would be observing dollars in the impact that are then used in the government spending pattern change scenario. 
    If you have any specific suggestions that could help me either back out these dollars before running a scenario or pulling the total estimated dollars out of the final impact observed (most likely that would be in the industry change scenario) I'd be happy to hear them. 
    One further note, I was skeptical of the list showing up in the Multiplier Specification including only the institutions (and capital and inventory changes) specifically being only the ones chosen for the build of the model but I didn't need to be. The GUI specifies exactly what I've included.
    So that's good.  Now I just need to test what the results of changing the model actually do to the numbers.
    And that brings up a question, suppose I want to remove the gov. dimension from the industry change scenario I am outlining above.  I think it would be relevant to also remove the state and local government institutions from the multiplier builds although I believe that won't take care of the indirect effects and induced effects that result from the taxes paid to government. Is that correct?
    Thank you again for your guidance.
  • Avatar

    Hi Dale,


    Apologies for the terseness of our previous response. We strive to ensure that all of our responses are prompt, cordial, and comprehensive.

    To your concern regarding double counting, our understanding is that you are worried that tax revenues generated by the direct, indirect, and induced levels of the industry change event will be representative of double counting when combined with a series of government expenditures. This will only be the case if your multiplier internalization is set to include the government accounts.

    Assuming the standard IMPLAN multiplier internalization (i.e., only the households are selected in the user preferences Multipliers screen), the tax dollars generated in the Industry Change scenario, while captured and presented in the tax impact report, do not generate further induced effects (i.e., the tax dollars are not spent).

    Assuming the multiplier internalization is changed to include government institutions, the tax dollars generated in all effect levels (direct, indirect, and induced) of the industry change will result in additional induced effects as those tax dollars are spent by the internalized government institutions. This is a legitimate approach to take when assuming that increased tax revenue will lead to increased local government spending.

    Therefore, if the model in which you run both portions of your analysis is built with government institutions internalized, then combining the two scenarios will double count the spending of the tax dollars generated by the industry, as the tax dollars generated in the industry change activity will drive additional government spending (increased induced effects) which you are also capturing through the additional government spending activity.

    However, if the model in which you run both portions of your analysis is built using the standard multiplier internalization (and you've accounted for spending on the target industry commodity as discussed previously), then there is no double counting of tax dollars. For the industry change, the tax revenue created is captured, but no government spending occurs. The only government spending impacting the model will be that of your government spending activity. In other words, those tax dollars are spent only once.

    Note that even though the tax dollars appear in one location (tax impact report generated by the industry change) and then also as part of your government spending pattern, we would not classify this as double counting. Instead, it's indicative of the circular nature of an I/O model. For example, consider how employee compensation (EC) is treated by the model with standard multiplier internalization. An Industry Change Activity will generate direct and indirect EC (and will be reported in the detail results as such). A portion of that EC goes to households and then gets spent locally, generating additional impacts (all induced in this case). We don't consider the spending of that EC as double-counting. 

    In your example, if you were to not internalize governments as part of your multiplier, and were to include both your industry change and your government spending, the result represents the same kind of circular nature and not double counting as we would define it. The only major departure is that impacting government spending directly will result in the government spending appearing as direct effects instead of induced (as would happen if the government accounts were internalized).

    Please let us know if we can provide further clarity.



    IMPLAN Team

Please sign in to leave a comment.