anomolous results whole and part

Comments

7 comments

  • Avatar
    DougO
    The software removes taxes, savings and imports from the total household spending before applying to the multipliers. The value actually applied to the multipliers is usually less than half of the original income, so that the total impact effect while greater than what was applied to the multiplier, can still be less than the original income. Larger areas that import less and have larger multipliers overcome the loss of the leakages to return a result that is greater than the original household income.
    0
    Comment actions Permalink
  • Avatar
    appold
    Doug, That is what worries me. Some of the regions apparently import less than the state taken as a whole. That doesn't sound right. I'd expect the parts to have greater leakage and thus lower multipliers. Some are, but some are higher. Also, I've entered the values as spending because I've already taken out a portion of the total funds as remittances, savings, and income taxes. The spending should need to pay sales taxes and whatever else comes down stream. Or am I entering the data incorrectly? Thank you.
    0
    Comment actions Permalink
  • Avatar
    jenny
    While there will be greater supply of most things at the state level than at a county level, there will also be great demand for most things at the state level than at the county level, so the LPP may or may not be higher at the state level for some items. This can happen when production of a commodity is highly concentrated in just one or a few counties in a state. The LPPs in those counties will be high but if there are many other counties that don’t produce that commodity, then the state as a whole may have a lower LPP for that commodity. If this commodity is one that is purchased a lot by households, then this can yield the kind of results you are seeing.
    0
    Comment actions Permalink
  • Avatar
    appold
    Thank you. I tried changing the option on the trade flows method to "econometric RPC" and that seemed to address the issue. I read somewhere in the knowledge base that the Implan national trade flows model is a new option. Is there documentation about this method and a "compare and contrast" somewhere in the documentation? I'm having a hard time finding my way around. Also, is there a mapping of v2 and v3 SAM multipliers? I've read some of the discussions in the documentation but am still unclear about how an option in v2 maps onto the options available in v3. I'm wondering in part because I'm trying to replicate a study on newer data. The resulting impacts per dollar spent are much smaller than they were in the past. I'm trying to figure out how much of that is due to real changes in the regional economy and how much is due to inadvertent differences in how the modelling is done. Thank you again.
    0
    Comment actions Permalink
  • Avatar
    DougO
    The econometrics are based on characteristics of the economy devoid of any consideration of geography. The gravity model has trade flows based on supply and competing demands that have no consideration for political boundaries. It is possible/likely that residents of some parts will satisfy their demands from local instate sources at a greater rate than other residents in other parts of the state and even at a greater rate than the average state resident (ie, the state model). Please tell me two of the counties/regions that show contradictory results and which household income class you used so I can see if there might be anything else going on.
    0
    Comment actions Permalink
  • Avatar
    appold
    Doug and Jenny, I've attached an Excel file which summarizes the input and part of the output. The first set of numbers contain the estimated buying power and estimated total (output) impact from an analysis done about 6 years ago on 2004 using Implan v 2 for the state, regions (aggregates of counties), and selected counties. The second set of numbers contains the same information for 2010 using v3 and the econometric RPCs. The third set of numbers uses the default trade model. I have two questions/concerns. First (in red), the newer analyses result in total output impacts which are relatively lower multiples of the input buying power than the older estimates. I'd like to figure out how much of that difference is due to a) changes in the world over the last several years, b) differences between the way v 2 and v3 make estimates, and c) differences in the way the information was entered at our end. Second (in blue), several of those multiples are larger for parts of the state (the central region and for Pulaski county) than they are for the whole state. This seems counter-intuitive and it goes away when the "econometric RPC" method is used in constructing the regions. I'd like to understand this better. In the recent analyses, the data were prepared and entered the same way for all runs. Estimates of income are derived and then reduced by estimated income taxes, remittances, savings, and interest payments. For the runs producing the second and third sets of numbers, the buying power was entered as Institutional Spending Patterns/Households 15-25k. I've tried entering data with somewhat different adjustments as Household and Labor income for the state as a whole. The impacts are somewhat different but comparable/consistent with the results I've sent. I haven't been able to fully reconstruct what was done with the 2004 data. I have some of the output files but don't see clues as to the choices of types of multiplier used at the time. I have tried experimenting with manipulating the user preferences but don't quite see how the options in v3 map onto the options in v2. When I add state and local government to the multiplier, there is only a small increase in the impact. Moreover, it is likely that default options were used in the earlier analysis. What could explain the decrease in measured impact per spent dollar? What could explain the higher multiples in several of the sub-parts of the state? Thank you.
    0
    Comment actions Permalink
  • Avatar
    jenny
    Multipliers can be expected to change every year due to changes in economic conditions (e.g., output per worker, income per worker, import rates, etc.). Our data development methods also change from time to time as we find better data sources and/or estimation processes. The most important thing is to make sure that the impact set-up is the same as previously, such that all the differences are simply due to a changing economy. 1. Use the Econometric RPCs (this paper describes the Trade Flow model: http://implan.com/V4/index.php?option=com_docman&task=doc_download&gid=138&Itemid=7). 2. Is the spending pattern the same (i.e., the proportion spent on each commodity)? Note that the sectoring scheme has changed since 2004 so if you are going simply by commodity code, the spending pattern will be off. 3. Were all retail purchases margined appropriately?
    0
    Comment actions Permalink

Please sign in to leave a comment.