Variance in Scenario Results following JTW updates

Comments

14 comments

  • Avatar
    IMPLAN Support
    Hello John, Thank you for verifying all those different factors. We were a little confused as to how you had gotten the original set of results with the overflow error. Was the original result set from a single region model? Likely the variances in the JTW is the cause of these differences since the Trade Flows would affect the Indirect as well as Induced (by changing the [url=http://implan.com/v4/index.php?option=com_glossary&id=237]RPC[/url]). The new JTW files must reflect a higher level of in-commuters (i.e. people who live outside the region, but who work in the model's region). You should be able to see this in the Employee Compensation column and row 28001 Domestic Trade in the Social Accounts>IxC SAM tab. Let us know if you have any further questions or concerns.
  • Avatar
    CalNeva
    Thanks. I am getting this discrepancy in single region models. Since I wrote to you initially the .ODT files were also updated, but the same discrepancy between the models persists. The JTW files do not seem to be the source of the difference. The models look identical across 1) Commodity Trade 2) Institutional Commodity Demand 3) HH Local Commodity Demand and 4)Govt Local Commodity Demand - I've tried to attached copies of the Commodity Trade details by way of illustration. The two models' SAMs also appear identical (aggregated spreadsheets from the two models are attached). Similarly the I/O accounts seem to be identical (also attached). The activities are identical (copies from both models attached), yet the impacts diverge considerably between the models (results from both models are attached as well). I would appreciate your clarify where the differences between the analyses comes from. We will be replicating the analysis and would like to ensure the results are dependable. Note: Model 1 is the original and Model 2 was created after the updates.
  • Avatar
    CalNeva
    I don't seem to be able to add the attachments - let me know if you would like to review them and how best to upload.
  • Avatar
    IMPLAN Support
    Hello John, Please email the models to implangroup@implan.com and write "#15922" in the subject line to reference this posting. Thanks!
  • Avatar
    CalNeva
    Thanks - I have emailed the models and the attachments.
  • Avatar
    CalNeva
    I was looking through the detailed impact results and noticed that in the earlier model (Model 1) Implan Sector 361 - [i]Imputed rental value for owner-occupied dwellings[/i] - had a direct impact value of zero in terms of its Value Added, Other Property Type Income effects, and Indirect Business Taxes. However, in the model I created after the update (Model 2) direct effects for sector 361 are listed as $711m, $611m and $100m. Was there/should there be a change in the modelling of Sector 361? Given that this is in the range of the difference between the two model's impacts I wanted to check. In our activities we have $1.1b in Sector 361's event value. We are using this sector to reflect expenditures by home owners for their dwelling ("Avg annual expenditures on Owned Dwellings from the BLS Consumer Expenditure Survey). We also allocated $375m to Sector 360 - [i]Real Estate[/i] - for expenditures on rent ("Avg annual expenditures on Renter Dwellings from the CES). Thanks.
  • Avatar
    IMPLAN Support
    Hello John, Thank you for sending in your models and spreadsheets, as this helped us determine where the variances are occurring. It appears that the primary variance is in your calculation of Employee Compensation and Proprietor Income are drastically different between the two models for your State activity. But the variances do not appear to be in the underlying model values. so for example, both of these Events were created directly in a model, the top one from the Old model, the bottom from the New model. Sector Industry Sales Employment Employee Compensation Proprietor Income Event Year Output Deflator GDP Deflator Local Purchase Percentage 320 Retail Stores - Motor vehicle and parts $255,046,336.00 2,537 $127,108,751.42 $20,189,398.19 2013 1.032 1.034 100.00% Sector Industry Sales Employment Employee Compensation Proprietor Income Event Year Output Deflator GDP Deflator Local Purchase Percentage 320 Retail Stores - Motor vehicle and parts $255,046,336.00 2,537 $127,108,751.42 $20,189,398.19 2013 1.032 1.034 100.00% Employment was used a proxy (as this produced the closest values to your old results) and you can see both Events generated identically. However, in the spreadsheets provided, the larger variance between Employee Compensation and Proprietor Income in the Models for sectors other than Retail was $11.81. For Retail sectors, however, the following variances were observed: Employee Compensation | Proprietor Income $99,239,896.65 $15,762,831.37 $7,238,918.94 $1,089,907.67 $28,234,028.94 $1,825,763.32 $12,546,117.66 $1,134,070.22 $194,319,085.99 $37,824,057.20 $44,112,335.63 $5,670,872.17 $40,387,067.53 $78,461,048.09 $14,283,901.90 $2,750,082.68 $32,745,586.90 $7,060,578.38 $145,795,596.81 $2,507,655.22 $5,694,266.56 $2,753,853.05 $10,130,621.55 $4,014,179.63 This seems to indicate an issue with margining. Did you use the same method for creating both sets of Events? Are you importing from these spreadsheets? If not, what process was used to generate these spending patterns in the Model? Can you also tell us what version of IMPLAN you are running found in the Help>About menu? Thanks!
  • Avatar
    CalNeva
    We are importing these activities from spreadsheets, but I copied the activities directly from the old model into the new model and verified that both were applying margining on the retail identically. We are using IMPLAN Version 3.0.17.2 Any idea why the value added effects for the Sector 361 (imputed rental value) would be so different between the two models? Thanks.
  • Avatar
    IMPLAN Support
    Hello John, We apologize this is taking so long. We are looking into your models. Would you be able to tell us what you are interested in when separating 361 out from Households, as we think you may be double counting.
  • Avatar
    CalNeva
    We are modelling the impact of retiree expenditures using the CES for persons age 65 and over.I have just sent you an email that maps out how we've allocated the CES expenditures to Implan. We placed the expenditures on "owned dwellings" in 361 and allocated the HH activity from cash payments made to other households. Thanks.
  • Avatar
    IMPLAN Support
    Thank you for that information. As regard 360 and 361, Model 2 has created an anomaly result. Unfortunately, we aren't sure why this has occurred because when reanalyzed a second time in your Model without making any modifications, the Direct Value Added goes away, which yields the expected Direct results (i.e. employment=0 Labor Income=0 Value Added =0 Output= industry sales) thus we think reanalyzing the scenario in your Model 2 will solve the discrepancy. Just to clarify, the Household Spending Pattern that you are impacting already contain 361 and many of the same expenditure that you have classified in the second tab of your spreadsheet. It appears that you are using this Household Income Change Activity to describe the spending of the transfer payment. Its that correct? Once thing to note if this is the case, is that you are assuming these transfer payment are being spent like all other form of standard household income. As regards your spending patterns experiments, it appears you are essentially creating your own specialized household spending pattern. So as long as you are restricting the households income change activities to the household transfers, you shouldn't have any double counting issues with this Activity. Please let us know if this does not address your question or if you have additional concerns.
  • Avatar
    CalNeva
    Thanks. I reanalyzed the model without any further modification, but the Direct Value Added is still there in the results. I then generated the model again and imported the activities and that analysis and re-analysis continues to generate Direct Value Added of $711m. We have the data running on another machine which is using Implan V3.1 but we also get the same $711m figure for Direct Value Added on that system. Any further suggestion to resolve this discrepancy? Thanks
  • Avatar
    IMPLAN Support
    Hello John, We apologize, the software technicians informed us to re-run it, but we also saw that it did not resolve the issue. We think there is an issue with the import, or with the original spending pattern being imported. We found that removing Event 361, and then re-adding Event 361 with the same in industry sales, that Direct VA for that sector disappears as it should. It will give you a notification stating that there is no employment, you can select OK and enter in the dollar value. We aren't sure what the issue is for that sector but this should resolve it. Let us know how it works. Thanks!
  • Avatar
    CalNeva
    Pulling 361 out as a separate activity and then running the analysis seems to have done the trick. Thanks.

Please sign in to leave a comment.