2010 PA Induced Impacts seem too large

Hi Tim, If you go to Customize > Study Area Data and compare a few of your directly-affected sectors between 2009 and 2010, you’ll notice that the Output per Worker has gone down for some of them, meaning that more employees are needed for the same amount of Output. If there are more employees, then there is more Labor Income to be spent – hence, higher induced effects. In addition to lower Output per Worker values, some of these sectors have higher Labor Income per Worker values – so workers are earning more – again leading to higher induced effects. Let us know if you have any further questions. Thanks

If more employees are required per unit of output then the direct and indirect employment impacts should be higher and they are not. While compensation per employee is higher the increases do not justify a doubling in the induced effects in one year. I spoke with Doug Olsen a few months ago about running the national model then the three states individually (PA, OH, and WVA) and then backing out the rest of the USA impacts. When you do this the high PA induced effects imply a negative induced impact for the rest of the nation, which makes no sense. See the attached spreadsheet. Tim C.

Are you running these as an aggregated model or just aggregating the results? Can you zip and email your models to info@implan.com?

I'm running the models separately and have thought about aggregating them. I would like to consider Pennsylvania, West Virginia, and Ohio as one region and then the rest of the U.S. as another. I would like to get impacts for these three states, th region, and the rest of the USA. Any advice would be most appreciated. On a separate but related note, I had a good exchange with one of your colleagues and he pointed me to the data, specifically output per employee and compensation per employee. His insight is correct, output per employee in Pennsylvania declined in 2010 compared with 2009 and compensation per employee increased, which would explain the increase in the Type II multipliers. But the resulting increase in the induced impacts, especially compared with those estimated using the 2009 data seems implausibly large. I'm wondering if there are some alternatives and would appreciate any suggestions. I will send you the models shortly.

I took another look at the Pennsylvania data. From my reading of the "study area data" aggregate gross output declined. In contrast, both the BEA and the IMPLAN data suggest gross domestic product and valued added increased during 2010 over 2009 levels. So I'm wondering if the decline in gross output is an artifact of the price deflators used to estimate real gross output.

Hi Tim, Thanks for sending the models. It appears that the 2010 PA model is built with the old Econometric RPC methodology instead of with the Trade Flow Model, and based on models we've built in house, we think this is where the problem lies. The 2009 model appears to be in Trade Method (as it matches our Trade Method results), as do the others. When we run the scenario with PA 2010 trade flows method and we get very different results then what you have, and the results are consistent with the 2009 results. It appears that if you rebuilt the 2010 PA model with the Trade Method this should resolve the issues you are seeing, and this will also allow you to do the comparison you are seeking to the U.S. results. Please let us know if this does not resolve the problem. Thanks

I sent this to your email address but just wanted to make sure you received it: Thank you, that did the trick but raises a question. What is the best way to estimate impacts occurring outside the region? Upon Doug Olsen's suggestion several months ago, I estimated the three individual state impacts and then ran the USA model using total spending equal to the sum of the totals for the region and outside the study area. I then took these impacts and subtracted the total regional impacts to estimate impacts outside the three state region. So as an alternative I ran the USA model using rest of the country spending levels and got economic impacts that seem more reasonable. The best method would be to aggregate the remaining states but I do not have the funds to purchase the entire state data file system right now. Attached is an Excel workbook summarizing the spending levels in the first worksheet and the employment estimates in the second. I would appreciate your thoughts.

Hello Tim, We think in this case, if you have spending outside your states, that the best bet is to run the analysis on the three states and then the rest of US spending on the US model as you suggested. You could do the subtraction of the run of all the spending at the US level and then subtract the state runs but you are kind of mixing single region and pseudo multi-region. I think it would be cleaner to just run the rest of US on the US model.