1. IMPLAN - Support
  2. Community
  3. General Discussion

Data Center Employment Estimates

Avatar
Morgan Erickson

Hello everybody!

I'm doing an analysis on the operational phase of a potential data center, but I'm having some trouble wrapping my head around some employment numbers that are coming out of it. It is a very large facility, but known direct employment is only 100 employees. My employment multiplier seems huge, it's over 26. My indirect employment is coming out as 1946 and my induced is coming out to 589.

The project uses the following variables.

Using impact analysis detailed

Sector: 418

W&S Employment: 100

Total Employment: 100

W&S Total Labor Income: 9,810,000

Proprietor Income: 0

Intermediate Inputs: 877,079,559.19

The spending pattern was also adjusted to reflect a more accurate budget for this data center. In this scenario, electricity spending was increased to nearly 50% and other categories were normalized.

Does anybody know of any changes that should be made to more accurately reflect employment when modeling data center operations? One particular study modeled a smaller data center using similar spending pattern changes, yet their employment multiplier was a little over 4.

Thank you!

 

Was this post helpful?
0 out of 0 found this helpful

Comments

13 comments

Sort by Date Votes
  • Official comment
    Avatar
    Whitney McKinzie

    Hi Morgan, 

    Thank you for providing all of this information, it was super helpful to dig into this analysis! 

     

    The large Indirect and Induced Employment results that you're seeing are because of the large Intermediate Inputs value being entered. Intermediate Inputs (II)—the non-durable goods, materials, and services an industry purchases from other businesses to operate—are the primary engine behind Indirect Effects and play a vital role in fueling subsequent Induced Effects. Indirect employment represents the business-to-business (B2B) jobs supported in the local supply chain. When a direct industry experiences some sort of shock, it must purchase intermediate inputs to satisfy that production. The portion of those inputs sourced within the region forms the first round of indirect effects. Supplier industries must scale up their production to meet this B2B demand. IMPLAN translates this increased supplier output into Indirect Employment. 


    Induced employment represents the jobs supported when employees (at both the direct and indirect supplier levels) spend their paychecks in the local economy. As the direct industry and its indirect suppliers ramp up production, they pay Labor Income (Employee Compensation and Proprietor Income) to their workforces. When these employees spend their wages locally, they create a surge in consumer demand at household-facing businesses (such as grocery stores, retail shops, or restaurants). To meet this household demand, these consumer-facing businesses must purchase their own Intermediate Inputs. Just like the direct supply chain, the purchasing of these inputs by consumer-oriented businesses triggers further rounds of B2B supply chain transactions. Under IMPLAN's modeling logic, once a dollar of output touches a household column, every subsequent round of purchasing is permanently labeled as Induced. Therefore, the intermediate inputs purchased by consumer-facing businesses—and the subsequent rounds of supplier purchasing those B2B transactions trigger—support Induced Employment rather than indirect employment. 

     

    One thing to keep in mind is that Industries with heavy intermediate input requirements will skew heavily toward supporting indirect employment.

     

    I looked at the underlying data for Industry 418 in St. Louis County, MN. The Intermediate Inputs value for Industry 418 in St. Louis County, MN is only about 3.5% of the Intermediate Inputs value that is in your analysis. 

    I hope this helps!

     

    Whitney McKinzie 

    Education Services Specialist  

  • Avatar
    Whitney McKinzie

    Hi Morgan, 

    Happy to look into this for you! What region are you using for your analysis? 

    Thanks!

    Whitney McKinzie

    0
  • Avatar
    Morgan Erickson

    Hi Whitney! We are using St. Louis County, MN.

    Thank you!

    0
  • Avatar
    Morgan Erickson

    Hi Whitney, thanks for the response!

    We really do appreciate the clarification. We do understand why the large indirect and induced employment numbers are as large as they are, but we are hesitant to continue with these numbers as we are skeptical of such a large multiplier and fear that it just isn't realistic. I guess as a follow-up, do you have any ideas for possible adjustments to the model, or maybe potential studies that have been done before than have encountered a similar issue? 

    0
  • Avatar
    Whitney McKinzie

    Hi Morgan, 

     

    Given that the underlying Intermediate Input data for Industry 418 in St. Louis County, MN is so much smaller than the Intermediate Input data being input into the analysis, I would leave Intermediate Inputs value blank in your analysis and let IMPLAN estimate for it. This will yield smaller values for Indirect and Induced Employment and hopefully provide multipliers more in line with what you were expecting. 

     

    Hope this helps! 

     

    Whitney McKinzie

    Education Services Specialist 

    0
  • Avatar
    mrhaynes

    Hi Whitney,

     

    I'm working with Morgan on this analysis, so I want to add some additional context to help explain the analysis.

     

    We know a handful of things (from public disclosures) about the proposed hyperscale data center here in St. Louis County:

    Employment

    Wages

    Electricity spending

    Water spending

    We also used IMPLAN's article and another academic paper  to assist us with modifying the industry spending pattern to be more in line with what is known about hyperscale data centers. These two articles suggest that electricity spending represents between 40-66% of intermediate inputs, so we adjusted the electricity commodity in the spending pattern to be 50%.

    If we leave the intermediate inputs field blank, the direct output that IMPLAN expects is about $34M. But that value is significantly lower (like, by a factor of ten) than what we know the electricity spending alone to be for the facility.

    The problem I have with the indirect effects is that IMPLAN assumes hundreds of jobs will be added to the electric power generation and distribution sectors, but I am skeptical that the local utility will actually require that level of increase in employment. I am wondering if there is any precedent for modifying the indirect effects? 

     

    Thanks for your assistance!

    Monica

     

    0
  • Avatar
    Whitney McKinzie

    Hi Monica, 

     

    Thank you for the additional clarification! Do you have electricity spend as a standalone figure separate from the $877M Intermediate Input value? Or do you just know the total Intermediate Input spend? 

     

    Best, 

    Whitney McKinzie

    Education Services Specialist

    0
  • Avatar
    mrhaynes

    The ALTERNATIVE URBAN AREAWIDE REVIEW for the project predicts that the "proposed data center is anticipated to consume 2,233,800,000 - 4,467,600,000 kilowatt-hours (kWh) of electricity" annually. Assuming an industrial electricity rate of $0.098/kWh, that equates to $218.9M - $437.8M on electricity costs annually. 

    One of the other reasons that I'm seeing such high indirect and induced effects is because the local electric utility for our region has its headquarters in St. Louis County. Therefore, all of the power generation and other ancillary sectors are being impacted, even though I don't think the power for the data center facility will be generated locally.

    One possible approach that I am considering is to customize the industry details in the model to zero out some of the power generation, knowing that the power will likely be generated in other areas. Similarly, I was considering modifying the output per worker and/or the employee compensation per worker for the electric transmission and distribution sector. Do you think that might be a good approach, and do you have any suggestions for how I might do that in a logical way?

     

    Thanks again!

    0
  • Avatar
    Whitney McKinzie

    Hi Monica, 

     

    Thanks for the additional information! Input-Output (IO) analysis shows the number of jobs supported by a given economic activity. IMPLAN calculates this by using industry data to see how much electricity $877 million in intermediate input spending represents, then applying the industry's average output per worker to figure out how many jobs are needed to produce that power. With that being said, I would not advise you to modify the output per worker and/or the employee compensation per worker for the electric transmission and distribution sector unless you have information explicitly stating otherwise. Additionally, I would also not advise any adjustments to power generation sectors unless you have data stating otherwise. 

     

    Even though this is a hyperscale data center, I am concerned that using the high end of the estimated energy usage range could be overstating effects, especially with the note in the document stating "The range in electricity consumption and associated GHG estimates range includes the amount of consumption provided for in the Electricity Service Agreement with Minnesota Power and provides design flexibility to account for possible fluctuation in usage, price, and energy mix over time". Other research I've found shows hyperscale data centers usually spend between $50-$100M on electricity, depending on local utility rates and load. 

    I think a more defensible option would be to run this analysis with the lower end of the estimated energy usage range, which would effectively cut your Intermediate Inputs value in half, as well as your Employment multipliers. The resulting multipliers would be more in line with what I would expect of a hyperscale data center for this particular region. 

     

    I hope this helps! 

     

    Best, 

    Whitney McKinzie

    Education Services Specialist 

    0
  • Avatar
    mrhaynes

    Thanks, Whitney!

    I really appreciate the additional guidance. I will look for some additional documentation myself, but would you have any sources you could share that would support the $50-$100M in electricity spending that you cited? I agree that the high range of the electric estimate seems overly high, but I want to make sure I'm backing up all of my assumptions with valid sources.

    Again, much appreciated,
    Monica

    0
  • Avatar
    Whitney McKinzie

    Hi Monica, 

     

    Completely understandable! I didn't dive too much to find published journal articles, but I'll provide a link for a resource I found. 

    https://www.texaselectricbroker.com/blog/data-center-energy-consumption

     

    Based on some quick research, it's looking like due to the nature of electricity transmission and distribution being localized, it's hard to pinpoint an exact figure. My personal preference when we're talking about economic impact studies is to err on the side of caution, and use conservative estimates as to not overstate impacts. 

     

    Running impacts on emerging technologies, especially with this new influx of data centers, can be tricky. Please do not hesitate to reach out should you have any other questions! 

     

    Best, 

    Whitney McKinzie 

    Education Services Specialist 

     

    0
  • Avatar
    mrhaynes

    Hi Whitney,

    I'm feeling more comfortable with the approach now. I am using the low end of the electricity estimate and I created three scenarios, each assuming a different share of IE goes to electricity: 40%, 50%, and 60%. The employment multipliers now range from 11-18. Still much higher than what I am used to, but more reasonable than the 26 estimate I was getting previously.

     

    I do have one additional question, however. You said you would not recommend making any adjustments to power generation sectors unless you have data stating otherwise. I am noticing that electric power generation from fossil fuels is one of the largest sectors experiencing indirect effects in my results. But public disclosures for the project indicate that the additional power supply that would be added to accommodate the increased demand from the data center would be entirely carbon-free: mostly from wind and solar. So, I wonder if I have justification to zero out that particular sector (fossil fuel generation) in a customized study area? Otherwise, it will continue to dominate the impacts and I know it is not anticipated to increase.

    0
  • Avatar
    Whitney McKinzie

    Hi Monica, 

     

    I think creating three scenarios with the lower end electricity estimate is a great approach! It's going to yield higher than normal employment multipliers, but with the size of this data center and the region in particular, I think this is a more appropriate estimate than what you were getting before. 

     

    If you know that the electricity generated for this data center is not going to fall under fossil fuels, then it would be okay to zero out that component. I would recommend running a second version of the project where you do zero out electricity generated from fossil fuels and compare your results between the two. 

     

    Best, 

    Whitney McKinzie

    Education Services Specialist 

    0

Please sign in to leave a comment.