Artificial Intelligence Community

I'm currently working on a new initiative to build a research expenditure strategic projection model that would allow us to feed in our existing data and generate more precise projections to support leadership decision-making. That said, I've found this to be especially challenging because there are so many variables that can influence spending. In many ways, it feels less like a purely technical or financial problem and more like a study of human behavior. Factors such as PI spending habits, award type and purpose, sponsor requirements, and even unexpected operational issues can significantly impact expenditures.

When I tried to find existing research specifically focused on projecting higher education research expenditures, I didn't come across much that felt directly applicable. I originally hoped to explore correlations, but I quickly realized how difficult it is to isolate clear relationships when so many variables interact simultaneously.

For my first attempt, I took a fairly straightforward approach. I used historical expenditure data from the past three fiscal years and combined it with actual expenditure data from July through November of the current fiscal year to project spending for the remainder of the FY. While this gives a rough estimate with seasonal forecasting, I'm not entirely confident it captures real-world behavior, mainly when unexpected changes occur mid-year.

I'm curious whether anyone else has worked on a similar effort. If so, what approach did you take? Where did you start, and what types of data did you find most useful? Did you rely more on historical expenditure trends, award-level details, staffing data, or anything else? I'd love to hear how you've navigated the complexity of projecting research expenditures. 

Thank you so much for sharing!

I'm currently working on a new initiative to build a research expenditure strategic projection model that would allow us to feed in our existing data and generate more precise projections to support leadership decision-making. That said, I've found this to be especially challenging because there are so many variables that can influence spending. In many ways, it feels less like a purely technical or financial problem and more like a study of human behavior. Factors such as PI spending habits, award type and purpose, sponsor requirements, and even unexpected operational issues can significantly impact expenditures.

When I tried to find existing research specifically focused on projecting higher education research expenditures, I didn't come across much that felt directly applicable. I originally hoped to explore correlations, but I quickly realized how difficult it is to isolate clear relationships when so many variables interact simultaneously.

For my first attempt, I took a fairly straightforward approach. I used historical expenditure data from the past three fiscal years and combined it with actual expenditure data from July through November of the current fiscal year to project spending for the remainder of the FY. While this gives a rough estimate with seasonal forecasting, I'm not entirely confident it captures real-world behavior, mainly when unexpected changes occur mid-year.

I'm curious whether anyone else has worked on a similar effort. If so, what approach did you take? Where did you start, and what types of data did you find most useful? Did you rely more on historical expenditure trends, award-level details, staffing data, or anything else? I'd love to hear how you've navigated the complexity of projecting research expenditures. 

Thank you so much for sharing!

Steven -- I have been asked this same or similar question many times when presenting my work in a variety of community groups. Surprisingly--it is not a question I have to answer at my own institution. We must have created a culture that focuses on discrete data-informed decisions rather than projections. I have not found a method or as you mention, any research, that supports a statistical model that works for projecting R&D expenditures within institutions that are federally funded. As you mention, there are so many variables that cannot be predicted and the time and resources to account from them all seem to out weigh the benefits. Instead, we look at three-year rolling averages of past actually expenditures to inform a potential one-year ahead. This is used to set budgets. Of course leadership is conservative and always ends up having more revenue than their budget...which is good. 

This may be a good topic for someone to submit a grant proposal to either NCURA research program or NSF GRANTED to study. I'd be happy to collaborate. 

I'm currently working on a new initiative to build a research expenditure strategic projection model that would allow us to feed in our existing data and generate more precise projections to support leadership decision-making. That said, I've found this to be especially challenging because there are so many variables that can influence spending. In many ways, it feels less like a purely technical or financial problem and more like a study of human behavior. Factors such as PI spending habits, award type and purpose, sponsor requirements, and even unexpected operational issues can significantly impact expenditures.

When I tried to find existing research specifically focused on projecting higher education research expenditures, I didn't come across much that felt directly applicable. I originally hoped to explore correlations, but I quickly realized how difficult it is to isolate clear relationships when so many variables interact simultaneously.

For my first attempt, I took a fairly straightforward approach. I used historical expenditure data from the past three fiscal years and combined it with actual expenditure data from July through November of the current fiscal year to project spending for the remainder of the FY. While this gives a rough estimate with seasonal forecasting, I'm not entirely confident it captures real-world behavior, mainly when unexpected changes occur mid-year.

I'm curious whether anyone else has worked on a similar effort. If so, what approach did you take? Where did you start, and what types of data did you find most useful? Did you rely more on historical expenditure trends, award-level details, staffing data, or anything else? I'd love to hear how you've navigated the complexity of projecting research expenditures. 

Thank you so much for sharing!