Examining the future of human capital
Intro
In Fall 2021, as part of the well-known class Analytics in Action at Columbia Business School our student team was paired with Citi Ventures Studio and asked to determine the key human capital drivers of financial performance for firms. We analyzed workforce composition, compensation and demographics for companies in the Financial, IT/Tech and Health Care industries. After 14 weeks of trial and error and many hours spent combing through data and testing different explanatory models, we demonstrated that workforce productivity is largely driven by the proportion of software development skills that a company has. We also showed that competition for this type of talent is increasing significantly.
Data
The data used in the analysis came from two sources:
revelio labs: a data set aggregating public employment records on 8,000 companies. Data included the count, attrition rate and salary of all job types, categories, skills and demographics.
CapIQ: a data set containing financial performance metrics for publicly listed companies.
Once we brought the revelio and CapIQ datasets together we would be able to relate workforce characteristics to financial performance metrics and were eager to start doing analysis. As we soon found out, the most important and longest part of any data project is the cleaning! A deeper look at our data set revealed some issues, including repeated companies and many columns of missing data. We agreed to remove the repeated companies and those with over 60% of their data missing. Ultimately, we wound up with data on 2,982 companies, a big change from what we initially started out with.
To preserve as many of the companies as possible we used a method employed by a number of academic papers on human capital which imputed missing values with the mean or median of companies based on their size. We created 3 quantiles of companies based on enterprise value and imputed the missing values with the median of each quantile. We were finally ready to start doing analysis.
Analysis
Our initial hypothesis was that an “ideal” proportion of job type existed. To test this theory, we created “buckets” of companies with different proportions of each job type (engineer, marketing, administrative, operations and finance) and plotted the median revenue/employee for these buckets. If there was an ideal proportion, we would see a “peak” around a certain bucket. For some, there was this peak shape, as pictured below.
However, when looking at the same analysis for Engineer, there didn’t seem to be a “peak”, rather, there seemed to be a positive relationship. We decided to investigate this at a more granular level of detail than just the high-level job type and look at specific employee skills and job categories instead.
After testing numerous regression models which produced insignificant results, we arrived at one with high explanatory power when we regressed all employee job categories and skills (rather than just employee type) against average revenue/employee growth. This implies that when identifying how to expand a workforce to an optimal balance, hiring for specific employee skills or categories may be more important than simply hiring more of a high-level job type. Skills related to programing and software development were among the positive, significant variables in the regression. We proceeded to take a closer look at the positive relationship between these skills and financial performance.
Focus on Software Development/Programming skills
Taking a deeper look at software engineers in the Financial, IT/Tech and Healthcare industries, we saw that on average, firms are clearly changing their workforce composition to add this type of employee. Especially within the Health Care sector, the average proportion of software engineers within a Health Care firm has increased 2.5% along with the average salary increasing over $16,000 since 2017.
Empirically, salary levels also seem to play a part in firm performance. Firms with both positive revenue/employee growth and positive change in proportion of software engineers pay more, on average, than others. Assuming that higher relative salary is correlated to the talent level of the employee, our analysis suggests that “paying-up” for talent is worth it. Interestingly, however, this salary relationship is most distinct in Health Care and does not hold up in Financials.
Why could this be? Through qualitative research, we found that technical career paths in many financial firms require employees to take on more managerial positions to keep advancing rather than developing more advanced technical skills. This could result in a talented software engineer leaving to go to a Healthcare or Tech firm where they can obtain the rank and salary of upper-level management through technical prowess alone. This may raise the ceiling of overall technical talent within firms as well as explain why Financial firms which are adding Software Engineers do not see equivalent revenue growth to similar Health Care and Tech firms.
______________________________________
This project was extremely rewarding for all of us. Shout out to our professors (Daniel and Brett), our TA’s (Elliot and Arunesh) and the team at Citi Ventures Studio (Aboli, Alex and Michael) for giving us the opportunity to work with this data and guiding us through the project. We learned how difficult it is to prepare data for analysis and that obtaining a perfect model isn’t always the end-all-be-all for drawing relevant conclusions from data. We wish Citi Venture Studio the best going forward and are starting to work on our software development skills immediately!
