Data Analytics

Making Data Work for Companies and Organizations

Global Economics Group’s Data Analytics Team helps companies make informed, data-driven decisions in order to improve profitability and reduce litigation risk. Our team includes data scientists, economists and mathematicians, with deep expertise in predictive statistical modeling, machine learning, and other powerful data science methods. We work collaboratively with our clients to help them achieve their specific goals using effective strategies that yield quantifiable results.

Our product offerings include, but are not limited to:

  • Predictive Modeling for Hiring Decisions. Improve hiring decisions and workforce outcomes related to punctuality, absences, disciplinary problems and retention to reduce labor costs. 
  • Disparate Impact Analysis for Machine Learning Models. Proactively identify and reduce inadvertent discrimination and associated litigation risk in machine learning models used in marketing campaigns and financial institutions’ lending decisions. 
  • Analysis of Diversity and Fairness in the Workforce. Proactively identify and reduce inadvertent discrimination and associated litigation risk related to HR decisions, such as hiring, firing, and reductions in force. 
  • Screening Models to Detect Market Manipulation. Detect price manipulation in various commodities markets with sophisticated econometric screening models. 

Predictive Modeling for Hiring Decisions

Our data analytics team helps companies and organizations improve hiring decisions and workforce outcomes. We build statistical models that relate employee performance metrics to pre-hire observables in order to strengthen hiring criteria, assess recruiting strategies, improve retention and reduce labor costs.

Our models predict employee punctuality, absence rates, disciplinary problems and retention rates as a function of pre‐hire observables. The models enable us to evaluate and refine hiring criteria and to quantify the bottom line impact of improved hiring standards. For example, given data on new‐hire onboarding and training costs, projected improvements in retention rates can be expressed in terms of labor cost savings and profitability gains.

Our models also let us project how recruitment methods that influence the number and quality of job applicants affect future labor costs. In addition, by relating employee performance metrics to labor costs, our models let clients quantify the contribution of high-achieving employees to the bottom line. Companies can use this information to intelligently award bonuses and raises.

Disparate Impact Analysis for Machine Learning Models

Many banks and other financial institutions are beginning to implement newer statistical techniques, generally known as machine learning, into their risk and marketing models. These models may determine underwriting outcomes, whether a person is targeted for marketing, credit card rates and offers, among other things. Regulators at both the state and federal level have become interested in potential disparate impact or disparate treatment issues arising from these models.

Our data analytics team has developed various methods to analyze the machine learning models from a potential discrimination standpoint. Because financial institutions often do not collect racial information about applicants, customers, and prospective customers, we use a statistical method based off of Bayes Theorem to determine racial, ethnic and gender probabilities of individuals and groups. We then use these probabilities to perform the statistical analyses that test for disparate impact and other measures of discrimination against protected classes. 

Analysis of Diversity and Fairness in the Workforce

HR decisions related to hiring, firing and reductions in force face similar discrimination risks. Our discrimination detection models can help our clients proactively determine whether they are in compliance with federal and state discrimination laws, such as equal pay for equal work. The early detection and avoidance of such unintentional discrimination can reduce litigation risk and expense. In addition, our models can allow a company to evaluate and balance its workforce in a way that optimizes its fairness and diversity profile.

Screening Models to Detect Market Manipulation

Our data analytics team works with regulators and law firms in the non- and pre-litigation space to develop screening models to detect manipulation in commodities markets, such as gold. The screening models are used to detect anomalies in futures and spot prices after controlling for market and macroeconomic fundamentals not affected by manipulation. In addition to the obvious application in the regulatory and legal environments, a large multinational company has used a variant of the screening models in the commercial space to reduce costs by improving pricing accuracy.  

Our Team

  • Steven J. Davis (bio)
  • Brendan Burke (bio)
  • Candice Rosevear (bio)
  • Rosa Abrantes-Metz (bio)
  • Chad Coffman (bio)
  • Roger Hickey (bio)