Study Shows Axiomatics Data Predicts Company Stock Returns


In a recent study published by Alan Kwan, PhD from the University of Hong Kong, company growth estimates based on employee benefits data from vendor Axiomatics Data outperformed the market during the Covid 19 pandemic.

Findings of the Study

In his recent research note called, Predicting Company Performance with Company Benefit Plans, Alan Kwan, PhD analyzed the value of using Axiomatic Data’s Form 5500 dataset for predicting company stock returns from January 2014 through September 2020. 

Form 5500 is collected by the US Department of Labor as part of the disclosure framework set by the Employee Retirement Income Security Act (ERISA) of 1974. This dataset includes, among other things, information on the different employee benefits offered by the firm, such as dental, medical, 401K, etc., the number of employees receiving benefits, and the funding status, total assets and employer and employee contributions of any defined contribution or defined benefit plan sponsored by the company. The vast majority of firms in the United States are required to file and submit the full-length Form 5500, and it reflects the largest store of public information on company’s employment practices in the United States.

Kwan conducted a back-test examining two signals based on Axiomatic’s dataset. In his baseline test, Kwan used Axiomatic Data’s proprietary ThriveScore metric, which is a score of a company’s growth in the benefits provided by the firm. The ThriveScore combines eighteen different growth rates (growth rate of employees, growth rate of employees in pension plans, employer contributions per employee, participant contributions per employee, growth rate of employer contributions in total, growth rate of participant contributions in total) using Axiomatic Data’s proprietary formula. The score can be interpreted as a growth rate of development in human capital at the firm.  Kwan also combined the ThriveScore with an amalgam of other data points from Form 5500, which he called the Composite Score.

First, Kwan constructed a portfolio sort splitting stocks each month into quintiles of scores. One would expect that on average, the stocks in quintile 1 underperform stocks in quintile 2, quintile 2 stocks underperform those in quintile 3, and so on. Each data point in the figure below is the average across my sample period of stocks in a given portfolio in a given month.

Second, Kwan created a hypothetical long-short strategy.  Here firms with positive scores over time (both the Composite Strategy and the ThriveScores) tend to outperform, especially in the prior year. One interpretation of this finding is that during the pandemic, firms who treated their employees better were equipped to function more effectively during the pandemic.

Kwan also examined the extent of whether these stock returns are attributable to market performance. He selected two benchmarks: the simple market portfolio and the Fama and French five factor portfolio. The five factors represent the five factors which best summarize asset pricing factors from prior research.  All strategies which were based on Axiomatic data ‘beat the market’. Applying a tougher benchmark, the value-weighted composite signal Kwan constructed beat the five-factor benchmark.

The findings of Kwan’s study suggests that companies whose benefit plans are generally growing tend to perform better than others, and the premium associated with these firms is not explainable by market risk.  Rather, Kwan believes these returns are related to the buildup of intangible capital at the firm, as increases in benefit plans enter into SG&A. This finding would be consistent with recent research that suggests firms that build intangible capital tend to outperform in the long run.

Our Take

Kwan’s study revealed that Axiomatic Data’s Form 5500 employee benefits data has predictive value when trying to forecast a company’s stock price.  What is also interesting about this dataset is that it could be seen to be a reasonable proxy for the “S” in ESG which addresses the interests of employees, customers, and other social stakeholders.

Axiomatic Data was founded in 2019 by financial services executive and serial entrepreneur Steve Goldstein.  The New York City-based startup aggregates pension plan level Form 5500 filings and creates a consistent, high quality, point-in-time database of company level benefits information. Form 5500 Filings are a disclosure tool used to satisfy annual reporting requirements by public and private companies in the U.S. for employee benefit plans under ERISA and the Internal Revenue Code. These employee benefit plans cover both pension and welfare plan benefits.

Axiomatic Data provides 7 years of monthly Form 5500 data (from January 2013 to April 2021) for all Russell 3000 companies, and for a total universe (both public and private companies) of 700,000 US companies.  In addition, Axiomatic provides their proprietary Thrivescore for all companies, based on employee growth and changes in benefits.

While the Axiomatic Data dataset should be an interesting input for both quantitative and fundamental asset managers looking for a unique measure of corporate value, it could also be interesting to the growing number of asset managers interested in looking at ESG factors and the impact these measures could have on long-term company performance.

For a complete copy of Alan Kwan’s study click here.  For more information about Axiomatic Data please contact Steve Goldstein at


About Author

Mike Mayhew is one of the leading experts on the investment research industry. In addition to founding Integrity Research, Mike is on the board of directors of Investorside Research Association, the non-profit trade association for the independent research industry, and a frequent speaker on research industry trends and developments. Mike has over thirty years of research industry experience. Email:

Leave A Reply