The authors test three hypotheses related to enterprise risk management (ERM) implementation: Firms adopting ERM experience a reduction in overall risk, this reduction will be lagged (and/or will become stronger) over time, and ERM adoption is associated with increasing risk-adjusted profits as evidenced by an increase in risk-adjusted return on assets (ROA).

These hypotheses are based on the premise that firms practicing ERM achieve greater cost efficiency of risk reduction compared with firms managing risk through a traditional siloed approach and thus have an economic incentive to further lower risk.

The authors offer strong support for ERM implementation as an effective method of reducing risk, as evidenced by the ongoing reduction in stock volatility after ERM adoption.

These findings are useful to a variety of stakeholders: firm management or board members considering ERM implementation, centralized risk managers justifying their role, and analysts evaluating the impact of ERM on share value.

The hypotheses related to risk reduction are based on the premise that firms implementing ERM are better able to recognize natural hedges and better able to evaluate and prioritize hedging activities and, therefore, optimize the selection of hedging instruments. This premise is intuitively sound because a centralized enterprise-level risk manager—for example, a chief risk officer—benefits from specializing in identifying and managing individual risks as a firm-wide portfolio, whereas understanding risks may be ancillary to the primary responsibility of managers of individual departments.

Regarding the relationship between ERM adoption and higher risk-adjusted returns, as measured by ROA/volatility, it should be noted that non-ERM firms in the sample have a higher mean and median ROA, although it is unclear whether this difference is statistically significant. The ERM firms are only superior when ROA is adjusted for stock volatility.

A sample consisting of 354 public US insurance companies from 1990 to 2008 is identified using the merged CRSP/COMPUSTAT database. Financial data are collected from COMPUSTAT, stock price data from CRSP, and institutional ownership information from Compact Disclosure. The authors determine the date of ERM adoption by searching Factivia, LexisNexis, Thomson Reuters, and Edgar for documents that provide evidence of ERM adoption by the sample insurance companies, yielding 69 unique firms that adopted ERM between 1995 and 2008.

The authors focus on the insurance industry because, compared with other industries, insurers are in the business of understanding and managing risk. Also, limiting the research to one industry controls for heterogeneity in regulatory and industry effects.

The sample is analyzed with a model that uses the two-step Heckman procedure. The authors control for variables deemed to influence risk and the decision to implement ERM—for example, firm size, number of business segments, presence of international operations, firm age, institutional ownership, firm leverage, stock volatility, growth opportunities (as measured by market-to-book ratio), impact of Sarbanes–Oxley, and impact of Standard & Poor’s inclusion of ERM analysis in its credit ratings. The rationale for the control variables is that ERM firms may be systematically different from non-ERM firms and characterized as generally larger, more diversified, more levered, less volatile, and having greater institutional ownership.

The actual—and thus marginal—costs of implementing ERM are difficult to ascertain because of the obvious scarcity of public data detailing these costs. Certainly given the stated complexity of ERM, it must require some investment to establish and carry out an ongoing ERM program. But it is apparent that ERM adoption successfully reduces stock volatility.

Additional research could further examine the relationship of ERM with unadjusted ROA. Does ERM drag on returns as a cost of doing business, or does more-optimal risk management and a lower cost of capital allow for more ambitious risk taking?