Wednesday, June 4, 2014

The AQM, Part 1: What is it looking at?

We would like to take a deeper look at the U.S. Securities and Exchange Comission’s Accounting Quality Model, informally known as RoboCop. We will do this in two parts, assessing what the AQM is using for its analysis, and then we will offer some thoughts on what it means for your company's filing.

There has been a lot of talk in the past year about the Accounting Quality Model (AQM)—nicknamed RoboCop—the U.S. Securities and Exchange Commission’s (SEC) tool designed to evaluate and analyze eXtensible Business Reporting Language (XBRL) filings from public filers. More specifically, the AQM has been designed to flag questionable or fraudulent filings based upon SEC-designed algorithms, including an evaluation of a company’s filing against its industry peer group. We thought it might be useful to dig a bit deeper into what the AQM actually is, along with its potential impact on the filing landscape.

A goal of the AQM and models like it is to find companies whose accounting practices show active “earnings management” by comparing filings from firms in similar industries. For example, these models hope to pinpoint discrepancies in the way a company chooses to account for accruals, which many regard as the most direct way that a firm might try to hide information from its shareholders. You can read more on this in the Forbes article by John Carney and Francesca Harker from BakerHostetler.

A firm’s total accruals are split into two sub-sections, non-discretionary and discretionary. Non-discretionary accruals are of less importance to the SEC, as they are recognized more or less automatically. On the other hand, the SEC hopes to pin down a firm’s discretionary accruals, which may be tampered with to potentially give misleading earnings numbers.

Typically, academics have gone about estimating discretionary accruals by building a regression of total accruals against factors that proxy for non-discretionary accruals. It is often easier to find factors that are drivers of non-discretionary accruals, and for this reason this method makes sense. Two common examples of factors that drive discretionary accruals are changes in revenues and the level of property, plant and equipment. (See Jones and modified Jones models).

Mathematically this is fairly simple; it is assumed that total accruals are equal to non-discretionary accruals plus error. Intuitively, this error should be the same as discretionary accruals if the model assumptions are correct. Firms with higher discretionary accruals than predicted by the model could then be flagged for review.

Unfortunately, this modeling has been likely to lead to a relatively large number of false positives (see article also for a more thorough explanation of the above modeling for the mathematically inclined). The SEC’s goal has been to take this modeling a few steps further to weed out the “noise” from the error term. They are now including factors that proxy for discretionary accruals into their regression.

More precisely, they have divided discretionary accrual factors into two camps, those that indicate earnings management and those that induce earnings management. Factors in the second camp hope to help investigators find market environments in which earnings management would be more plausible.

While we may not have access to the factors the SEC uses to estimate non-discretionary and discretionary accrual practices, assuring that your XBRL filing contains the proper balance sheet concepts and accounting policies for a firm in your industry is important. If you are relying on a third-party for your XBRL, an emphasis should be placed on partnering with experienced companies with individuals knowledgeable in XBRL best practices.

This concludes part 1 of this two-part series. Check back soon for the next part, where we will take a look at what all of this means to public filers.