Has Goodwill Accounting Gone Bad?
- Kevin Li and Richard Sloan
- A version of the paper can be found here.
- Live implementation data can be found at Empirical Finance Data™ (Expected Q2 2011)
“SFAS 142 replaces the systematic amortization of goodwill with a periodic impairment test. We examine the impact of this standard on the accounting for and valuation of goodwill. Our results indicate that the new accounting results in inflated goodwill balances and untimely impairments. Goodwill impairments lag deteriorating operating performance and stock returns by at least two years. Moreover, investors do not appear to fully anticipate the delayed goodwill impairments, since firms with deteriorating operating performance and large goodwill balances have both predictable future impairments and negative abnormal stock returns. Overall, our results suggest that management exploits the discretion afforded by SFAS 142 to delay goodwill impairments and that this causes earnings and stock prices to be temporarily inflated.”
The authors use Compustat Xpressfeed to obtain accounting data including annual goodwill data and quarterly goodwill impairment. Return data are obtained from CRSP. The time period analyzed for the trading strategy is from 2003-2009.
CEOs love to play games with other people’s money.
Standardized accounting rules are theoretically supposed to increase transparency and help investors identify managers who are engaging in nefarious activity. Unfortunately, the system isn’t perfect.
In this paper, the authors look at the accounting of goodwill to determine if the market can identify when managers are using the accounting rules to manipulate their earnings.
There are really three ways in which a manager can account for goodwill:
- Expense the cost in the current period–this will kill earnings in the short-term, and overstate earnings in the future.
- Amortize the cost of the intangible asset over the asset’s anticipated lifetime–this method assumes each time period gets equal benefits from the asset.
- Periodically test the asset’s value and impair accordingly–if the asset is no longer valuable, write it off, if the asset has not been impaired, don’t take a charge.
The first the thing the authors examine is how amortization expense reacts to the introduction of SFAS 142 in 2001. Not surprisingly, the figures shows that managers immediately move to lower amortization expense when given the opportunity. One could argue that every manager simply felt their goodwill asset was worth more tomorrow than today, but that seems far-fetched.
Next, the authors look at the post SFAS 142 period and see how often firms engage in goodwill “write-off baths.” Under SFAS 142, one would expect goodwill write-offs to come gradually because goodwill assets typically do not die off in a single year–it takes a while for a brand or intellectual edge to be impaired. Of course, if managers decided to not take amortization expense over the previous years to increase their pay, and instead take a “bath” all at once when goodwill game could no longer be played, we’d see a situation where managers rarely take goodwill impairments, and then suddenly write-off goodwill in large chunks. This is exactly what the authors find–46% (28+3+4+5+6) of the time goodwill impairments are 50% or greater!
So how can we make some money?
The authors first estimate a model that can be used to predict whether or not goodwill will be impaired. They find four variables (based on their previous tests) that are associated with goodwill impairment. The variables are total goodwill scaled by total assets (GTA), return on assets (ROA), change in return on assets (Change ROA), and a dummy variable equal to 1 if the firm is in the highest quintile of GTA and lowest quintile of ROA (lots of goodwill, no profitability).
The authors then put together a relatively complicated strategy called IPROB, which uses the estimated coefficients from the IMPDUM model above to predict future goodwill impairments.
Let’s keep it simple, stupid. Shoot. Move. Communicate.
Instead of calibrating model coefficients and trying to create an IPROB, how about we focus on two simple measures–GTA and ROA. The authors do just that and show us the trading strategy performance. Table 5 verifies that the combo of GTA and ROA actually predict future impairment. For example, ROA/GTA(1,10) shows a 28.8% probability of impairment.
Now on to the returns:
Looks like the combination of low profitability and high goodwill lead to crappy returns via the future goodwill impairment channel: -15.6% to be exact! Managers simply do not want to release bad news until they absolutely must!
- Identify firms that are in the highest GTA and lowest ROA deciles.
- Short these firms, because there is likely going to be a goodwill impairment announcement that will be unanticipated by the market.
- Hedge with a size-adjusted benchmark (e.g., Russell 2000 etf) of long-side alpha strategy.
- Make money.
Having read 3 or 4 perturbations of this paper, I can tell you that the results have shifted around quite a bit–originally, the High GTA, low ROA, hit a -23% mark, and now it is banging out -15.6%. So one definitely has to think about robustness when interpreting these results. That said, -23%, or -15.6% are still profound numbers that cannot be ignored. We’ll be churning the data on this strategy and posting it to our data site in the near future–stay tuned.
No related posts.