After writing the post titled “Complexity is the Work of the Devil” across on CNBC came the story “Readability of Annual Reports Affects Accuracy of Analyst Forecasts” or in shorter terms complexity leads to inaccuracy. As Reuven Lehavy along with his Business Ross School colleagues Feng Li and Kenneth Merkley took on the challenge to measure the readability of more than 33,700 observations from the 10-K filings of firms from 1995 to 2006. To make this uniform, the team utilized the (Gunning) Fog Index developed in the computational linguistics literature models; which allowed them to empirically determine the written complexity of 10-K reports by “counting” the number of syllables per word and therefore the number of complex words per sentence.
From the results of this statistical analysis, the researchers made an interesting discovery in that they found more financial analysts followed firms with less readable (complex) 10-K reports. This they determined translated to a greater demand for analyst reports for these firms as it also took these analysts on average, two days longer to issue a first forecast revision following a 10-K filing. This in turn also suggests more effort had to be put forth by them; and provide earnings forecasts that result in proportionally higher firm returns associated with their reports which also suggested that investors found these analysts’ reports more informative. How to summarize all of this then? Well that’s an easy one, as the more complex the 10-k filing the more likely the return right? As if their “that good” then they should be able to take this to the bottom line one could say.
Well here is the kicker, as what Reuven Lehavy and friends soon discovered is that this was not the case at all. In fact what the researchers found their study showed that analyst earnings forecasts for companies with less readable 10-K reports have greater dispersion, thus are less accurate and are associated with greater overall analyst uncertainty. While Lehavy says “Our results are consistent with the prediction of increasing demand for analyst services for firms with less readable communication and a greater collective effort by analysts for firms with less readable disclosures“. It seems he is missing the obvious bug-a-boo which is complexity brings to the table a greater likely hood of error.
As we talked about in “Complexity is the Work of the Devil”, the idea of the greater number of potential values lead the analysts down the golden path of greater error in their forecasts and in fact (they) need to simplify their summary rather than attempt distill the complexities “fully” into their own models as more than likely, the complexity was not a fact of intended obscurity by the filing corporation. However their own failure to come to grips with their own inability to overcome their “complexity issues” thus unknowingly propagating the error throughout the value chain.
One of the interesting things is complexity is a kin to “system noise” and in as such will be subject to the power laws as discussed in prior posts. This means as the “complexity” (noise) moves through the system, it will be subject to logarithmic growth. Thus the complex 10-K filings were created by misunderstood complexities of the business model as a “tell” that the filing business lacked a firm grasp upon its objectives…