Table 1 displays earnings estimates correlation statistics for Quarter 1, Fiscal Year 1, and Fiscal Year2 from IBES and FirstCall for the S&P500 and Full Universe.\u00a0 The values highlight a strong relationship between these two sources with correlations ranging from .97 to .99.\u00a0 The relationship, at the high-end of this range, is consistent to the level of being virtually identical.\u00a0\u00a0 Moreover, both datasets cover on average more than 3000 securities.<\/p>\n
Although our examination is on a two and half-year period ending in June 2004, the same testsconducted on longer periods going back to 1998 revealed no material difference between the two time periods.\u00a0 The short period is selected to facilitate optimal comparison of IBES and FirstCall data as well as model results based on business calendar month end[2]<\/a>.<\/p>\n Having established the similarity of the IBES and FirstCall earnings data, we next explored the relationship of SEF implementation on these sources.\u00a0 Figure 1 illustrates SEF Q1\u2019s correlation is high. The average correlation of 99% indicates the two models\u2019 ability to predict future earnings is consistent. Given the model\u2019s identical construction on the two databases, the graphed values are evidence of near perfect association between the earnings data used as input. There\u2019s no material difference in the Fiscal Year1 and Fiscal Year2 relationship over the same period.<\/p>\n The SEF performance statistics in Table 2 is based on IBES and FirstCall data for two broadly followed universes \u2013 S&P500 and the S&P Small-Cap universe.\u00a0 While there are some minor differences between the two predictive earnings models\u2019 performance on the indicated data sets, the similarities dominate.\u00a0 There\u2019s very low deviation in the number of securities that are covered in both universes.\u00a0Moreover, the annualized quintile spreads for the three forecasted (Q1, FY1, FY2) measures averages 4.07 percent on IBES versus 3.69 percent on Firstcall for the S&P500 universe.\u00a0 For the S&P Small-Cap, the model posted 11.99 percent as compared to 12.36 on the respective databases. The average difference is less than 40 basis points.\u00a0 The risk profile, as characterized by the information ratio (IR), of the two model implementations paints a similar picture.\u00a0 The alpha per unit of risk of each forecasted item indicates no material difference on the two databases.\u00a0 For example, the S&P500 universe\u2019s IR for SEF Q1 is 1.19 on IBES versus 1.14 on Firstcall.\u00a0 While SEFY2 posted a .50 difference on the S&P Small-cap universe (1.38 vs. 1.88), the other measures generated IRs with significantly less distinctions.<\/p>\n The foregoing analysis provides convincing evidence that there is no material performance difference for implementing a predictive earnings model on the IBES QFS or Firstcall databases[3]<\/a>.\u00a0 As such, QRG is offering its suite of Smart Earnings Forecast products on the IBES QFS database.\u00a0 This implementation provides clients a number of benefits. \u00a0First, with its nearly perfect correlation, the SEF implementation on IBES maintains continuity for clients who are accustomed to using the Firstcall version we implemented during our tenure at DAIS Group.\u00a0 Second, the deeper IBES data history enables us to back fill the model scores to include several more years.\u00a0 This expansion guarantees more coverage of the Markets\u2019 historical variations. Lastly, IBES\u2019 standard production process and consistent availability enables us to deliver our SEF product in a much more timely and predictable fashion.\u00a0 This provides clients a wider pre-market-open processing window for incorporating our SEF rankings into their internal investment process.<\/p>\n In summary, the resulting performance of QRG\u2019s SEF on two distinct detail earnings database products indicates there\u2019s no content advantage in using one versus the other.\u00a0 This should come as no surprise to most practitioners.\u00a0 Competitive, technological, and regulatory forces have promoted efficiency in the earnings collection and distribution landscape over the past decade.\u00a0\u00a0 Therefore, clients have a high probability of being equally well served by most earnings database choices.\u00a0 The foregoing examination suggests that the earnings content field is fairly leveled.\u00a0 Consequently, quality issues relating to delivery, production, and customer service are bound to be the driving forces of differential advantage.<\/p>\n Smart Earnings Forecast on IBES Earnings Data We have expanded our suite of predictive earnings model to use Thomson\u2019s IBES earnings data products.\u00a0 Dating back to 1998 at the DAIS Group, we have been developing predictive earnings models using FirstCall\u2019s customized earnings databases.\u00a0 Nevertheless, recent vendor consolidation coupled with technology advances has eliminated most of<\/p>\n![\"Table1\"](\"http:\/\/s1094408.instanturl.net\/quantitativegroup.com\/wp-content\/uploads\/2016\/07\/Table1-1.png\")
<\/a><\/p>\n![\"Figure1\"](\"http:\/\/s1094408.instanturl.net\/quantitativegroup.com\/wp-content\/uploads\/2016\/07\/Figure1-1.png\")
<\/a><\/p>\n![\"Table2\"](\"http:\/\/s1094408.instanturl.net\/quantitativegroup.com\/wp-content\/uploads\/2016\/07\/Table2.png\")
<\/a><\/p>\n[1]<\/a>\u00a0 SEF uses Thompson\u2019s raw data feeds consisting of Consensus, Reported Earnings, Individual Analyst Forecasts and Pre-announcements to: first generate a conditioned database; Second, build a series of component models that captures systematic characteristic of earnings and analysts from said database, including updated forecast, frequency, recency, track records; Third, optimally combine these sub-models to generated forecasts that are more accurate then the Consensus and also provide improved stock selection.<\/h6>\n
[2]<\/a>\u00a0 According to Thompson Financial, monthly production cycle for IBES is \u201cThe Thursday before the third Friday of the month\u201d while the FirstCall data is based on the US Markets\u2019 business month end.<\/h6>\n
[3]<\/a>\u00a0 While QRG\u2019s SEF utilizes the detail earnings data from these sources as input, as indicated at the beginning of the paper, the data input to the SEF algorithm is much more enhanced than these raw feeds.\u00a0 Generally, QRG applies a series of quality control procedures to correct for a variety of errors and omissions to make sure the data reflects the \u201cas was\u201d market conditions while minimizing non-apparent biases and misalignments.<\/h6>\n","protected":false},"excerpt":{"rendered":"