eThesis - School of Business electronic theses

PURPOSE OF THE STUDY

The main objective of this thesis is to develop and test a method for selecting comparable firms on the basis of common analysts. The method is based on a Monte Carlo simulation on analyst choices, and it produces firm- and time-specific analyst-defined peer groups for sample firms. The analyst-defined peer groups are based entirely on analysts’ choices, and they are independent of conventional industry classifications.

As part of the thesis, I also use the analyst-defined peer groups to study the effect of peer firms’ dividend initiations and stock splits on the propensity to initiate dividends and execute stock splits.

DATA

The sample of the thesis consists of firms listed in the NYSE and individual sell-side analysts who provide coverage for them. The analyst data is based on estimates in the IBES database, and the sources for company data are Compustat and Datastream databases. The sample covers years 1990 to 2007.

RESULTS

The results of the thesis indicate that, compared to conventional industry classifications, the analyst-defined peer groups are significantly better at explaining cross-sectional variation in valuation multiples and financial statement ratios. The explanatory power of the analyst-defined peer method outperforms the industry classifications on most tested variables also when firms without analyst-defined peers are excluded from the analysis. The method produces peer groups for approximately 50 % of the sample firms, and a limitation compared to conventional industry classifications is that peer firms can only be found for firms with sufficient analyst coverage.

The results on dividend initiations and stock splits indicate that firms tend to initiate dividends and execute stock splits in the same years as their peers. The effects persist also after controlling for firm-level variables explaining dividend initiations and stock splits, which indicates that herding behavior is a possible explanation for the findings.