Strategy, Valuation, and Forecast Accuracy: Evidence from Italian Strategic Plan Disclosures

Abstract

Using a sample of 264 strategic plan presentations by Milan Stock Exchange firms during 2001–2012, we present evidence of both a security price reaction and an increase in the accuracy of analysts’ earnings forecasts pursuant to plan disclosure. In the cross-section, the information content of the plan disclosures and the accuracy increase are incrementally associated with the extent of forward-looking narrative disclosures in the plan, after controlling for other disclosures within and outside the plan presentation and the fact that the firm has self-selected into the sample. Both quantitative and qualitative narrative disclosures are informative to investors and analysts. The results are driven by narrative disclosures about company strategy and action plans rather than about the business environment in which the company operates. Our study informs the current debate on the use of voluntary comprehensive, integrated, long-run-oriented strategic plan disclosure as a potential complement for disclosures such as quarterly earnings forecasts that have been described as an example of ‘short-termism’.

Acknowledgement

We acknowledge the contributions of participants at the XI Financial Reporting and Business Communication Conference and at the XXXI European Accounting Conference and of Sasson Bar Yosef Michel Magnan, Chris Cornwell, and Bill Lastrapes, who provided comments and suggestions on earlier drafts of this paper.

Supplementary Material

Supplemental data for this article can be accessed at 10.1080/09638180.2016.1152905.

Notes

¹During this period, the presence and activism of institutional investors on the MSE increased competition among firms with regard to fundraising, suggesting one reason why the disclosure benefits might be high. Choi (1973) finds that firms competing for scarce capital upgrade and increase financial disclosure. Also, the disclosure of longer run strategic plans is particularly salient in Italy given the longer term orientation of shareholders (i.e. the state in the utilities industry, foundations in the banking sector, and families in the manufacturing and other industries). The condition in Italian markets that firms choose to disclose the plan raises issues of self-selection, for which we control.

²We discuss specific measurement rules for the narrative later. We classify a narrative disclosure as quantitative if it contains a number, and qualitative otherwise.

³Appendix 1 summarizes the calculation of all variables in the study. In Section 3 and Appendix 2, we discuss the strategic plan data coding process in greater detail and we present specific examples of coded sections of the strategic plans.

⁴Precision refers to the disclosure form. For example, a continuum of disclosure form precision from least precise to most precise is nondisclosure, qualitative disclosure, open range, bounded range, and point.

⁵The use of absolute values of security prices to measure the information content of the release when expectations are not available is consistent with Cready and Hurtt (2002), Francis, Schipper, and Vincent (2002), Bushee et al. (2011), and Kirk and Markov (2013). The latter two articles use absolute values of security price changes to measure information content in the related contexts of corporate presentations and investor days.

⁶We employ Standardized |AR_i,t | when we test the mean effect of the strategic plan release. We use |AR_i,t | to examine cross-sectional differences in price reaction because dividing by the standard deviation of absolute returns may also reduce Standardized |AR_i,t | disproportionately for firms who have more firm-specific news days in the estimation period and hence greater variability in returns. Our results are robust, however, to using Standardized |AR_i,t | in cross-sectional regressions. The use of standardized and unstandardized absolute abnormal returns to assess mean and cross-sectional differences in information content is consistent with other disclosure studies (e.g. Francis et al., 2002; Bushee et al., 2011; Kirk & Markov, 2013).

⁷Even though the day +1 median standardized absolute abnormal return is near zero, it is substantially more positive than surrounding days, consistent with an upward spike in absolute security price change. The persistent negative median standardized abnormal return for all of the other trading days is also apparent in Bushee et al. (2011, Table 3). Francis et al. (1997) document a mean (median) market model based signed (not absolute) day 0 abnormal return for corporate plan presentations of 0.0027 (−0.0003); the mean is significant, the median is not. We document similar signed reactions (not tabulated). Our mean (median) day 0 abnormal returns are 0.0036 (−0.0004); the mean is significant, the median is not.

⁸Other potential proxies which we did not include are analyst following which is highly correlated with firm size in Italian markets (and in the US as well) and institutional ownership which is available only for a portion of our sample.

⁹Exclusion is justified by theoretical argument (e.g. Lennox et al., 2012) because exclusion restrictions are inherently untestable (e.g. Larker & Rusticus, 2010; Bertomeu, Beyer, & Taylor, 2015). For example, our price change (i.e. ex post return) dependent variable is theoretically determined by systematic risk (i.e. beta), potentially size and book-to-market, and cash flow news. Therefore, we do use firm size, market-to-book, and performance (i.e. ROA change) as exclusion restrictions. We are not aware of asset pricing models that argue that our exclusion restrictions (family ownership, demand for capital, and the sophistication of shareholder) are theoretical determinants of expected return, and we have no reason to expect them to be. Our increase in analyst forecast accuracy dependent variables is a change in accuracy. Therefore, the essentially fixed nature of the variables we use for exclusion restrictions and the fact that they are not earnings information suggest that they are not determinants of a change in analyst forecast accuracy except (possibly) through disclosure of a strategic plan.

¹⁰We use the two-day absolute abnormal returns given the results in Table 4.This approach is similar to the approach used by Francis et al. (2002) to first examine when the price reaction occurs and then use the days of price reaction to measure the dependent variable for subsequent cross-sectional analysis. We report the analysis with unstandardized rather than standardized returns to reduce the influence of firms with fewer non-event period disclosures and thus a much smaller mean absolute residual in the estimation period. However, results are similar using standardized returns (not tabulated).

¹¹Recall that LogSize is one of the exclusion restrictions for Equation (7) for reasons discussed earlier. The often documented relationship between firm size and price reactions prohibits excluding it from Equation (6).

¹²We also perform several other specification tests (reported in an online appendix) including replicating our results using one exclusion restriction at a time (our inferences remain the same) and demonstrating that the exclusion restrictions are not associated with either of our second stage dependent variables in selected non-event periods.

¹³We double cluster on firm and year and use heteroscedasticity-adjusted standard errors. 

¹⁴The narratives in our study are generally right skewed (see Table 2). We report our main results without a skewness-related data transformation for several reasons. First, in the univariate tests, we do not want to suppress larger values of the dependent variables because they are the most meaningful reactions to the event. For statistical purposes, we recognize the skewness in analyst forecast accuracy improvement by supplementing with a nonparametric test. In the regression tests, the strategy plan narratives are skewed independent variables, a condition which does not indicate bias or inefficiency or imply non-normality of regression residuals and the associated problems in inference. Further, our sample size is large enough to suggest that departures from regression residual normality are not likely to cause inferential problems. However, as a robustness check, we log transform the strategy plan-related independent variables, earnings volatility, and analyst following and repeat our second-stage Heckman tests. Our conclusion that narratives are incrementally useful (H2a and H2b) is not affected by the log transformation. One variable that did not achieve significance in the tabulated regression results (Environment) achieves marginal significance in one regression (with absolute stock price as the dependent variable). However, EnvQuant and EnvQual are not individually significant, similar to the tabulated regressions.

Additional information

Funding

We gratefully acknowledge the Research Division (DIR) of the SDA Bocconi – School of Management for financial support and are especially grateful to Marialaura Ricchetti, Laura Doria, and Elia Ferracuti for research support.