The Effect of Geographic Diversity on Managerial Earnings Forecasts

Abstract

We examine whether geographic diversity – a salient characteristic of the firm’s organizational structure – affects the timing and quality of voluntary disclosure. We find that firms with higher geographic diversity issue earnings forecasts that are more pessimistic, less precise, and less accurate. We also find that firms with higher geographic diversity are more likely to bundle managerial earnings forecasts with the prior quarter’s earnings announcement and less likely to issue forecasts during the quarter. These results are consistent with geographic diversity increasing information acquisition costs associated with providing managerial earnings forecasts. Consistent with these findings, we provide evidence consistent with managers substituting managerial earnings forecasts with firm-initiated non-earnings press releases, which require less information acquisition, and that managerial earnings forecasts are less useful to analysts and investors when geographic diversity is higher. Overall, our findings suggest that a firm’s organizational complexity is a factor that shapes the information environment of the firm.

Keywords:

• Organizational complexity
• Geographic diversity
• Management earnings forecasts
• Press releases
• Information acquisition costs

JEL codes:

• M40
• M41

Acknowledgements

We thank Thorsten Sellhorn (the Editor), two anonymous reviewers, Bob Bowen, Dave Burgstahler, Elizabeth Chuk, Ed deHaan, Frank Hodge, Weili Ge, Amy Hutton, Michael Kimbrough, Valerie Li, Dawn Matsumoto, Sarah McVay, D. Shores, Jacob Thornock, and Joshua Lee for helpful comments. We also thank seminar participants at Boston College, Harvard University, University of Washington, University of Wisconsin–Madison, the International Symposium on Forecasting, and the American Accounting Association Annual Meeting for helpful comments and suggestions. Earlier versions of this paper were circulated under the title ‘The effect of organizational complexity on earnings forecasting behavior.’ We would also like to thank the research assistance of Robert Stoumbos. We are grateful for financial support from the Olin Business School, Krannert School of Management, and Williams School of Commerce, Economics, and Politics. All errors remain our responsibility.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental Data and Research Materials

Supplemental data for this article can be accessed on the Taylor & Francis website, https://doi.org/10.1080/09638180.2022.2139738.

Online Appendix A – Descriptive Statistics

Online Appendix B – Segment Data

Table OA1 Descriptive Statistics

Table OA2 Correlations

Table OB1 Descriptive Statistics – S&P Capital IQ

Table OB2 Geographic Diversity and Management Forecast Timing

Table OB3 Geographic Diversity and Management Forecast Revisions

Table OB4 Geographic Diversity and Management Forecast Properties

Table OB5 Market Responses to Management Forecast News

Table OB6 Geographic Diversity and Firm-Initiated Press Releases

Table OB7 Cross-Sectional Tests: Cultural Tightness – Disclosure quantity

Table OB8 Cross-Sectional Tests: Cultural Tightness – Disclosure quality

Notes

1 For example, lawyers frequently bring class action lawsuits against firms for failing to update existing disclosure. Cornerstone (2013) provides evidence that approximately 54% of lawsuits between 2009 and 2013 have allegations of ‘false forward-looking information.’ Rogers and Van Buskirk (2009) suggest that plaintiff’s attorneys frequently argue that the Private Securities Litigation Reform Act of 1995, which purportedly protected forward-looking disclosures, does not protect the defendant’s forward-looking disclosures.

2 Drawing causal inferences from non-experimental data is challenging (Gow et al., 2016) due to endogeneity concerns. To mitigate the possibility of correlated omitted variables, we include control variables for many firm characteristics and disclosure determinants. We also include industry and time fixed effects to control for unobservable industry-specific characteristics and time-varying shocks. While we believe that reverse causality concerns are less likely (e.g., managerial quarterly earnings forecasts alter the firm’s decision to expand or divest its international operations), we cannot eliminate the possibility of reverse causality. Overall, we acknowledge that, due to the nature of non-experimental data, our study cannot provide causal evidence on the relation between geographic diversity and managerial forecasts. Generally, prior research does not examine whether geographic diversity affects management’s communication with external market participants with two exceptions. First, Herrmann et al. (2010) provide evidence that the quality and frequency of earnings guidance is higher for firms with a higher percentage of foreign sales prior to the passage of Reg FD. Second, Feng et al. (2009) find evidence consistent with ineffective internal controls decreasing the accuracy of management forecasts. Unlike these prior studies, we examine what disclosures managers use to substitute for managerial earnings forecasts when geographic diversity increases. We also provide evidence on how geographic diversity affects investor and analyst reactions to news communicated through managerial forecasts.

3 The following studies provide evidence that diversification is positively associated with firm performance and growth: Ghoshal (1987); Hitt et al. (1997); Rugman (1976); Kim et al. (1993); Markides and Ittner (1994). For example, international diversification allows firms to increase their reach to other international markets and to tap new opportunities to raise capital (Hitt et al., 2006).

4 It is possible that advances in technology, communication, travel, and information systems reduce or eliminate the information acquisition costs that corporate headquarters face when operating in multiple geographic regions, which in turn would attenuate the negative association between geographic diversity and the quality/timeliness of management’s earnings forecasts.

5 We assign value of zero to missing values of the geographic diversity measure. Our results are qualitatively similar when we delete observations with missing values of the measure.

6 The number of segments, discontinuation of geographic segments, and the geographic location of subsidiaries are alternative measures for geographic diversity. These alternative measures are subject to the same self-reporting issues and potential measurement error in Geo Diversity_i,t. Contrary to these alternative proxies, our measure weights the firm’s geographical segments using the sales generated by each segment, which allows us to measure the importance of the geographic segment to the firm. Using these alternative measures, we find weaker but qualitatively similar results (untabulated).

7 If firms choose to make more precise segment disclosures, it may reduce information asymmetry and reduce investor demand for managerial earnings forecasts. However, we do not believe this to be the case for two reasons. First, segment disclosures provide historical information while earnings forecasts provide forward-looking information. Segment disclosures and earnings forecasts are likely to be complements because firms with high quality mandatory disclosure likely provide high quality voluntary disclosure (Ball et al., 2012). Second, in Table 7, we find that firms with higher geographic diversity are more likely to issue firm-initiated non-earnings press releases, suggesting that segment disclosures do not necessarily substitute for managerial voluntary disclosure.

8 The extant research, predominantly in the finance literature, speaks to the challenges that diversified firms face regarding how the market views and values business diversification (Habib et al., 1997; Krishnaswami & Subramaniam, 1999; Nanda & Narayanan, 1999).

9 Nagar et al. (2003) include the number of business segments as a control in their model and also find no relation between the number of business segments and management forecast frequency.

10 F. S. Zhou (2021) compares the I/B/E/S guidance database with the CIG database. He finds that the ‘I/B/E/S guidance is superior to CIG in the coverage comprehensiveness.’

11 Following prior studies (Drake et al., 2014), we use business press data from RavenPack, an aggregator of business press articles. The dataset includes the Dow Jones (DJ) news archives – consisting of all DJ Newswires and Wall Street Journal articles. Tetlock (2007) argues that the DJ news archives is a natural choice for a business press data source that impacts the market, due to its large circulation and influence.

12 We do not use a logit/probit model (i.e., non-linear model) with fixed effects to avoid the incidental parameters problem (Neyman & Scott, 1948). In untabulated tests, we find that our results are robust to using a logit or a probit model with or without fixed effects.

13 Several control variables are likely associated with the sophistication of the firm’s information system. For example, larger firms, more volatile firms, and more mature firms are more likely to have more sophisticated information systems. We also control for industry and institutional ownership, which is likely associated with the demand for information. Lastly, we also control for the lagged dependent variable in many empirical tests, which also controls for the impact of the information system to the degree that voluntary disclosure is associated with the information system.

14 We use the standard deviation in revenues as opposed to earnings to more directly capture demand uncertainty facing the organization. However, we also use earnings volatility (which is highly correlated with the volatility in revenues) in unreported tests and find similar results.

15 The provision/properties of managerial forecasts and the disclosure of geographical segments in the financial statements could be influenced by the sensitivity of managers’ compensation to stock market fluctuations. In additional robustness tests, we control for the sensitivity of managers’ compensation to stock market fluctuations. We note that the results are weaker; however, the inferences are qualitatively similar. We believe that the results are slightly weaker because the sample size is smaller, potentially reducing the statistical power of the empirical tests. Compensation data in ExecuComp is only available for S&P 1,500 firms.

16 Bundled Last MF_i,t is equal to 1 when the last managerial forecast is issued in the 5-day window surrounding the earnings announcement, and Bundled MF_i,t is equal to 1 when there is a managerial forecast issued during the 5-day window surrounding the earnings announcement. The regression including Bundled Last MF_i,t as the dependent variable includes firm-quarter observations for which a managerial forecast is provided by management. The regression including Bundled MF_i,t as the dependent variable includes all firm-quarter observations.

17 The standard deviation of Geo Diversity_i,t for this reduced sample with non-missing values of the dependent variable is equal to 0.282 (untabulated).

18 We use the last earnings forecast for firm i during quarter t to compute the variables. We find that the results are unaffected by using the first earnings forecast issued during the quarter.

19 In untabulated tests, we run an ordered logit with forecast form as the dependent variable. Forecast form is equal to two when the forecast is a point forecast, one when the forecast is a closed-ended range forecast, and zero when the forecast is an open-ended forecast. We find a negative but insignificant coefficient on Geo Diversity_i,t. However, we do not believe the insignificant relation nullifies or diminishes the results in Table 5. The analysis in Table 5 allows us to identify continuous variation in the range forecasts, which we cannot do with the ordered logit test. The insignificant results using forecast form could be because there is less variation in the dependent variable. In our sample, among all the last management forecasts issued during the quarter, only 2.09% ( = 684 / 32,714) of the forecasts are open-ended range forecasts while 85.97% ( = 28,125 / 32,714) and 11.94% ( = 3,905 / 32,714) are closed-range range and point forecasts, respectively. These descriptive statistics are consistent with prior research that provides evidence that closed-range forecasts are the most prevalent form of management earnings forecasts (Ciconte et al., 2014).

20 Based on Ravenpack’s recommendation, we identify firm-initiated press releases by using a relevance score of higher than 75, an event novelty score of 100, and news type of ‘press release.’ A press release is defined by Ravenpack as ‘a corporate announcement originated by an entity and distributed via a news provider.’

21 In case where the geographic segment definition is not specific to an individual country (e.g., Europe, Asia, Latin America, and etc.), we use the region’s GDP weighted average of cultural tightness score. We obtain GDP data and the regional information from the World Bank (https://data.worldbank.org/indicator/).

22 Bradshaw and Sloan (2002) suggest that ‘the quarterly earnings announcement season has become a closely watched ritual.’

23 https://www.capitaliq.com/.

24 We use the following dependent variables to test whether firms with higher geographic diversity have less timely managerial earnings forecasts during the quarter: Unbundled MF_i,t (Table 3), Bundled Last MF_i,t (Table 4), MF Revision_i,t (Table 4), and lnNR_Earn_i,t (Table 7). The robustness tests using S&P Capital IQ data to construct the geographic diversity measures are consistent with the results reported in the paper when using the following dependent variables: MF Revision_i,t (Table OB3 in the Online Appendix), Bundled Last MF_i,t (Table OB3 in the Online Appendix), and lnNR_Earn_i,t (Table OB6 in the Online Appendix). However, the geographic diversity measures are not significant when Unbundled MF_i,t, is the dependent variable (Table OB2 in the Online Appendix). Therefore, in summary, the results in the robustness tests support our inference that geographic diversity leads to less timely earnings forecasts during the quarter.