Voluntary disclosure of investment forecasts and the cost of capital: evidence from the treatment effect estimates model

ABSTRACT

This study examines the economic consequences of voluntary disclosures for investment forecasts. Using Japanese data, I examine whether voluntary disclosure of management forecasts of capital investments and research and development investments is related to the cost of capital in the same and subsequent years. The results indicate that firms that disclose investment forecasts realize a greater reduction in the cost of capital in the same and subsequent accounting periods than firms that do not disclose. Another finding is that the initial investment forecast disclosures affect the reduction in the cost of capital in the year subsequent to the initial disclosure. This suggests that information effects of investment forecasts on the cost of capital appear over time.

KEYWORDS:

• Investment forecast disclosure
• voluntary disclosure
• implied cost of equity capital

Acknowledgement

I am grateful to the Editor and the anonymous referee(s) for their helpful and constructive comments. I would like to thank Associate Professor Tomohiro Suzuki and the participants at 28th Asian-Pacific conference on international accounting issues and workshop at Hitotsubashi University for their valuable advice.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. For example, Apple Inc. describes in a Form 10-K report in 2016 that ‘The Company anticipates utilizing approximately $16.0 billion for capital expenditures during 2017, which includes product tooling and manufacturing process equipment; data centers; corporate facilities and infrastructure, including information systems hardware, software and enhancements; and retail store facilities’. In addition, Vodafone Group PLC describes in the Chief Executive’s strategic review in 2016 that ‘We expect free cash flow of at least €4 billion. Total capital expenditure is now targeted to be in the mid-teens as a percentage of annual revenue; this is higher than the 13%-14% range that we previously anticipated, as we believe that there are attractive investment opportunities available to further accelerate our growth and improve our long-term strategic positioning’.

2. The TSE states that the forward-looking information to be disclosed is not limited to earnings forecasts; a wide range of contents are included in forward-looking information, including a descriptive explanation of the future vision, principle management index forecasts such as ROE, and the financial indicators that influence future business performance (e.g. outlay of capital investment and R&D investment, as well as depreciation).

3. The percentages of TSE listing firms that disclosed actual and forecast values for capital investment and R&D investment by industry between 2002 and 2011. For actual figures, the average disclosure ratio for manufacturing industries is 96.86%, whereas that for non-manufacturing industries is 88.21%. For forecast figures, the average disclosure ratio for manufacturing industries is 80.27%, whilst that for non-manufacturing industries is 59.93%.

4. One possible explanation is that voluntary disclosure of investment forecasts causes the potential loss of competitive advantage or core business competencies (Diamond and Verrecchia 1991). Indeed, Depoers (2000) empirically finds that some disclosures give competitors too much information and cause financial harm in France.

5. For example, Microsoft Inc. provides detailed information on R&D investment forecasts in its Form 10-K, whilst Suncor Energy Inc., a Canadian firm, descriptively discloses its future prospects for capital expenditures in its Management Discussion and Analysis (MD&A) . Although firms around the world consider investment forecast information to be of concern for investors and spontaneously disclose such information, the venues through which particular firms offer disclosures and disclosed content are diverse.

6. In this study, it should be stressed that information quality (information content) is not the point in question because I focus on the information amount and whether firms disclose investment forecasts. I compare firms that disclose investment forecasts with those that do not and examine the information quantity firms provide to investors.

7. See Mercer (2004) for more discussion about information supporting earnings forecast.

8. The estimated implied cost of capital at the end of the month when financial statement information for year t is disclosed is used as the implied cost of capital for year t.

9. In both cases, DISC_t will be correlated with the error term and the estimated parameters from OLS regression will be biased and inconsistent.

10. Other estimation models are also developed to measure discretionary accruals in previous studies. In this study, I adopt Kasznik’s (1999) estimation model because it is proposed under the context of voluntary disclosure and earnings management.

11. The Nikkei middle classification is used.

12. Healy and Palepu (2001) show six hypotheses – capital markets transactions hypothesis, corporate control contest hypothesis, stock compensation hypothesis, litigation cost hypothesis, management talent signalling hypothesis, and proprietary cost hypothesis.

13. Other hypotheses can also be considered. For example, Depoers (2000) indicates that internationality (foreign activities), auditor firms, and labour pressure are possible determinants of voluntary disclosure, and Lang and Lundholm (1993) suggest that analyst following is also a determinant of disclosure levels. However, in this paper, these factors are not taken into account because of the problem of data availability.

14. ROA is earnings before interest, taxes, depreciation and amortization (EBITDA) in year t–1, divided by total assets at the end of year t–1.

15. The numbers in the income statements cannot be compared equally when the accounting period changes. Mergers largely affect both balance sheets and income statements, and thus firms that merge or are merged are excluded.

16. Industry classifications are based on the Nikkei middle classification.

17. The mean and median values of the cost of capital in 2004 are 6.31% and 4.31%, whilst those in 2011 are 15.76% and 6.77%.

18. White’s t-tests and rank sum tests are employed to test the mean and median differences.

19. The differences in mean values are tested using White’s t-tests, because F-tests rejected the null hypothesis that the variances of the two groups are equal.

20. Similar to the result for mean value, the median value for firms that disclose capital investment forecasts (0.0020) is smaller than that for firms that do not disclose them (0.0033). The difference between the two detected by a rank sum test is statistically significant at the 1% level (z-value is –4.93). In addition, the median value for firms disclosing R&D investment forecasts is smaller than that for firms not disclosing them; the former is 0.0020 and the latter is 0.0029. The difference is statistically significant at the 1% level.

21. The significance of ρ in models (1) and (2) suggests that applying the treatment effect estimates models using DISC_FCAP_t and DISC_FRD_t as selection variables is appropriate, and that this estimation model is well specified.

22. From the results in the lower part of the table (first-stage regressions) in model (1), it can be concluded that the determinants of capital investment forecast disclosures are market capitalisation, misvaluation, entry barrier and board ownership. The results in model (2) suggest that the determinants of R&D investment forecast disclosures are firm size and new financing from direct markets.

23. I follow Dhaliwal et al.’s (2011) model and take the differences (Δ) for each control variable (CNTL_t).

24. The results of t-tests of mean differences show that there is no mean difference between the sample group and the matching group for most of the independent variables in Equation (3)(3) . This implies that the matching process based on the propensity scores calculated from these independent variables makes sense.

25. For INT_FCAP_t, the coefficient and t-value are –1.0887 and –3.85, respectively. The coefficient of INT_FRD_t is also negative (–2.6354) and significant at the 1% level (t-value is –9.53).

Additional information

Funding

This work was supported by the Grant-in-Aids for Science Research (26885004 and 16K17204) funded by the Japan Society for the Promotion of Science.