The impacts of product market competition on the quantity and quality of voluntary disclosures

The only exception is Shin (2002), which differentiates capacity competition from price competition and shows that firms engaged in capacity competition disclose relatively more information than those engaged in price competition.

The only exception is Rogers and Stocken (2005), who use Herfindahl–Hirschman Index as a proxy for industry concentration. However, as Karuna (2007) points out, industry concentration alone could be a poor proxy for competition due to endogeneity.

To the best of my knowledge, Brown et al. (2006) and Jones and Cole (2008) are the only papers that study management forecasts on future capital expenditures.

Dye (1985) and Jung and Kwon (1988) argue that partial disclosure could also be achieved if investors are unsure about the manager’s endowment of private information. However, all theoretical models discussed in this section assume that the manager is always endowed with private information and it is certain to investors that the manager has such information endowment.

Using equally-weighted PP&E, capital expenditures, and R&D in the analysis does not change the results.

For example, Sutton (1991) regards R&D outlays as endogenous sunk costs firms incur at stage 1 to enhance the demand for their products at stage 2. In contrast, setup costs are regarded as exogenous sunk costs at stage 1.

Note that to compute the value of principal components, original variables are standardized. Therefore, principal components could have negative values.

Ali et al. (2009) argue that Compustat-based industry concentration measures are subject to measurement error, as most of the private firms are not covered by Compustat and high Compustat-based concentration ratio is likely to be due to the declining of the industry, which is left with only a few large, public firms relative to private firms. Alternatively, they suggest that researchers should use concentration ratios from US Census data. In this paper, I choose to use Compustat concentration measures for the following reasons. First, the US Census measure of concentration is only available for the year 2002 of my sample period and only available for manufacturing industries. Hence, using US Census data would reduce the sample size, thereby contradicting the aim of this paper to provide large sample evidence. Second, using a Compustat-based concentration measure in this paper is a conservative approach. Previous literature suggests that firms with poor performance usually provide less voluntary disclosures (Miller 2002; Kothari et al. 2009). Therefore, if the Compustat-based concentration ratio is capturing the declining of the industry, using it will work against me finding the results for H1B that existing competition (industry concentration) is negatively (positively) associated disclosure quantity. Nevertheless, I further address the measurement error problem with Compustat-based competition measures in the robustness analysis by using exploratory factor analysis.

Under Safe Harbor, it is more difficult to prove the defendant guilty, because plaintiffs must identify the specific statement or statements that are misleading when they file the lawsuit rather than undertaking a “fishing expedition” for supporting documentation during the discovery process (Johnson et al. 2001).

The results are unchanged if total assets are used as the deflator.

Management forecasts are mainly issued through conference calls, conferences, analyst meetings, shareholder presentations and press releases. To obtain information about management investment forecasts, I first use key words to search in Factiva and download all output articles. The search period starts 24 months before the forecasting fiscal year-end. Then, I use Perl to extract relevant information from downloaded articles. Finally, I manually read and code the extracted information.

The purpose of this data requirement is to have meaningful industry-average measures. Nevertheless, the results are not sensitive to this data requirement (unreported).

Two possible reasons may cause the data on management forecasts to be missing: (1) the firm did not issue any forecast or (2) the firm was outside the coverage of First Call database, which primarily covers only firms followed by analysts (Anilowski et al. 2007). I, therefore, limit the sample to firms that have data available on analyst estimates. In other words, a firm with missing data on management forecasts is regarded as a nonforecaster only if it has non-missing data on analyst estimates. Firm-years that are not covered by analysts are excluded from the sample.

Note that the numbers for ACCURACY and SURPRS in this paper are multiplied by 100.

The results are qualitatively similar but with lower statistical significance if the standard errors are clustered at industry-level, due to the small degrees of freedom for industry-level regressions.

The results are qualitatively similar but with lower statistical significance if the standard errors are clustered at industry level, due to the small degrees of freedom for industry level regressions.

The results are similar if the standard errors are clustered by both industry and calendar year.

Although the medians are negative, the magnitudes are much smaller, indicating a long right tail. The reason why investment forecasts are pessimistic on average is unclear and outside the scope of this paper. The empirical evidence on the market reaction to investment announcements is mixed. For example, McConnell and Muscarella (1985) find that announcements of increases in capital expenditures lead to significant positive stock returns for industrial firms, but such association does not exist for public utility firms. Chung et al. (1998) find that share price reaction to a firm’s capital expenditure decisions depends critically on the capital market’s assessment of the quality of its investment opportunities.

Note that competition measures based on common factors, such as POTENT-COMP, EXIST-COMP, POTENT-COMP_EFA, and EXIST-COMP_EFA, have been multiplied by −1, so that higher value indicates higher competition level. Therefore, the coefficients on some of the original variables, such as IND-PPE, IND-CPX, and IND-CON4, in this table have the opposite signs as those on POTENT-COMP and EXIST-COMP in Table 6.

So far, this measure has only been empirically applied in banking industries.

In unreported analysis, results from regressing POTENT-COMP_j,t on POTENT-COMP_j,t−1, POTENT-COMP_j,t−2, FORECASTER_j,t−1%, FORECASTER_j,t−2%, and control variables indicate that disclosure does not increase potential competition in a Granger sense. Similarly, results from regressing EXIST-COMP_j,t on EXIST-COMP_j,t−1, EXIST-COMP_j,t−2, FORECASTER_j,t−1%, FORECASTER_j,t−2%, and control variables indicate that disclosure does not decrease existing competition in a Granger sense.

I thank the referee for pointing this out and suggesting the alternative measure. The results are qualitatively similar if I use the average forecasting accuracy from years t − 3, t − 2, and t − 1 as an alternative.

For example, For example, Harris (1998) and Botosan and Stanford (2005) find that competition encourages disclosure, while other empirical studies generally find that competition discourages disclosure.

This is consistent with the methodology that SIC uses to assign primary SIC code to each firm.