Incentives for Non-Disclosure by Corporate Groups

Reduction in Reporting and Auditing Costs

The major purported reason for the adoption of the deed is the reduction of the firm's direct (out-of-pocket) reporting costs. Subsidiaries in the closed group do not have to prepare, have audited and file an annual report with the corporate regulator. Dean and Clarke (2005) pose the counter view that each subsidiary in the closed group will still produce financial statements, as these are necessary for internal reporting purposes (i.e., managerial control) and external reporting purposes (i.e., as the foundations for the preparation of the economic group's consolidated statements). ASIC stated that the ‘no filing’ requirement would eliminate the need for certain statutory disclosures at the individual firm level (e.g., related party disclosures and cash flow statements) and the audit may be undertaken using group materiality considerations, thereby reducing the scope and cost of the audit.

The reporting and audit cost savings are likely to be related to lower audit fees. In our cross-sectional design, we are unable to identify the audit fees prior to the deed being in place. Consistent with other empirical studies, firm size has been used as a proxy for the ‘economies of scale’ related to voluntary disclosure (e.g., Firth, 1979) and voluntary interim reporting (Bradbury, 1992b). However, drawing on the regulator's primary reasoning for deed adoption, the cost savings are likely to be related to the number of subsidiaries, rather than size per se. Therefore, the log of the number of subsidiaries (NSUB) is used as a proxy for the potential audit and filing savings arising from adoption of the deed.

Proprietary Costs (and Competition)

One likely reason managers object to providing segment data is that their disclosure would give competitors commercially sensitive information (e.g., see Sprouse, 1969, and Ettredge et al., 2002). Verrecchia (2001) notes, for example, that results from the analytical literature are highly sensitive to specific, often stylized, assumptions. For example, Darrough and Stoughton (1990) argue that greater competition will increase disclosure to discourage entry into the product market. In contrast, the existence of proprietary costs provides a theoretical rationale for less disclosure (e.g., Verrecchia, 1983; Wagenhofer, 1990). Darrough (1993) shows that both the type of information (demand versus cost) and the nature of competition (Cournot versus Bertrand) influence managers' disclosure preferences. Hence, the relationship between voluntary disclosure and competition is likely to be context specific.

The decision to eliminate the level of sub-entity disclosures by adopting the deed is a different setting from the assumptions employed in the theoretical models of voluntary disclosure. For example, theoretical models often assume that managers' disclosure decisions are conditional on the observed signals of firm performance. On the other hand, the decision to adopt the deed (generally) produces a more permanent arrangement and is unlikely to be conditional on the immediate financial performance of subsidiaries.⁹ We therefore turn to the empirical literature that relates competition to accounting policy choice.

One of the first studies to include a competition explanatory variable in an empirical analysis is Hagerman and Zmijewski (1979). They examine the impact of competition, with other factors, on accounting policy choices for depreciation, inventory, pension costs and investment credit. Their hypothesis is that the more concentrated the industry, the greater the likelihood of either anti-trust action (a political cost) or rivals' entry into the industry and hence the greater incentive to choose an accounting method that reduces reported profits. Evidence of a negative relationship between the concentration ratio (a proxy for market power) and income increasing accounting policy choices was found.

Bradbury (1991) also uses a concentration ratio to explain the level of voluntary interim reporting, arguing that the costs of accounting disclosure include the impact of entry decisions by potential competitors, and the short-run decisions of existing competitors. As the number of participants increases, the market becomes more competitive and the share of the industry supplied by a firm decreases. Where an industry comprises numerous firms, it is likely that producers will ignore the effect of rivals' actions on their pricing and output decisions. Hence, the cost of disclosing competitive information is likely to be higher in the more concentrated industries. Bradbury finds univariate evidence of a weak negative relationship between the concentration ratio and the level of disclosure.

Scott (1994) argues that information in pension plan disclosures by Canadian firms is proprietary with respect to labour unions. He uses strike incidence, pay rate and return on assets relative to the industry to measure proprietary costs of pension disclosures. The results are consistent with the view that disclosure is a trade-off between proprietary costs and capital market benefits.

Harris (1998) and Leuz (2004) examine the relationship between competition and segment disclosures. Harris shows that the decision to report operations in a given industry as a segment is a function of two measures of industry competition: the four-firm concentration ratio and the speed of abnormal profit adjustment. Her results, consistent with Hayes and Lundholm (1996), suggest that firms with high-profit operations will hide these from competitors using the discretion in segment reporting standards. Leuz finds that capital intensity, as a proxy for entry barriers, is positively related to the level of segment disclosure.

Bhoraj et al. (2004) and Bradbury and Hooks (2005) examine voluntary disclosures in a reorganization of the electric utility industry. Bhoraj et al. measure disclosure during the electric utility industry's transition towards deregulation. They find evidence of capital market incentives positively related to disclosure. Product market incentives are perceived to play a role in disclosure, but only after regulatory concerns have been resolved. Bradbury and Hooks show that disclosure levels decreased when the electric utility industry in New Zealand was reorganized to introduce product market competition. In particular, they find lower disclosures for segment data, forward-looking data and current operating data relating to the firm's derivative positions. They also note that the lower disclosures are temporary and are partially offset from a demand for higher disclosures by capital market agents.

Further empirical evidence of a relationship between competition and proprietary disclosure is provided by Verrecchia and Weber (2006). Investigating ‘redacted proprietary information’ when the SEC grants a firm's request to withhold information from its material contract filings, they find there is an increase in adverse selection (measured by bid-ask spread, market depth and share turnover). Firms are found to be less likely to withhold information when issuing long-term debt and when they are in a competitive industry or experience losses.

In summary, the theoretical models imply that proprietary costs are a major influence on the level of voluntary disclosure. The empirical literature indicates that other incentives for increased disclosure (e.g., capital market, agency costs and regulatory costs) are likely to dominate. However, most of the empirical research is concerned with an increase in voluntary disclosure. The closest scenario to our study of non-disclosure is Verrecchia and Weber (2006). Based on their results, we hypothesize, ceteris paribus, that decisions to reduce proprietary information are likely to be more frequent for firms in competitive industries (i.e., those industries, generally, with lower concentration ratios).

We measure market concentration (and the proprietary costs of disclosure) using the ranks of a four-firm concentration ratio (rankCR4) estimated from all firms listed on the Australian Stock Exchange. The industry sectors comprise between nine and 376 firms with a median of 47.¹⁰

There is doubt over whether the industry concentration ratio variables proxy for the proprietary costs of disclosure (i.e., non-disclosure of segment data to hide high margin activities) or management deceit (i.e., non-disclosure of segment data to hide low margin activities). Botosan and Stanford (2005) and Berger and Hann (2007) exploit the change in the segment reporting standard (from SFAS 14 to SFAS 131). Botosan and Stanford examine a sample of firms that had previously reported as a single-segment firm and initiated segment disclosures with SFAS 131. They find these firms use the latitude in SFAS 14 to hide profitable segments operating in industries that are less competitive than the primary industry. There is no evidence that firms use SFAS 14 to mask poor performance. Berger and Hann use segment profitability (rather than industry or firm level profitability). ‘Old’ (SFAS 14) and ‘new’ (SFAS 131) segment reports yielded opposite predictions for the agency and proprietary costs motives to disclose segment information. Firms were found to withhold segment disclosures of segments with low abnormal profits (the agency cost hypothesis), and there was mixed evidence supporting the withholding of abnormal profit segment data (the proprietary cost hypothesis).

We do not have the data to address the issue of conflicting agency cost and proprietary cost hypotheses for non-disclosure. As discussed in note 9, the deed is arguably a more permanent arrangement, and therefore current performance (either good or bad) is less likely to be a determinant of non-disclosure. Furthermore, management is unlikely to put loss making subsidiaries in the closed group because, with cross guarantees in place, such a strategy might bring down other companies within the closed group. Hence, our competition variable is considered to proxy for proprietary costs of disclosure.

Finally, the more operating segments the firm has, the greater its ability to mask the performance of competitive businesses within the closed-group companies, either through segment identification or cost allocation (see Hayes and Lundholm, 1996, for a theoretical analysis). That is, for the non-disclosure of proprietary costs to be successful under the deed there must also be non-disclosure under segment reporting. This suggests a need to capture the interaction between the proprietary costs (rankCR4) and the number of segments (NSEG).

Capital Market Incentives

Several analytical papers model the voluntary disclosure decision as a trade-off between capital market benefits and proprietary costs (e.g., Verrecchia, 1983; Hayes and Lundholm, 1996). Pressure from analysts may influence a firm's reporting choices. For example, Lang and Lundholm (1996) and Healy et al. (1999) find an increase in analysts' following after increased disclosure. Several studies (e.g., Choi, 1973; Gibbins et al., 1990; Lang and Lundholm, 1993) support the view that firms increase disclosures prior to issuing new securities. Leuz (2004) employs a dummy variable to indicate a forward-looking equity offering, which is not significant in explaining the level of segment reporting. Verrecchia and Weber (2006) investigate a firm's decision to redact information (presumably proprietary information) from its Securities and Exchange Commission (SEC) material contract filings. They argue redaction exacerbates adverse selection, thereby increasing the extent to which market participants will price-protect, finding that firms are more likely to redact information when they are in a competitive industry or suffering losses, less likely to redact information when they issue debt.

This study employs a cross-sectional sample likely to capture the more long-term characteristics of capital market incentives to disclose information. Evidence of long-term capital market incentives for increased disclosure is provided by Frankel et al. (1995). There is a long-term association between the issuance of management forecasts and the raising of external finance. However, forces such as legal liability deter firms from issuing more frequent forecasts around the time of finance offerings. To capture the long-term characteristics of capital market incentives for increased disclosure, market to book equity ratio (MKTBK) is used. MKTBK is a measure of growth, proxying for future financing needs, and is expected to be negatively related to the decision to reduce disclosures (Leuz, 2004). Leuz et al. (2007) provide evidence that ‘going dark’ (i.e., those companies that deregister common stock, thereby suspending or terminating their obligation to comply with SEC reporting requirements leaving shareholders without access to information) is a response to financial difficulties and deteriorating growth opportunities.

Agency Cost Incentives

Watts (1977) uses an agency cost framework to explain the production, and the cross-sectional variation in the content of financial statements. Empirical evidence supports this claim for interim reporting frequency (Leftwich et al., 1981) and the level of interim reporting disclosures (Bradbury, 1991, 1992b).We consider the potential increase in agency costs of equity and debt on the decision to reduce disclosures.

Jensen and Meckling (1976) argue that agency costs increase as the level of outside equity increases. Furthermore, firms with dispersed ownership are more likely to use public disclosures to communicate information. Following Leftwich et al. (1981), Bradbury (1991) and Leuz (2004), the log of the number of outstanding shares (NSHARES) is used to capture the agency costs of equity. Theory suggests there is a negative relationship between NSHARES and the decision to reduce disclosures by adopting the deed. Given that shareholders have access to the consolidated financial statement of the whole group, it is unlikely that the agency costs of equity will be important for the deed adopting firms.

The agency problems of debt are likely to increase as leverage increases. Voluntary disclosures are often hypothesized to be positively associated with leverage. The results of studies are mixed—with some finding a positive relationship between leverage and segment disclosures (e.g., Bradbury, 1992a; Mitchell et al., 1995), and others a negative one (e.g., McKinnon and Dalimunthe, 1993) or none (Leuz, 2004).

An additional benefit of the mutual cross guarantee under the deed is that it potentially facilitates debt contracting. First, the assets of the closed group are pooled in the event of a liquidation of a closed-group company (i.e., where a closed-group company is unable to meet its debts and is put into formal liquidation), and therefore there is more ex ante certainty and diversification of security for creditors.¹¹ Second, it allows the entity to contract with a single set of lenders to the closed group thereby reducing the financing costs of negotiating with multiple lenders to multiple subsidiaries. While we do not have data to estimate the negotiation costs of financing incurred by the entity, if the cost of debt negotiation under the deed is lower, then firms with the deed in place would likely have more debt than non-deed companies (Whittred, 1987). Leverage (LEV) is estimated by the ratio of debt to total assets.

Directors' Liability

Early commentaries predicted a declining popularity of the deed, as the extent of directors’ legal obligation to certify the financial integrity of the closed group (as described earlier) became more apparent (Hill, 1992, 1995). The additional directors’ liability is created by the additional solvency declarations required by the deed. However, if the deed is a more efficient business decision that lowers costs, then the probability of failure will be lower.

Prior research also suggests that managers' discretionary disclosure policy is likely to be affected by litigation risk (Richardson, 2001). Empirical evidence on litigation risk affecting the disclosure of bad news is provided by Skinner (1994). While measuring directors' liability is difficult, we consider it is important to include a proxy variable for directors' liability to avoid problems with an omitted variable and the strength of prior concerns over directors' liability. Arguably, the directors' liability is more likely to affect independent non-executive directors who would be reasonably expected to have less knowledge of the firm's internal activities than inside directors. Hence, ceteris paribus, a negative relationship is expected between the existence of the deed and the proportion of non-executive directors on the board (PNED).

Impact of Organizational Structure

The non-disclosure and cross guarantee effects from the deed could be achieved by simply transferring the net assets of subsidiaries to a single company structure (without subsidiaries).¹² In this scenario cross guarantees would be unnecessary.¹³ Clearly this solution would entail likely considerable transaction costs (e.g., legal costs and taxes) of reconstructing the group into a single entity. Futhermore, if a single entity is an optimal operating structure, in all likelihood the firm would already be organized in this manner. Hence, the existing group structure (if optimal) constrains the ability of the firm to realize the benefits of non-disclosure by adopting the deed. In particular, we consider the impact of organizational factors relating to central control, vertical integration and complexity on deed adoption.

Chowdhry and Nanda (1994) provide an analytical model to explain why firms will bail out struggling subsidiaries (despite no legal obligation to do so). On the other hand, Robbins and Stobaugh (1973) found that most firms prefer to not guarantee the external debt of their subsidiaries, though firms often provide a ‘moral guarantee’ in the form a comfort letter.¹⁴ It is also significant that parent companies rarely provide statements of non-guarantee, preferring to leave intentions ambiguous to facilitate maximum flexibility (Beranek and Clayton, 1985). Clarke, et al. (2003) report several examples of Australian parent companies ‘letting go’ subsidiaries.

Data

The initial sample comprises the top 225 companies (by market capitalization) listed on the Australian Stock Exchange. Deed adoption is found to be more likely by large firms (Dean and Clarke, 2005; Ramsay and Stapleton, 2001). We eliminate fifty-six firms in the financial sector, as these have different disclosure requirements and financial structures to non-financial firms. Seven overseas listed firms that are unlikely to be eligible to apply the deed are also eliminated. Following Ramsay and Stapledon (2001), two property trusts and five other firms that do not have subsidiaries are eliminated. Accordingly, the final sample comprises 155 firms, 82 (52.9 per cent) of which have executed a deed. Table 1 summarizes the sampling procedures.

Table 1.
SAMPLE SELECTION

Table 2 analyses deed usage by industry. The greatest industry uptake is in the Automobile & Components industry, with 100 per cent of the sample adopting the deed, although only two firms comprise this industry. The next highest is the Food & Staples Retailing sector with an 87.5 per cent uptake (seven out of eight firms). The Utilities industry, with 16.7 per cent uptake (one out of six firms) is the lowest. Table 2 reveals a variety of deed use across a range of industries.¹⁸

Organizational Variables

We hypothesize that deeds are most likely to be executed by firms that are centrally controlled, vertically integrated and interdependent, and have few foreign operations. We measure central control by the proportion of parent activities to group activities in terms of assets (PASSET) or debt (PDEBT). Compared to horizontally and conglomerate-based firms, a vertically integrated firm is likely to have its business assets concentrated in fewer business segments. To measure vertical integration, we use the proportion of the major business segment to the total size of the group measured in terms of assets (MSEGA) or revenue (MSEGR). The size of foreign operations is measured by the proportion of foreign segment revenue (FSEGR) or assets (FSEGA).

Arguably, whether revenues, assets or debt measures best capture underlying organizational structure is equivocal.¹⁹ Furthermore, there are likely to be strong correlations among segment data. Thus, factor analysis is used to classify variables according to their underlying structure.²⁰

Table 3, Panel A reports the descriptive statistics for each organizational variable. Panel B reports Pearson and Spearman correlations. As expected, there are some high correlations among the variables.

Panel C of Table 3 shows the results of the factor analysis. There are three factors with eigenvalues greater than 1. These factors explain 76 per cent of the variation in the organizational variables. For each factor only factor loadings greater than |0.3| are reported. The proportion of the major segment assets and revenue load on factor one and we label this factor vertical integration.²¹ The proportion of foreign operations in terms of revenues and assets load on factor two. Consistent with our hypothesis, this factor is labelled complexity. The proportion of parent company debt to group total debt and parent company total assets to group total assets load on to the third factor. This factor is labelled control by the parent company over the financing, operating, and information interdependence of the group.²²

Panel C in Table 3 also reports the estimated communalities (which are equivalent to the squared multiple correlations obtained from the regression of each organizational measure with other organizational measures). The high correlations suggest that the use of factor analysis is appropriate.

Descriptive Statistics and Univariate Tests

Table 4 reports the descriptive statistics and univariate tests of the explanatory variables. A Mann-Whitney U, Z statistic is used to test for differences in hypothesized firm characteristics between those firms with deeds and those without.

There are four variables (NSUB, NSEG, LEV and the vertical integration factor) where the difference between firms with and without the deed is significant (at the 0.001 level). Deed firms have more subsidiaries, more reported segments, higher leverage and a higher vertical integration factor. This is consistent with the view that deed adoption is easier for firms with high levels of operating, information and financial interdependence (i.e., they are vertically integrated). Rank of CR4 is significantly different at the 0.10 level. DEED firms are in industries with a lower concentration ratio. This is consistent with firms in competitive industries adopting the deed to lower disclosures because of proprietary costs. Firms adopting the deed also have more shares outstanding (significant at 0.05 level), which is opposite to the predicted hypothesis that firms with higher agency costs will have higher levels of disclosures. We consider that logNSHARES is likely to be proxying for size rather than agency costs of equity. There is no univariate support for deed adoption relating to market to book (capital market incentives for disclosures), the proportion of non-executive directors (a proxy for directors' liability), and organizational factors relating to complexity and to control.

A correlation matrix is provided in Table 5. As expected, factor analysis reduces the correlations among the organizational variables. The number of subsidiaries (NSUB) is highly correlated with several variables (NSEG, MKTBK, LEV, NSHARES, vertical integration factor and the complexity factor). This raises a possible multicollinearity problem, which is addressed in the sensitivity analysis below.

Multivariate Tests

(DEED), the binary dependent variable indicates that a logit regression is an appropriate multivariate model. Table 6 reports two models. Model 1 includes all variables and model 2 omits the number of subsidiaries (NSUB). Model 1 explains 77 per cent of deed usage. The results show that the logNSUB variable (proxying for accounting and auditing benefits) is significant at the 0.01 level and the main driver of the deed use. The rank of the concentration ratio (rankCR4), the interaction between rankCR4 and the number of segments (NSEG), the complexity factor are significant at the 0.05 level. Market to book ratio (MKTBK) and the number of shares outstanding (logNSHARES) are significant at the 0.10 level. The coefficient on the MKTBK ratio is, however, opposite to the hypothesized direction. Rather than picking up the need for external financing, the growth option variable is possibly proxying for ‘knowledge’ interdependence (Mian and Smith, 1990). The proportion of non-executive directors (PNED) and the organizational factors relating to complexity and control are not significantly associated with deed use.

Further insight into non-disclosure is to be gleaned from Table 7, which reports the estimated probabilities of adopting the deed from model 1 at specific points in the distribution of selected explanatory variables. For each variable, observations are ranked and grouped in quintiles. For each quintile the median value for the variable and the estimated probability from model 1 are reported.

For the number of subsidiaries (logNSUB), the probability of deed adoption is monotonic across all quintiles. Hence, the benefits of non-disclosure by adopting the deed are linear. For the concentration ratio (rankCR4) the probability of non-disclosure is constant across the first four quintiles (probability range of 0.521 to 0.638) but drops to 0.383 in the fifth quintile. Hence, it is only when competition is very strong (i.e., where the concentration ratio is low) that non-disclosure is an important determinant. This suggests that future research ought to consider converting the concentration ratio into a dummy variable in order to capture proprietary costs. The number of segments (NSEG) is also nearly monotonic across all quintiles.²³ Hence, it is not simply the existence of competition that drives firms to reduce disclosures at the subsidiary financial statement level, but also the ability to retain non-disclosure in the segment report at the consolidated financial statement level. For leverage (LEV) the impact on deed adoption is evident at the tails. That is, the probability of deed adoption is almost constant from the second to the fourth quintiles (between 0.511 to 0.682). However, when there is no debt there is very little incentive to adopt the deed (0.239) and with high debt the fifth quintile probability is 0.704. For market to book (MKTBK), shares outstanding (logNSHARES) and the organizational factors no obvious pattern exists in the probability of deed adoption (and therefore the results are untabulated).

Sensitivity Analysis

A number of alternative specifications for the explanatory variables are explored. We examine several measures of market concentration; a two-firm concentration ratio based on a four-digit industry classification and equivalent two-firm and four-firm Herfindahl index were estimated. There is a high correlation between these measures (i.e., greater than 0.90) and the results are insensitive to the type of industry concentration measure employed or whether ranks or raw measures are used.²⁴ We also replaced the concentration ratio with a capital intensity variable as a proxy for barrier to entry. Following Leuz (2004), capital intensity is measured as the ratio of property, plant and equipment to total assets. The capital intensity variable, and its interaction with the number of segments, is not statistically significant at the 0.10 level. In this regression market to book is no longer significant, but leverage is significant at the 0.05 level.

Scott (1994) and Leuz (2004) use trading volume to capture the firm's benefits from providing disclosures to the capital market. Hence, in model 1 we replaced the market to book ratio (MKTBK) with the log of the total number of shares traded in the year 2002 scaled by market capitalization.²⁵ The trading volume variable is not significant in either the univariate or multivariate tests. The MKTBK variable is a proxy for the financing needs of the firm. That is, a firm that has financing needs will not reduce disclosures because of pressure from capital markets to increase disclosure. In model 1, MKTBK is replaced with ex post indicator variables for the increase in debt and increase in shares outstanding. Neither of these variables is significant at conventional levels. However, their inclusion does increase the significance of the leverage variable to the 0.05 level. There is little change in the significance of other variables.

As leverage (debt capacity) is related to assets in place, which might be industry specific, leverage in relation to industry averages is also measured to show whether the leverage of an economic entity with a closed group is greater than the median leverage for firms in the same industry. The results are insensitive to whether industry adjusted or book denominators are used in estimating leverage.

Each of the organizational factors was replaced with the highest loading variable in each factor. That is, the organizational factors were replaced with the proportion of major segment revenue (MSEGR), foreign segment revenue (FSEGR) and parent company debt (PDEBT). The logit regression has a lower explanatory power and the percentage of correct predications falls to 72.3 per cent. The market to book ratio (MKTBK) and foreign segment revenue (FSEGR) are no longer significant at the 0.10 level.

A concern over multicollinearity was also raised because several strong bivariate correlations among the explanatory variables were observed in Table 5. In particular, logNSUB was correlated with a number other variables. An OLS regression using the same variables as model 1 yielded similar levels of significance as in Table 6 for each of the variables. Variance inflation factors in the OLS analysis indicate that, as expected, there is high collinearity among rankCR4, NSEG and the interaction term. The highest VIF on the remaining variables is 1.7 for logNSUB. Model 2, Table 6 reports the logit regression results with logNSUB omitted. The result for model 2 is similar to model 1, except that leverage is significant and MKTBK is not. The model has a lower explanatory power and classification accuracy and hence is robust with regard to concerns of multicollinearity.