BOARD STRUCTURE AND SUPPLEMENTARY COMMENTARY ON THE PRIMARY FINANCIAL STATEMENTS

Purpose This research investigates the relationship between the extent and focus of supplementary narrative commentary (SNC) on amounts reported in the primary financial statements and board structure variables. Design/Methodology/Approach The study uses the disclosure index methodology to measure the extent of SNC in annual reports of 167 FTSE 250 companies. Ordinary Least Squares (OLS) regression analysis is employed to examine the association between the extent and focus of SNC and board structure variables. Findings The findings show that the extent of SNC on amounts reported in the primary financial statements is about 30%, suggesting that companies provide commentary on a small number of amounts reported in the financial statements. In terms of focus of SNC, companies provide greater SNC on amounts in the income statement relative to the balance sheet. The regression results indicate that the extent of SNC is negatively associated with board size, and positively associated with audit committee independence and financial expertise. Focus of SNC is negatively related to audit committee independence and finance expertise. Originality/Value The research contributes to both the voluntary disclosure and impression management literature streams. The findings provide evidence of the extent and focus of SNC on amounts in the financial statements. They also demonstrate that board structure variables are related to the extent and focus of SNC on amounts in primary financial statements. These findings have implications for policy makers who have responsibilities for ensuring that users of annual reports receive adequate information to make decisions.


Introduction
In this paper, we investigate supplementary narrative commentary (SNC) on amounts reported in the primary financial statements 1 using a sample of UK listed companies. We Similarly, the academic literature also argues that SNC is important to investors as it provides a means of clarifying the financial statements numbers (Cole and Jones, 2005). Abarbanell and Bushee (1998) and Cole and Jones (2005) argue that explanations on changes to amounts reported in the primary financial statements compared to previous year provide market participants with better insights into whether the changes represent good or bad news. For example, an increase in debt might either be good news (Harris and Raviv, 1990) or bad news (Miller and Rock, 1985). To the extent that no supplementary commentary is provided on the increase in debt, investors would not know the reasons for such an increase. Other studies show that supplementary commentary affects share price returns (e.g., Kyeyune, 2010;Tauringana et al., 2011) and helps private unsophisticated users to understand the financial results better because it is couched in non-technical language (Bartlett and Chandler, 1997). 2 In spite of this, in the UK, management has considerable discretion regarding supplementary commentary provision. The only requirement by the Companies Act 2006 (section 417) states that the business review must, 'where appropriate, include references to, and additional explanations of, amounts included in the company's annual accounts' without further guidance on which or how many of the amounts should be commented on. This discretion provides an interesting setting to investigate because there are likely to be greater variations in commentary among companies.
There is much literature investigating disclosure in different media, and the impact of corporate governance and company-specific factors on such disclosures. For example, some studies have examined aggregate disclosure (e.g., Barako et al., 2006;Haniffa and Cooke, 2002). Others have focused on specific disclosure types such as management earnings forecasts (Karamanou and Vafeas, 2005); social information (Said et al., 2009); environmental information (Liu and Anbumozhi, 2008); and intellectual capital information (Li et al., 2008). These studies provide important insights into disclosure practices and their relation with corporate governance factors, but the results are mixed and suggest that different types of disclosure may be influenced by different factors (see Barako, 2007).
Our study extends the literature by focusing on SNC, an important type of disclosure which is useful to the capital markets (ASB, 2006;Tauringana et al., 2011), but for which evidence is limited. Although the impression management studies examined supplementary commentary, they have focused mainly on revenues, profit and earnings per share (see Aerts, 1994;Clatworthy and Jones, 2003) so that there is no comprehensive evidence on SNC. Additionally, these studies have focused on examining the self-serving behaviour of managers in explaining performance. In this regard, our study has three objectives. First, we examine the extent to which UK companies provide SNC on amounts reported in the primary financial statements.
Second, we analyse the extent to which there is SNC focus in the commentaries on the primary financial statements. We define SNC focus as the propensity to provide SNC on more amounts reported in one primary financial statement relative to the other. This is supported by the impression management literature (Abrahamson and Park, 1994;Clatworthy and Jones, 2003), which suggests that managers have incentives to use corporate reports to 'manipulate the perceptions and decisions of stakeholders' (Yuthas et al., 2002, p. 142). In this context, we argue that managers may focus their commentary on amounts in the financial statement that shows good performance, thus portraying themselves in good light. Finally, we examine whether differences in the extent and focus of SNC can be explained by differences in corporate governance mechanisms, in particular board structure variables. In this context, we investigate whether boards of directors take actions that increase SNC, but curtail SNC focus.
Our sample consists of 167 companies selected from the London Stock Exchange (LSE)'s FTSE 250 index. We measure SNC using the disclosure index methodology (e.g., Haniffa and Cooke, 2002;Mangena and Pike, 2005). We find that approximately 30% of the amounts are commented on. This implies that the absence of specific requirements by the Companies Act 2006 and/or guidance by the ASB (2006) means that users of annual reports may be denied useful information. In terms of SNC focus, we observe that companies comment more on income statement amounts relative to balance sheet amounts, suggesting that there is propensity to explain financial performance more than financial position. After controlling for company size, age, analyst following and industry, the regression results indicate a positive relation between the extent of SNC and audit committee independence and financial expertise, but negative relation with board size. We find no significant relationship between SNC and the proportion of non-executive directors. Our results show that SNC focus is negatively related to audit committee independence and financial expertise, but has no significant relation with board size and proportion of non-executive directors. Overall, these results are consistent with previous research suggesting effective boards improve the quality of financial reporting (e.g., Mangena and Pike, 2005;Karamanou and Vafeas, 2005).
Our study contributes to the literature in a number of ways. First, we provide the first time evidence on the extent and focus of SNC on amounts reported in the primary financial statements. Second, we contribute to the literature by providing evidence on the relationship between the extent and focus of SNC and board structure variables. These issues have not been addressed in prior literature.
The rest of the paper is organised as follows. In Section 2, we summarise the literature and develop the hypotheses in Section 3. We discuss the research methods in Section 4, and in Section 5, we present and discuss the empirical findings. The final section is a summary and conclusion. Barton and Mercer (2005) suggest that SNC is an important element of a company's disclosure policy. There are many reasons why managers might want to provide supplementary narrative commentary. Merkl-Davies and Brennan (2007) group the reasons into two perspectives. One perspective is that SNC overcomes information asymmetries between managers and outsiders (Healy and Palepu, 2001;Merkl-Davies and Brennan, 2007). Information asymmetry arises from the separation of ownership from control in companies (Jensen and Meckling, 1976), which increases agency costs (Healy and Palepu, 2001). Diamond and Verrecchia (1991) suggest that enhanced disclosure reduces information asymmetry. In this context, the provision of SNC can be seen as an attempt by managers to reduce information asymmetry, and helping improve investor understanding of the results, thus increasing the liquidity of the shares and lowering the cost of financing (Barton and Mercer, 2005).

Literature Review
The second perspective is that managers have incentives to provide supplementary commentary to exploit the information asymmetries and engage in impression management (Clatworthy and Jones, 2003;Barton and Mercer, 2005).
Impression management refers to the process by which individuals attempt to control the impression of others (Merkl-Davies and Brennan, 2007). In corporate reporting, impression management occurs when management selects information to display and presents the information in a manner that distorts readers' perceptions of corporate achievements (Merkl-Davies and Brennan, 2007). This implies that management behave opportunistically and provide self-serving information to enhance the capital market perceptions of their abilities and company's prospects (e.g., Clatworthy and Jones, 2003). In the context of SNC, managers may have a propensity to provide more SNC on amounts in the financial statement that show them in good light.

Hypotheses development
The board of directors is responsible for monitoring managerial performance in general (Karamanou and Vafeas, 2005), and in particular, the financial reporting processes (Mangena and Pike, 2005). In the context of financial reporting, the UK Corporate Governance Code (FRC, 2012, p. 17) Schaffer (2002) and Barton and Mercer (2005) also suggest that SNC can help the board of directors in their monitoring and evaluation of managerial performance. For example, the board might use them to assess managerial performance compared to previous year as well as monitor or judge achievement of objectives set out. Consequently, we argue that effective board structures (i.e., board size, non-executive directors and audit committees) will enhance the extent of SNC and reduce the extent of SNC focus.

Board size
The link between board size and disclosure has been investigated by only a few studies (e.g. Cheng and Courtenay, 2006;Said et al., 2009). The argument is that larger boards are more likely to be endowed with a greater range of expertise, thus enhance effective monitoring of management actions (Karamanou and Vafeas, 2005).
In contrast, the benefits of large boards may be offset by the incremental cost of poorer communication and decision making inefficiencies that are associated with large groups (Dalton et al., 1999). This means that with dispersed opinions and noncohesiveness in viewpoints, boards that are too large may actually have diminished monitoring capabilities (Cheng and Courtenay, 2006). These contrasting views suggest that the effects of board size on the extent and focus of SNC are not clear.
However, the limited empirical evidence suggests that there is no significant relation between board size and voluntary disclosure (see Karamanou and Vafeas, 2005;Said et al., 2009). Nevertheless, given the contrasting views, we make no directional prediction on the relation between board size and SNC. Thus:

H1:
There is an association between board size and the extent of SNC on amounts reported in the primary financial statements.

H2:
There is an association between board size and the extent of SNC focus on amounts reported in the primary financial statements.

Proportion of Non-executive Directors
A number of studies have linked the proportion of non-executive directors (NEDs) to disclosure (e.g., Eng and Mak, 2003;Mangena and Tauringana, 2007) and management earnings forecasts (e.g., Ajinkya et al., 2005;Karamanou and Vafeas, 2005). These studies suggest that NEDs have incentives to protect shareholder interests against managerial opportunism. In this respect, we suggest that NEDs may take actions that encourage more SNC on amounts in the primary financial statements (1) to comply with the requirements of the Companies Act 2006 and the recommendations of the ASB (2006) and (2) to aid their assessment of managerial performance (Schaffer, 2002;Barton and Mercer, 2005). In addition, given their responsibility to provide a balanced assessment of the financial performance and position, NEDs are more likely to discourage SNC focus.
Empirically, some studies support a positive relation with disclosures (see Karamanou and Vafeas, 2005;Ajinkya et al., 2005). Others show no significant relation with interim disclosures and statements of best practice (e.g., Mangena and Pike, 2005;Mangena and Tauringana, 2007). Nevertheless, we hypothesise the following:

H3:
There is a positive association between the proportion of non-executive directors and the extent of SNC on amounts reported in the primary financial statements

H4:
There is a negative association between the proportion of non-executive directors and the extent of SNC focus on amounts reported in the primary financial statements.

Audit Committee
The board of directors performs its monitoring activities by delegating important oversight duties to relevant board committees (Mangena and Pike, 2005;Karamanou and Vafeas, 2005). In terms of financial reporting, the audit committee (AC) has the delegated responsibility to ensure the quality of the financial reporting processes (Smith Committee, 2003;Mangena and Pike, 2005). The Smith Committee (2003) suggests that the effectiveness of the AC is enhanced when the AC is independent and has financial expertise. The issue of AC independence draws from the widely accepted notion that independent directors are more likely to be effective monitors of management actions (e.g., Fama and Jensen, 1983). Carcello and Neal (2003) argue that an independent AC is more likely to be free from management influence in ensuring the quality and credibility of the financial reporting process. Given the importance of SNC to capital market participants (see ASB, 2006;IASB, 2010), we suggest that an independent AC would enhance the provision of this type of information and reduce the level of SNC focus. Empirical evidence shows a positive relationship between AC independence and disclosure (Mangena and Pike, 2005;Mangena and Tauringana, 2007) and management earnings forecasts (Karamanou and Vafeas, 2005;Ajinkya et al., 2005). Thus we hypothesise:

H5:
There is a positive association between audit committee independence and the extent of SNC on amounts reported in the primary financial statements

H6:
There is a negative association between audit committee independence and the extent of SNC focus on amounts reported in the primary financial statements.
The UK Corporate Governance Code (FRC, 2012) recommends the AC should comprise members with knowledge of the business environment, and, at least one AC member should have recent and relevant financial experience. Knowledgeable AC members are in a better position to understand (1) the capital market implications of providing quality disclosures (Karamanou and Vafeas, 2005;Mangena and Pike, 2005) and (2) the implications of compliance with regulatory requirements and best reporting practices (Mangena and Tauringana, 2007). This should lead to improvement in disclosure, including the provision of SNC in the financial statements. Prior empirical studies indicate a negative relation between financial expertise and financial statements fraud (e.g., Abbott et al., 2000), earnings management (Klein, 2002); dismissal of auditors after issuing a going concern report (Carcello and Neal, 2003), and a positive relationship with disclosure (Mangena and Tauringana, 2007). Hence:

H7:
There is a positive association between AC finance expertise and the extent of SNC on amounts reported in the primary financial statements

H8:
There is a negative association between AC financial expertise and the extent of SNC focus on amounts reported in the primary financial statements.

Model specification
To examine the association between the board structure variables and the extent and focus of SNC in the primary financial statements, we specify the following two multiple regression models: We define all the variables in Table I. Drawing from prior research, we control for a number of other variables that are linked to disclosure. These are company size (Haniffa and Cooke, 2002); analyst following (Hope, 2003); gearing (Schwartz and Soo, 1996); profitability and liquidity (Barako, 2007) and company age (Li et al., 2008) INSERT  (Karamanou and Vafeas, 2005). In these companies, voluntary disclosures are likely to be more critical in reducing information asymmetries between management and shareholders. Consistent with previous research (e.g., Mangena and Tauringana, 2007), we eliminated 77 financial companies (such as banks, insurance) because they face additional disclosure and corporate governance requirements, which may not be faced by non-financial companies. We then downloaded annual reports for the 173 remaining companies from the individual companies' websites. Following this, we eliminated two companies that were taken over and three that changed their accounting period during the period covered by our study. We expect these events to trigger the provision of more SNC to explain their effects on the amounts in the primary financial statements. We further eliminated one more company because it has negative equity and we considered it to be an outlier. The final sample used in this study is 167 companies.

Measuring supplementary narrative commentary
We measure the extent of SNC using a disclosure index methodology (e.g., Haniffa and Cooke, 2002;Owusu-Ansah, 2005). We first developed a checklist of all the items required to be disclosed on the face of the primary financial statements from an analysis of the requirements of UK and international accounting standards. 3 This process yielded a total of 44 items, as indicated in Table II.

INSERT TABLE II ABOUT HERE
To measure the extent of SNC, a company is awarded a score of '1' for the presence of SNC on the relevant amounts, and '0' otherwise. In identifying the SNC, we read the narrative section of the annual report. Unlike previous studies, the issue of item applicability is not a problem in our study because SNC relates to specific amounts in the primary financial statements. For example, if the item, 'finance costs', is included in the income statement, then the expectation is that it should be commented on and hence applicable to the company. This means that if a narrative commentary is not provided, the company is awarded a '0' score. On the other hand if the item, 'finance costs', is not included in the income statement, then the item is considered not applicable and scored accordingly. After scoring all the items, the SNC score for each company is computed as an index by dividing the company's total score (i.e., the sum of all the 1's) by the maximum possible score for the company. For each company, we create four SNC indices: overall SNC index, income statement SNC index, balance sheet SNC index and the SNC focus index (see Table I for definitions). 4 The scoring of all annual reports was conducted by one well-experienced coder. However, a second coder randomly selected 10% of the sample (17 annual reports) and independently scored the reports. The correlation coefficient between the scores of the two coders was 0.99, suggesting that the scores are reliable.

Descriptive statistics
The summary descriptive statistics for SNC, board and control variables are presented in With regard to the individual primary financial statements, Table III shows that companies provide more SNC on amounts in the income statement relative to the balance sheet. The mean SNC in the income statement is 42.3% compared to 20.1% in the balance sheet. These results are also reflected in our SNC focus measure with a mean of 2.659, reflecting the propensity by managers to explain more income statement amounts relative to balance sheet amounts. This suggests that companies may not be providing a balanced assessment of financial performance and financial position. There are two possible explanations for this focus. One, from the perspective of the impression management literature, management might be commenting on income statement amounts because performance is good, and wants to portray themselves in good light (Clatworthy and Jones, 2003). Alternatively, managers know that the most read financial statement by both analysts and individual investors alike is the income statement (e.g. Bartlett and Chandler, 1997;Black and White, 2003) and therefore are responding to this need for information Table III also presents summary statistics for the board structure and control variables.

INSERT TABLE III ABOUT HERE
The mean board size consists of about nine directors and the mean proportion of nonexecutive directors is 56%. The results also show that on average 81% of AC members are independent and 38% have financial expertise. The percentage of AC members with financial expertise is slightly higher than the 30% reported by Mangena and Tauringana (2007) (using data collected in the 2002 interim reports), suggesting the number of AC members with financial expertise has increased over time. As for the control variables, the mean company size (market capitalisation) is £1,048.9 million and the mean number of analysts is five. The means for the other variables are as follows: gearing ratio (1.07), average profitability (15.4%), liquidity ratio (1.58) and company age (21 years).

Regression results and discussion
In Table IV, we provide the Pearson product-moment correlation matrix among the independent variables.

INSERT TABLE IV ABOUT HERE
An examination of the correlations among our independent variables suggests no multicollinearity problems exist. Field (2005) suggests that multicollinearity becomes a problem only when the correlations exceed 0.80 or 0.90. As Table IV shows, all the correlations are below these threshold values. We also examine the variance inflation factors (VIFs) (not tabulated) to further test for multicollinearity and all are well below the threshold value of 10 (see Field, 2005).
In Table V, we present the results of the four multiple regression models. 5 Our data-set is not normally distributed, and therefore we run these regressions using normal scores (Cooke, 1998). Column 2 of

INSERT TABLE V ABOUT HERE
We first focus on the extent of SNC (columns 2 to 4). With regard to the overall SNC (Column 2), the results in Table V show that board size, AC independence and finance expertise are associated with SNC at the 5% level or better. Thus our hypotheses 1 (H1), 5 (H5) and 7 (H7) are supported. The coefficient of proportion of non-executive directors is, however, not significant, hence hypothesis 3 (H3) is not supported. In terms of the individual primary financial statements, we find that AC independence and financial expertise are positively associated with commentary in both the income statement (Column 3) and the balance sheet (Column 4) at the 5% level or better. The coefficients of board size and proportion of non-executive directors are not significant. Taken overall, our results suggest that companies with effective board structures provide SNC on more amounts in the primary financial statements.
The findings of a negative association between board size and the extent of SNC contradict Cheng and Courtenay (2006) and Said et al. (2009) who report no significant association. Nevertheless, our results support the argument that larger boards are not effective monitors of management (Dalton et al., 1999), and may not enforce high level of commentary on amounts in the primary financial statements. Our results indicating that AC independence and financial expertise are positively associated with SNC are consistent with previous studies (e.g., Kelton and Yang, 2008). This implies that AC independence and expertise are important in ensuring transparency and improve monitoring of financial reporting (Carcello and Neal, 2003). Additionally, the results support the recommendations contained in the UK Corporate Governance Code (FRC, 2012) that ACs should be composed of independent non-executive directors and members with financial expertise.
For the control variables, we find that company size is related to overall, income statement and balance sheet SNC consistent with previous studies (e.g. Mangena and Pike, 2005;Li et al., 2008). Analyst following is not related to overall supplementary commentary, but is negatively related to the balance sheet statement and positively related to income statement supplementary commentary. Gearing ratio and company age are related to balance sheet supplementary commentary at the 10% level or better and 1% level or better, respectively. We find no significant relation with profitability and liquidity.
Finally, in respect of SNC focus (Column 5), we find that the coefficients of AC independence and financial expertise are negative and significant at the 5% level or better. Hence, our hypotheses 6 (H6) and 8 (H8) are supported. However, both board size and proportion of non-executive directors are not significantly related to SNC focus, thus hypothesis 2 (H2) and 4 (H4) are not supported. These results are interesting and suggest that when the AC is independent and has financial expertise, the propensity for managers to provide commentary on more amounts in the income statement relative to balance sheet amounts may be curtailed. In this case, the results suggest that the presence of AC members who are independent and with financial expertise might result in a balanced commentary on amounts in the primary financial statement. In terms of the control variables, SNC focus is negatively related to company size and company age at the 5% and 1% level or better, respectively. The relation between analyst following and supplementary narrative focus is positive at the 1% level or better, implying that companies with large analysts following provide more commentary on amounts in the income statement. This perhaps reflects the attempt by management to provide information to meet the demands of analysts (see Barker, 1998;Mangena et al., 2007). Overall, these results suggest that the propensity to provide SNC on more income statement amounts relative to balance sheet amounts is lower in companies that (1) have effective AC structures, (2) are large, and (3) have a long history of stock exchange listing, but higher in companies with greater analyst following.

Summary and Conclusion
The focus of this study was to investigate the extent and focus of SNC on the amounts reported in the primary financial statements in the annual reports. We also investigated the relation between the extent and focus of SNC and board structure variables, in particular board size, proportion of non-executive directors, audit committee independence and financial expertise. We find that companies provide SNC on a small number of amounts in the primary financial statements. Our results show that companies are more inclined to provide SNC on amounts reported in the income statement relative to the balance sheet amounts, raising doubts about whether companies provide a balanced assessment of company performance and position. We find that SNC is negatively related to board size, but positively related to AC independence and financial expertise. We also find that AC independence and financial expertise are negatively related to SNC focus, suggesting that effective audit committees may encourage a more balanced supplementary commentary.
These results should be interpreted in the light of the fact that our model has explained only a very small amount of the variation in the extent and focus of SNC.
Also while the use of SNC mitigates the problem of item applicability associated with previous disclosure studies, the limitation of our approach is that we assume that companies have to provide commentary on all amounts in the primary financial statements. It is possible that an amount in the financial statements has remained largely unchanged, and with no material changes, managers might perhaps find little need to comment. Further, the study examined the extent and focus of supplementary narrative commentary using annual reports only. Potentially, some of the changes to amounts in the financial statements may have been explained through other media.
In spite of these limitations, our study makes some important contributions to disclosure literature. First, the study provides the first evidence on the extent and focus of SNC on amounts in the primary financial statement. To the best of our knowledge, there is no existing study that specifically examines supplementary Notes 1 ASB (1999), Chapter 7, identifies primary financial statements as the income statement, the statement of recognised gains and losses, balance sheet and cash flow statement. 2 Unlike sophisticated investors such as analysts and institutional investors, who have access to management (see Holland, 1998), unsophisticated investors have no access to management to ask for explanations of changes to amounts in the primary financial statements. Hence they rely on supplementary commentary given in the annual report to appreciate the reasons for changes in company financial position and performance. 3 Although UK FTSE250 companies adopted IFRS on 1 January 2005, they continued to prepare a statement of recognised gains and losses. This was in line with the ASB (1999) Statement of Principles for Financial Reporting. For the period covered by the study, there were no requirements by IAS to prepare a statement of gains and losses. However, effective January 2009, IAS 1 (revised) requires a statement of comprehensive income. 4 In our analyses, we incorporated the items in the statement of recognised gains and losses into the income statement and those in the cash flow statement into the balance sheet. Our reason for this is that the number of items in each of these statements is very small. 5 We re-run these regressions including industry dummies following Mangena and Tauringana (2007) and our results remain similar. Defined as the propensity to provide supplementary narrative commentary more on amounts in one primary financial statement than the other, and is measured as the income statement supplementary narrative commentary score scaled by the balance sheet supplementary narrative commentary score for the company.

BDSZ
Board size Board size is measured by the total number of directors on the board as at the end of the year.

NEDS Proportion of non-executive directors
Proportion of non-executive directors is measured by the number of non-executive directors on the board scaled by total number of directors on the board (as a percentage). We do not distinguish between independent and non-independent non-executive directors.

ACIN AC independence
Audit committee independence is measured by the number of independent non-executive directors on the audit committee scaled by the total number of directors on the audit committee (as a percentage) ACFE AC financial expertise Audit committee financial expertise is measured by the proportion of the members of the audit committee who are identified as having financial expertise scaled by the total number of directors on the audit committee (as a percentage). We consider audit committee members to have financial expertise if they hold an accounting or finance qualification.

COSZ
Company Size The market capitalisation of the company at the financial year end. Market capitalisation is computed by multiplying the outstanding ordinary shares by the market share price at the balance sheet date.  Gross profit from continuing operations 4

ANFG
Other Income, e.g. profit from sale of plant and equipment 5 Operating costs -administrative and selling 6 Other Expense -e.g. loss from sale of plant and equipment 7 Finance income 8 Finance costs 9 Profit before taxation 10 Taxation 11 Profit from continuing operations 12 Profit attributed to equity holders 13 Profit attributed to minority interests 14 Earnings per share -basic and diluted 15 Dividend per share Statement of recognised gains and losses   Table 1 ***. Correlation is significant at the 1% level or better **.Correlation is significant at the 5% level or better *.Correlation is significant at the 10% level or better .001 (2.866***) -.022 (-3.180***) *** Significant at the 1% level or better ** Significant at the 5% level or better * Significant at the 10% level or better ¹ All variables are defined in Table I ² t-statistics are in parentheses