Econometric methods

As noted in McInish and Wood (1992), Madhaven (2000) and Stoll (2003), it is established in some of the prior studies that cross-sectional variations in spreads can be explained by economic variables, and that the relationship between the spread of a security and the trading characteristics of that security is one of the strongest and most robust relationships in finance.

Demsetz (1968), for example, find that spreads are positively related to price and volume, whilst Stoll (1978) documents that spreads are positively related to volatility. These studies have also

suggested that firm size can be used as a control variable (Stoll, 2000; 2003).

Following Stoll (2000), the average of each of the underlying variables is taken across all of the days in the overall sample period in order to reduce the errors associated with a single day. This paper investigates the following regression model in order to control for the factors that might be important in determining the spreads:

i earnings management calculated using the method referred to in the previous subsection;

SDRETi is the standard deviation of daily stock returns without dividends. LNTVi is the natural log of average daily trading dollar volume; LNCLPi is the natural log of the mean daily closing stock price; LNMVi is the natural log of the market value of firm i (i.e. the closing price at the end of sample period multiplied by the company's common shares outstanding), and LNTRi is the natural log of the average daily total number of trades.

The accounting accruals are adjustments to cash flows, and should sum to zero over the life of a corporation. If managers manipulate the earnings of a firm, the effects of the manipulation will ultimately unwind and such corrections will subsequently be reversed, or realized, as the same account; as such, upwardly managed accruals are expected to be followed by downwardly managed accruals. Chan, Jegadeesh, and Sougiannis, (2004)

demonstrate that earnings management causes a negative relationship between current accruals and future earnings; hence, the absolute value of discretionary accruals will be positively autocorrelated.

When information asymmetry is high, shareholders do not have sufficient resources to monitor managers’ actions; thus, earnings management can occur. Richardson (2000) shows that the proxy variable for information asymmetric costs (the bid-ask spread) could affect earnings management; hence, in the model, earnings management is endogenous³, which demonstrates that an instrumental variable estimation procedure is more appropriate for the model structure. Following the prior studies, this paper adopt ‘debt to asset ratio’, ‘quarterly operating cash flow volatility’, ‘firm size’, ‘market-to-book ratio’, ‘net revenue growth’ and

‘return on assets’ as the instrumental variables used to estimate Equation (7).

The empirical research design of the relationship between earnings management and equity liquidity comprises of two parts. Firstly, since the main research interest of this research focuses on the effects of earnings management on equity liquidity especially in the periods of corporate financial crises, this paper considers the estimation of Equation (7) for the period from October to December 2001. In order to ensure the timely disclosure of operating results and financial position, firms must announce interim financial report and any unscheduled material events or corporate changes that are deemed to be of importance to

3 To test for endogenous relationship, this study performs the Hausman-Test. The test results reject the null hypothesis and instrument variable is suitable for Equation (7).

shareholders or the SEC during the accounting year. For example, listed firms must announce quarterly financial reports and monthly sales. Firms may also hold investor conference and earnings conference calls which provide a setting to study the implications of public information’s complementary role. The sample periods of our study are last quarters of 2001 and 2002 which are close to the end of the two fiscal years. Much accounting information of the accounting year (for example: the 3^rd quarterly accounting report and earnings conference calls) have been released by firms and been interpreted by market participant.

Furthermore, to some extent, earnings management is an overall accounting arrangement, and some time is required for the discretionary accruals to be adjusted. For example, if managers manipulate their earnings, the effects of such manipulation will ultimately unwind and eventually be reversed at the same amount, albeit coming into play during subsequent periods. Consequently, those firms with higher earnings management in the previous period may still have relatively higher earnings management in the near future.

Accordingly, although market makers and liquidity suppliers do not have any access to explicit financial reports of firms and any public information on abnormal discretionary accruals during the sample period, they can use interim financial information to conclude the degree of earnings management. Since simultaneity may well exist between earnings management and the liquidity of a firm, this study conducts a simultaneous equation model and estimate Equation (7) by the three stage least squared (3SLS), which uses ‘debt to assets

ratio’, ‘quarterly operating cash flow volatility’, ‘firm size’, ‘market-to-book ratio’, ‘growth opportunity’ and ‘return on assets’ as the instrumental variables.

Finally, in order to investigate the impact of earnings management on the adverse selection components of the equity bid-ask spread, this paper also tests the relationship between earnings management and the information asymmetry component using the GKN approach.

3.5 Instrumental variables

3.5.1 Leverage

Firms facing financial constraints or debt covenants may have incentives to manage earnings in order to avoid financial problems and potential penalties. DeAngelo, DeAngelo, and Skinner (1994) report evidence of abnormal accruals when firms are faced with debt covenants, whilst Dechow, Sloan, and Sweeney (1996) also demonstrate the need for the avoidance of debt covenants as a motivation for earnings manipulation.

Although this debt-covenant scenario will predict a positive relationship between earnings management and financial distress, if lenders are to closely monitor the earnings of highly indebted firms, the degree of earnings management may decline with financial leverage. This paper therefore adopts financial leverage, determined by the ratio of total debt to total assets, as a proxy for the proximity to covenants and the association with the

existence and rigidity of such covenants.

3.5.2 Operating cash flow volatility

When firms are faced with operating risk or economic shocks, there may be a tendency amongst managers to conceal the underlying performance of such firms by using their accounting discretion to buffer cash flow shocks and to smooth the reported operating earnings; i.e., to effectively reduce the volatility of the firm’s reported earnings (Leuz et al., 2003). Managers may decide to accelerate future revenues or to delay current costs in order to hide poor pre-management earnings. Conversely, they may underreport current revenues or accelerate current costs when pre-management earnings are high. Accordingly, high volatility in pre-management earnings is accompanied by high abnormal accruals. Following Peasnell, Pope, and Young (2005), this paper uses operating cash flow as the instrument for pre-management earnings in order to capture the economic performance of the firms, and then use the volatility of operating cash flow as a proxy for the firms’ risk (Richardson, 2000).

3.5.3 Firm size

Zmijewski and Hagerman (1981) provide evidence to show that political costs increase with firm size. Managers of large firms may have greater incentives to manipulate earnings in order to reduce costs; on the other hand, since they are actively followed by outside capital

markets, such firms may be less able to hide earnings management behavior, since the level of information transparency and disclosure increases with firm size. This study therefore uses the logarithm of total assets as a proxy to capture the size of the firm and its information environment.

3.5.4 Growth opportunity

Given that it is much more difficult to scrutinize the activities of rapidly-growing firms, it is much easier for such firms to manage their earnings than slower-growing firms. Dechow et al.

(1996) demonstrate that those firms which are alleged to have violated GAAP by overstating their reported earnings have higher market-to-book ratios vis-à-vis a control group, and suggest that investors expect these firms to have higher growth opportunities. Park and Shin (2004) also find that earnings management has a positive correlation with the growth opportunities for a firm, whilst Richardson (2000) posits that earnings management has a positive correlation with sales growth. Accordingly, this study measures the current and future growth opportunities for a firm by net revenue growth and market-to-book ratio, respectively.

3.5.5 Operating performance

Given that, in many companies, the stock price and managers’ compensation are tied to earnings performance (Holthausen et al., 1995; Lee, Li, and Yue, 2006), this may motivate managers to

engage in earnings manipulation; a positive relationship between discretionary accruals and the profitability of firms is also found by McNichols (2000) and Lee et al. (2006). Following that, this study adopts ROA as a proxy to capture the performance of firms.

Chapter 4 Empirical results

4.1 Descriptive statistics

Table 1 reports the average values and other summary statistics for percentage spread (PSP), the measure of earnings management (EM) and other firm-specific trading and financial variables. The final sample in pre-SOX period comprised of a total of 999 firms in 44 industries. Among the sample firms, 537 corporations are listed on the NASDAQ and 457 (5) corporations are listed on the NYSE (AMEX). EM is approximately 6.72 per cent of total assets, ranging between 61.53 per cent and 0.01 per cent. The highest average percentage spread value is 4.27 per cent, while the lowest percentage spread is 0.04 per cent.

Table 2 presents the correlation coefficient matrix of the variables, thereby providing some basic analysis of the correlation between variables. A preliminary analysis of the details provided in Table 2 indicates that percentage spread (PSP) has a positive correlation with EM.

Furthermore, this study finds that percentage spread (PSP) has a negative correlation with trading characteristic variables, such as trading dollar volume (LNTV), number of trades (LNTR), closing price (LNCLP) and market value (LNMV), but that it has a positive correlation with standard deviation of returns (SDRET). EM has a negative correlation with firm leverage (LEV) and profitability (ROA), but that it has a positive correlation with cash flow volatility (CFVAR) and market-to-book ratio (MB).

Table 1 Descriptive statistics of the variables

a EM is the absolute value of the discretionary accruals for the year 2001; PSP is the average percentage spread for the

sample period, 1 October to 31 December 2001.; SDRET is the standard deviation of daily stock returns; LNTV is the natural log of average daily trading dollar volume; LNCLP represents the natural log of the average closing stock price; LNTR is the natural log of the average daily total number of trades; LNMV is the natural log of the market value of the firm at the end of 2001; LEV is the debt to total asset ratio; LTA represents the natural log of total assets at the end of 2001; MB is the market to book ratio (i.e., the closing price at the end of 2001 multiplied by the company’s common shares outstanding, divided by common equity as reported); CFVAR represents the standard deviation of quarterly operating cash flows over the past three years divided by the average quarterly operating cash flows over the period; ROA is the return on assets for 2001; and GROWTH is the net revenue for 2001 less the net revenue for 2000, divided by the net revenue for 2000. Total number of sample = 999.

Table 2 Variable correlations

Correlation Coefficients ^b Variables ^a

PSP SDRET LNTV LNCLP LNTR LNMV LEV LTA MB CFVAR ROA GROWTH EM 0.054* 0.201*** 0.045 -0.114*** 0.093*** -0.016 -0.063** -0.028 0.102*** 0.091*** -0.278*** -0.015

PSP 0.419*** -0.696*** -0.679*** -0.531*** -0.668*** -0.044 -0.556*** -0.200*** 0.132*** -0.279*** -0.008 SDRET -0.096*** -0.540*** 0.159*** -0.387*** -0.146*** -0.394*** -0.029 0.191*** -0.443*** 0.023 LNTV 0.592*** 0.909*** 0.8759*** -0.019 0.678*** 0.403*** -0.133*** 0.144*** 0.063**

LNCLP 0.374*** 0.643*** 0.018 0.465*** 0.303*** -0.183*** 0.356*** 0.088***

LNTR 0.693*** -0.128*** 0.469*** 0.401*** -0.086*** -0.009 0.055*

LNMV 0.042 0.803*** 0.356*** -0.215*** 0.285*** 0.015

LEV 0.376*** -0.077** -0.070*** -0.02415 0.013

LTA 0.002 -0.150*** 0.189*** -0.107***

MB -0.119*** 0.020 0.191***

CFVAR -0.129*** 0.019

ROA(%) 0.055*

Note:

*** indicates significance at the 1 per cent level; ** indicates significance at the 5 per cent level; and * indicates significance at the 10 per cent level.

4.2 Effects of earnings management on equity liquidity

4.2.1 Evidence from the Enron crisis period

The first sample period in this study covers the period from 1 October 2001 to 31 December 2001. This study examines the effects of earnings management on equity liquidity based on the EM for the year 2001, which could provide tests on the information content of the EM for the present year.

Although the main proposition in this paper is that earnings management positively affects the bid-ask spread, Richardson (2000) demonstrates that the bid-ask spread has a positive effect on the earnings management behavior of a firm; in other words, discretionary accruals could be simultaneous. This paper therefore estimates the regression model using the three-stage least squared (3SLS) method. Table 3 (Table 4) presents the regression results for the simultaneous equation model on EM and the percentage spread (stock turnover) for the Enron crisis period.

Table 3 3SLS regression results of EM and PSP, October-December 2001

a The dependent variable in the first equation is the average percentage spread (PSP) for the sample period, 1

October to 31 December 2001. The dependent variable in the second equation is the absolute value of the discretionary accruals for the year 2001. The variables are defined as in Table 1.

Table 4 3SLS regression results of EM and TURNOVER, October-December 2001

a The dependent variable in the first equation is stock turnover (TURNOVER) for the sample period, 1

October to 31 December 2001. The dependent variable in the second equation is the absolute value of the discretionary accruals for the year 2001. The variables are defined as in Table 1.

Stoll (2000) notes that several empirical studies had demonstrated that market structure appeared to have an effect on spreads. Van Ness, Van Ness, and Warr, (2002) also find the adverse selection is actually higher for NASDAQ stocks than for NYSE and AMEX stocks. This paper therefore further tests the relationship between earnings management and equity liquidity by adding in an extra market structure dummy variable, DEXCHi which is equal to 1 if firm i is listed on the NASDAQ and zero if the firm is listed on the NYSE or the AMEX.

As reported in Table 3, the positive relationship between earnings management and percentage spread persists, with the coefficient of DEXCH being significantly positive, indicating a significant difference in the percentage bid-ask spreads between the two market structures. Table 4 presents the results of the effects of EM on the TURNOVER. The coefficient of EM is significantly positive at the 0.01 level in

pre-SOX period, indicating that firms with high EM will have reduced stock trading turnover. The regression results for earnings management, which explain the variations in absolute discretionary accruals, are presented in Panel B of Table 3 and Table 4. Earnings management is positively (negatively) related to percentage spread (stock turnover) and significantly different from zero at less than the 0.01 level.

Table 3 also finds that earnings management has a negative correlation with firm leverage (LEV) and a positive correlation with growth opportunities, which is

measured by the market-to-book ratio; these results are consistent with those of Park and Shin (2004). Finally, the coefficient on cash flow volatility (CFVAR) is significantly positive, as in Richardson (2000). All of these results reveal that the simultaneity existing between earnings management and the average percentage spread is statistically significant.

In general, this paper find that after controlling for cross sectional differences in firms’ trading characteristic variables, such as price, volatility, trading value, number of trades, market value and stock exchange differences, the liquidity amongst those companies with higher absolute discretionary accruals are lower. Therefore, the results in Table 3 and Table 4 support the proposition that firms with high abnormal discretionary accruals induce higher equity trading costs.

Table 5 presents the results of the effects of earnings management (EM) on the asymmetric information component of percentage spread. There are fewer observations in Table 5 than in Table 3, essentially because the GKN method fails to produce reliable estimates of the information asymmetry coefficients for some of the companies.

Table 5 3SLS regression results EM and IA, October-December 2001

a The dependent variable in the first equation is the information asymmetry component of percentage spread (IA)

for the sample period, 1 October to 31 December 2001. The dependent variable in the second equation is the absolute value of the discretionary accruals (EM) for the year 2001.IA is the information asymmetry components of the percentage spread. The other variables are defined as in Table 1.

Following Lin, Sanger, and Booth, (1995), the information asymmetry component of percentage spread is calculated as the estimated coefficient of the information asymmetry (1 ) times the percentage spread. The coefficient of EM is significantly ₁

positive at the 5 per cent level, indicating that firms with a high EM will incur higher equity liquidity costs due to the higher degree of asymmetric information that may be perceived by market makers.

The case in which the NASDAQ market dummy variable is added is considered in Model 2 of Table 5, from which it is find that the NASDAQ stocks in the sample appear to have higher information asymmetry costs. The results of the asymmetric information cost effects are similar to those in Model 1.

In summary, the results presented in Tables 3, Table 4 and Table 5 support the argument that earnings manipulation provides a clear signal of aggressive accounting practices with the sole intention of the managers being to obtain certain private benefits, and the rational response of liquidity providers being to widen the bid-ask spreads so as to afford themselves some measure of price-protection.

4.2.2 Evidence from the post-SOX period

The Sarbanes-Oxley (SOX) Act was promulgated on 30 July 2002, largely in response to a number of major corporate and accounting scandals, the effects of which are still being felt throughout the US economy. The Act requires executives, boards of

directors and auditors to take specific measures to bring about greater corporate accountability and transparency. Jain et al. (2008) demonstrate that the above provisions improved market liquidity and the improvements were more conspicuous for the firms with better financial report quality. This indicates that, after the implementation of the SOX Act, market participants attach greater importance to the quality of financial report.

Since section 302 of the SOX Act requires that ‘the CEO and CFO of each issuer shall prepare a statement to accompany the audit report to certify the appropriateness of the financial statements and disclosures contained in the periodic report, and that those financial statements and disclosures fairly represent, in all material respects, the operations and financial condition of the issuer’, there may, as a result, be an increase in the adverse selection costs of earnings management. The effects of the cross-sectional differences in earnings management on equity liquidity are tested for the post-SOX period from 1 October 2002 to 31 December 2002; the results are presented in Tables 6, Table 7 and Table 8.

Table 6 to Table 8 present the results for the post-SOX period, showing that percentage spread and information asymmetry components of the percent spread are both significantly affected by the EM of the same year and that the reverse effect holds. However, the effect of EM on stock turnover is insignificant in Table 7. As

shown in Table 6 and Table 8, percentage spread is positively related to return volatility and market value, while it is negatively related to trading value and closing price (largely in line with the results presented in Table 3 and Table 5). The total number of firms in the sample is larger than the pre-SOX period, and the exchange dummy controlling for the NASDAQ stocks is also significantly positive, indicating that NASDAQ stocks have higher trading costs. The signs of the coefficients across the two periods are the same, with the coefficients of EM and PSP both being

shown in Table 6 and Table 8, percentage spread is positively related to return volatility and market value, while it is negatively related to trading value and closing price (largely in line with the results presented in Table 3 and Table 5). The total number of firms in the sample is larger than the pre-SOX period, and the exchange dummy controlling for the NASDAQ stocks is also significantly positive, indicating that NASDAQ stocks have higher trading costs. The signs of the coefficients across the two periods are the same, with the coefficients of EM and PSP both being