Online Appendix to Voluntary Disclosure and Information Asymmetry: Evidence from the 2005 Securities Offering Reform

Transcription

1 Online Appendix to Voluntary Disclosure and Information Asymmetry: Evidence from the 2005 Securities Offering Reform This document contains several additional results that are untabulated but referenced in our paper. The additional tests are as follows: 1. Table OA1 presents additional tests to identify market conditioning behavior. We examine (i) whether analyst forecasts issued during the pre-seo period are systematically more optimistic than they are in the neighboring windows and those issued for control firms, and (ii) whether management forecasts are systematically greater than preceding analyst forecasts in the pre-seo period relative to the neighboring periods and that of control firms. The idea is that if managers are trying to condition the market before SEOs, they are likely to bias their disclosures upwards to influence analysts to be more optimistic about the firms future prospects. We find no evidence of forecast bias. 2. Table OA2 also presents additional robustness tests related to identifying market conditioning behavior. In the paper, we link subsequent returns (AR) with the relative frequency of good news disclosed and the aggregate news disclosed in the pre-seo window (i.e., Proportion of GN and Sum of SRET) to capture the extent to which managers engage in hyping behavior before the SEO. However, it is plausible that the relation between pre-seo good news disclosures and post- SEO abnormal returns is non-linear. To allow for non-linearity, we verify the robustness of our inferences to using (i) the Sum of SRET, conditional on SRET being positive, (ii) the decile ranks of the Sum of SRET, and (iii) the total frequency of good news disclosure (rather than the proportion of good news disclosures). We find no evidence of market conditioning in any of these tests. 3. Table OA3 presents our market conditioning analyses without using a control sample. To the extent our proxy for abnormal returns (i.e., the firm s return minus the return of the CRSP valueweighted index) eliminates any systematic component of stock returns that is expected given the level of risk, there is no benefit to including a control sample. Table OA3 shows that our inferences are unaffected if we re-estimate our regressions without a control sample. 4. Table OA4 presents the return reversal results and the cost of raising capital results after partitioning the data into WKSIs and non-wksis. Although, we do not have differential 1

2 predictions for the WKSI and non-wksi firms in these tests, for the sake of completeness, we do conduct these tests. Table OA4, Panels A and B show that there is no evidence that either WKSIs or non-wksis condition the market after the Reform. In Panel C, we find that both WKSIs and non-wksis observe a negative stock price reaction to SEO announcements prior to the Reform and a reduction in the negative reaction after the Reform, consistent with a cost of capital reduction for both sets of firms after the Reform. 5. Table OA5 presents the information asymmetry and cost of capital tests after partitioning firms into those that increase good news disclosure before SEOs and those that increase bad news disclosures before SEOs. We find that both the good news and the bad news disclosures are associated with lower information asymmetry and a reduced cost of raising capital, consistent with our main inferences and inconsistent with hyping. 6. Table OA6 examines whether our analyses of good news disclosure prior to SEOs is robust to using two additional measures of good news: (i) the frequency of good news relative to total news and (ii) the (unscaled) frequency of good news. We find that our inferences discussed in the paper are unchanged when we use these alternative measures of pre-seo good news. 7. Figure OA1 presents the frequency of SEOs relative to quarterly earnings announcements. This test follows Korajczyk, Lucas and McDonald [1991], who show that firms make offerings in periods of low information asymmetry, such as after earnings announcements, to reduce the cost of raising capital. The figure shows that the Reform does not lead to changes in the timing of SEOs relative to earnings announcements. 2

3 TABLE OA1 Alternative Tests to Detect pre-seo Hyping This table is analogous to Table 5 in the paper, where the dependent variable is replaced with Analyst Forecast Errors in Panel A, and Management Forecasts Relative to Prevailing Analyst Consensus Forecasts in Panel B. Analyst Forecast Errors is computed as the consensus analyst earnings forecast issued in a given period minus the actual earnings, scaled by the absolute value of actual earnings. We remove observations with zero actual earnings. Management Forecasts Relative to Prevailing Analyst Consensus Forecasts is computed as the difference between management s forecast and the consensus analyst forecast of earnings, scaled by the absolute value of actual earnings. All other variables are as described in the paper (see the notes above Table 5). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Analyses of Analyst Forecast Errors Analyst Forecast Errors Disclosure Type Management Forecasts Press Releases Disclosure Measure (DISC): Proportion of GN Sum of SRET Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM DISC POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET * * LSIZE ** ** ** ** MTB ROA *** *** ** *** 2.73 PDA FREQ Year & Industry Indicators 9.9% 9.9% 9.8% 9.9% 1,484 1,484 1,484 1,484 3

4 TABLE OA1 continued Panel B: Analyses of Management Forecasts Relative to the Prevailing Analyst Forecasts Management Forecasts Relative to Prevailing Analyst Consensus Forecasts Disclosure Type Management Forecasts Press Releases Disclosure Measure (DISC): Proportion of GN Sum of SRET Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM DISC * *** 2.81 POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET LSIZE MTB ROA PDA FREQ *** *** Year & Industry Indicators 7.2% 7.5% 5.8% 5.9% 1,484 1,484 1,484 1,484 4

5 TABLE OA2 Addressing Potential Non-Linearity in the Relation between Pre-SEO Returns and Post-SEO Return Reversals This table is analogous to Table 5 in the paper with the main independent variables of interest replaced with Sum of Positive SRET, Sum of SRET Deciles, and Sum of GN. Sum of Positive SRET is the aggregate cumulative abnormal returns in the three-day window [-1, 1] around disclosure announcements in the pre-seo window, and in the neighboring windows, conditional on the return being positive. Sum of SRET Deciles is the decile rank of the aggregate cumulative abnormal returns in the three-day window [-1, 1] around disclosure announcements in the pre-seo window and in the neighboring windows. Sum of GN is the annualized number of good news disclosures in the pre-seo window and in the neighboring windows, where a disclosure is considered to provide good news if it induces a positive cumulative abnormal return in the three-day window [-1, 1] around the disclosure announcement. DISC is Sum of Positive SRET, Sum of SRET Deciles, or Sum of GN. We demean DISC to center the variable on zero. All other variables are as described in the paper (see the notes above Table 5). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Management Forecasts 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Sum of Positive SRET Sum of SRET Deciles Sum of GN Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM ** ** ** DISC POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET *** *** *** LSIZE MTB ROA * * * 1.66 PDA * * * FREQ Year Indicators Industry Indicators 11.2% 11.2% 11.3% 1,484 1,484 1,484 5

6 TABLE OA2 continued Panel B: Press Releases 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Sum of Positive SRET Sum of SRET Deciles Sum of GN Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM *** ** DISC ** POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET *** *** *** LSIZE MTB ROA *** *** *** 3.54 PDA * * * FREQ Year Indicators Industry Indicators 12.8% 14.3% 12.9% 1,484 1,484 1,484 6

7 TABLE OA3 Return Reversal Tests and SEO Announcement Tests without Benchmark Samples Panels A and B in this table are analogous to Panels A and B in Table 5 in the paper. However, unlike in the paper, the analyses below do not have a control sample. Similarly, Panel C in this table is analogous to Table 9 in the paper with the sole difference that the results below are not benchmarked to a control sample. All variables are as described in the paper (see the notes above Table 5 and 9). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Management Forecast 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic DISC POSTREFORM DISC POSTREFORM QABRET LSIZE MTB ROA * * 1.85 PDA *** *** FREQ Year & Industry Indicators 18.3% 18.1% Panel B: Press Releases 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic DISC POSTREFORM DISC POSTREFORM QABRET * LSIZE MTB ROA ** ** 2.07 PDA * * FREQ Year & Industry Indicators 19.0% 19.0%

8 Panel C: SEO Announcement Returns TABLE OA3 continued Coefficient t -Statistic Coefficient t -Statistic INTERCEPT * * POSTREFORM ** ** 1.98 MTB ANALYST FOLLOWING LSIZE ROA LNOWN INST_HOLDING LNPRC *** ** 2.26 SALES_GR SALE LEVERAGE * 1.93 CAPEX *** 3.16 CASH PRE_RETURN SHARE_PER *** SEC_OFFER Year Indicators Industry Indicators Abnormal Return around SEO Announcement 13.1% 12.7%

9 TABLE OA4 Analyses of Abnormal Returns following SEOs and SEO Announcement Returns for WKSIs and non-wksis Panels A and B in this table are analogous to Panels A and B in Table 5 in the paper. However, unlike in the paper, the analyses below partition the data into WKSIs and non-wksis. Similarly, Panel C in this table is analogous to Table 9 in the paper with the sole difference that the results below are partitioned into WKSIs and non-wksis. All variables are as described in the paper (see the notes above Table 5 and 9). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Return Reversals tests using Management Forecast as Disclosure Measure 18 Month Abnormal Returns (AR ) Type of Firms: WKSIs non-wksis Disclosure Measure (DISC): Proportion of GN Sum of SRET Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM * * DISC POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET ** ** ** ** LSIZE MTB *** *** ** ** ROA * * ** *** 2.59 PDA ** ** FREQ Year Indicators Industry Indicators 14.5% 14.3% 12.1% 12.2%

10 TABLE OA4 continued Panel B: Return Reversals tests using Press Releases as Disclosure Measure 18 Month Abnormal Returns (AR) Type of Firms: WKSIs non-wksis Disclosure Measure (DISC): Proportion of GN Sum of SRET Proportion of GN Sum of SRET Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM ** ** *** DISC POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET *** *** LSIZE * * MTB ** ** ROA ** *** * 1.71 PDA ** *** FREQ Year Indicators Industry Indicators 16.2% 16.1% 13.7% 13.5%

11 TABLE OA4 continued Panel C: SEO Announcement Returns Test Three Days Abnormal Returns around SEO Filing Date Type of Firms: WKSIs non-wksis Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM *** *** *** *** SEOFIRM POSTREFORM ** ** ** ** 2.12 POSTREFORM MTB ANALYST FOLLOWING LSIZE ** ** ROA LNOWN INST_HOLDING * 1.76 LNPRC *** ** SALEGROW * SALE LEVERAGE CAPEX ** 2.02 CASH PRE_RETURN SHARE_PERC * SEC_OFFER ** Year & Industry Indicators 9.5% 10.5% 9.7% 11.4%

12 TABLE OA5 Analyses of Information Asymmetry prior to SEOs and SEO Announcement Returns Partitioned by Good News and Bad News Disclosures Panels A, B, and C in this table are analogous to Table 6 in the paper. However, unlike in the paper, the analyses below partition the data into firms disclosing more pre-seo good news and firms disclosing more pre-seo bad news. Similarly, Panels D and E in this table are analogous to Table 9 in the paper with the sole difference that the results below are partitioned into good news and bad news disclosing firms. To classify firms disclosing good/bad news, we use the median values of our disclosure metrics (i.e., Proportion of GN and Sum of SRET). All variables are as described in the paper (see the notes above Table 6 and 9). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Using the Adverse Selection Component of Bid-Ask Spreads as the proxy for Information Asymmetry ASC_Spread Disclosure Metric for Good/Bad News: Proportion of GN Sum of SRET Type of Firms: Good News Bad News Good News Bad News Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEO SEOFIRM SEOFIRM SEO ** ** ** *** 2.71 SEOFIRM SEO POSTREFORM ** ** ** *** SEO POSTREFORM SEOFIRM POSTREFORM POSTREFORM MTB ANALYST FOLLOWING *** *** 2.85 LSIZE *** *** *** *** ROA *** ** *** * LNOWN INST_HOLDING *** *** *** *** LNPRC *** *** *** *** Year & Industry Indicators 52.4% 55.4% 52.3% 56.4% 2,299 2,352 2,353 2,298 12

13 TABLE OA5 continued Panel B: Using Market Depth as the proxy for Information Asymmetry Market Depth Disclosure Metric for Good/Bad News: Proportion of GN Sum of SRET Type of Firms: Good News Bad News Good News Bad News Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEO SEOFIRM SEOFIRM SEO ** ** ** *** SEOFIRM SEO POSTREFORM *** ** ** ** 2.05 SEO POSTREFORM SEOFIRM POSTREFORM * POSTREFORM MTB ANALYST FOLLOWING 3.59 *** *** *** *** 2.84 LSIZE *** *** *** *** ROA LNOWN * ** 2.17 INST_HOLDING LNPRC ** ** ** ** 2.50 Year & Industry Indicators 67.0% 61.4% 66.0% 62.7% 2,299 2,352 2,353 2,298 13

14 TABLE OA5 continued Panel C: Using Analyst Forecast Accuracy as the proxy for Information Asymmetry Analyst Forecast Accuracy Disclosure Metric for Good/Bad News: Proportion of GN Sum of SRET Type of Firms: Good News Bad News Good News Bad News Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEO SEOFIRM SEOFIRM SEO ** ** ** ** SEOFIRM SEO POSTREFORM ** ** ** ** 2.24 SEO POSTREFORM *** *** 2.88 SEOFIRM POSTREFORM POSTREFORM MTB ** ANALYST FOLLOWING ** ** LSIZE ** 2.36 ROA LNOWN INST_HOLDING LNPRC *** *** *** *** 2.76 Year & Industry Indicators 9.4% 10.6% 7.6% 11.4% 1,830 1,911 1,924 1,817 14

15 TABLE OA5 continued Panel D: SEO Announcement Returns by Good/Bad News Firms (Parsimonious Model) Three Days Abnormal Returns around SEO Filing Date Disclosure Metric for Good/Bad News: Proportion of GN Sum of SRET Type of Firms: Good News Bad News Good News Bad News Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM *** *** *** *** SEOFIRM POSTREFORM ** ** ** ** 2.13 POSTREFORM MTB ** ANALYST FOLLOWING LSIZE ROA * * LNOWN INST_HOLDING LNPRC * * ** 2.13 Year & Industry Indicators 10.7% 10.1% 10.5% 9.6%

16 TABLE OA5 continued Panel E: SEO Announcement Returns by Good/Bad News Firms (Expanded Model) Three Days Abnormal Returns around SEO Filing Date Disclosure Metric for Good/Bad News: Proportion of GN Sum of SRET Type of Firms: Good News Bad News Good News Bad News Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic Coefficient t -Statistic SEOFIRM *** *** *** *** SEOFIRM POSTREFORM *** ** ** ** 2.32 POSTREFORM * 1.82 MTB ANALYST FOLLOWING LSIZE ROA * * LNOWN INST_HOLDING LNPRC * * * ** 2.38 SALEGROW SALE LEVERAGE CAPEX *** *** CASH PRE_RETURN *** *** SHARE_PERC SEC_OFFER Year & Industry Indicators 12.0% 12.0% 11.7% 11.3%

17 TABLE OA6 Good News Disclosures before SEOs Panels A and B (C and D) in this table are analogous to Table 3 (Table 5) in the paper with the dependent (main independent) variables of interest replaced with Proportion of GN to Total News and Sum of GN. Proportion of GN to Total News is the annualized number of good news forecasts (press releases) relative to the total number of forecasts (press releases) in the pre-seo window, and in the neighboring windows, where a disclosure is considered to provide good news if it induces a positive cumulative abnormal return in the three-day window [-1, 1] around the disclosure date. Sum of GN is the annualized number of good news disclosures in the pre-seo window and in the neighboring windows, where a disclosure is considered to provide good news if it induces a positive cumulative abnormal return in the three-day window [-1, 1] around the disclosure announcement. In Panels C and D, DISC is Proportion of GN to Total News or Sum of GN. All other variables are as described in the paper (see the notes above Table 3 and 5). The standard errors are clustered at the firm level. ***, **, and * denote statistical significance at the two-tailed 1%, 5%, and 10% levels, respectively. Panel A: Analysis of Good News Forecast Frequency before SEOs Disclosure Measure: Management Forecasts Proportion of GN to Total News Sum of GN Coefficient t -Statistic Coefficient t -Statistic SEO *** SEOFIRM *** SEOFIRM SEO SEOFIRM SEO POSTREFORM *** *** 3.89 SEO POSTREFORM *** *** SEOFIRM POSTREFORM POSTREFORM MTB * ANALYST FOLLOWING * ** 2.22 LSIZE * 1.88 ROA *** *** 4.55 LNOWN * 1.83 INST_HOLDING *** *** 3.59 LNPRC Year Indicators Industry Indicators 16.0% 17.4% 4,651 4,651 17

18 TABLE OA6 continued Panel B: Analysis of Good News Press Release Frequency before SEOs Disclosure Measure: Firm Initiated Press Releases Proportion of GN to Total News Sum of GN Coefficient t -Statistic Coefficient t -Statistic SEO *** SEOFIRM ** SEOFIRM SEO SEOFIRM SEO POSTREFORM ** ** 2.11 SEO POSTREFORM *** SEOFIRM POSTREFORM POSTREFORM MTB ANALYST FOLLOWING *** *** 4.48 LSIZE * *** 3.62 ROA ** LNOWN INST_HOLDING ** 2.47 LNPRC *** Year Indicators Industry Indicators 8.4% 18.1% 4,651 4,651 18

19 TABLE OA6 continued Panel C: Analysis of Abnormal Returns following SEOs and Good News Management Forecasting Frequency 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Proportion of GN to Total News Sum of GN Coefficient t -Statistic Coefficient t -Statistic SEOFIRM ** ** DISC POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET *** *** LSIZE MTB ROA * * 1.66 PDA * * FREQ Year Indicators Industry Indicators 11.3% 11.3% 1,484 1,484 19

20 TABLE OA6 continued Panel D: Analysis of Abnormal Returns following SEOs and Good News Press Release Frequency 18 Month Abnormal Returns (AR ) Disclosure Measure (DISC): Proportion of GN to Total News Sum of GN Coefficient t -Statistic Coefficient t -Statistic SEOFIRM *** ** DISC ** POSTREFORM SEOFIRM DISC SEOFIRM POSTREFORM DISC POSTREFORM SEOFIRM DISC POSTREFORM QABRET *** *** LSIZE MTB ROA *** *** 3.54 PDA * * FREQ Year Indicators Industry Indicators 12.7% 12.9% 1,484 1,484 20

21 FIGURE OA1 Distribution of SEOs relative to Earnings Announcements This figure presents the frequency of SEOs relative to quarterly earnings announcements. The x-axis represents the number of days after a quarterly earnings announcement and the y-axis represents the percentage of SEOs. 21