Internet Appendix to Broad-based Employee Stock Ownership: Motives and Outcomes *

Transcription

1 Internet Appendix to Broad-based Employee Stock Ownership: Motives and Outcomes * E. Han Kim and Paige Ouimet This appendix contains 10 tables reporting estimation results mentioned in the paper but not reported. We briefly describe the estimation results in each of the tables. The tables are at the end of the document. *Kim and Ouimet, 2013, Internet Appendix to Broad-based Employee Stock Ownership: Motives and Outcomes, Journal of Finance [DOI STRING]. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the authors of the article. 1

2 1. Sensitivity to separating firms into not-so-numerous- and numerous-employee firms by the top quartile in the number of employees. We document a significant wage increase following the adoption of a small ESOP by not-so-numerousemployee firms and no significant changes in wages following large ESOP adoptions at not-so-numerousemployee firms. For numerous-employee firms, neither ESOP indicator variable is significant. To check the sensitivity of these results to separating firms by the top quartile in the number of employees, we re-estimate the regressions using the number (or the log) of employees at the year of the ESOP initiation (or the match year for control firms). The results are reported in Table IA.I. The regression specification is the same as in Table V, except we drop the control variable proxying for firm size, because of its high correlation with the number of employees. 1 In column 1, we observe a negative and significant coefficient on the interaction of ESOP and firm employment at year 0, Emp_y0. As the number of employees at the firm increases, we find smaller wage gains associated with a small ESOP. This result is consistent with the subsample results in Table V, which shows a positive and significant coefficient on ESOP for not-so-numerous-employee firms and an insignificant coefficient with a point estimate near zero on ESOP for numerous-employee firms. Column 1 also shows a positive and significant coefficient on the interaction of and Emp_y0. However, given the wage gains associated with a small ESOP are declining in the size of the firm s employment and the coefficient on captures incremental differences compared to EOSP, the net effect is no significant net relation. 2. Robustness to Q or industry sales changes as a proxy for growth opportunities at the time of ESOP initiation. We argue the wage gains observed following the adoption of a small ESOP at a not-so-numerous firm are consistent with the incentive hypothesis, however, differences in growth opportunities could explain the same results. Suppose a firm with strong growth opportunities prefers a small ESOP to a large ESOP. As 1 The correlation between firm Sales and Emp_y0 is 73% (significant at 001). 2

3 this firm grows, worker wages are likely to grow rapidly, creating higher ex-post wages at firms which adopt small ESOPs vis-à-vis firms which adopt large ESOPs. In Appendix Table AII, we control for additional proxies for growth and find robust results. As a further robustness check, here we add Q at the time of the ESOP initiation, Q_y0, and interact it with ESOP variables. (For matched firms, Q_y0 is measured at the year of the match). If Q is missing, we use the sample median to avoid changing sample. The results are reported in Table IA.II. If strong growth opportunities are driving our results, then we should find greater wage gains at firms with higher Q at the time of the ESOP initiation. Instead, column 1 shows the interaction term, ESOP* Q_y0, is insignificant with a negative sign. Column 2 uses industry sales change at the time of the ESOP adoption, Industry sales change_y0, as an alternative proxy for growth opportunities at the year of the ESOP initiation. The result again shows no significant interactive effects with ESOP variables. 3. Selection story that firms with superior past performance may prefer small ESOPs to large ESOPs. If firms which adopt small ESOPs have prior superior performance and the superior performance continues after ESOP adoptions, it will give the appearance of small ESOPs leading to higher wages and shareholder value. We proxy for pre-esop performance by the pre-esop run-up in stock price and calculate raw and market-adjusted returns over one, two, and three years preceding the ESOP adoptions. This is done separately for small ESOPs, s, and small ESOPs at not-so-numerous-employee firms. The results are presented in Table IA.III and show similar patterns of pre-esop run-up in prices for all three groups. 4. Correlations between run-up in stock prices with changes in cash wages and industryadjusted Q. We also estimate the correlations for each of the six measures of run-up in prices with changes in cash wages and industry-adjusted Q. All correlations are estimated separately for small ESOPs, large ESOPs, and small ESOPs at not-so-numerous-employee firms. We calculate changes in cash wages and industry- 3

4 adjusted Q as the differences between the terminal year and year 0, the year of ESOP adoption: (wages terminal wages 0 )/wages 0 and industry adjusted Q terminal industry adjusted Q 0. 2 We use year 0 as the baseline year because if we go back any further it overlaps with the pre-esop period used to measure the price run-ups. Terminal years are defined as the earlier of the last year an establishment is observed or +10 years. 3 We use this definition of terminal year as opposed to a specific year 5 or 10 cut-off points to include the most observations and avoid a bias of looking at only surviving firms/establishments. For wages, we take an employee-weighted average to obtain firm level values. Missing values are replaced with the median to keep sample size constant within a column, across the rows. We winsorize run-ups at 1% on each tail. The estimated correlations are mostly insignificant and reported in Table IA.IV. 5. Replacing missing WBP with zero. For establishments located in non-msa areas, we are unable to calculate the WBP and use the sample median value to avoid changing the sample. As mentioned in the paper, we make this compromise because of the Census disclosure policy to protect privacy. The Census screens all regression coefficients and sample sizes submitted for disclosure to prevent the release of any information which can be traced to one or a small group of firms. So we cannot disclose results using two samples which are very similar or where a dummy variable captures similar but not identical sets of firms across samples. This is an acute problem with our data, especially when we cut the data across ESOP size and firm characteristics, because the number of ESOP firms is limited. As an alternative to replacing missing WBP observations with the sample median, we assign zero to missing observations. We refer to this alternative definition of WBP as WBPm0. Establishments with missing WBP are typically located in non-msa areas, which tend to be rural areas with few alternative 2 We do not divide industry-adjusted-q-by-year 0 values because year 0 values can be very close to 0 or negative. 3 For example, if an establishment is closed 5 years after an ESOP, the terminal year is year +5; if the establishment continues to operate beyond the cutoff year +10, then the terminal year is year

5 employers. We re-estimate Table VI, columns (1), (5), and (6) in the paper with this alternative definition of WBP. The results, reported in Table IA.V, are robust. 6. The full set of coefficients in Table VI. In columns (3) and (4) of Table VI reported in the paper, we do not report the coefficients on the standalone industry immobility variables and their interactions with WBP or ESOP variables to conserve space. The full set of coefficient values are reproduced in Table IA.VI. 7. Reporting additional estimation results for the cash constraint hypothesis. We use the Kaplan and Zingales (KZ) index at the time of the ESOP adoption for ESOP firms (or at the time of the match for control firms) as our proxy for cash constraints. We find post- wage gains are lower, the more cash constrained an -initiating firm is at the time of the initiation. In column (1) of Table IA.VII, we add the interaction of CashConst_y0*ESOP and find an insignificant coefficient on the interaction term, consistent with our argument in the paper that the cash conservation motive applies primarily to large ESOPs. In column (2), we show the results are robust to restricting the sample to not-so-numerous-employee firms. In column (3), we use an alternative proxy for cash constrained firms based on Hadlock and Pierce (2010), who find that young and small firms are more likely to be financially constrained. We do not include the size component of the age-size index of Hadlock and Pierce (2010) because of the high correlation between the size as measured by sales or total assets and the number of employees, which is used to separate the sample according to the susceptibility to free-rider problems. Age is defined as the number of years the firm has been in Compustat as of the year of the ESOP initiation for ESOP firms (and as of the year of the match for the set of control firms.) Age_y0 is a proxy for less financial constraint; hence, the positive signs of the * age_y0 coefficient is consistent with the results using the KZ index. Columns (4)-(6) repeat the same set of estimations with industry adjusted Q as the dependent variable. As noted in the paper, we find less robust results. 5

6 8. Employment changes at the establishment level. Table X in the paper shows the relation between firm-level employment and ESOPs. Here we examine employment changes at the establishment level by regressing log (1 + employment) at a given establishment on ESOP and with the same set of control variables as in the wage regressions, but we now replace the state-year and industry-year average wages by state-year and industry-year average employment. Estimation results are reported in Table IA.VIII. We find no significant relation between ESOP adoptions and the level of employment at the establishment level. This is true even when we divide the sample into numerous- and not-so-numerous-employee firms. 9. Re-estimation of wage regressions while controlling for 401K employee share ownership. The incentive effect of share ownership through a 401K plan is likely to be weaker than that through an ESOP plan. With 401K share ownership, there is no long term commitment on the part of employees because they can sell company shares held in their 401K. This ability to sell company shares weakens the incentive effect. In contrast, employees with ESOP shares do not have that option unless they quit the job or reach certain near-retirement age thresholds. Furthermore, ESOPs have to be given to all employees, subject to a few small exceptions. Participation in a 401K plan is optional for employees, and when firms offer a match in the form of company stock, the employee can immediately sell these shares for diversification and invest the proceeds elsewhere. As such, employee share ownership through 401K is unlikely to be as broad-based as an ESOP plan. The productivity-improving peer pressure and comonitoring, which arise from being a part of a broad based group, is likely to be weak in 401K ownership. With ESOP plans, we observe no employee share ownership for a number of years and then see a sharp jump in employee ownership. There is a true before and after. With 401K plans, there are gradual increases in employee ownership and then decreases in the later part of our sample period, showing gradual and small within-firm variation in ownership over time. All of our wage regressions control for establishment fixed effects. As such, regression coefficients associated with 401K employee ownership will reflect changes in small amounts of ownership for example, 1% employee ownership to 6

7 2% over several years. With our ESOP data, we are able to test the difference between zero and significant employee ownership. Testing a relation between 401K and Q will be especially difficult because we will be looking at changes in actual ownership, not when the market anticipates such changes in ownership. Furthermore, employees can always rebalance their allocation, either selling or buying shares in their own company. They may time the rebalancing based on the current stock price and/or their anticipation of future stock price, which may be systematically related to current market value of the firm. Such rebalancing activities will confound the estimated relation, making it difficult to draw any meaningful inferences. To estimate 401K ownership we use a combination of our ESOP data with those from IRS Form We thank Josh Rauh for sharing his Form 5500 data. Rauh s data extends from 1985 to 1998 and excludes: firms with missing information on incorporation, firms whose plans report zero assets, firms whose plans have improperly coded employee ownership shares, and firms who report assets in common, master, or collective trusts. The employee ownership data obtained from Rauh does not break down the ownership into 401K and ESOPs. To distinguish ownership via 401K plans from ownership via ESOPs, we assume 401K ownership (as a % of total shares outstanding) is the total ownership reported in the Form 5500 minus 1%, if the firm has a small ESOP; and the total ownership reported in the Form 5500 minus 5%, if the firm has a large ESOP. These 1 % and 5% estimates are the most conservative of the current actual employee share ownership via ESOPs, assigning greater share ownership to 401K. Table IA.IX reports the re-estimation results. The results concerning ESOP variables are consistent with those reported in the text, while the effects of ownership stake via 401K is mostly insignificant. 10. Re-estimation of wage and Q regressions while simultaneously including the cash constraint index, the BCS variables, and WBP. In the paper, we test sequentially and separately the three different hypotheses. To test the three hypotheses simultaneously, we re-estimate both the wage and industry-adjusted Q regressions while 7

8 including all three factors simultaneously the cash constraint index, the BCS variables, and WBP. The re-estimation is performed for all firms, not-so-numerous and numerous-employee firms. The results, reported in Table IA.X, are robust. 8

9 Table IA.I. Wage Changes Following ESOP Initiations Interacted with Total Firm Employee Count. The dependent variable is log wages per employee. The set of observations includes a sample of ESOP firms and a sample of matched control firms, as described in the text. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stock at any given time. Emp_y0 is the total firm employee count in the year the ESOP was adopted (ESOP firms) or the year the firm was matched (control firms.) The following control variables are included in all regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are clustered at the firm level. "*", "**", and "***" indicate statistical significance at 10%, 5%, and 1%, respectively. 1 2 Sample All All ESOP ** (0.069) ** (0.074) ** (0.080) *** (0.089) Emp_y *** (0.074) ESOP * emp_y * (0.076) * emp_y ** (0.086) Log Emp_y (0.138) ESOP * log emp_y * (0.122) * log emp_y ** (0.155) Adjusted R-squared N

10 Table IA.II: ESOP Interacted with Growth Options at Year 0. The dependent variable is log wages per employee. The set of observations includes a sample of ESOP firms and a sample of matched control firms, at firms with not-so-numerous employment. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP controlling at least 5% of the firm's outstanding common stocks at any given time. The following control variables are included in all regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively. 1 2 ESOP 0.21*** (0.07) 0.12*** -0.29*** -0.18*** (0.05) Q_y0 0.06** Q_y0 * ESOP Q_y0* - (0.07) Ind sales change_y Ind sales change_y0 * ESOP 0.04 Ind sales change_y0 * N Adjusted R-squared

11 Table IA.III. Average Price Run-up by Type of ESOP. The table reports the mean pre-esop run-up in stock prices using a 1-year, 2-year and 3-year measurement window. Raw and market adjusted returns are reported. Market-adjusted returns are calculated using the 12 months prior to the estimation window to estimate the market model, using the CRSP value weighted index with dividends as the market return. Column 1 includes firms with small ESOPs which never control 5% or more of the firm s outstanding stock. Column 2 includes firms with large ESOPs which control more than 5% of the firm s outstanding stock for at least one year. Column 3 includes firms with small ESOPs and firm employment in the bottom 3 sample quartiles. Missing observations are replaced with the sample median Measure of run-up Small ESOP Small ESOP at not-sonumerous employment firm Raw 1-year returns Raw 2-year returns Raw 3-year returns Market model 1-year returns Market model 2-year returns Market model 3-year returns

12 Table IA.IV. Correlations between Pre-ESOP Stock Price Run-ups and Post-ESOP Changes in Wages and Industry-adjusted Q. Panel A. Correlation between price run-up and wage change. The table reports the correlation between wage changes and pre-esop run-up in stock prices using a 1- year, 2-year and 3-year measurement window. Correlations are reported with p-values in parentheses. Change in wages is calculated at the establishment level as (wages termainal wages 0 )/wages 0. Then we take an employee-weighted average to obtain their values at the firm level. Year 0 is the year of the ESOP initiation. The terminal year is defined as min {the last year the establishment is observed in the sample, 10 years after the ESOP initiation}. Market-adjusted returns are calculated using the 12 months prior to the estimation window to estimate the market model, using the CRSP value weighted index with dividends as the market return. Missing values are replaced with the median to keep sample size constant within a column, across the rows. Column 1 includes firms with small ESOPs which never control 5% or more of the firm s outstanding stock. Column 2 includes firms with large ESOPs which control more than 5% of the firm s outstanding stock for at least one year. Column 3 includes firms with small ESOPs and firm employment in the bottom 3 sample quartiles. Correlations that are significant at 10% are noted in boldface Measure of run-up Small ESOP Small ESOP at not-sonumerous employment firm Raw 1-year returns - (0.98) (0.80) Raw 2-year returns (0.47) (0.61) Raw 3-year returns (0.61) (0.68) Market model 1-year 0.11 (0.27) 0.08 (0.36) 0.10 (0.39) returns Market model 2-year 0.01 (0.92) 0.01 (0.93) - (0.97) returns Market model 3-year returns (0.48) 0.17 (0.15) 12

13 Panel B. Correlation between price run-up and industry- adjusted Q. The table reports the correlation between changes in industry adjusted Q and pre-esop run-up in stock prices using a 1-year, 2-year and 3-year measurement window. Correlations are reported with p-values in parentheses. Change in industry adjusted Q is calculated as industry adjusted Q terminal industry adjusted Q 0. Year 0 is the year of the ESOP initiation. The terminal year is defined as min {the last year the establishment is observed in the sample, 10 years after the ESOP initiation}. Market-adjusted returns are calculated using the 12 months prior to the estimation window to estimate the market model, using the CRSP value weighted index with dividends as the market return. Missing values are replaced with the median to keep sample size constant within a column, across the rows. Column 1 includes firms with small ESOPs which never control 5% or more of the firm s outstanding stock. Column 2 includes firms with large ESOPs which control more than 5% of the firm s outstanding stock for at least one year. Column 3 includes firms with small ESOPs and firm employment in the bottom 3 sample quartiles. Correlations that are significant at 10% are noted in boldface Measure of run-up Small ESOP Small ESOP at not-sonumerous employment firm Raw 1-year returns 0.06 (0.57) (0.83) 0.07 (0.56) Raw 2-year returns 0.03 (0.78) (0.89) Raw 3-year returns (0.68) (0.55) Market model 1-year 0.04 (0.72) 0.09 (0.32) 0.09 (0.44) returns Market model 2-year 0.03 (0.75) (0.51) returns Market model 3-year returns 0.06 (0.58) 0.11 (0.22) 0.10 (0.39) 13

14 Table IA.V: Wage Changes following ESOP Initiation Interactive Effects of Worker Bargaining Power based on an Alternative Definition of WBP(WBPm0). The dependent variable is log wages per employee. The set of observations includes a sample of ESOP firms and a sample of matched control firms at not-so-numerous-employee firms. Column 3 excludes observations during the first 5 years after an ESOP initiation (for ESOP firms) or after the year of match (for control firms) and observations where the industry year mean wage at year 0 is below the sample median. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stock at any given time. WBPm0 is similar to WBP except missing observations are replaced with 0. The following control variables are included in all regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are clustered at the firm level. "*", "**", and "***" indicate statistical significance at 10%, 5%, and 1%, respectively ESOP (0.15) (0.24) -0.25** (0.12) (0.24) WBPm (0.18) (0.18) (0.19) WBPm0*ESOP 0.18** 0.19** 0.21 ** WBPm0* -0.18** -0.19** WBPm0_y (0.05) WBPm0_y0*ESOP -0.20* (0.12) WBPm0_y0* 0.48*** (0.15) WBPm0 _y5*esop WBPm0_y5* N Adjusted R

15 Table IA.VI: Wage Changes following ESOP Initiation Interactive Effects of Worker Bargaining Power (WBP). Reporting coefficients not reported in Table VI of Kim and Ouimet (2013) to conserve space. The dependent variable is log wages per employee. The set of observations includes a sample of ESOP firms and a sample of matched control firms at not-so-numerous-employee firms. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stock at any given time. The following control variables are included in all regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are clustered at the firm level. "*", "**", and "***" indicate statistical significance at 10%, 5%, and 1%, respectively. 1 2 ESOP 1.82* (1.10) 0.01 (0.12) -0.95* (0.57) WBP WBP*ESOP 0.02 (0.13) WBP* (0.15) Ind_immobility 0.29 (0.21) Ind_immobility*WBP (0.26) Ind_immobility*ESOP Ind_immobility* (0.14) Ind_immobility*WBP* ESOP 0.23** (0.12) Ind_immobility*WBP* (0.14) Indywages_y (0.54) Indywages_y0 * WBP (0.61) Indywages_y0* ESOP (0.32) Indywages_y0 * 0.18 (0.15) Indywages_y0 * WBP*ESOP 0.62* (0.37) Indywages_y0 * WBP * (0.17) N Adjusted R

16 Table IA.VII: Wage Changes following ESOP Initiations Interactive Effects of Cash Constraints: Reporting Additional Estimation Results. The dependent variable in columns 1-3 is log wages per employee. The dependent variable in columns 4-6 is industry adjusted Q, winsorized at 1%. The set of observations includes a sample of ESOP firms and a sample of matched control firms, as described in the text. Columns 1, 3, 4, and 6 include all observations. Columns 2 and 5 include observations at firms with not-numerous employment. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given time. The following control variables are included in the regressions in columns 1-3: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. The following control variables are included in the regressions in columns 4-6: Assets, Sales, R&D, R&D_missing, Capex, Sigma, Sigma dummy, and firm- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are clustered at the firm level. "*", "**", and "***" indicate statistical significance at 10%, 5%, and 1%, respectively Sample All Not-sonumerounumerous All All Not-so- All ESOP 0.06 (0.05) 0.19*** ** 0.23*** -0.13** (0.07) -0.31*** -0.22*** *** (0.14) CashConst_y0 -* -0.01*** * CashConst_y0 *ESOP CashConst_y0 * -0.03* -0.03*** 0.08* 0.01 Age_y0 Age_y0 * 0.01*** 0.01*** N Adjusted R- squared

17 Table IA.VIII. Establishment-Level Employment Changes Following ESOP Initiations. The dependent variable is log employees per establishment. The set of observations includes a sample of ESOP firms and a sample of matched control firms, as described in the text. Columns 1 and 2 include all observations. Column 3 includes observations at firms with not-so-numerous employment. Column 4 includes observations at firms with numerous employment. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given time. See Appendix Table AI for a description of variables. Establishment and year fixed effects are included. Coefficients are reported with standard errors in parentheses. All standard errors are corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively Firm employment All All Not-so-numerous Numerous ESOP (0.05) 0.08 (0.07) (0.07) S-Y mean employment ** 0.11 (0.17) I-Y mean employment 0.13*** 0.13*** 0.19* 0.15*** (0.05) Establishment age 0.07*** 0.07*** 0.05* 0.05 Sales *** 0.10*** Leverage (0.05) Ad 0.33 (0.34) 0.31 (0.33) 0.02 (0.14) -2.17* (1.21) Ad_missing R&D 1.49 (1.81) R&D_missing (0.05) 0.05** Tangibility 0.40*** 0.39*** *** (0.18) Cap_labor N Adjusted R-squared

18 Table IA.IX: Wage Changes Following ESOP Initiations controlling for 401K Ownership. The dependent variable is log wages per employee. The set of observations includes a sample of establishments at not-so-numerous employee ESOP firms and a sample of establishments at matched notso-numerous employee firms. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given time. The following control variables are included in all regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively ESOP 0.16* 0.16* 0.16* 0.16* -0.22** -0.21** -0.24** -0.23** CashConst_y0 - - CashConst _y0 * -0.01** -0.10** 401K own (0.23) (0.25) (0.26) 0.01 (0.30) 401K own 2 (0.56) (0.54) ** 401k own * CashConst _y k own 2 * CashConst _y N Adjusted R-squared

19 Table IA.X, Panel A. Wage Changes Following ESOP Initiations by Firm Characteristics. The dependent variable is log wages per employee. The set of observations includes a sample of ESOP firms and a sample of matched control firms, as described in the text. Column 1 includes all observations. Column 2 includes observations at firms with not-so-large employment. Column 3 includes observations at firms with large employment. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given time. The following control variables are included in regressions: S-Y mean wages, I-Y mean wages, Establishment age, Sales, Leverage, Ad, Ad_missing, R&D, R&D_missing, Tangibility, Cap_labor, and establishment- and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively All Not-so-numerous Numerous ESOP (0.15) 0.18*** -0.29*** -0.27** (0.12) -0.22*** CashConst_y0 -* -0.01*** CashConst_y0 * -0.03*** -0.03*** -0.02* BCS_Blk BCS_Board 0.07*** 0.10*** 0.05* Maxesop - BCS_Blk *esop BCS_Blk *maxesop0 - * WBP (0.15) (0.32) 0.01 WBP * ESOP ** (0.15) -0.18*** (0.07) WBP * 0.18* *** (0.07) N Adjusted R-squared

20 Table IA.X, Panel B. Q Changes Following ESOP Initiations by Firm Characteristics. The dependent variable is industry-adjusted Q, windorized at 1%. Q it is fiscal year-end market value of equity plus market value of preferred stock plus total liabilities divided by total assets. Industry-adjusted Q is obtained by subtracting the median Q matched by industry (3-digit SIC code) and year. The set of observations includes a sample of ESOP firms and a sample of matched control firms, as described in the text. Columns 1 and 4 include all observations. Columns 2 and 5 include firms with not-so-large employment. Columns 3 and 6 include firms with large employment. ESOP is a dummy variable which takes the value of 1 if the firm has an ESOP. is a dummy variable which takes a value of 1 if the firm has an ESOP that controls at least 5% of the firm's outstanding common stocks at any given time. The following control variables are included in the regressions: Assets, Sales, R&D, R&D_missing, Capex, Adjusted sigma, Adjusted sigma dummy, firm and year fixed effects. See Appendix Table AI for a description of variables. Coefficients are reported with standard errors in parentheses. All standard errors are corrected for clustering at the firm level. "*", "**", and "***" reflect statistical significance at 10%, 5%, and 1%, respectively All Not-sonumerous Numerou s All Not-sonumerous Numerous ESOP 0.77** (0.34) 0.84*** (0.34) (0.73) 0.83* (0.49) 0.71* (0.42) 0.26 (0.87) -0.87* (0.48) -0.97** (0.48) (0.98) -0.93* (0.53) -0.94** (0.48) (0.89) CashConst_y0 CashConst_y0 * BCS_Blk BCS_Board Maxesop BCS_Blk *esop BCS_Blk *maxesop0 Equal weighted WBP Equal weighted WBP * ESOP Equal weighted WBP * 0.01* (0.07) 0.07 (0.05) (0.14) -0.73* (0.41) 0.87 (0.58) (0.12) (0.15) -0.82** (0.42) -0.97* (0.58) (0.12) *** 0.06 (0.18) -0.01* 0.13 (0.45) 0.36 (0.85) 0.51 (1.19) 0.01* (0.07) 0.06 (0.05) (0.12) ** (0.05) 0.07 (0.13) *** -0.01* 0.04 (0.18) Payroll weighted WBP (0.15) (0.15) Payroll weighted WBP * ESOP (0.54) (0.45) Payroll weighted WBP * * (0.62) (0.54) N Adjusted R-squared (0.60) (1.04) 0.72 (1.13) 20