Wealth and Income Consequences of Employee Ownership: A Comparative Study from Washington State

Peter A. Kardas

Adria L. Scharf

Jim Keogh

The authors wish to thank the Ford Foundation, Richard Freeman from the National Bureau of Economic Research, and Chris Mackin of Ownership Associates for funding this research. The authors also thank Janelle Edwards and Nick Lee from the Washington State Department of Community, Trade, and Economic Development for their careful work with data input and for other assistance they provided to the project. A special thanks also to Douglas Kruse for his comments and questions on an earlier draft of the paper, and for his excellent summary of the paper at the Shared Capitalism Conference in Washington, DC, May 22 and 23, 1998.

In addition to being involved in occasional research projects, Peter Kardas works with unions, workers, and businesses in Washington State on economic development, employee ownership, and worker participation projects. Adria Scharf is a doctoral student at the University of Washington concentrating on organizational sociology and the sociology of work. She has a particular research interest in employee ownership. Jim Keogh is a business retention specialist for the Washington State Department of Community, Trade, and Economic Development, and managed the agency's Employee Ownership Program for nine years..

This study attempts to answer questions about the success of companies with employee stock ownership plans (ESOPs) in getting more wealth and income into the hands of employees. . Using 1995 employment and wage data from the Washington State Employment Security Department and 1995 data on retirement benefits from a survey of companies and from IRS Form 5500, the study matched up 102 ESOP companies with 499 comparison companies in terms of industrial classification and employment size. The data allowed the authors to estimate the value of retirement assets in ESOP companies compared to the value of retirement assets in other companies; to compare ESOP and control companies in terms of wages and other benefits (such as health care and other insurance); and to analyze whether the distribution of wealth and wages is more egalitarian in ESOP companies. Independent variables included company size, industrial sector, unionization, percentage of ownership by the ESOP, age of the retirement plan, and company participation programs.

This study By comparing retirement assets and wages in Washington State ESOP companies with those in matched similar non-ESOP firms, the attempts

analysis shows The study foundthat ESOP companies provide significantly higher retirement benefits than comparison firms. T the average value (per participant) of all retirement benefits in ESOP companies (in 1995) wasequal to approximately $32,000,, with an whereas the average value in the comparison companies was of about $12,500. None of the independent variables in the analysis eliminated or significantly diminished the ESOP as an explanation for higher asset values. A large percentage of comparison companies (between 58% and 71%) had no retirement plan at all, and for in those that did, employee participation rates in the plans were lower than than in the ESOP companies. Further, About 60% of the retirement assets in ESOP companies was in the form of company stock, while in those comparison companies that have plans, approximately 70% of the value of the assets was in stock offered through 401k plans. Ccompanies with ESOPs contributed on average about 10% of pay to all retirement plans, while the comparison companies contributed on average about three percent. Whereas in those comparison companies that have retirement plans, approximately 70% of the value of the assets was in stock offered through 401k plans (and presumably diversified), whereas in ESOP companies about 60% of retirement assets take the form of company stock .

In terms of wages,The company stock held in the ESOP does not appear to come at the cost of wages. Tthe median hourly wage of $14.72 in the ESOP firms was 8% higher than the median hourly wage in the comparison companies. At the 10^th percentile of wages, hourly wages were 4% higher in the ESOP firms, while at the 90^th percentile, ESOP wages were 18% higher than in the comparison firms. Therefore, the ratio between the 90^th and 10^th percentiles was higher in the ESOP companies than in the comparison firms. Unions, in both ESOP and control companies, had the effect of raising wages at the 10^th percentile and lowering them at the 90^th, with the result that median wages for unionized control companiess are significantly higher than for non-union controls. On average, the ESOP firms in this study provide a significantly higher total compensation to their employees than do their competitors,, though but the ratio of 90^th to 10^th percentile wages suggests that they do so withinthey do so withinreplicate the framework of rewards already established in the economy.

In writings during the 1950s and 19’60s, Louis Kelso argued that ownership by employees of company stock is necessary to create a society in which affluence is broadly shared and extremes of economicy inequality of wealth and poverty reducedreduced. If the ultimate goal of employee stock ownership is to achieve a society that has both greater greater equality of economic condition as well as equality of opportunity for economic gain, broad capital ownership equality of economic condition as well as equality of opportunity for economic gain, without sacrificing employee wages, then it may be of interest to examine the distribution of wealth and income wages distribution in companies that have established employee stock ownership plans (ESOPs) using the tax incentives that Kelso helped inspire. Nearly 24 years after the establishment of ERISA, have ESOPs lived up to their promise to more broadly share the gains of stock ownership?redistribute wealth? What are the financial benefits of ESOPs to company employees and to ESOP participants? Scott--Maybe it's just my interpretation of Kelso, but it seems to me that he's arguing for greater equality of economic condition and greater equality of opportunity for gain, which of course is not the same thing as arguing for absolute equality. In Two Factor Theory he writes that the economic objective of universal capitalism ". . .is the production and enjoyment of the highest level of affluence (humanly useful goods and services) for every family, consistent with optimum use of the economy's resources and productive power, and the desire of its people to consume. . . ." (p. 8). One of the political objectives of universal capitalism is the "broad diffusion of privately owned economic power." That sounds like an interest in greater equality of economic condition, an interpretation that is reinforced by a statement on page 11 that "any society seriously caring about freedom must structure its economic institutions so as to widely diffuse economic power while keeping it in the hands of individual citizens," and another statement on page 121 that "every individual's human dignity requires that he enjoy general affluence, and that he produce it." Your rephrasings don't capture the moral and political intent that I see in his writings, so I request that our original formulation remain. Scott, you wrote: It may seem a quibble, but Kelso was not primarily interested in equality per se—communism did that, he’d note; rather, he was interested in greater wealth for more people. In his radical moments, he’d say that having more money than you or your heirs could consume was theft, but he probably didn’t believe that (his widow has more than she and her heirs can consume). Perhaps you should say “If the ultimate goal of employee ownership was a broader distribution of substantive capital ownership” or “a society in which everyone could participate substantively in capital ownership.” Also, Kelso never opine

d about income distribution within companies, did he?

The study reported here begins to address these questions with data on wages and retirement plan assets in Washington State companies. We combine government wage data on ESOP companies and comparison companies with retirement plan information from a survey of those companies and from Internal Revenue Service (IRS) Form 5500 filings to estimate:

· How the value of retirement assets in ESOP companies compares to the value of retirement assets in other companies;

· How wages in ESOP companies compare to wages in comparable non-ESOP companies;

· How ESOP and control companies compare on the provision of other benefits, such as health care insurance; and

· Whether the distribution of wealth and wages is more egalitarian in the ESOP companies.

In addition, we investigate the effects of a number of independent variables, including company size, industrial sector, percentage of ownership by the ESOP, years that the plan has been in place, unionization, and company participation programs.

Methodology and Description of Companies

The Sample

The sample of 102 ESOP companies includes nearly every such company in Washington State that we were able to identify. We used Form 5500 data and records from the Washington State Employee Ownership Program to generate a list of all definite and potential ESOP companies in the state. We then made phone calls to those companies to confirm whether they have an ESOP or a KSOP (i.e., a combined 401(k) plan and ESOP).[1]

The 499 control companies were selected by random match. For each employee stock ownership company confirmed to have an ESOP, three to seven control companies of the same employment size and industrial sector (based on the four-digit Standard Industrial Classification [SIC] code) were randomly selected from a database of all companies in the state provided by the Washington State Employment Security Department. All but three ESOP companies were matched with between three and seven controls. For three ESOP companies, there were only two possible comparison companies of the same sector available. This resulted in an average number of control companies per ESOP company of five.

Wage and employment data for 1995 for all 601 companies—102 ESOP companies and 499 matched controls—were obtained from the Employment Security Department. Wages included all gross wages[2] for employees covered by unemployment insurance.[3] Tables 1 and 2 show the distribution of companies in the wage data sample by size and industrial sector, with a column included in each table for the distribution of ESOPs nationwide. A comparison of the nationwide and Washington State data indicates that the Washington companies are fairly representative of other ESOPs in terms of size and industrial sector, except that there is a smaller percentage of companies in Washington with over 500 employees.

Table 1. Company Size in 1995

Company Size in 1995 (Number of Employees per Company)	Count		Percentage (of Sample)			ESOPs nationwide[4]
	ESOP	Non-ESOP	ESOP		Non-ESOP
1–49	38	244	37%	49%		34%
50–99	16	101	16%	20%		19%
100–199	23	58	23%	12%		16%
200–299	3	26	3%	5%		14% (200 – 500 employees)200 to 500 empl. = 14%
300–399	5	13	5%	3%
400+	9	21	9%	4%
Missing	8	36	8%	7%		17 % (500+ employees)Over 500= 17 %
Total	102	499	100	100

Table 2. Industrial Sector

Industrial Category (SIC Code)	Count		Percentage (of sample)		ESOPS nation-wide[5]
	ESOP	Non-ESOP	ESOP	Non-ESOP
SIC1: Forestry, Fishing, Mining, Construction	10	58	10%	12%	11%
SIC 2: Manufacturing (food, lumber, printing, chemicals)	9	47	9%	9%	25% (SIC 2 & 3)
SIC 3: Manufacturing (metal, industrial machinery, transportation equipment)	20	108	20%	22%
SIC4: Transport, Communications, Utilities	2	6	2%	1%	4%
SIC 5: Wholesale and Retail Trade	24	127	24%	26%	19%
SIC 6: Finance, Insurance, Real Estate	23	78	23%	16%	22%
SIC 7: Services (taxable companies)	7	40	7%	8%	18% (SIC 7 & 8)
SIC 8: Services (health, legal, social, engineering)	7	35	7%	7%
Total	102	499	100%	100%	99%

The Survey

We sent surveys to all 601 companies and made follow-up phone calls to 400 of those, obtaining usable responses from 148 companies—47 companies with ESOPs and 101 comparison companies (see the appendix for a copy of the survey). Out of these usable responses, we were able to match up 37 ESOPs with 68 control companies. From survey respondents we have detailed information on the value of assets held by retirement plans, the formula by which benefit assets are allocated to employees, and the number of employees in different wage categories covered by each benefit plan. In addition, survey respondents were asked the value of salary and non-salary compensation for highly compensated employees who are not covered by unemployment insurance, whether the company is public or private, whether its employees are covered by a collective bargaining agreement, the company’s age, the types of participatory management techniques used, and the degree of employee influence in various decision-making areas. For ESOP companies, the survey asked for information about the ESOP plan, including the percentage of company stock held in the ESOP trust, the percentage of payroll contributed to the plan in 1995, the basis on which stock is allocated to employee accounts, whether the plan was leveraged, and the reason(s) the ESOP was implemented.

According to survey data from ESOP companies that provided this information, the percentage of company stock owned by the ESOP trust in 1995 ranged from 0% to 100% (as summarized in table 3).[6] Of those, the ESOP owned a majority of the stock in 15 companies (39%), with four of those ESOPs owning 100% of the stock. The average percentage of ownership by the stock plans was 42% and the median was 35%. Seven ESOPs for which we had this information were publicly traded, and 35 were privately held. Four comparison companies were publicly traded, and 93 were privately held.

Table 3. Distribution of ESOP Companies by Percentage of Company Stock Held by ESOP Trust

Percentage of Stock Held by ESOP Trust	Count	Percentage (of ESOP Companies for Which % of Ownership Is Known)
0-24%	13	33%
25-49%	11	28%
50-74%	8	21%
75-100%	7	18%
Total	39	100%

IRS Form 5500

In addition to data from the survey, we procured IRS Form 5500 data[7] for tax year 1995 for 250 companies in our sample, out of which we were able to match and use 202 cases—66 ESOP companies and 136 controls. All companies that provide qualified retirement plans subject to ERISA, including ESOPs, 401(k) plans, defined benefit pension plans, and profit sharing plans, must file Form 5500 with the IRS. Companies with 100 or more participants must file Form 5500 every year. Companies with fewer than 100 participants must file Form 5500-C at least every three years; for years in which Form 5500-C is not filed, such companies must file the abbreviated version, Form 5500-R. The responses to Form 5500 identify all qualified retirement plans provided per company and give the total value of assets, net value of assets, and employer contribution for each plan. We used this data both as an accuracy check for our survey information and as a supplemental source of information.

Table 4 summarizes mean and median company size for all companies in the study, for ESOP and comparison companies that were both matched with each other and that returned surveys, and for ESOP and comparison companies that were matched and for which Form 5500 data was available. In general, the ESOPs had higher employment than the comparison companies, with the differences being greatest for companies that returned surveys and smallest for companies in the 5500 data. Of the 37 ESOP companies that were matched with comparison companies and for which we had survey data, there was information for 31 of them in the Form 5500 data. Of the 68 comparison companies that were matched with ESOPs and for which we had survey data, there was information for 30 in the 5500 data.

Table 4. Median and Mean Company Size in 1995

Data Source	Mean Number of Employees		Median Number of Employees
	ESOP	Non-ESOP	ESOP	Non-ESOP
All Companies in Study	171	133	71	45
Matched Companies That Returned Survey	225	84	74	31
Matched Companies That Filed 5500 Forms	194	185	85	65

A couple of other important pieces of information about the ESOP companies and matched comparisons are: the mean number of hours worked per quarter in 1995 was 392 for the ESOP companies and 371 for the control companies. The average start date for all retirement plans in both ESOPs and controls was 1984, with a median start date of 1986 for both groups of companies. The mean start date for ESOP plans only was 1985.

Retirement Assets

In comparing ESOPs to the matched comparison companies on benefits and income, we will examine first the value of retirement assets (including company stock), then wages, and finally the provision of other benefits. We will look at retirement assets in three ways. First, we will compare per-participant assets held in all plans in ESOP and comparison companies. Second, we will compare the percentage of payroll contributed to retirement plans by ESOP and non-ESOP companies. Finally, we will estimate the value of assets held on behalf of ESOP participants in different wage categories.

Because both ESOP and comparison companies often have more than one retirement plan (e.g., 401(k) and profit sharing plans), the task of comparing the wealth holdings of participants in the two types of companies is somewhat complex. To accurately compare employees' retirement assets in ESOP companies with retirement assets in control companies, we need a measure of benefits that pulls together the value per participant of each plan in a company. To have an accurate understanding of how ESOPs compare to typical competitors, we also must take into account the percentage of companies that do not file Form 5500.

Table 5 presents average assets per covered employee for ESOP companies and matched controls that returned surveys, and for ESOPs and matched controls that filed Form 5500. The top row gives the sum of the average assets per participant for all plans listed in rows three through seven (401(k) plan, ESOP, etc.). This measure assumes that a participant in one plan is also a participant in every other plan, so the sum (“Sum of Average Assets per Participant”) equals the total value of an individual’s assets from all the different plans. But what if the participants in any one plan do not participate in any other plans? In that case, to get an average value of retirement assets per participant for the whole company, we must sum up the asset values of the various retirement plans (401(k), ESOP, etc.), then divide the result by the sum of participants for each plan. That is the value represented in the second row of the table (“Total Assets All Plans/Sum of Participants”).

Table 5. Assets per Participant for Several Plans, Using Survey and Form 5500 Data^a

Assets Per Participant, Different Plans	ESOP Companies, from Survey n=37	Control Companies, from Survey n=37 (weighted)	ESOP Companies, from 5500s n=66	Control Companies, from 5500s n=66 (weighted)
Sum of Avg. Assets per Participant, All Plans	$32,213*	$12,735*	$47,680***	$24,946***
Total Assets All Plans / Sum of Participants	$21,634**	$7,739**	$31,967*	$21,020* $12,612^b
401(k) Assets / Participant	$3,796	$8,890	$13,021	$14,720
ESOP Assets / Participant	$24,260	$0	$20,396	$136^c
Defined Benefit Assets / Particip.	$1,254	$410	$3,148	$2,203
Profit Sharing Assets / Particip.	$607	$1,464	$10,466	$5,013
Other Assets / Participant	$2,295	$1,971	$650	$2,873

^a Results for the control companies are weighted so that the sum of control companies for each ESOP company equals one, thus eliminating the bias that results from there being more controls for some ESOP companies than for others. Numbers in the table represent average assets per participant for all plans for matched companies that either returned a survey or filed a Form 5500, even if the companies did not use one of the plans listed. Therefore, a zero for average assets in any plan is treated as a number and averaged together with other numbers.

^bThis number is weighted by .6 to take into account control companies that did not return the 5500 Form, and to bring the number into line with survey data. See main text for more explanation.

^cOne control company reports ESOP assets on the 5500 form, though to the best of our knowledge the company did not have an ESOP trust in place in 1995.

* p < .05

**p < .03

***p < .003

Both measures of assets per participant (rows 1 and 2 of table 5) indicate that ESOP companies provide substantially higher assets per participant than matched comparison companies. The differences s are are statistically significant, meaning that they are very unlikely to be the result of random sampling errorchance. (Throughout this article, when we say “significant,” we refer to statistical significance, not to the magnitude of difference between the two samples.) However, the two techniques of determining assets per participant yield very different results. For companies that returned surveys, the first row seems to yield the more accurate result because the figure in the second row (for total assets divided by the sum of participants) gives us a figure for the overall average that is lower than the figure just for the ESOP plans (row 4). Use of the first measure, which assumes a participant in one plan to be a participant in all others, is given some support by tables 6 and 7, in which the participation rate in different plans (using data from the surveys) can be compared to the percentage of total employment represented by different wage categories.

Table 6, which for the ESOP companies provides retirement plan participation rates (plan participants per wage category as a percentage of total employment) for the ESOP companies, (plan participants per wage category as a percentage of total employment), indicates that the highest overall rates are in the ESOP, 401(k), and profit sharing plans, with those percentageshich either approaching or exceeding 100%.[8] The highest rates in the control companies, as indicated in table 7, are for the 401(k), defined benefit, and profit sharing plans, with the percentagesrates in the 70% to 90% range. While these numbers cannot tell us that that, on average, a participant in one plan is guaranteed to be a participant in all others, they do give us some confidence that there is broad participation in most plans in ESOP companies, and that the participation rate in ESOP plans is probably higher than the rate in control companies. Therefore, these figures indicate that when using the survey data, comparing the sum of assets in ESOP and control companies is a reasonable thing to do. Because participation rates appear to be lower in the control companies than in the ESOP companies—meaning that employees in the control companies are less likely to be participants in all plans—any bias will be more in the direction of inflating the control company numbers. The per-participant retirement asset calculations for ESOP companies and controls may not be accurate to the dollar, but the relationship between the values appears to be reasonable.[9]

Table 6. Wage Sector as Percentage of Total Employment, and Plan Participants as Percentage of Total Employment, for ESOP Companies (Data from Surveys)

Hourly Wage Category	Employees in Wage Category as % of Total Employment n=37	ESOP Participation Rate* n**=36	401(k) Participation Rate* n**=10	Defined Benefit Participation Rate* n**=2	Profit Sharing Participation Rate* n**=2	Other Participation Rate* n**=1
Under $6/hour	4%	2%	1%	0%	0%	0%
$6.01 to $10/hour	23%	11%	11%	1%	3%	0%
$10.01 to $14/hour	23%	32%	34%	12%	18%	1%
$14.01 to $20/hour	23%	29%	55%	37%	75%	11%
$20.01 to $40/hour	22%	17%	29%	16%	35%	14%
Over $40/hour	5%	4%	15%	0%	0%	1%
TOTALS	100%	95%	145%	66%	131%	27%

*Percentages are only for those companies which have the designated plans. Participation rates are derived by dividing the number of plan participants in each wage category by total company employment, then averaging for all ESOP companies.

**n equals number of companies with this plan.

HOW CAN TOTALS IN BOTTOM ROW OF TABLE 6 EXCEED 100%? (SCOTT--endnote 8 explains this. Former employees in some wage brackets may still be participants in a plan, even though they no longer work for the company.
Table 7. Wage Sector as Percentage of Total Employment and Plan Participants as Percentage of Total Employment for Control Companies (Data from Surveys)

Hourly Wage Category	Employees in Wage Category as % of Total Employment n=67	401(k) Parti- cipation Rate* n**=18	Defined Benefit Participation Rate* n**=6	Profit Sharing Participa- tion Rate* n**=8	Other Participa- Rate* n**=1
Under $6/hour	6%	1%	1%	.4%	0%
$6.01 to $10/hour	31%	9%	15%	10%	0%
$10.01 to $14/hour	22%	15%	39%	17%	0%
$14.01 to $20/hour	21%	22%	27%	18%	14%
$20.01 to $40/hour	17%	21%	5%	18%	0%
Over $40/hour	3%	3%	.4%	11%	9%
TOTALS	100%	71%	87.4%	74.4%	23%

* Percentages are only for those companies which have the designated plans. Participation rates are derived by dividing the number of plan participants in each wage category by total company employment, then averaging for all control companies.

**n equals number of companies with this plan.

The measure that is most accurate for the survey results may not be most accurate for the Form 5500 results, however. As we can see from row 1 of table 5, the dollar figure from the 5500 forms for the sum of assets per participant is significantly higher than the equivalent figure from the surveys. This is true even if we look only at the 5500 results for those companies that also returned surveys. The average sum of assets per participant for surveyed ESOP companies that were matched with controls is $45,317 (n=36). The discrepancy between the $45,317 (frorm 5500 data) and the $32,213 (from survey data) for ESOPs may be due to companies that filed the 5500 double reporting asset totals for plans that have more than one feature, e.g., a KSOP, or a profit-sharing ESOP. This assumption is supported by the other rows in Table 5. For In ESOP companies, 401(k) plan assets are significantly higher in the 5500 column than in the survey column, as are profit sharing assets. The figures for For ESOP plans, , on the other hand, the form 5500 data and the survey data are not that differentmore consistent for the 5500 versus the survey results.[10]

So which numbers should we use? The survey data in row 1 of table 5 appears to accurately represent benefits for both ESOP and control companies, while the 5500 data in row 2 appears to be accurate for ESOPs but not for controls. We can, however, weight the control company responses in row 2 to take into account those comparison companies that do not have plans. Assuming that the figure for control companies should be $12,500 (close to the sum of average assets number from the survey), the weight would be approximately 0.6 ($12,500 ¸ $21020). In all analyses in the report using 5500 data, control company responses for the second measure of assets (total of all assets divided by the sum of participants) will be weighted by 0.6. HOW IS THIS “0.6” WEIGHT DERIVED? SCOTT--see three lines above.

Interpreting the Results

The numbers in table 5 indicate that the average value of assets per participant is significantly higher in the ESOP companies than in the controls. Looking at the first two columns, representing data from the surveys, we see that the average value in the ESOP companies is $32,213, while the average value in the control companies is $12,735. The composition of the numbers differs significantly as well. For the typical ESOP participant, the ESOP represents 75% of the combined asset value of his or her retirement accounts. Of the 75% that the ESOP holds, 80% is in company stock,[11] meaning that 60% (.75 × .80) of the asset value represented by the ESOP is in company stock. Of the remaining value in the typical ESOP participant’s retirement accounts, 12% is from 401(k) assets, 4% from defined benefit assets, and 2% from profit sharing plans. In the control companies, 70% of the value of the assets is from 401(k) plans, while 3% is from defined benefit plans and 11% from profit sharing plans. So while the value of the ESOP company assets is approximately $20,000 higher than the value of the control company assets, the ESOP investment is heavily concentrated in the stock of the employing company and thus carries more risk. On the other hand, the diversified piece of the ESOP participant’s retirement assets (40% of 32,000) is almost identical to the total assets of non-ESOP participants.

What do these per participant assets mean to employees at different wage levels? Looking at ESOP companies that allocate stock to employee accounts either on the basis of payroll (28 out of the 40 companies for which we have data) or payroll to a cap (another 5 companies, for a total of 33 out of 40 who responded),[12] we can calculate a number representing assets per participant per wage category.[13] The results in table 8 should be taken as suggestive only, since we are estimating what the value is of assets per employee in each company --; we do not know the actual number. Furthermore, the value for the wage category between $6.01 and $10 an hour is out of line with the other wage categories, indicating that something unusual may be going on in a few companies. Also, the number for each wage category is derived from the sum of values for the various plans, and we cannot be sure that assets for the 401(k) plan, defined plans, and so on are allocated by W-2. Given these caveats, the table still gives us a sense of how people at

different wage brackets benefit from an average retirement asset valued at a little over $30,000.

Table 8. Asset Values for Individuals in Different Wage Categories

Wage Level	Total Benefit Payment/Employee	N of Companies*
Under $6/hour	$6,203	3
$6.01 to $10/hour	$37,668	17
$10.01 to $14/hour	$18,220	23
$14.01 to $20/hour	$30,810	22
$20.01 to $40/hour	$62,744	22
Over $40/hour	$158,593	14

*Each wage category includes data only for those companies for which we were able to make calculations—i.e., zeros are treated as missing data.

Translated to monthly payments, if an average asset value of $32,000 earning a 5.5% interest rate were paid out monthly for 20 years, with the principal declining to zero at the end of that period, the monthly payments would equal $220.[14] If $18,200, the estimated value of retirement benefits for an employee in the $10 to $14 an hour range, were paid out in the same manner, the monthly payment to the individual would be $125. The monthly value to an individual in the over-$40 per hour category would be $1091. By contrast, 70% of the monthly income for a full-time worker in the $10–$14 per hour bracket would be approximately $1,456 (before taxes). For employer funded defined benefit pension plans, the rule has traditionally been that a covered employee could count on 70% of the last three years’ salary as a retirement benefit (see Blasi and Kruse 1991, p. 94).

The average value of $32,000 is based on the current value of the assets. If the company continues to make contributions to company stock or to other retirement plans, and/or the value of the stock increases, the value of the assets will increase. It is therefore of interest to know how much of payroll the company is putting into retirement assets on an ongoing annual basis. The percentages in table 9 are derived by dividing a company’s total compensation for 1995 (data from the Employment Security Department’s database) into the amount the company reported contributing to the different plans for that year (data from the survey of companies and from Form 5500). In terms of the total contributed to all plans, the percentages are very close for both the Form 5500 and the survey data. ESOP companies in 1995 contributed between 9.6% and 10.8% of payroll to all plans, while the control companies contributed between 2.8% and 3.0% (although the percentages from the 5500 database for control companies should probably be reduced by about half because this data does not include companies that had no retirement plans). The composition of the ESOP company totals is different for the 5500 data and the survey data, with the survey data showing almost all contributions coming from contributions to the ESOP, and the 5500 data showing significant percentages for 401(k) and profit sharing plans. This difference is probably a reflection of some ESOP plans also being profit sharing and 401(k) plans. The 5500 data does not include control companies that have no plans, while the survey data does. The end result of these levels of contribution, if continued annually, would be ESOP company employees seeing the value of their retirement assets increase at three to four times the rate of comparison companies due to the increased rate of company investment alone, all other things (e.g. relative stock values) being equal. What’s the assumed rate of stock growth in this calculation? Assets increase from both contributions and growth in value, and more from stock value growth than contribution level as accounts mature. SCOTT--but the value of the 401k stock for comparison companies will increase as well, so we hold the increase in stock values constant and just look at the rate of contributions.

Table 9. Percentage of Pay Contributed to Plans

Percentage of Pay Contributed To:	ESOP, from Survey Data^a n=37, unless noted	Controls, from Survey Data n=37 (weighted)	ESOP, from Form 5500 Data n=61	Controls, from Form 5500 Data n=130 (unweighted)
All Plans	10.8%*	2.8%*	9.6% *	3.0% *
401(k) Plans	1.0%	1.5%	2.5%	1.9%
ESOP Plans	10.0% (n=36)	0%	5.6%	0%
Defined Benefit Plans	0%	.3%	.1%	.1%
Profit Sharing Plans	.04%	1.0%	1.4%	.7%
Other Plans	.2%	.1%	0%	.3%

In this table, a zero is included in the final percentage; it is not treated as missing data. Data from the 5500 forms for the control companies is unweighted because two controls and one ESOP were dropped from the calculations due to missing data or percentage of wages that appeared unreasonably high (over 200% of payroll).[15]

^aNumbers for the individual plans in this column do not quite add to the number for “All Plans” due to missing data for ESOP plans for one company.

*p< .01

Independent Variable Analysis: SIC Code, Unionization, Percentage of Ownership, and Participation

For independent variable analysis we will look first at industrial sector (measured by one-digit Standard Industrial Classification codes), then at unionization, percentage of ownership, and participation. Table 10 presents retirement assets by one-digit SIC code,[16] with averages both from survey data and from 5500 data. The survey shows higher asset values for ESOP companies in five out of the seven SIC codes (3 and 5–8) and lower values in SIC codes 1 and 2. The 5500 data shows higher ESOP values in all seven SIC codes. While we will explore this a bit more vigorously later when we present regression runs, it appears that SIC codes by themselves do not explain the difference in values between ESOP and control companies; i.e., the higher ESOP values are not loaded up in only a few SIC codes.

Table 10. Retirement Assets by One-Digit SIC Code

Assets per Participant by 1-Digit SIC Codes	ESOP Companies	Control Companies
SIC Code 1 Assets/Participant, Survey Assets/Participant, 5500	$12,489 (n=8)^a $50,852 (n=8)	$22,148 (n=13) $15,927 (n=15)
SIC Code 2 Assets/Participant, Survey Assets/Participant, 5500	$9,099 (n=5) $17,465 (n=7)	$10,443 (n=12) $10,217 (n=12)
SIC Code 3 Assets/Participant, Survey Assets/Participant, 5500	$43,389 (n=7) $30,104 (n=12)	$23,150 (n=14) $12,634 (n=32)
SIC Code 5 Assets/Participant, Survey Assets/Participant, 5500	$37,872 (n=6) $29,048 (n=15)	$3,222 (n=12) $9,828 (n=33)
SIC Code 6 Assets/Participant, Survey Assets/Participant, 5500	$87,692 (n=4) $33,963 (n=13)	$20,807 (n=5) $14,894 (n=22)
SIC Code 7 Assets/Participant, Survey Assets/Participant, 5500	$10,552 (n=4) $30,645 (n=5)	$3,780 (n=7) $4,945 (n=8)
SIC Code 8 Assets/Participant, Survey Assets/Participant, 5500	$40,844 (n=3) $31,503 (n=6)	$15,034 (n=5) $18,719 (n=12)

Control company numbers for 5500 data are weighted by .6 to reflect the companies that do not have retirement plans, and therefore do not file Form 5500.

^aWhile the n is eight for both survey and 5500 data in this cell, only six of the companies actually overlap.

What about the independent effect of unions?[17] Table 11 presents asset per participant values just for ownership and unions as the independent variables, and table 12 looks at ownership and unionization for SIC codes 2 and 3.[18] The apparently large difference between union and nonunion control companies represented by the data from the surveys largely evaporates in the data from the 5500 forms, though Table 12 indicates that there is still a difference between union and nonunion control and ESOP companies in SIC Codes 2 and 3 (though the difference is not statistically significant). According to the regression analysis summarized in Table 13, the presence of an ESOP increases per participant asset value by $20,298.72 when sector, company size, and unionization are held constant. The presence of a union does not have a statistically significant effect on asset values.

Table 11. Assets per Participant, Union and Nonunion Companies

Assets per Participant	ESOP Companies with Unions	ESOP Companies Without Unions	Control Companies with Unions	Control Companies Without Unions
Assets per Participant, Survey Data	$23,612 (n=6)*	$33,877 (n=31)*	$87,498 (n=7)*	$6,052 (n=61)*
Assets per Participant, 5500 Data	$21,990 (n=9)**	$33,542 (n=57)**	$15,315 (n=12)**	$12,280 (n=122)**

*Using survey data, difference between union and nonunion companies, ignoring ESOP vs. control, is significant at the .001 level. Difference between ESOPs and controls for survey data significant at the .05 level. Difference between the union and nonunion controls significant at the .0000 level.

**Difference between ESOP and controls significant at the .0000 level.

Table 12. Assets per Participant for SIC Codes 2 and 3

Assets per Participant for:	ESOP Companies With Unions	ESOP Companies Without Unions	Control Companies with Unions	Control Companies Without Unions
SIC Codes 2 & 3 Assets/Participant, 5500 Data	$30,274 (n=6)	$23,220 (n=13)	$14,758 (n=11)	$11,047 (n=33)

Table 13. Regression of ESOP on Asset Value, Controlling for Unionization and SIC Code

Variable	Assets per Participant
	B	(SE)
ESOP (ESOP = 1)	20298.72	(3915.64)	*
Unionized (1=union)	-6.16	(4.97)
SIC 1	7730.63	(6628.40)
SIC 2	-5349.49	(7354.37)
SIC 5	-1699.03	(5565.99)
SIC 6	3026.62	(6017.08)
SIC 7	-2412.08	(8518.09)
SIC 8	3226.97	(7462.38)
Company size	-1063.39	(6773.68)
Constant	12715.43	(4205.78)	*
R^2	0.15
Adjusted R^2	0.11
N	198
* = p < 0.01

What is the effect of majority ownership? The average value of assets per participant for 12 majority-owned ESOP companies is $30,694 using the survey data or $36,369 using the 5500 data, while the average value for 21 minority-owned ESOP companies is $37,000 using the survey data or (for 19 companies) $42,632 using the 5500 data. In either case the difference is around $5,600. But while the majority-owned companies appear to fare worse than those that are minority-owned, their per-person asset values are still significantly higher than the values of their matched controls, as can be seen in table 14.

Table 14. Asset Values for Majority and Minority-Owned Companies and Their Respective Controls

Asset Data From:	Majority ESOP Companies n=12	Controls for Majority ESOP Companies n=12 (Weighted)	Minority ESOP Company	Controls for Minority ESOP Company
Survey Data	$30,894	$14,803	$36,700* (n=21)	$7,259* (n=21 weighted)
Form 5500 Data	$35,847*	$13,978*	$42,632 (n=19)**	$10,835 (n=19 weighted)**

*p < .01

**p <.05

It is curious why, despite the higher average values of retirement assets in the majority-owned ESOP companies compared to their matched controls, the average assets held by plans in majority-owned ESOP companies are lower in value than the average assets held by plans in minority-owned ESOPs. The expectation would be that the more stock allocated to employee accounts, the higher the value of the assets per person. Are the minority-owned companies contributing more to 401(k) or profit sharing plans than the majority-owned companies, and driving the values higher that way? The data refutes this, indicating that the contribution percentages for majority- and minority-owned companies to 401(k) plans are about the same. Is there something going on at the 1-digit SIC code level that could explain the difference? We cannot find anything in regression analysis. Are there fewer participants relative to total employees in the minority-owned companies, thus giving those participants more value per person? On the contrary, the ratio of participants to employees for majority-owned companies is 67%, and for minority-owned companies 89%. Does unionization have something to do with it? There is a higher percentage of unionization in the majority-owned companies, but there is no statistically significant effect of unionization on the relationship between percentage of ownership and asset value. Is there a difference in how old the plans are in the two categories of companies? No There is no difference that matters. Well then, what about the fact that minority-owned companies pay better on average than the majority-owned companies? Is there a relationship between lower asset values and lower pay?

There is, in fact, a correlation of .33 between median pay and asset value (p < .01), but a curious thing happens when one breaks out this relationship by majority-owned and minority-owned companies. For the majority-owned companies, the correlation is .66 (p < .01), while for minority-owned, the correlation is -.06 (no significance). The scatterplots in figures 1 and 2 illustrate the difference.

Figure 1. Assets by median wage, majority-owned ESOPs

Figure 2. Assets by median wage, minority-owned ESOPs

With the majority-owned plot, it is easy to see that the asset per participant values generally increase as wages increase, but with the minority-owned companies it looks like a couple of lower-paying ESOP companies have very high asset values, which interrupts an otherwise general tendency for higher-paying companies to have higher assets. In fact, when we remove the two outliers, the correlation increases to r=.26, p < .01. The mean value of the assets then drops to $26,063 (Form 5500 data), nearly $10,000 below the mean value of the majority-owned companies. The correlation is still higher for the majority-owned companies (higher median pay, higher benefit levels), and there is no statistical significance in the difference between the majority and minority-owned ESOPs, so removing the two outliers does not completely clarify the relationship between percentage of ownership and asset value. But their removal does bring results more in line with expectations.

What about might be the effect of participation in workplace decisions on retirement wealth? We measured participation with two sets of questions in our survey, by asking companies first what employee participation techniques they use and second what degree of influence employees have over various decisions (section V of the survey in the appendix). In a 1992 study of company growth, Kardas et. al. had asked companies the same question about participation techniques, and found that the more techniques a company used, the more likely they were to have positive growth relative to their competitors. Keogh had found the same things in an earlier study of Washington State companies (see Keogh n.d.). The supposition for adding up techniques is that the use of a larger number of techniques can be taken as an indicator of a company’s commitment to experimentation, with varied opportunities for feedback about the process, that would result in productive employee participation. Support for this theory came from the fact that majority-owned companies in the 1992 study that used more of the participative practices had higher growth rates than any other category of companies. Being participatory and majority-owned, these companies presumably would be even more likely to seek out those participation practices that both provide for employee satisfaction and lead to more productive practices.

Given our previous findings, we expected to see a similar pattern in this study: more participatory practices linked to higher growth would mean either higher wages for the employees or higher stock values that reflected the companies’ success. In table 15 we can see the mean asset values from survey and 5500 data for more participatory and less participatory ESOP companies and controls, with more participation defined as the use of five or more participation techniques, and less participatory the use of four or fewer. The survey data appears to indicate that more participatory ESOP companies have asset values nearly twice as large as the less participatory ESOPs (though that relationship is without statistical significance), and that both more and less participatory ESOPs have values higher than the more participatory control companies, which themselves are higher than the less participatory controls. However, this greater advantage of more participatory ESOPs disappears when we look at the 5500 data, with the less participatory ESOPs in fact having values nearly $12,000 higher than the more participatory ESOPs. Removal of one outlier from the less participatory ESOPs drops the value for that group to $28,083, less than the value for the more participatory ESOPs but not as dramatically less as with the survey data. Regression analysis confirms that little is going, on (at least in based on the 5500 data, ) between participation has little effect on and asset value when controlling for employee ownership and SIC codeindustrial sector. There remains a strong association, however, between employee ownership itself and asset value.[19] We get essentially the same results when using the other variable that measures employee participation by the degree of employee influence on various kinds of decisions.

Table 15. Asset Values for ESOPs and Controls, Broken Out by Participation

Asset Data From:	ESOP Companies with More Participation	ESOP Companies With Less Participation	Control Companies with More Participation	Control Companies with Less Participation
Survey*	$43,981 (n=18)	$23,541 (n=17)	$19,306 (n=26)	$13,481 (n=35)
Form 5500**	$32,662 (n=18)	$44,411 (n=17)	$19,304 (n=15)	$8,773 (n=21)

*Overall difference between ESOP and control companies significant at the .1 level.

**Overall difference between ESOP and control companies significant at the .01 level.

What about other possible explanations for the differences in asset values between employee owned and control companies on asset values? As with the wage data, company size is negatively correlated with asset values (-.11, p < .1), as is the mean start date for retirement plans at the company (-.22, p < .01). Bigger companies or companies with earlier start dates do not have higher asset levels.[20]

Summary of Findings on Retirement Assets

Using two data sources (a survey of companies and Form 5500 records), we found asset values in ESOP companies to be roughly 2.5 times the value of assets in the matched controls. Overall, there also appeared to be higher rates of employee participation in ESOP retirement plans than in the controls, while a high percentage (between 58% and 71%) of control companies appeared to have no retirement plan at all. Within the ESOP companies, some 60% of the value of the retirement assets was in the form of company stock, while in those control companies that have retirement plans, approximately 70% of the value of the assets was in stock offered through 401(k) plans. Most ESOP companies reported allocating their stock on the basis of payroll, which meant that the value of total shares held per person varied greatly depending on a participant’s wage or salary bracket. In 1995, ESOP companies contributed at least three times the percentage of payroll to retirement plans as did the control companies.

In trying to sort out the contribution made by several independent variables to the differences in asset values, we found that company categorization by one-digit SIC code did not diminish the independent explanatory power of ESOP versus non-ESOP ownership. Within ESOP companies, average asset values were lower in unionized ESOP companies than in nonunion ESOP companies. On the other hand, asset values in the unionized controls were higher than in the nonunion controls, with the greatest differences showing up in the survey data. Minority-owned ESOP companies had higher asset values than majority-owned ESOPs, though the removal of two outliers dropped the values found in minority-owned companies below the values found in majority-owned. In either case, the value of assets in majority-owned ESOP companies was significantly higher than the value of assets in the matched controls. In terms of participation, there was no clear pattern. On the one hand, Whereas survey data indicated that more participatory ESOPs had much higher average asset values than less participatory ESOPs, while 5500 data indicated the opposite. With both data sources, more participatory control companies had higher asset values than less participatory controls. As with wage data, company size is negatively correlated with asset values, as is the mean start date for a companies’y’s retirement plans.

Wages

Given that the value of retirement benefits is significantly higher in ESOP companies than in comparison companies, do employees at ESOP companies typically take lower wages to make purchase of company stock possible? The simple comparison of means summarized in table 16 suggests otherwise. Here, mean and median wages, wages at the 10th and 90th percentiles, and the ratio between wages at the 90th and 10th percentiles are presented for ESOP companies and their matched controls. The results show that ESOP companies pay both higher average as well as higher median wages. The average ESOP company wage of $19.09 is 12% higher than the average control company wage of $17, and the median ESOP company wage of $14.72 is 8% higher than the median control company wage of $13.58. At the 10th percentile, wages in the ESOP companies are 4% higher than in the controls. Chance cannot be ruled out as an explanation for the differences in the mean, median, or 10th percentile wages. At the 90th percentile, ESOP wages are 18% higher than comparison wages, causing the ratio of 90th to 10th percentile wages to be 11% higher in ESOP companies.

Table 16. Hourly Wages for ESOP and Control Companies

Hourly Wage	ESOP Companies	Control Companies
Mean Hourly Wage	$19.09 (n=90)	$17 (weighted n=90)
Median Hourly Wage	$14.72 (n=90)	$13.58 (weighted n=90)
Hourly Wage at 10th Percentile	$8.85 (n=90)	$8.47 (weighted n=90)
Hourly Wage at 90th Percentile	$30.91 (n=89)*	$26.12 (weighted n=89)*
Ratio of 90th to 10th Percentile (Average of all companies)	3.49 (n=89)	3.15 (n=89)

*p<.1

Results for the control companies are weighted so that the sum of control companies for each ESOP company equals one, thus eliminating the bias that results from there being more controls for some ESOP companies than for others. Companies included in the table are all but one of the ESOP companies in the Washington State Employment Security Department’s database for which we have at least one match, plus, of course, the matches themselves. (One ESOP company was eliminated because it had a median hourly wage of $96, more than four times the median wage for its matched controls. Two others were removed because the companies only reported wages for the employees, not hours worked, thus making it impossible to calculate wages per hour.)

How do these preliminary results address the equality of material condition and equality of opportunity questions raised at the beginning of the paper? Compared to their competitors, Washington State ESOP companies typically have higher pay, so employees at the middle of the pay scale are better off in terms of take-home pay working in an ESOP company than in a comparable conventional company. On the other hand, workers at the bottom of the pay scale in ESOP companies do not make much more than comparable workers in competing companies, and there is a greater distance between those at the bottom of the wage scale and those at the top than in conventional companies. There is some greater distance between the employee at the median pay level and the employee at the 90th percentile in an ESOP compared to a conventional company (2.06 versus 1.96), but the difference is not so great as the distance at the 10th percentile.[21]

What happens to these results when we control for other factors, such as industrial sector, unionization, majority ownership, company size, and workplace participation? Table 17 presents the results, for both ESOP and control companies, for the median hourly wage and hourly wages at the 10th and 90th percentiles broken out by one-digit SIC codes. While the differences vary between different SIC codes, in every case but one the median wage for the ESOP companies is higher than for the control companies. The exception is SIC code 5 (Wholesale and Retail Trade), where the ESOP median wage is nearly equal to the control company median wage. The greatest difference between ESOPs and controls is in SIC code 6 (Finance, Insurance, and Real Estate), where the difference is 15%.

It is a similar story for wages at the 90th percentile. In every case the wages in the ESOP company are higher, with the difference ranging from 8% in SIC codes 7 (Services) and 8 (Services, including legal, engineering, and educational services) to 43% in SIC code 2 (Manufacturing, including lumber, paper, printing, and petroleum refining). Wages at the 10th percentile are much closer, but again, in five out of the seven SIC codesindustrial sectors for which we have data, wages are higher in the ESOP companies are higher.

Table 17. Median, Tenth, and Ninetieth Percentile Wages Broken Down by One-Digit SIC Code

Hourly Wages By 1-Digit SIC Code	ESOP Companies	Control Companies
SIC Code 1 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$19.27 (n=10) $11.08 (n=10) $35.02 (n=10)	$18.85 (n=51) $10.56 (n=50) $27.83 (n=50)
SIC Code 2 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$14.62 (n=9) $8.90 (n=9) $39.19 (n=9)	$12.96 (n=44) $8.90 (n=44) $27.35 (n=43)
SIC Code 3 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$14.87 (n=19) $9.37 (n=19) $26.72 (n=19)	$13.15 (n=99) $8.54 (n=99) $24.15 (n=98)
SIC Code 4 (no companies in this SIC code for this analysis)
SIC Code 5 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$12.41 (n=21) $7.46 (n=21) $25.89 (n=21)	$12.52 (n=108) $7.71 (n=108) $23.50 (n=106)
SIC Code 6 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$13.92 (n=19) $8.23 (n=19) $34.51 (n=18)	$12.14 (n=65) $7.82 (n=65) $28.44 (n=64)
SIC Code 7 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$12.36 (n=6) $7.62 (n=6) $25.91 (n=6)	$11.59 (n=27) $7.24 (n=27) $23.89 (n=27)
SIC Code 8 Median Wage* Wage at 10th Percentile* Wage at 90th Percentile**	$19.76 (n=6) $11.40 (n=6) $36.77 (n=6)	$18.91 (n=26) $11.86 (n=26) $33.95 (n=25)

*Differences for median wage and 10th percentile wage by SIC code significant at the .0000 level.

**Differences for 90th percentile wage by SIC code significant at the .01 level.

Since a few ESOP companies, because of the nature of their business, were matched with control companies from more than one SIC code, control companies in the above analysis are unweighted.

What happens to these numbers when we take unionization into account? Table 18 shows median wages and wages at the 10th and 90th percentiles by ESOP and comparison companies, controlling for unionization.

Table 18. Median, 10th, and 90th Percentile Wages Broken Down by Unionization

Hourly Wage for:	ESOP Companies with Unions	ESOP Companies Without Unions	Control Companies with Unions	Control Companies Without Unions
Median Wage	$15.79 (n=11)	$14.57 (n=79)	$15.79 (n=27)*	$13.62 (n=393)*
10th Percentile Wage	$10.05 (n=11)	$8.69 (n=79)	$10.99 (n=27)**	$8.45 (n=392)**
90th Percentile Wage	$23.69 (n=11)	$31.93 (n=78)	$24.40 (n=26)	$26.11 (n=387)

*p< .1

**p<.01

The Median Wage row in table 18 suggests that median pay in unionized control companies is 16% higher than in nonunionized controls and that median pay in unionized ESOP companies is 8% higher than in nonunion ESOPs. Furthermore, union wages for ESOP companies are 16% higher at the 10th percentile and 26% lower at the 90th percentile than are wages in nonunion ESOPs, while the equivalent percentages for the unionized control companies are 30% higher (10th percentile) and 7% lower (90th percentile). These differences hold up, for the most part, when the data is are broken out by one-digit SIC code. For three out of the four SIC codes (SIC codes 1 through 3) where we know union companies to be present, the wage at the 10th percentile is higher for union control companies than for nonunion controls, and in all four SIC codes the wage at the 90th percentile is lower for the union controls. For ESOP companies, the wage at the 90th percentile is lower in every case in unionized companies, while in two out of the four SIC codes, the wage at the 10th percentile is higher.

Overall, these numbers suggest that in non-ESOP companies, unions have the effect of raising the median wage as well as the wage at the 10th percentile, and that to some extent they may do the latter by holding down wages or salaries at the upper percentiles. Unions have a smaller effect in pushing up the median wage in ESOP companies, which already have a higher median wage than do the control companies. Unions in ESOP companies also appear to raise wages at the 10th percentile in some industrial sectors and to lower wages at the 90th percentile in all industrial sectors. The wage at the 90th percentile is much lower in unionized ESOP companies than in nonunion ESOP companies, while the difference between union and nonunion controls is not nearly so great. The overall median wage in unionized controls is higher than in nonunion controls due to hHigher wages at the 10th percentile and with wages onlyslightly lower wages (6% lower) at the 90th percentile would explain. why the overall median wage for unionized controls is higher than for nonunion controls, while h Higher wages at the 10th percentile combined with wages 26% lower at the 90th percentile would explain why the union ESOP company median wage is not higher than nonunion ESOP companies. While the numbers are only suggestive, they do conform to evidence about the effect of unions on wages presented by other researchers (see, e.g., Freeman 1994, chapter 2).

Are there other independent variables that explain some of the wage differences between ESOPs and their matched controls? Do larger companies tend to pay better than smaller firms? What about majority-owned ESOP companies and more participatory ESOP companies? Do we find the effect that we saw in the 1993 study of employment and sales growth in Washington State ESOP companies (see Kardas et. al. 1994), where the combination of apparent company commitment to participation and majority ownership typically meant higher levels of growth?

One might assume that company size would be positively associated with pay (large companies have more resources, are more likely to have unions, and so on). However, the correlation between employment size and median pay is -.1413 (p=.001), with similar negative numbers for the correlation between employment size and both 10th and 90th percentile wages. This may be a reflection of the fact that some of the larger companies in the study are in retail trade and business services (temporary employment agencies, for example), which have low-paying jobs, while some of the smaller are in computer software, finance, insurance, or real estate, some of which tend to pay better. The negative relationship between employment size and pay holds up when controlling for ownership, SIC code, and unionization.

The independent effect of ownership, type of industry, and unionization can be seen in the results of a regression analysis that examines median, 10th percentile wages, and 90th percentile wages as dependent variables, and ownership, company size, industrial sector, and unionization as independent variables. As table 19 shows, statistically significant determinants of variation in median wage are the presence of an ESOP; location in sectors SIC1 and SIC 8; and unionization. Company size has an inverse relationship to median wages: smaller company size correlates with higher wages. Location in SIC 1 increases the median wage by about $5.40, compared to manufacturing work and when controlling for other factors. Location in one of the service sectors (SIC 8) similarly increases the median wage by about $5.77. The presence of an ESOP increases the median wage by $.99, and the presence of a union drives up the median wage by $1.51, controlling for other factors.

Statistically significant determinants of variation in 10th percentile wages, when controlling for other factors, are company size; location in sectors SIC1, SIC 5, and SIC 8; and unionization. Factors contributing to variation in 90th percentile wages are presence of an ESOP, company size, location in SIC 1, SIC 2, SIC 5, SIC 6, and SIC 8; and determinants of company level inequality as measured by the ratio of 90th to 10th percentile wages are ESOP; company size; location in sectors SIC 2, SIC 6, SIC 7, and SIC 8; and unionization.

Table 19. Determinants of Median Wage, 10th and 90th Percentile, and Ratio of Wages in ESOP Companies and Comparison Companies
Variable	Median Wage			10th Percentile			90th Percentile			Ratio 90th/10th
	B	(SE)		B	(SE)		B	(SE)		B	(SE)
ESOP (ESOP = 1)	0.99	(0.57)	^x	0.16	(0.33)		5.07	(1.81)	**	0.40	(0.20)	*
Company size	-0.002	(0.00)	**	-0.001	(0.00)	*	-0.004	(0.00)	*	0.000	(0.00)	^x
SIC 1	5.40	(0.77)	**	1.86	(0.45)	**	4.54	(2.44)	^x	0.05	(0.27)
SIC 2	-0.36	(0.82)		-0.04	(0.47)		5.39	(2.58)	*	0.78	(0.28)	**
SIC 5	-0.64	(0.63)		-0.78	(0.36)	*	-0.54	(1.99)		0.21	(0.22)
SIC 6	-0.83	(0.70)		-0.61