Union status and merger activity in the U.S.

I. Introduction

Extensive research has examined the effects of industry characteristics on union status [Farber, 1983; Hirsch and Berger, 1984; Belman, 1988]. While these studies specify several industry determinants of union status, they have not considered the possibility that corporate merger activity may influence a worker's union status. Nonetheless, unions have reported substantial membership losses following nonconglomerate mergers in the petroleum, airline, and publishing industries during the 1980s. Contrary to these reports, this study proposes that the potential for workers receiving high union wages and the lower cost of unionization enhances the probability of union membership in industries composed of firms formed from nonconglomerate mergers.

Industry information from the Federal Trade Commission (FTC) is used to examine the separate union status effects of horizontal, vertical, and conglomerate mergers. Distinguishing merger types avoids confounding their effects on union status. This study also considers the possibility that merger activity is not determined exogenously since acquiring employers may target highly organized firms with the objective of extracting union rent. To address the estimation bias caused by merger endogeneity, predicted values for each merger type are included in the status equation. Most importantly, the approach used in this study allows testing of the hypothesis that nonconglomerate merger activity is associated with a greater probability of an employee being a union member. Such a test represents an extension over previous research on the determinants of union status and highlights the potential role of unionization in determining the pattern of mergers.

II. Union Membership Across Merger Type

Theory suggests that union representation can be expected to develop as long as the marginal benefits of unions are greater than the marginal costs of organizing [Hirsch and Berger, 1984]. Within this framework, mergers may influence the benefits of union membership by increasing the rents that unions are able to redistribute to workers. Given the adherence to antitrust laws that prohibit mergers creating market dominance, it is most likely that rents are generated by enhanced production efficiency. For instance, the horizontal merging of companies that produce the same good can engender efficiency gains through synergy and x-efficiency [Pratten, 1971]. The vertical merging of suppliers with their buyers can improve efficiency by eliminating input choice distortion, which in turn reduces the costs of transaction for these vertically merged employers [Pratten, 1971; Scherer, 1980]. Even the conglomerate merging of completely dissimilar firms can generate increased efficiency by introducing economies of scope into the production process. Thus, the potential for paying high union wages exists for all merger types.

Mergers certainly influence union membership by changing the union's cost of organizing. Indeed, the amalgamation of once separately employed groups of workers reduces the number of employers a union faces in its organizing effort. However, the extent to which organizing costs will decline partly hinges on the merged employer's ability to divert operations away from organized plants or even reduce employment at these plants. The effectiveness of such tactics is likely to differ by diversification. As a consequence, newly-merged employers may be more or less successful in such tactics depending on whether the merger was horizontal, vertical or conglomerate [Hendricks, 1976; Rose, 1991; Heywood and Peoples, 1994].

Since a union's membership usually consists of workers from the same industry, the merging of companies that produce identical products may limit the ability of merged employers to repudiate unions. These horizontal mergers are most likely to occur within the jurisdiction of a single union, thus not enhancing the employers ability to divert operations away from union plants. Even if this type of merger does not cross union jurisdiction, the low per-member-organization costs associated with employment concentration should allow for greater membership by reducing barriers to unionization. While this view seems to contradict accounts of union membership losses resulting from these types of mergers, it is consistent with the contention that reports of declining membership are typically of a highly unrepresentative set of these merger types [Brown and Medoff, 1988].(1)

The extent of union jurisdiction has less importance for vertical mergers where workers at every plant are necessary in the production process. The importance of each plant, organized or not, in the production process strengthens the staying power of the union by diminishing the merged employer's ability to transfer operations to non-union plants or to substantially reduce production at union plants. Despite a merger between unionized and a non-unionized firm, having a buyer and seller relationship requires production from both to provide the final good or service. Hence, a firm formed by vertical integration should face relative difficulty reducing union membership in either its downstream or upstream operations.

A merger involving the consolidation of two essentially unrelated firms is the most likely to allow the repudiation of unions. These pure conglomerates are unlikely to be covered by a single union jurisdiction. Further, union staying power is far weaker since pure conglomerates do not rely on production from other acquired plants. Thus, the strong likelihood that a pure conglomerate has an enhanced ability to continue operations while reducing union production suggests that these firms are in a stronger position to reduce union coverage.

In sum, the potential for receiving high union wages and the low organizing costs suggests greater probability of union membership for individuals working in industries composed of firms formed from nonconglomerate mergers. The high cost of organizing conglomerates together with the lower benefits (due to the weakened effectiveness of a strike by workers) suggests a lower probability of unionization despite the possibility that these type of mergers may allow employers to pay high union wages.

Past research focuses mainly on the employment effects of mergers rather than any union membership effects. Despite the lack; of attention given to union status, these studies provide some insight on this issue by revealing the effect merger activity has on the extent of labor's role in the production process. Guerard [1982] shows that merger activity is associated with small declines in employment.(2) This finding may suggest that union members and other workers do not encounter severe job losses due to merger activity. However, a highly aggregate measure of employment was used in that analysis thus obscuring the effect of mergers across occupations. Taking this shortcoming into consideration, Lichtenberg and Siegel [1989] examine the separate employment effects on production, research and development, and central office employees. Except for central office employees, they do not find appreciable reductions in employment following merger activity. For example, mergers are associated with only a 5 percent employment reduction for production workers. Since a disproportionate share of union members are employed as production workers, these findings seem to provide further support for the notion that unions do not have much to fear from mergers. Despite this lack of support for a negative employment effect, it remains possible that merged employers substitute union members with low wage non-union workers once merger types are separately examined. Thus, estimating the probability of union membership could provide new information on the effects of merger activity.

III. Data and Estimation Technique

The analysis requires combining industry and individual information since individual companies do not report the union status of individual workers. Industry information from 1977-79 Federal Trade Commission Mergers and Acquisitions files specifies the type of merger and the asset value of large publicly owned acquisitions in mining and manufacturing. The 1977 Census of Manufacturing and Mining Industries supplies other industry information. Current Population Survey (CPS) files provide individual characteristic information for union and non-union members for the years from 1979-85, excluding 1982, since information on union status was not reported for that year. This sample period allows the investigation of union membership patterns following merger activity. The selected sample consists of full-time, male production workers across 75 3-digit census manufacturing industries. It is limited to production workers to allow analysis of occupations that have a tradition of high union membership levels. Women and part-time workers are omitted because their small sample size prohibits separately predicting the effect of mergers on these workers' propensity to join a union. The choice of industries is dictated by the limitations of the available data sources. This limitation of sampling is also common in the literature.

This study models union status as a discrete choice made by workers where the decision to join a union is considered dependent on a set of worker and industry characteristics. Other than the inclusion of merger types as an explanatory variable, the initial specification of the model, as depicted by equation (1), is similar to that used in previous studies of union status [Farber, 1983; Hirsch and Berger, 1984; Belman, 1988].

[Mathematical Expression Omitted],

where [Phi] is a normal probability function and [Union.sub.i] is a binary variable with a value of 1 if the individual is classified as a union member, and zero otherwise. X is the vector of individual and industry variables. Years of schooling, potential years of experience, marital status, race, occupation, geographic region of employment, and status as an armed services veteran make up the set of individual variables.(3)

The individual variables measuring human capital are included to account for the possibility that unionized employers choose highly productive workers in response to high union wages. The race of individual workers is considered to allow for the likelihood that union enforced wage standardization policies increase the probability of nonwhites choosing union employment. The marital status of workers is identified since the reduced mobility of married individuals may increase their demand for collective voice. The occupation and region of workplace of individual workers is considered to reflect the historical preference towards unionization by skilled production workers and individuals located in the Northeast and North Central regions of the U.S.

The four-firm concentration ratio and the capital-labor ratio are the other industry variables. Industry concentration is included in the union status equation since the likelihood of greater economic profits flowing from a lack of competition in concentrated industries enhances the ability of employers to meet union wage demands. The cost of organizing in such an industry is also low since they are likely to have a high concentration of workers across employers. The capital-labor ratio is included in the union status equation since productivity gains from industries employing relatively high levels of capital can generate profits that unions can redistribute to their members. In addition, greater capital intensity can lower organizing costs because the reduced mobility of the firm in the short-run leads to a more inelastic labor demand curve.

Merger is the merged industry asset share for each of the three types of mergers. Its coefficient C reflects the effect of merger activity on the probability of an individual being a union member.(4) The vector t is a set of year dummies and ([lw.sub.i,u] - [w.sub.i, nu]) measures the estimated union-nonunion log-wage gap in 1985 dollars for each individual where the subscripts u and nu indicate the individual's union status.

The inclusion of the union-nonunion wage gap in the status equation requires the use of a simultaneous equation procedure since the wage gap is determined endogenously. The two-stage process proposed by Heckman [1976] is used to address this problem. The initial stage of this procedure involves calculating the union-nonunion log-wage gap by using the estimates from the following wage equations for union and nonunion production workers:

[lw.sub.i,u] = [A.sub.u] + [B.sub.u] Z + [C.sub.u] (merger) + [D.sub.u][lambda.sub.u] + [E.sub.u]t + [[Epsilon].sub.u,i] and (2)

[lw.sub.i,nu] = [A.sub.nu] + [B.sub.nu]Z + [C.sub.nu] (merger) + [D.sub.nu][lambda.sub.nu] + [E.sub.nu]t + [[Epsilon].sub.nu,i], (3)

where Z is a vector of individual and industry characteristics including all variables in equation (1) along with hours worked per week and the industry unemployment rate.(5) Other than lambda, the remaining notation is the same as in equation (1). The inverse Mills ratio (lambda) is included in the wage equation to account for the possibility of selectivity bias. This variable is calculated using a first stage probit procedure for the union status equation that excludes the union-nonunion wage gap. The same probit procedure is used in the final stage to then estimate the structural equation (1).(6)

IV. Results

Table 1 provides means, standard deviations, estimated coefficients, and t-statistics for the union status determinants. The descriptive statistics reveal conglomerates as the merger type with the largest share of industry assets. The results for the control variables show worker characteristics, industry characteristics, and union-nonunion wage differentials with conventional signs and significance. Specifically, compared to their non-union counterparts, union members are more likely to be married, highly experienced, work in the northeast as non-transportation operative, and work in highly capital intensive industries having higher levels of product market concentration. Lastly, the finding on the time dummies suggests a declining probability of union membership from 1980-85.

[TABULAR DATA FOR TABLE 1 OMITTED]

The estimated coefficient for the different types of mergers support the hypothesis that workers are least likely to belong to a union if employed in industries with greater levels of conglomerate merger activity. Indeed, the estimated coefficient for the conglomerate merger variable is negative and statistically significant, while the sign of the other merger variables is positive and statistically significant.

The significance of the estimated coefficients on the merger variables permits the estimation of merger activity's influence on the probability that a worker is a union member. These results are revealed in Table 2. For mean worker and industry characteristics, the predicted probability of being a union member is 0.4443. Row (1) of Table 2 shows that this measure increases to 0.5199 and 0.4880 with an increase of one standard deviation in horizontal and vertical merger activity, respectively. The probability of being a union member falls to 0.4247 in the event of an increase equaling one standard deviation in conglomerate merger activity. Thus, as revealed in row (2), there is a .0633 probability difference in the conglomerate and vertical merger effect on union status.(7) The probability difference between the conglomerate and horizontal effect is 0.0872.

[TABULAR DATA FOR TABLE 2 OMITTED]

Since the union membership predictions are non-linear, the merger effect on union status is calculated for a one standard deviation reduction in merger activity. Column (3) on Table 2 reveals that the magnitude of these effects does not change appreciably when observing a decline in the three types of mergers. In sum, these estimates show non-trivial differences in the union effect of conglomerate merger and other types of mergers.

Thus far, the specification of the union status equation assumes that merger activity is determined exogenously. If this is inaccurate, finding a positive association between non-conglomerate mergers and union membership may be attributable to acquiring firms taking over unionized firms that pay premium wages. Therefore, the direction of causality may well run from unionization to mergers, thus biasing the estimation results. Indeed, past research on corporate mergers and union premiums shows that the level of union rents is positively associated with horizontal and vertical mergers but does not influence conglomerate mergers [Heywood and Peoples, 1994].

To address the possibility of merger endogeneity, predicted values for merger activity are calculated by controlling for interindustry differences in union rent. These predicted merger values are then used as determinants of union status.

A log-odds estimation procedure is used to predict the level of merger activity in an industry. This approach is sensible because it permits the inclusion of information on industries that were not involved in merger activity for the sample period.(8) To account for merged asset share having values from zero to one, this variable is converted into the following formulation as done by Voos and Michel [1986]:

Lmerger = ln [(merger + 1/2 n) / (1 - (merger + 1/2 n))], (4)

where n measures industry assets.

The prediction of merger activity is then calculated by first estimating equation (5):

[Lmerger.sub.j] = [[Beta].sub.0] + [[Beta].sub.1]V + [[Beta].sub.2]t + [[Mu].sub.t,j], (5)

where j indexes individual industries and V is a vector of industry characteristics. This set of variables includes individual industries' price cost margin, concentration ratio, capital/sales ratio, import/sales ratio, export/sales ratio, growth rate, and expected union rent. Excluding union rent, these variables are the same as those used by Scherer and Ravenscraft [1987] to predict merger activity.(9) Union rent is included in this study's merger equation to control for the possibility that unionization may influence merger activity. This variable, as calculated by equation (6), is the product of the average per worker union wage premium and the percentage of union workers in an industry.(10)

[Rent.sub.j] = ((1/[m.sub.j]) [center dot] ([[Sigma].sub.i] ([w.sub.u,i] - [w.sub.nu,i])) [center dot] (%[union.sub.j]), (6)

where i indexes individual union members in industry j, [m.sub.j] is the number of such members, and %[union.sub.j] is the percentage of workers who are union members in industry j. After estimating equation (5), the predicted value of the jth industry's merged asset share is retrieved using the following transformation:

merger = [[Epsilon].sup.[Beta]0 + [Beta]1V + [Beta]it)] / [1 + [[Epsilon].sup.([Beta]0 + [Beta]1V + [Beta]it)]]. (7)

The predicted merged asset share values derived from this calculation then replace the actual values in the original union status equation. Although this technique allows for addressing the possibility of union endogeneity, its use could introduce some estimation problems. For example, including a union type measure to estimate the merger variable could result in estimation bias. This potential problem suggests caution when interpreting the results.

The union status results derived after addressing the possibility of merger endogeneity are presented in Table 3. These findings reveal that while the estimated coefficients for the status equation are slightly altered when controlling for merger endogeneity, they still retain the same sign. Most importantly, the high degree of statistical significance persists for the estimated coefficient on the merger variables. As expected, the value of these coefficients are altered due to the use of the log-odds transformation to estimate the level of merger activity.

The predicted probabilities obtained from using the new approach to estimate the union status equation are presented in Table 4. When compared to the results from estimating the original status equation, the differences between the effect of conglomerate mergers and other types of mergers is slightly larger, as conglomerate mergers are now associated with an even smaller probability of being a union member. Indeed, the results in row (2) show the probability difference in the conglomerate and vertical merger effect on union status is now .0940. The probability difference in the effect on union status is .0960 for conglomerate and horizontal mergers. These results are not substantially different from the original findings on probability differences of .0633 and .0872 for the respective merger types. Hence, the pattern of merger activity's influence on union status remains intact regardless of the specification of the status equation or of the potential problem from addressing merger endogeneity.

[TABULAR DATA FOR TABLE 3 OMITTED]

[TABULAR DATA FOR TABLE 4 OMITTED]

VI. Conclusion

Combining merger information from the FTC with individual worker information allowed for analysis of the effect of merger activity on union status. Such a determinant has not been included in past studies, even though unions have complained that the current low membership levels are partly due to high levels of merger activity. Furthermore, these data sources permitted distinguishing the specific effects of horizontal, vertical and conglomerate mergers.

Initially this study used a standard sample-selection derived union status equation to estimate merger activity's influence on union status. After using this approach, predicted values for merged asset shares were incorporated to address the possibility of merger endogeneity. Regardless of the specification, this study detects a strong pattern of lower union membership flowing from conglomerate merger activity. For horizontal and vertical mergers, the probability of a worker being a union member is positively correlated with merger activity. Thus, this study reveals new information on union activity supporting the notion that only conglomerate mergers are associated with lower union membership.

1 The erosion of union power during stepped-up merger activity in the airline industry have received the most attention. However, this activity coincides with the influx of non-union airlines as deregulation lowered barriers to entry.

2 Guerard finds that employment declines more for non-merging firms. He suggests that this outcome is indicative of the economies of scale experienced by merging.

3 Individuals who are single, white, civilian, non-transportation operatives residing in the Western U.S. are the benchmark comparison group.

4 The probability of being a union member is derived by using the following transformation: [Phi] ([[Sigma].sub.i] ([Beta][X.sub.i])), where i is the sample observation and [Phi] is the standard normal density [Greene, 1990].

5 Hours worked per week is included in the wage equation to capture the earnings effect of working overtime, while the industry unemployment rate is included to control for the earnings effect from changes in the equilibrium quantity of labor. Consistent with past studies on union member [Farber, 1983; Belman, 1988], these variables are not considered as determinants of union status.

6 While this system is technically identified, the dearth of instruments included in the wage equations may suggest fragile estimates. Despite this possibility, the estimation results are not appreciably different from an ordinary least squares (OLS) result. (The OLS results are available from the author upon request).

7 The percentage point differences in the merger effects are derived by taking the difference in the predicted probability of being a union member for the respective merger types.

8 For the data set used in this study, 27 percent of the individuals work in industries that do not report conglomerate merger activity. This measure is 44 and 71 percent respectively for horizontal and vertical mergers.

9 Scherer and Ravenscraft [1987] include research and development and advertising expenses in the merger equation. These determinants are excluded from the merger equation due to insufficient information on these variables at the 3-digit, SIC level.

10 While the equation specified by equation (3) allows the author to retain the 0 value observations, it does not account for the possibility that such observations may represent censoring. However, estimation results from using a tobit technique reveal no appreciable change in merger activity's influence on union status. (The tobit results are available from the author upon request).

REFERENCES

Belman, D. "Concentration, Unionism, and Labor Earnings: A Sample Selection Approach," The Review of Economics and Statistics, 70, 1988, pp. 391-7.

Brown, C.; Medoff, J. L. "The Impact of Firm Acquisitions on Labor, in Corporate Takeovers," in A. J. Auerbach eds., University of Chicago Press, Chicago, 1988.

Farber, H. S. "The Determination of the Union Status of Workers," Econometrica, 51. 1983, pp. 1417-38.

Greene, W. G. Econometric Analysis, MacMillian Publishing Company, New York, NY, 1990.

Guerard, J. B. "The Role of Employment and Capital Expenditure in the Merger and Acquisition Process," Mergers and Acquisitions, in M. Keenan and L. J. White eds., Lexington Books, Lexington, MA, 1982.

Heckman, J. J. "The Common Structure of Statistical Models of Truncation, Sample Selections, and Limited Dependent Variables and a Simple Estimator for Such Models," Annals of Economic and Social Measurement, 5, 1976, pp. 475-92.

Hendricks, W. "Conglomerate Mergers and Collective Bargaining," Industrial Relations, 15, 1976, pp. 75-87.

Heywood, J.; Peoples, J. "Unions and the Pattern of Corporate Mergers: U.S. Evidence," Labour Economics, 1, 1994, pp. 203-21.

Hirsch, B. T.; Berger, M. C. "Union Membership Determination and Industry Characteristics," Southern Economic Journal, 50, 1983, 665-89.

Lichtenberg, F.; Siegal, D. "The Effect of Takeovers on the Employment and Wages of Central-Office and Other Personnel," National Bureau of Economic Research, working paper 2985.

Pratten, C. F. Economics of Scale in Manufacturing Industry, Cambridge University Press: Cambridge, England, 1971.

Rose, D. "The Effect of Changes in Firm Diversification on Union Wage Settlements for 15 Major U.S. Firms," Southern Economic Journal, 55, 1989, pp. 896-907.

-----. "Are Strikes Less Effective in Conglomerate Firms?" Industrial and Labor Relations Review, 1991 pp. 131-44.

Scherer, F. M. Industrial Market Structure and Economic Performance, Rand McNally: Chicago, IL, 1980.

Scherer, F. M.; Ravenscraft, David. Mergers, Sell-Offs, and Economic Efficiency, The Brookings Institute: Washington D.C., 1987.

Voos, P.; Michel, L. "The Union Rent Impact on Policies: Economics for Industry Price-Cost Margins," Journal of Labor Economics, 4, 1986, pp. 105-34.

Article Details

Printer friendly Cite/link Email Feedback
Author:	Peoples, James
Publication:	Atlantic Economic Journal
Date:	Mar 1, 1995
Words:	4081
Previous Article:	On black market exchange rate and demand for money in developing countries: the case of Nigeria.
Next Article:	Golden parachutes, information and shareholder value.
Topics:	Acquisitions and mergers Analysis Labor union recognition Labor unions Recognition