New evidence on unions and plant closings: Britain in the 1990s.

• Author: Addison, John T.

1. Introduction

   The effect of unions on establishment closings is an important but poorly developed theme in empirical labor economics. This is unfortunate because the association may assist our understanding of union impact on economic performance while serving as a potential explanation for the ongoing decline in union membership. Both topics remain controversial in part because of the dearth of evidence of union impact on establishment closings.

   In the present paper, we use British workplace data for the 1990s to investigate the issue. We obtain very different results from those reported in earlier U.S. and British studies. Specifically, we find a positive and well-determined association between unionization and establishment closings in the private sector that is driven by plants that are part of multiestablishment enterprises. In single independent establishments, a more conventional statistically insignificant association is reported. We argue that the empirical distinction between types of establishment reflects not only theoretical considerations but also former Prime Minister Margaret Thatcher's sustained legal challenge to the union movement during the decade of the 1980s (see Addison and Siebert, in press).

   The plan of the paper is as follows. In the next section, we review the sparse empirical literature on unions and plant closings, which we preface with material on other performance indicators. After that, we offer some applied theoretical conjectures on the expected association(s) between a union presence and closures. A brief review of the data set is then followed by presentation of our detailed findings. An interpretative section concludes.

2. The Empirical Literature

   In both the United States and Britain, there is a large literature documenting union effects on firm and establishment performance. This literature offers a somewhat pessimistic view of union impact that, when taken in the round, might lead one to anticipate a positive association between union presence and plant closings.

   In their review of U.S. productivity studies, Addison and Hirsch (1989) conclude that no compelling case emerges for a statistically or economically significant positive or negative union productivity effect. For its part, the British evidence points consistently to negative union productivity effects, subject to the important caveat that these have been much attenuated in recent years (see, e.g., Metcalf 1994, 2003). As far as productivity growth is concerned, the U.S. evidence points to a negative although not always robust association between unionism and changes in productivity (e.g., Hirsch 1991). The British evidence points to lower productivity growth in union regimes in the 1970s, although in later decades this result is sensitive to type of union (e.g., Gregg, Machin, and Metcalf 1993; Addison and Belfield 2001).

   The evidence on profitability is both stronger and weaker than that for productivity: stronger in the sense that almost without exception the union effect is negative and well determined in both countries, albeit again subject to some erosion through time in the British case, (1) and weaker to the extent that there is lingering controversy as to the source of the profit effect. For the United Kingdom, although not it seems the United States, there are at least some indications that the union wage premium may accrue from product market power (e.g., Machin and Stewart 1996).

   The U.S. evidence on investments in physical capital generally matches the productivity and profitability results for that country in indicating consistently lower investment in union regimes (e.g., Hirsch 1991, 1992; Fallick and Hassett 1999). Time-series studies for Britain also indicate some negative union effects (e.g., Denny and Nickell 1992), even if there is again evidence of a weakening in such effects in the 1980s compared with previous decades (see Metcalf 1994, p. 151). Research in the United States has also found rather strong evidence that unionization adversely affects investment in research and development (R&D) (Hirsch 1991). But this indication of union rent seeking acting as a tax on intangible capital is not replicated for the United Kingdom. In particular, Menezes-Filbo, Ulph, and Van Reenen (1998a) report that the negative correlation between union density/recognition and R&D intensity that is observed in cross-section and panel data for plants and companies, respectively, disappears when one accounts for age effects and the availability of innovative activity in the relevant industry.

   Finally, when one turns to the evidence on employment change, U.S. findings generally suggest that employment growth is lower (and employment contraction is greater) in the union sector (e.g., Linnemann, Wachter, and Carter 1990; Leonard 1992; Dunne and Macpherson 1994). British studies consistently point to the same result regardless of the time period examined (e.g., Blanchflower, Millward, and Oswald 1991; Femie and Metcalf 1995; Booth and McCulloch 1999).

   In sum, although the evidence is not always easy to interpret statistically (some theoretical ambiguities are aired in the next section), there are few grounds for optimism in the data concerning union impact on firm performance. The bright spot is the seeming improvement in the U.K. situation in recent years. Observed changes in the impact of British unions through time have been attributed to the Thatcher union reforms, which are widely seen as having reduced the "disadvantages of unionism (Oulton 1990). Nevertheless, the union effects on profits, R&D, and productivity growth are causes for concern. Lower profitability may be expected to lower investment in vulnerable forms of capital both directly because of the union "tax" (most clearly evident for the United States) and indirectly because profits are a pool of resources from which investments are frequently financed. And even if many firms do not depend on formal R&D for their survival (but rather on organizational innovation), all will rely on their pro ductivity growth if they are to prosper in the long run. Thus, it appears reasonable in these circumstances to anticipate higher rates of business failure in union regimes in both countries, even if in the case of the United Kingdom such effects might be more obvious in the pre-Thatcher era.

   In any event, there is little concrete evidence of this in the few published studies of plant closings for the two countries. Beginning with the U.S. evidence, although there are case studies pointing to high union labor costs as a major reason for plant closings during the 1980s (see Gerhart 1987), the two best-known econometric studies fail to detect higher rates of attrition in unionized regimes. Thus, in a sectoral analysis based on establishment data from the 1977 and 1982 Census of Manufactures microdata files, Dunne and Macpherson (1994) report that sectors with high union membership did not experience significantly greater employment loss due to plant closings than their less unionized counterparts, controlling for the price-cost margin and establishment size. As noted earlier, however, the authors' separate analysis of sectoral contraction and expansion rates suggests that more heavily unionized sectors do downsize more and have (marginally) lower growth, other things being equal.

   The more comprehensive U.S. analysis by Freeman and Kleiner (1999) uses three data sets: a sample of firms/business lines from the COMPUSTAT I and II files, 1983-1990, supplemented with union density information from Hirsch (1991); information on displaced workers from the Current Population Survey (CPS) Displaced Worker Supplements for 1994 and 1996, linked to control samples from the CPS outgoing rotation group, 1991-1995; and information on the rate of plant closure in union and nonunion regimes from the files of the Federal Mediation and Conciliation Service (FMCS).

   Of most interest is Freeman and Kleiner's analysis of the union impact on insolvencies using the COMPUSTAT data files. Their final sample comprises 633 business units (319 firms and 314 business lines), of which 126 became insolvent (67 firms and 59 terminated business lines) between 1983 and 1990. For a probit specification that includes just a union dummy and union density, plus dummies for whether the observation was a firm or a line of business, it is found that union presence is negatively associated with the likelihood of failure but that as the union density increases the firm/business line is more likely to fail. Adding in controls for firm age, sales, and a vector of industry characteristics (namely, concentration ratios, growth rate of sales, import penetration rates, and bankruptcy rates) reduces the effect of union presence but not that of union density, each remaining statistically significant. Finally, replacing the union variable(s) with the categorical measures of low, medium, and high density indicates that only the last category is associated with greater rates of insolvency, compared with no union presence. Regardless of specification, however, insolvencies are higher than in nonunion regimes only where union density is at or above 60%, a level that is twice the average unionization rate of the sample.

   Since even their sample of insolvent firms had been in business for some time, Freeman and Kleiner use the FMCS data to informally address the issue of whether newly unionized firms close more frequently than other companies. For the interval 1986-1993, it is found that just 341 out of 10,783 certification elections (and 3009 out of 168,945 dispute cases) resulted in plant closures. Expressed as rates of closure, these values compare favorably with estimates of business failures from the Annual Surveys of Manufactures, averaging 3.4% per year between 1974 and 1978. Equating the latter with average rates of plant closure in the absence of new unionism, Freeman and Kleiner (1999, p. 525)...