Strategic Management Journal Strat. Mgmt. J., 30: 25–44 (2009) Published online 19 September 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.725 Received 29 July 2004; Final revision received 9 July 2008

CAPACITY RATIONALIZATION AND EXIT STRATEGIES ANDREW WOOD* Essex Business School, University of Essex, Colchester, UK

A case study of the response to chronic excess capacity in a small competitive industry (the manufacturing of clay bricks) permits a generalization of Bower’s concentration hypothesis. Barriers to exit produced a free rider problem where only smaller and lower quality brick plants were shut when the efficient solution demanded major closures. The exit logjam was resolved by the strategic actions of growth-maximizing managers. They used major acquisitions as the basis for substantial reductions in firm and industry capacity while growing their own market share. The fall in industry capacity enabled other firms to follow suit while maintaining their market share as predicted by prospect theory. Copyright  2008 John Wiley & Sons, Ltd.

INTRODUCTION Capacity decisions are among the most important a firm makes. This is reflected in a vast literature on investment decisions, but substantially less work has been undertaken on the decision to disinvest. This would be understandable if there was symmetry between the two decisions, with the triggers to disinvest paralleling those to invest. However, it is not at all clear that this is the case. Neoclassical economic models of firm exit assume capacity closure decisions are implemented by reference to a simple threshold trigger. They have either resulted in predictions that have not been supported by subsequent empirical tests (Ghemawat and Nalebuff’s [1985, 1990] size-exit hypothesis) or have not produced any testable predictions (Whinston, 1988). By contrast, the evidence-based literature infers theories from case studies of industries with a Keywords: capacity rationalization; barriers to exit; prospect theory; mergers and acquisitions *Correspondence to: Andrew Wood, Essex Business School, University of Essex, Wivenhoe Park, Colchester, CO4 3SQ, UK. E-mail: [email protected]

Copyright  2008 John Wiley & Sons, Ltd.

known excess capacity problem. This shifts the focus from market signals and instead highlights a variety of structural, political, institutional, and agency factors that act as barriers to exit and prevent smooth adjustments of capacity. It therefore allows for the possibility that capacity rationalization is chronically inefficient with closures being delayed and not necessarily involving the least efficient plants or firms (Bower, 1986; Baden-Fuller, 1989; Jensen, 1993; Harrigan, 1980). The Bower (1986) case study of the petrochemical industry typifies the latter literature. Bower highlighted a variety of barriers to exit that prevented individual firms from responding to market pressures. These ranged from the reluctance of management to make cuts that would benefit their rivals to political interference in the decision about whether to close a plant. The barriers were such that a substantial restructuring of the industry was necessary before significant progress could be made in removing the excess capacity. In particular Bower argued that concentration—a reduction in the number of competitors—is a key prerequisite to solving the problems of an industry

26

A. Wood

facing excess capacity. When contrasted with the neoclassical view of capacity exit, Bower’s conclusions, along with those of the barriers to exit literature, are somewhat pessimistic. The reluctance of individual firms to voluntarily implement the market solution of cutting productive capacity as plants become uneconomic implies that the industry experiences prolonged pain before embarking on the necessary path of radical rationalization. This study makes two contributions to the literature. First, we examine whether the Bower (1986) hypothesis that concentration through takeovers is a necessary prerequisite for rationalization generalizes to other industries. His conclusions arose from a study of an industry that, due to its size and political significance, would likely face more exit barriers than many other industries facing similar problems of excess capacity. We test Bower’s concentration hypothesis in the context of a smaller competitive industry with a lower political profile. More specifically, we examine the rationalization process in the British brick industry following the collapse of the construction boom of the late-1980s to address these questions. This industry makes for an interesting case study because it is relatively uncomplicated: the product is homogeneous, its production process is well understood and had not undergone radical innovation, and the industry is not subject to the political intervention that has been witnessed in many other strategically important industries facing excess capacity. Most importantly, the brick production process requires a very high level of energy input that does not vary with output, which causes it to be uneconomic to operate plants at below full capacity. Since it is unprofitable for firms to operate plants substantially below full capacity, they are faced with the stark choice of having to close plants to downsize their capacity. In fact, conditions in the industry were ideal for a swift, market-based solution to the excess capacity problem, but our evidence shows that the initial solution involves just small marginal cuts to capacity when a more fundamental cut was required. The second contribution is that we provide a more general theoretical basis for the concentration of ownership and the extent and nature of rationalizations that follow whereas Bower’s (1986) framework is specific to large, politically sensitive industries. We use a variant of the managerial theory of the firm—which posits that managers maximize growth or market share—to explain why Copyright  2008 John Wiley & Sons, Ltd.

some firms engage in mergers and acquisitions (M&As) in an industry facing barriers to exit. This theory has a natural appeal under growing market conditions, but maximizing growth is a less obvious strategy in markets with excess capacity such as the brick industry in our case study. In this latter context, prospect theory suggests that the bottom line for managers is to maintain their market share, which is regarded as a fixed reference point. This explains the reluctance of firms to close plants as dictated by the market solution. Some firms can grow market share but only through M&As and not by building new plants under these conditions. Since the acquired plants may duplicate some of their existing facilities, firms will also typically engage in post-acquisition plant closures. Thus, large dominant firms can employ an M&A strategy and reconcile simultaneously the desire to grow market share and rationalize capacity.1 Using a dataset that consists of plant level capacity data supplemented by qualitative survey evidence, we find evidence of a range of barriers to exit that prevent a rapid voluntary removal of excess capacity as the neoclassical model might predict. Instead, consistent with Bower’s (1986) concentration hypothesis, we find that major advances in the removal of excess capacity are only achieved following three major acquisitions that allow the remaining firms, particularly the acquiring firms, both to grow market share and overcome barriers to exit. Does concentration of ownership through major M&As facilitate an efficient rationalization with the least efficient plants being closed first, or do the exit barriers continue to skew the rationalization program? We find that the burden of closures is disproportionately taken by the acquiring firm, and, consequently, the capacity selected for closure is not necessarily the least efficient in the industry. This is a refinement of the Bower hypothesis and strong evidence against the neoclassical hypothesis. The article proceeds with a brief review of the exit literature followed by an outline of the case study with a detailed analysis of the rationalization process that draws on both interview evidence and an empirical analysis of the pattern of plant closures. This is followed by the reporting of panel regression results before a final section concludes with a discussion. 1 There is also evidence that acquiring managers receive special bonuses for successful M&As.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies RELATED LITERATURE AND HYPOTHESES Neoclassical hypotheses on exit The capacity exit literature can be divided into two strands. The first uses conventional neoclassical economics to identify the order of exit. This framework is premised on the assumption that capacity is closed as soon as it ceases to be profitable. As such, less profitable plants are likely to be closed first with the pressure to close intensified by higher levels of competition. Hypothesis 1a: Less efficient plants and plants that produce inferior products are first to close.

Hypothesis 1b: Firms are more likely to close plants when confronted by more intense competition. A less obvious hypothesis arising from this literature is the prediction of a size-exit relationship in which firms with the largest market shares are the first to close capacity (Ghemawat and Nalebuff, 1985, 1990). The intuition for this is straightforward. Smaller firms remain profitable for longer and therefore outlast their larger competitors in a monotonically declining industry. In addition, while all surviving firms benefit from reductions in excess capacity, larger firms gain most and therefore have a greater incentive to commence closing capacity. Subsequent studies have cast doubt over the existence of a simple size-exit relationship. Whinston (1988) has shown that the theoretical result is dependent on very restrictive assumptions. More importantly, the empirical evidence indicates that firm size is not positively related to the probability of exit (Deily, 1988, 1991; Lieberman, 1990; Baden-Fuller, 1989; Gibson and Harris, 1996). Hypothesis 1c: Large firms are the first to close capacity. Barriers to a neoclassical solution The second strand of literature identifies a number of barriers to exit that may prevent an industry from achieving an orderly reduction in capacity. In his study of the petrochemical industry, Bower (1986) outlines obstacles to exit relating to the Copyright  2008 John Wiley & Sons, Ltd.

27

emotional and strategic considerations of management as well as political considerations and government intervention that reflected the strategic importance of the industry and a concern for employment. In her more wide-ranging study, Harrigan (1980) refers to a variety of exit barriers that influence what she describes as the endgame strategy. Employment concerns are likely to be exacerbated for plants that are geographically isolated. In such cases there are likely to be limited relocation opportunities and poor alternative employment prospects for staff made redundant (Lieberman, 1990). This problem is most acute when decision makers feel closure will threaten their own personal prospects; a situation which is less likely for large, diversified firms whose decision makers are likely to be both more dispassionate and more experienced at making tough decisions regarding uneconomic production facilities (Baden-Fuller, 1989). Hypothesis 2a: Isolated or rural plants that are significant local employers are less likely to be closed.

Hypothesis 2b: Diversified firms are more likely to close capacity. In sharp contrast to the neoclassical approach, excessive competition can be part of the problem rather than the cure (Bower, 1986). Jensen (1993) notes that it is difficult for the culture of organizations and the mindset of management to adjust to the need for rationalization, particularly if the industry has experienced rapid growth: ‘managers fail to recognize that they themselves must downsize; instead they leave the exit to others’ (Jensen, 1993: 847). Indeed, firms may continue to invest to ensure they ‘have a chair when the music stops’ (Jensen, 1993: 847). In the same vein, Bower (1986) argues that some form of cooperation is required to overcome excessive competitive urges. A common theme underlying Jensen (1993) and Bower (1986) is the reluctance of firms to concede ground to rivals. This reluctance was highlighted during interviews with managers of brick firms indicating their priority was to increase, or at least maintain, their market share. Within a contracting market this objective deters individual firms Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

28

A. Wood

from taking a lead in closing capacity without assurances that rival firms will follow suit. In the absence of explicit cooperation, the problem can be overcome either by one firm taking a lead in the hope that others will follow, or with a reduction in the number of rivals. Taken together, the barriers to exit literature suggest that closures are less likely when competition is fierce but can be triggered by a relaxation of competition brought about by the actions of a single or a group of firms. Hypothesis 2c: Firms are less likely to close plants at times of intense competition and more likely to close capacity following a relaxation of competitive pressure. The role of acquisitions In situations where there is a standoff between rival firms, the most effective way of reducing the competitive logjam is for the withdrawal of one firm either as a result of a merger or an acquisition. This is emphasized in Bower’s (1986) three phases of restructuring: preparation, concentration, and rationalization. During the first phase individual firms lay the groundwork for preparing for rationalization by revising their corporate strategy and restructuring the organization so that it is ready to confront and deal with the reality of excess capacity. The second phase sees a concentration within the industry resulting from mergers, acquisitions, and takeovers. Bower identifies this phase as the key for achieving rationalization as the reduction in the number of rival firms effectively removes the competitive barriers that prevent capacity exit. Once the concentration phase is underway, the remaining firms, particularly the acquiring firm, are able to implement their rationalization programs. Subsequent work by Jensen (1993) identifies M&As as part of the remedy to excess capacity that was evident in a range of industries during the 1980s. Three separate hypotheses linking M&As to closures are suggested by the barrier to exit literature. As suggested by Bower (1986), M&As overcome competitive or strategic standoffs and are consequently followed by a general industry-wide closure program, with the acquiring firm being the most likely to close. In addition, following an M&A, the acquiring firm is in possession of newly Copyright  2008 John Wiley & Sons, Ltd.

acquired plants for which there is no great emotional attachment, such as described in Hypothesis 2a. Accordingly, such plants are more likely to be closed than those previously owned by the acquiring firm. Hypothesis 2d: M&As lead to industry-wide plant closures.

Hypothesis 2e: Acquiring firms are more likely to close plants than non-acquiring firms.

Hypothesis 2f: The acquiring firm is more likely to close newly acquired plants than those owned prior to the M&A. The neoclassical literature also suggests a role for M&As. Extending Ghemawat and Nalebuff’s (1990) multiplant model, Reynolds (1988) identifies a motive for merger between two firms operating in a declining industry. He notes that a merger can raise joint profits for the firms because a joint enterprise will take into account the revenue gain on all remaining plants that result from the closure of a single plant. Consequently a large firm may have an incentive to acquire and then retire a smaller firm. Hypothesis 1d: Firms that acquire rivals are more likely to close plants if they have a large market share as a result of the acquisition. A problem with much of the existing empirical evidence relating to capacity exit is that it is dominated by studies of large strategically important industries—such as steel, steel castings, and chemicals—that are known to be politically sensitive and are therefore not surprisingly subject to barriers to exit. By contrast, the current study focuses on an industry that operates from small geographically dispersed plants that are not subject to foreign competition, the closure of which may have highly localized employment ramifications that will not register at the national level. Moreover, high fixed costs of production effectively remove the option of responding to downturns in demand by reducing capacity utilization. Combined, these industry characteristics make the brick industry an ideal case study for market-based explanations of capacity exit. Indeed, it can be argued that there are Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies few other manufacturing industries that are more suitable for an efficient, competitive/market-based solution to a problem of excess capacity.

THE CASE STUDY Historical background

2004

2000

1996

1992

1988

1984

1980

1976

1972

1968

1964

1960

1956

1952

9000 8000 7000 6000 5000 4000 3000 2000 1000 0 1948

millions

Demand for British clay bricks has been in long run decline since the end of the long postwar construction boom. Figure 1 shows that the 1964 peak in deliveries was followed by a pronounced downward trend that is punctuated by sharp cyclical swings. Rather than looking at the entire post1964 period, this case study examines the capacity reductions carried out between 1989, when the 1980s construction boom ended resulting in chronic excess capacity, and 1996, when the last significant spate of post-boom plant closures took place. Interviews and contemporary reports indicate a consensus view that the reduction in demand during 1989–1991 was considered to be permanent and required a substantial cut in capacity, but that the post-boom level of demand represented a likely floor—and this indeed has proved to be the case.2 The sample period is of particular interest because it follows a relatively long and sustained

Source: Office of National Statistics

Figure 1.

Brick deliveries, 1948–2004

2 Numerous contemporary news reports confirm the consensus regarding the permanency of the decline in demand. Brick deliveries fell from an annual peak of 4.7bn in 1988 to a low of 2.9bn in 1992. Most of the decline in deliveries occurred during the early stages of the recession, with declines of 16 percent in 1989, 11 percent in 1990, 13 percent in 1991, 7 percent in 1992. The post-boom annual deliveries (1990–2004) averaged 3bn with little variation around this level with the exception of one year in which deliveries rose briefly to 3.5bn (the standard deviation during this period is 174.3).

Copyright  2008 John Wiley & Sons, Ltd.

29

construction boom during which brick deliveries increased by 32 percent. Profit margins increased significantly during the 1980s, encouraging firms to expand by investing in existing plants, rebuilding old plants, and even constructing new capacity on greenfield sites. The latter were the first new plants to be commissioned in nearly two decades, indicating a belief that the secular decline of the industry had come to an end. By the peak of the boom, total annual capacity had been increased to approximately 5 billion (bn) bricks per year, broadly in line with peak deliveries. By the 1992 trough, deliveries had contracted to less than 3bn, a reduction of 39 percent, but it was not until the end of 1996 that capacity came into line with deliveries. To understand the implications for brick capacity of such demand swings and why the gap between output and deliveries is indicative of unsustainable excess capacity, it is necessary to examine the process of manufacturing clay bricks. The manufacture of bricks can be divided into four stages: 1) winning and preparing the clay; 2) shaping the bricks; 3) drying; and 4) firing. The key stage is the fourth, since it is the size and technology of the kiln that determines the efficiency of a brickwork and the lumpiness of capacity decisions. When discussing the size of brickworks, one should distinguish between factories that produce fletton bricks (made by London Brick Company (LBC)) from those that produce non-fletton bricks.3 Historically, fletton kilns and brickworks have been substantially larger than those that produce non-fletton bricks. For the sample period of this case study, the average fletton brickworks could produce 140 million (mn) bricks per year. By comparison, the average annual capacity for

3 Fletton bricks are made from Lower Oxford clay, found primarily in Bedfordshire and Cambridgeshire. Fletton bricks have a cost advantage over non-fletton bricks since their high carbonaceous content means they can be fired with the addition of relatively small quantities of fuel. This cost advantage has facilitated their production on very large scales. A typical fletton kiln can produce 60mn bricks per year, and the largest fletton brickworks had a peak capacity of over 800 mn bricks, representing approximately 20 percent of the entire national market. These advantages have been offset by the low durability and poor aesthetic quality of fletton bricks that have resulted in the long-term decline in demand for flettons and caused LBC to develop new processes to improve the performance and appearance of flettons. While this has allowed LBC to compete in the same markets as that of non-fletton producers, it has come at the cost of effectively removing the fletton cost advantage.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

30

A. Wood

% of plants

non-fletton brickworks was 33mn bricks. The primary determinant of the size of a non-fletton kiln is the technology available at the time of its construction. Non-fletton kilns constructed during the 1950s have a typical capacity of 10–12 mn bricks per year, while kilns made in the 1960s have a capacity of up to 20–25 mn per year, with this figure rising to approximately 50 mn for kilns built during the 1980s. The tendency to operate from larger plants to benefit from economies of scale is demonstrated in Figure 2. Since most modern kilns are continuous (permanently firing) and consume large amounts of fuel irrespective of their output, it is essential for brick firms to ensure that kilns are operating at or near full capacity in order to keep unit costs low. When demand declines, brick firms confront a stark choice: they can continue to produce at full capacity, allowing stocks to increase in the hope that either demand will recover or other firms will make the necessary rationalizations;4 or, the firm can cut production by switching kilns off. This latter option amounts to effective closure of capacity since once kilns are allowed to cool, conditions within the plant quickly deteriorate and equipment, ranging from kiln cars to computerized control facilities, rust in the damp conditions of an unused brickworks. In addition, the structure of a kiln that is allowed to cool can deteriorate and can require substantial investment before it can be relit. Consequently, restarting production after a break involves potentially large and unpredictable investments to restore the productive potential of the plant. The need for an aggregate reduction in capacity and the determination of individual firms to 40% 35% 30% 25% 20% 15% 10% 5% 0% >100

50-99

Figure 2.

40-49

30-39

1985

1988

20-29

10-19

<10

1996

Size of non-fletton plants

4 Firms are constrained in their ability to let stocks increase by cash flow considerations, space constraints and the risk that bricks that spend a long time stacked in open yards may lose value.

Copyright  2008 John Wiley & Sons, Ltd.

avoid plant closures provides the foundations for a short-lived but costly war of attrition. In the Netherlands this tension was resolved by a legally binding agreement between Dutch brick manufacturers to coordinate the necessary rationalizations over a five-year period from 1993 to 1998.5 British based production directors interviewed for this study universally condemned the Dutch arrangement as uncompetitive and inefficient and argued that while the British system is brutal, British plants are very modern and efficient. Reflecting this Darwinian approach, the British brick industry has been characterized by frequent acquisitions. The period covered by this case study sees three major acquisitions that are shown to be linked to the rationalization process.

DATA AND ECONOMETRIC MODEL The dataset The dataset includes plant histories for all brick firms that had an annual capacity of at least 50mn bricks. The exclusion of smaller firms is justified on the grounds that they are predominantly smallscale concerns producing handmade bricks for a specialist market. Plant level capacity data was compiled from a variety of sources. A list of plants operating during 1981 was obtained from Ridgway (1982) while the Brick Development Association (http://www.brick.org.uk/) supplied a list of brickworks that were operating in 1996. The differences between the two lists were plants that were either commissioned or closed during the intervening years. The next stage was to identify the capacity of each plant, date the openings and closures of the plants that took place during the sample period, and date and quantify any incremental changes in plant capacities. Company reports and media news articles proved to be a source of surprisingly detailed capacity information.6 The reason for this 5 Agreement was obtained from 16 firms to close a total of seven plants. The agreements required the permanent closure of the seven plants and signatories to the agreement were not allowed to expand capacity during the life of the agreement, which was set at five years or shorter, depending on the market situation. A compensation fund was established with each firm contributing to this fund according to their production levels. The fund was used to contribute to the costs of closure of the seven plants. 6 Contemporary media reports were obtained from FT-Profile and back copies of Building and Construction News.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies is that brick producers often find it in their interest to publicize their investment decisions so that large customers are kept informed about changes in product design and availability. Gaps in the dataset were filled by requesting the information directly from firms. This stage of the data collection process went hand in hand with the collection of the primary qualitative data that has been used in this and related studies (Scheibl and Wood, 2005; Wood, 2005). In all, 10 directors, 17 works managers, two marketing or sales managers, and two company secretaries were interviewed, often on condition of anonymity. Contact was maintained with key interviewees allowing follow-up interviews to take place. Modeling capacity decisions must take account of the regional dimension of the industry. High transport costs due to the high weight-to-value ratio result in a large proportion of the output of a brick plant being delivered within a radius of 100–150 miles, depending on the road networks. Table 1 provides a geographical snapshot of the location of plants at the beginning of the sample period. A minority of firms start with a good regional spread of locations (Butterley, Ibstock, and Tarmac), while of the other large manufacturers, Redland and LBC are concentrated in southern England, and Steetley is underrepresented in the south (the location of the largest market). Table 2 ranks the major firms by capacity size as measured at the beginning of the sample period and gives the number of plants each closed during the sample period.7 Three aspects summarized in this table are worth emphasizing. First, the capacity of five of the firms included in the sample was acquired by rivals, enabling some firms to grow by acquisition. Second, three small- to mediumsized firms (Baggeridge, Chelwood/Salvesen and Marshalls) survived the period without having to close any plants. Third, most closures were made by the largest firm at the beginning of the period (LBC and Butterley, jointly owned by Hanson) and those that grew by acquisition (Ibstock, Redland, and Tarmac).

7 For a variety of reasons, the number of plants operating in 1996 may not equal the number of plants operating in 1989 less the number closed. Some firms increased their number of plants by acquiring capacity, or by commissioning new plants. The figures are also influenced by recommissioning plants. For example, Blockleys closed their Heritage plant during 1992 and then recommissioned it in 1995.

Copyright  2008 John Wiley & Sons, Ltd.

31

Table 1. Location of plants, beginning of 1989: number of plants (percent of firm’s total capacity)9 Southern Midlands Northern Scotland England and Wales England Baggeridge Blockley Blue Circle 2 (100) Butterley 4 (18) GISCOL Ibstock 4 (33) LBC 10 (100) Marley Marshalls Raeburn Redland 14 (69) Rudgwick 1 (100) Salvesen 1 (2) SBC Steetley 1 (5) Tarmac 4 (50) Wemyss

5 (100) 3 (100) 9 (66)

4 (16)

5 (48)

2 (19)

2 (72)

1 (21) 3 (100)

3 (22)

1 (9)

1 (35)

2 (64)

4 (39) 1 (4)

6 (53) 4 (45)

7 (100) 1 (3) 2 (100)

2 (100) 1 (3) 1 (100)

Model specification Our panel consists of annual data from 1989–1996 for 120 plants belonging to 18 different firms with a dichotomous dependent variable yit , t = 1989, 1990, . . . , 1996; i = 1, 2, . . . , Nt , zero if open throughout the year, one if closed during the year, otherwise the plant does not appear in the dataset. Nt is the number of plants operating at the beginning of year t. The independent variables, xit , are observed at the beginning of year t.8 The resulting probit model with random effects is defined in the following equation. yit∗ = xit β + νit 8 Strict observance of this definition has the undesirable consequence of neglecting plants that are opened or reopened during the period and ignoring the consequences of acquisitions of plants during the period. This issue is particularly important because as we have seen a number of plant closures were undertaken by firms that recently acquired new capacity. It would therefore be misleading to condition the probability of closure on the original owning firm when it is the acquiring firm that undertakes the closure. There is no ideal solution to this problem. The compromise adopted here involves the independent variables being defined at the beginning of the period for all plants except for those plants that are either commissioned or are acquired during the period. For these plants, xit is defined at the time of commissioning or at the time of acquisition. 9 Four regions are defined: the south (south-east, south-west and East Anglia), the midlands (west midlands, east midlands and Wales), the north (north-west, Yorkshire and Humber, and northeast) and Scotland.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

32

A. Wood

Table 2. Summary data for sampled companies

LBC13 Butterley Steetley Ibstock Redland Tarmac Baggeridge Chelwood/Salvesen14 Marley Marshalls GISCOL SBC Blockley Blue Circle/Ockley Raeburn Wemyss Hepworth AmbionF

Capacity share end 1988 (%)12

Capacity share end 1996 (%)

Number of plants in 1989

Number of plants in 1996

Number of plants closed

36.2 12.1 10.0 7.0 6.2 4.9 4.7 3.7 3.1 2.4 2.0 1.9 1.8 1.1 1.0 1.0 0.6 NA

20.1 13.2 A 35.5 B C 7.3 5.7 D 4.2 1.1 E 1.2 0.8 0.7 1.5 0.8 5.6

10 17 12 11 18 9 5 4 4 3 7 2 3 2 2 1 1 NA

5 11 A 29 B C 5 4 D 4 2 E 2 2 1 1 1 5

5 7 0 8 15 4 0 0 2 0 5 2 2 0 1 0 1 0

A: acquired by Redland, April 1992 B: acquired by Ibstock, August 1996 C: acquired by Ibstock, May 1995 D: acquired by Tarmac, October 1993 E: acquired by Ibstock, 1994 F: New firm, result of DTI ruling following Ibstock’s acquisition of Redland during 1996

where νit = ci + eit and

acquisitions facilitate the size-exit relationship, we include SHARE interacting with a dummy variable, ACQUIRE, that signifies whether the plant is owned by a firm involved in an acquisition.11

yit = 1 if yit∗ > 0, else 0 For robustness, we also estimate the following logit model with firm fixed effects, zj . yit∗ = xit β + zj δ + uit Explanatory variables: market based model The main prediction arising from conventional economic models is that larger firms are the first to close capacity. The variable SHARE measures the firm’s regional operational capacity at the beginning of the period as a proportion of the total capacity of plants within range of that region.10 To test Reynolds’s (1988) prediction that 10 When defining the region in which a plant is situated, use is made of the 11 standard regions of mainland Britain. If a plant is located adjacent to an efficient road network that makes for easy access to major metropolises of neighboring regions, that

Copyright  2008 John Wiley & Sons, Ltd.

plant is classified as being a player in each of those regions. For example, there is a concentration of plants situated in and around the West Midlands town of Stoke-on-Trent, which is adjacent to the M6 motorway that provides easy access to the major markets of Birmingham in the West Midlands and Manchester and Liverpool in the Northwest. For this reason these plants are considered to be competing with other firms that supply both the West Midlands markets and the markets of the Northwest. 11 In the next section, separate dummy variables are defined as plants owned by the acquiring firm prior to the acquisition (ACQUIRER) and plants acquired in the acquisition (ACQUIRED). The dummy variable ACQUIRE combines both previously owned and newly acquired plants. 12 These figures slightly overstate the actual shares since they refer to the firms’ share of capacity amongst those included in the sample. The sample excludes a number of very small firms that primarily produce from small sites that supply specialist markets. 13 LBC and Butterley are both owned by Hanson brick. They are kept as separate entities because of the distinction between fletton and non-fletton bricks, and more importantly they have, until 1996, been operated as distinct companies in competition with each other. 14 Salvesen Brick became Chelwood Brick following a management buyout in 1995. Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies Table 3.

33

Plant characteristics (end of period) (A) All plants

Number Average capacity Average plants per firm Low durability (%) High durability Low aesthetic High aesthetic

1988

1989

1990

1991

1992

1993

1994

1995

1996

110 42.11 10.4 0.45 0.15 0.59 0.14

108 39.14 10.2 0.44 0.16 0.57 0.15

98 37.70 9.0 0.43 0.16 0.59 0.13

95 37.11 8.8 0.43 0.16 0.56 0.13

75 40.16 7.0 0.45 0.13 0.43 0.13

72 39.02 8.5 0.44 0.14 0.54 0.11

75 39.68 9.0 0.44 0.13 0.53 0.12

75 42.17 11.3 0.43 0.15 0.52 0.12

70 43.34 13.9 0.43 0.14 0.51 0.13

(B) Non-fletton plants

Number Average capacity Average plants per firm Low durability (%) High durability Low aesthetic High aesthetic

1988

1989

1990

1991

1992

1993

1994

1995

1996

100 29.57 10.5 0.40 0.17 0.55 0.15

100 28.94 10.4 0.4 0.17 0.54 0.16

91 30.53 9.13 0.38 0.18 0.56 0.14

89 31.24 8.96 0.39 0.17 0.43 0.13

69 34.60 8.13 0.41 0.14 0.49 0.14

67 35.00 8.7 0.40 0.15 0.51 0.12

70 35.51 9.3 0.40 0.14 0.50 0.13

70 36.46 11.8 0.39 0.16 0.49 0.13

65 37.28 14.6 0.38 0.15 0.48 0.14

A positive coefficient for each of these variables would support Hypotheses 1c and 1d respectively. Direct measures of efficiency are not available, but we do have access to some plant-specific data that relates to their efficiency and profitability. The most obvious indicator of a plant’s efficiency is its capacity. Not only do larger plants have significant cost advantages over smaller plants, but size is also an indicator of age, with large plants tending to be relatively young and therefore incorporating modern facilities. Interviews with production managers confirm that exploiting kiln and plant level economies of scale are crucial to maximizing efficiency and profitability, giving rise to the longterm trend toward operating from larger plants (Figure 2). Accordingly, we include two variables that are based on the plant’s capacity, CAP FIRM measures capacity relative to the owning firm’s average plant capacity while CAP IND measures capacity relative to the average of all plants of a corresponding clay type (fletton or non-fletton) operating within the region. Using these two variables we can define a closure program as efficient at the industry level if the closures involve plants that are small relative to the industry average. Alternatively, a closure program may not be efficient at the industry level but is nevertheless Copyright  2008 John Wiley & Sons, Ltd.

efficient for the firm that is carrying out the closures if the plants that are closed are small relative to the firm’s own portfolio of plants. In addition to capacity, individual plants vary according to the differing durability and aesthetic qualities of the bricks they produce.15 While there does not appear to be any evidence of declining demand for low durability bricks, there has been a long-term shift in demand away from bricks of low aesthetic quality to bricks of higher aesthetic quality, with the proportion of plants that produce bricks of low aesthetic quality falling from 59 15 With advice from one of the country’s leading brick consultants, data for frost resistance, soluble salt content, average compressive strength, and average water absorption, was used to categorize each plant as a producer of high, medium, or low durability bricks. Plant output was also graded according to its aesthetic quality. As with durability, the main determinant of the aesthetic quality of a brick is the clay, although additives and the technique for forming the brick are also important. While this may be regarded as a highly subjective venture, there are certain aesthetic qualities relating to the color and texture of a brick that can be recognized as more or less desirable, and therefore reflected in the price. The output of each plant was given a grade ranging from A for bricks that possess an inherent beauty in terms of both color and texture, B for bricks that can be described as less handsome but nevertheless capable of being used for the construction of an attractive house, to C for the more utilitarian bricks that may be used for mass housing but require mixing with some bricks of different colors to avoid a barrack-block effect. The final classifications were shown to production directors for corroboration.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

34

A. Wood

percent of all plants operating at the end of 1988 to 51 percent by the end of 1996 (Table 3). To formally test the role of quality in determining the closure programs, we define two dummy variables, ASTHLOW and ASTHHIGH, indicating whether the aesthetic appearance of the brick is of low or high quality. Three variables are used to control for interaction effects and the level of competition: RIVALDEC measures the log absolute value of any decrease in capacity undertaken by rival firms in the same region during the previous year; RIVALINC is the corresponding measure of rival increases in capacity; and RIVAL measures the regional Herfindahl index. In a market-based solution to the excess capacity problem we would expect closures in regions in which rivalry is high, implying a negative coefficient for RIVAL, and negative and positive coefficients for RIVALDEC and RIVALINC respectively. Explanatory variables: barriers to exit Interviewees spoke of their personal commitment and loyalty to their employees. In the context of making decisions to close plants, this is often exacerbated by the rural setting of many brick plants and is demonstrated by the comments of a director of a large brick firm interviewed for this study who spoke of the difficulty of closing brick plants due to the fact that although small, a brick plant was often the only significant employer in the locality and that all employees were known personally and were often second-generation employees of the company.16 The interviewee was also aware that following plant closures, redundant employees who were not ‘work shy’ nevertheless often found it extremely difficult to even consider commuting 10–20 miles to a neighboring town for employment. For this reason, interviewees suggested that firms were reluctant to close geographically isolated plants because of poor alternative job opportunities. This employee loyalty hypothesis is tested with the inclusion of the variable, RURAL, that takes the value one if the firm is located in a rural or village setting. Employee loyalty is less of an obstacle to plant closures if the firm has other 16 This consideration is also recognized by Harrigan (1980) as a managerial barrier to exit.

Copyright  2008 John Wiley & Sons, Ltd.

plants in the vicinity, because staff can be relocated or offered employment at the remaining site. Being able to transfer employees to nearby plants also enables firms to overcome redundancy costs while retaining the skilled labor force. Accordingly, FRIEND is a dummy variable that takes the value one if the plant is located near (within travel to work distance) plants owned by the same firm. Besides having to lay off employees, firms may also be concerned with the possibility of losing valuable customers who may worry about the reliability of supply in the event of plant closures. This latter effect may be offset if the firm operates more plants within supply range of the plant that is to be closed. The variable PLANTS measures the number of plants owned by the same firm within delivery range to capture this effect. To test Hypothesis 2b, we include a dummy variable, DIV, that takes the value of one for plants owned by a diversified firm that does not manufacture bricks as its core business. A striking finding from interviews was the very strong mindset of numerous company directors to maintaining their market share. Consistent with managerial theories of the firm, agents prioritized the maximization of market share above all other objectives. An insight into this mindset was provided by an interviewee based at the industry’s trade association, the Brick Development Association, who spoke of the intense rivalry between brick firms regarding their respective market shares and of how they would attend meetings held by the Association armed with their estimates of their market shares, sometimes calculated to two decimal points. This obsession with market share potentially undermines the industry’s chances of achieving a speedy resolution to the excess capacity by exacerbating the coordination problem. This hypothesis would be supported by positive and negative signs respectively on the previously defined variables, RIVALDEC and RIVALINC. We would also expect a positive coefficient for RIVAL as firms are more reluctant to close capacity and concede market share when rivalry is high. Bower’s (1986) concentration hypothesis suggests that plant closure programs follow major acquisitions that effectively remove many barriers to exit. To test for a general industry-wide effect we include time dummies for each year in which a major acquisition took place: 1992, 1995, Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Copyright  2008 John Wiley & Sons, Ltd.

− + − −

ASTHHIGH

RIVALINC

RIVALDEC

RIVAL

+

+

ASTHLOW

ACQUIRED

+

−

CAP FIRM

+

+

−

CAP IND

ACQUIRER

−

+

ACQUIRE∗ SHARE

Barrier to exit

+

Market based

Predicted signs

Panel regressions predicting plant closures

SHARE

Table 4.

0.0281 (0.243) 0.0511∗∗ (0.027) −0.0244∗ (0.061) −0.0266∗ (0.078) 0.0194∗∗ (0.040) −0.0090 (0.319) −0.0054 (0.197) 0.0088∗∗∗ (0.006) −0.0670∗ (0.070)

Probit (1a) 0.0086 (0.720) 0.0490∗ (0.058) −0.0284∗ (0.089) −0.0130 (0.401) 0.0114 (0.158) −0.0116 (0.169) −0.0030 (0.490) 0.0086∗∗ (0.011) −0.0568 (0.125)

Logit (1b)

1. Market based

−0.0172∗∗ (0.033) 0.0170∗∗∗ (0.008) −0.0649 (0.301) 0.0127 (0.846) 0.2107∗∗∗ (0.001)

Probit (2a)

−0.0079 (0.166) 0.0141∗∗∗ (0.006) −0.0231 (0.571) 0.0161 (0.733) 0.1521∗∗∗ (0.002)

Logit (2b)

2. Barriers to exit

0.0006 (0.980) −0.0212 (0.544) −0.0104 (0.305) −0.0255∗∗ (0.025) 0.0155∗∗ (0.021) −0.0072 (0.267) −0.0068∗∗ (0.044) 0.0078∗∗∗ (0.005) −0.0090 (0.766) 0.0138 (0.719) 0.2586∗∗∗ (0.005)

Probit (3a)

−0.0059 (0.798) −0.0125 (0.709) −0.0192 (0.125) −0.0112 (0.346) 0.0098∗ (0.091) −0.0063 (0.342) −0.0047 (0.147) 0.0078∗∗∗ (0.006) −0.0014 (0.962) 0.0113 (0.761) 0.1789∗∗∗ (0.007)

Logit (3b)

3. Combined

Capacity Rationalization and Exit Strategies 35

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Copyright  2008 John Wiley & Sons, Ltd.

− + + + +

RURAL

FRIEND

T92

T95

T96

0.1939

−0.2530∗∗∗ (0.002) −0.0006∗∗ (0.021) 0.0105 (0.514) −151.342 8.08∗∗∗

Probit (1a)

26.79∗∗∗ 0.2433

−0.2019∗∗ (0.010) −0.0007∗∗ (0.013) 0.1443∗∗∗ (0.005) −142.631

Logit (1b)

1. Market based

0.1868

0.0097∗∗∗ (0.001) 0.0103 (0.534) −0.0216 (0.113) −0.0123 (0.387) 0.0996∗∗∗ (0.005) −0.0041 (0.892) 0.0082 (0.771) −0.0664 (0.727) −0.0001 (0.868) 0.0173 (0.553) −158.194 4.26∗∗

Probit (2a)

25.72∗∗∗ 0.2419

−0.0182∗∗ (0.047) −0.0131 (0.178) 0.0488∗∗ (0.027) −0.0053 (0.808) −0.0023 (0.910) −0.0663 (0.645) −0.0003 (0.435) 0.1538∗∗∗ (0.008) −142.911

0.0065∗∗∗ (0.002)

Logit (2b)

2. Barriers to exit

0.2899

0.0039∗∗∗ (0.003) 0.0026 (0.696) −0.0048 (0.431) −0.0034 (0.564) 0.0554∗∗∗ (0.004) −0.0028 (0.806) 0.0013 (0.905) −0.0556 (0.455) 0.0000 (0.947) −0.0029 (0.724) −133.316 5.31∗∗

Probit (3a)

20.24∗∗∗ 0.3261

−0.0050 (0.359) −0.0028 (0.609) 0.0369∗∗∗ (0.009) −0.0025 (0.839) −0.0005 (0.965) −0.0361 (0.634) −0.0001 (0.77) 0.0332 (0.123) −127.038

0.0037∗∗∗ (0.004)

Logit (3b)

3. Combined

Reported coefficients are marginal effects calculated at the variables’ mean values, with the exception of coefficients for dummy variables that measure the effect of a change of the dummy variable from 0 to 1 (calculated at the mean values for all other variables). p-values are reported in parentheses. ∗ ∗∗ ∗∗∗ , , indicate significance at the 10%, 5% and 1% levels respectively.

Log likelihood LR test random effects LR test fixed effects Pseudo R2

FLETTON

STARTS

REALPRICE

+

DIV

Barrier to exit +

Market based

Predicted signs

PLANTS

Table 4. (Continued )

36 A. Wood

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies and 1996. Two dummy variables are defined to test Hypotheses 2e and 2f. During the year of acquisition, ACQUIRER takes the value one for plants owned by an acquiring firm prior to the acquisition and ACQUIRED takes the value one for newly acquired plants. Positive coefficients for these dummies will indicate that the acquiring firm carries out a disproportionate number of plant closures. Control variables Demand is proxied by the price of bricks deflated by the gross domestic product deflator, labeled REALPRICE, and the percentage change in regional housing starts, STARTS. A number of other demand proxies such as brick stocks and brick deliveries were tried but were found to have very low explanatory power. A dummy variable titled FLETTON takes the value one for plants producing fletton bricks, zero otherwise. Dummy variables for other clay types were included during initial regressions but were found to be not significant and are not reported here.

EMPIRICAL RESULTS The results for the probit random-effects and logit fixed-effects models are presented in Table 4. The results for the market-based model are presented in Column 1, those for a pure barrier to exit model are reported in Column 2, and those for a combined model with variables relevant to both market-based decisions and exit barriers in Column 3. Control variables Both the demand variables are correctly signed, with coefficients for REALPRICE and STARTS significant in the market-based model, but the inclusion of the T92 dummy variable dominates in the models reported in Columns 2 and 3 causing both demand variables to be insignificant. The FLETTON dummy is significant in each of the logit models indicating that low-quality fletton plants are more likely to be closed once we control for firm fixed effects. Market-based variables The central prediction of the market-based exit literature, that there is a positive relationship between Copyright  2008 John Wiley & Sons, Ltd.

37

capacity share and the probability of closure, is not supported. Only when estimated separately for plants involved in an acquisition (ACQUIRE × SHARE) is there any support for a positive size-exit relationship. However, once we include separate dummy variables for ACQUIRED and ACQUIRER in Column 3 the acquisition firm-size effect ceases to be significant. This suggests that the positive coefficient for ACQUIRE × SHARE reported in Column 1 is a consequence of a pure acquisition effect rather than a size effect. Turning to the plant-specific variables used to test Hypothesis 1a, plants that produce bricks of low aesthetic quality are more likely to be closed, with three of the four coefficients for ASTHLOW significant to the 10 percent significance level or better. However, plants that produce bricks of especially high aesthetic quality are no less likely to close than those that produce standard house bricks. With the exception of the results reported in Column 3b, either CAP IND or CAP FIRM is negative and significant at the 10 percent level or better. The former is significant in Columns 1a and 1b, while the latter is significant in Columns 1a and 3a. The lack of consistency of this result can be explained by the relatively high correlation between the two variables. Indeed, in unreported results we repeat the regressions excluding either variable and find the remaining measure of plant size is significant at the one percent level. We conduct some simple simulations using the results reported in Table 4 to confirm the importance of plant efficiency as proxied by size. The overall effect is large; the probit regressions indicate that a plant that is half the size of both the industry average and the owning firm’s average has a five to seven percent higher probability of closure during any given year. Although the level of competition (RIVAL) does not have a robust effect on the probability of closure, there is evidence to suggest that expansionary activities of rival firms (RIVALINC) reduces the likelihood of closure. This effect is, however, only evident in the results for the probit regressions reported in Columns 2a and 3a. A more robust finding is the significantly positive coefficients for RIVALDEC reported in each of the six specifications, five of which are significant at the one percent level. This indicates a cascade effect in that plants are more likely to close following a reduction in capacity by local rivals. These results Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

38

A. Wood

are contrary to the market-based solution since the reduction in capacity signals an alleviation of the excess capacity problem in that region that should reduce the probability of further closures. To sum up the findings for the market based approach to exit, the results indicate that less efficient plants and plants that produced inferior products are more likely to be closed (Hypothesis 1a) but reject the hypotheses that large firms or firms that face more intense competition are more likely to close plants (Hypotheses 1c and 1b, respectively). Although there is some evidence in favor of the acquisition-size hypothesis (1d), this result is not robust to the inclusion of barrier variables and the absence of support for Hypothesis 1c leads us to reject Hypothesis 1d.

coefficient for RIVAL is insignificant in all specifications, the hypothesis that a reduction in competitive pressure leads to a greater probability of plant closure (Hypothesis 2c) is nevertheless supported by the positive relationship between RIVALDEC and the probability of closure since this indicates that firms are more likely to close plants following a reduction in capacity by local rivals. This suggests a degree of tacit cooperation among firms as closures by one firm enables rivals to follow suit. This is consistent with the concern of firms to maintain their market shares and their reluctance to take the lead with capacity closures. But once one firm leads the way others follow giving rise to a clustering effect with periods during which closures by one firm are followed by closures by rival firms.

Barriers to exit Despite interview evidence suggesting firms care about the employment implications of closing isolated rural plants (Hypothesis 2a), little support is found for this concern influencing firms’ closure decisions in a systematic way since the coefficient on FRIEND is insignificant and we find a significant coefficient for RURAL in just one specification. There is, however, a robust positive relationship between the number of plants operated by a firm in a given region, PLANTS, and the probability of one of those plants being closed. This finding corresponds with previous studies by Lieberman (1990) and Gibson and Harris (1996), with Lieberman suggesting that the number of plants captures the ability of firms to avoid large redundancies. While this may be part of the explanation, interviews conducted for the current study suggest that another explanation is when a firm operates a large number of plants in a given region it is easier for that firm to close individual plants while reassuring customers that supply can be maintained from the remaining plants. This not only helps the firm to retain loyalty from large and long-term customers, but also leaves the firm in a good position to take advantage should there be a recovery in demand. From the probit regressions we find that highly diversified firms are no more likely to close plants than firms with a narrow focus on the brick industry, thus rejecting Hypothesis 2b.17 Although the 17 The variable DIV is not included in the logit regressions to avoid perfect colinearity with firm fixed effects.

Copyright  2008 John Wiley & Sons, Ltd.

Acquisitions The same regressions show newly acquired (ACQUIRED) plants were upwards of 15 percent more likely to be closed during the year of acquisition, while plants that were already owned by the acquiring firm (ACQUIRER) were no more likely to be closed. The reason for the greater likelihood of acquired plants being closed was explored during interviews. Several on-site interviews were conducted and were followed up by a tour of the plant conducted by a senior manager. During these tours it was obvious that there was a strong emotional bond between the tour guide (production director or factory manager) and the factory. The guide had taken part in the life history of the plant and could provide the dates and details of expansions and even minor modifications to the plant and machinery and anecdotes about teething problems with new equipment, just as a parent might fondly recount the developments of a child. These data provide evidence that senior management, and especially production directors, had a commitment to individual plants. In practice this meant that they were determined to defend the survival of plants, and that this determination was driven by a deep personal involvement in the design and development of those plants; hence comparable acquired plants are chosen for closure ahead of previously owned plants. Closer inspection of the underlying data reveals that two of the acquisition-based closure programs were large, not just in terms of number of plants Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies Table 5.

Rationalization following mergers and takeovers Acquiring firm

1992 1993 1995 1996

39

Redland Tarmac Ibstock Ibstock

Selling firm

Steetley Marley Tarmac Redland

Number of plants before acquisition

Number of plants closed since acquisition

Acquiring firm

Selling firm

Originally part of acquiring firm

Originally part of selling firm

18 7 13 22

13 2 10 17

4a 0 1c 0d,e

4a 0 3b 2d

a

Does not include two plants sold as a result of a DTI ruling The three former Tarmac plants were not closed until after the acquisition of Redland c This plant was one of the two plants acquired from SBC during 1994. d Excludes five plants sold to Ambion, as part of DTI/MMC requirement e Not including the three former Tarmac plants that closed after the Redland acquisition b

closed, but also in the magnitude of actual capacity. Summary information in Table 5 shows the four major acquisitions were directly associated with a total of 14 plant closures, with the largest single closure program linked to the acquisition of Steetley Brick by Redland when Redland closed eight plants during the year of acquisition. The overall capacity of these eight plants was equivalent to 29 percent of the total capacity purchased from Steetley, while four of the eight plants closed by Redland were former Steetley plants, representing 19 percent of the acquired capacity. Two of the three subsequent acquisitions were also linked with several plant closures. Ibstock followed its acquisition of Tarmac Brick by closing four plants, including three former Tarmac plants, amounting to approximately 30 percent of the newly purchased capacity. Then in 1996 Ibstock acquired Redland Brick and immediately closed two former Redland plants.18 The coefficients for the time dummies indicate the first acquisition heralded an industry-wide program of rationalization in a way the later acquisitions did not. The highly significant coefficients for T92 reveal that all plants, irrespective of whether they were directly involved in the acquisition, were more likely to be closed during the year of acquisition. The importance of this acquisition in achieving a breakthrough is further revealed by an analysis of the temporal distribution of closures as 18

Although these represented a relatively small proportion of the acquired capacity, it is interesting to note that a condition of the acquisition imposed by the Secretary of State for Trade and Industry was the sale of five further brickworks, with a total capacity of approximately 170mn bricks a year (over 30% of Redland’s capacity prior to acquisition), as a going concern. Copyright  2008 John Wiley & Sons, Ltd.

illustrated in Figure 3 and Table 6. Prior to 1992, closures were restricted to small non-fletton plants and a handful of large but very inefficient fletton plants. Despite the need for brick manufacturers to respond quickly to the fall in demand19 there was a standoff between rival firms as each firm waited for its rival to move first before undertaking major closures. The non-fletton plants closed during the first three years of the recession represented just 24 percent of the total non-fletton capacity closed over the whole sample period. Interestingly, Redland was not the only plant looking to make the breakthrough by increasing concentration. During interviews, the managing director of Steetley’s brick division (at the time of their acquisition by Redland) confirmed that an aborted SteetleyTarmac joint venture proposed during the winter of 1991/1992 was also intended to rationalize production capacity. The effect of concentration was immediate. In addition to the eight plants closed by Redland, a further 12 non-fletton plants were closed by Redland’s rivals. The total of 20 plants closed during this year represented 42 percent of all non-fletton capacity closed throughout the whole 1989–1996 period (Figure 3). This demonstrates that Redland’s acquisition of Steetley represented a general breaking of the logjam with rival firms following the lead of Redland with its own closures, albeit 19 This point is emphasized by the fact that even during the very early stages of the recession, unions were sympathetic to the need for rationalization. Following the announcement of 400 redundancies by LBC, Brian Cox, a national officer of the Transport and General Workers’ Union, commented: ‘The industry is reducing its capacity too in keeping with the reduced demand. It is ridiculous not to do so.’ (Building, 1989).

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

capacity as% of total closed 1989-96

40

A. Wood 50% 40% 30% 20% 10% 0% 1989 1990 1991 1992 1993 1994 1995 1996 total

Figure 3.

non-fletton

Distribution of capacity reductions

Table 6. Summary of plant closures: number and average size by year Number of plant closures

1989 1990 1991 1992 1993 1994 1995 1996 Total sample

Average size of plant closures

total

non-fletton

total

non-fletton

3 10 6 20 4 2 1 6 52

1 9 5 20 3 2 1 6 47

36.7 12.1 39.8 19.2 63.0 9.5 23.0 30.2 25.5

10.0 10.1 17.8 19.2 38.0 9.5 23.0 30.2 19.4

on a smaller scale.20 This link between acquisitions and the long-awaited rationalization can be explained within the framework of prospect theory. According to prospect theory, an individuals’ well-being depends on changes to a given prospect in relation to a reference point or frame (Kahneman and Tversky, 1979). Applying prospect theory to the brick industry, firms are concerned about the prospect of their market share. Their reference point is their prevailing market share, with negative changes to this prospect resulting from closures of their own plants and positive changes

20 It is worth noting that by no means did all of Redland’s rivals undertake closures during this period. Indeed, only two major brick manufacturers, Butterley and Tarmac, closed plants in 1992. Other major producers, including Baggeridge and Salvesen who managed to avoid any closures throughout the whole postboom period, reaped the benefit of the 1992 closures without bearing any of the cost.

Copyright  2008 John Wiley & Sons, Ltd.

resulting from the closure of rival plants. Kahneman (1992) has used prospect theory to demonstrate how conflict resolution that involves concessions is particularly hard to achieve.21 Because losses weigh heavier than gains, the gains to the resolution must significantly outweigh the loss due to the concession, otherwise the rivals will be averse to concessions and the status quo or exit logjam prevails. Applying this to the coordination of plant closures subject the constraint of maintaining market share, an acquisition first enables the acquiring firm to make significant closures that coincide with the acquisition without experiencing this as a loss in market share. The resulting decline in overall capacity allow rival firms to follow suit on a smaller scale, also without experiencing this as a loss in market share relative to their preacquisition reference point. Subsample analysis In order to further explore the impact of concentration by acquisitions, we report in Table 7 results for regressions using the preconcentration sample (1989–1991) in Panel A, the initial concentration of 1992 in Panel B, and the subsequent period of concentration (1993–1996) in Panel C. For ease of exposition we only report results for the probit regressions using those variables that were found to be significant for the full sample.22 The first column of each panel excludes CAP FIRM, while the second column excludes CAP IND to allow for the correlation between our plant capacity variables. Two results stand out. First, RIVALDEC is only significant in Panel A. This indicates that the cooperative follow-my-neighbor effect is only evident during the initial years following the collapse in demand when firms were tentatively taking turns to close small plants. The initial period of concentration saw substantial closures by the acquiring firm, with acquired plants in excess of 38 percent more likely to be closed, which was followed by reciprocative closures by rivals that took place during the same year of the initial closures and is therefore not picked up by the RIVALDEC variable in Panel B. Plant closures during the final period were primarily undertaken by the acquiring 21 Kahneman (1992) suggested applications to international conflict, with rival countries negotiating over missile reductions, and to labor negotiations. 22 Qualitatively similar results are obtained using logit estimations.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies Table 7.

41

Subsample probit regressions predicting plant closures Panel A: 1989–91 (1)

(2)

(1)

−0.0005 (0.517) 0.0018∗∗∗ (0.006) 0.0017 (0.901)

0.4207∗∗ (0.034) −0.0146 (0.823) 0.0677 (0.190) 0.1818 (0.621) −0.2714∗∗∗ (0.002)

ACQUIRED RIVALINC RIVALDEC RIVAL CAP IND

−0.0006 (0.823) 0.0058∗∗∗ (0.007) 0.0054 (0.900) −0.0164∗∗∗ (0.006)

CAP FIRM 0.0109∗∗ (0.050) PLANTS 0.0020∗∗ (0.012) DIV 0.0110 (0.216) RURAL −0.0007 (0.873) REALPRICE −0.0696 (0.322) STARTS 0.0000 (0.839) FLETTON −0.0041 (0.306) Log likelihood −51.845 LR test random effects 0.54 Pseudo R2 0.284 Sample 325 ASTHLOW

−0.0079∗∗∗ (0.001) 0.0065∗∗ (0.011) 0.0008∗∗∗ (0.002) 0.0029 (0.295) 0.0000 (0.998) −0.0215 (0.317) 0.0000 (0.918) −0.0011 (0.323) −46.880 1.20 0.352 325

firm, with acquired plants some 18 percent more likely to be closed, and were not reciprocated by rivals. The second result relates to CAP IND and CAP FIRM, as both variables are highly significant in Panels A and B but only CAP FIRM is significant in Panel C with a p-value of 0.045. This indicates that the plants closed during the initial period and around the time of the 1992 acquisition were the least efficient plants both in respect of the closing firm and the industry as a whole, but those closed during the final period were not the least efficient when compared with the industry as a whole. The significant coefficient for CAP FIRM suggests that, while the closed plants were not the least efficient in the industry, they were among the least efficient plants owned by the closing firm, once we control for the greater probability of closing acquired plants. This discrepancy can be explained by the relatively small number of firms Copyright  2008 John Wiley & Sons, Ltd.

Panel B: 1992

0.1149 (0.139) 0.0212 (0.196) 0.0313 (0.723) −0.0439 (0.571) −0.0140∗ (0.068) −0.1108 (0.150) −37.253 NA 0.255 100

(2) 0.3845∗∗ (0.050) −0.0364 (0.631) 0.0563 (0.309) 0.1173 (0.782) −0.2493∗∗∗ (0.006) 0.1032 (0.224) 0.0327∗ (0.084) 0.0196 (0.842) −0.0593 (0.498) −0.0169∗ (0.054) −0.1429 (0.118) −40.015 NA 0.200 100

Panel C: 1993–96 (1)

(2)

0.1811∗∗∗ (0.001) −0.0049 (0.183) 0.0037 (0.217) −0.0130 (0.396) −0.0063 (0.153)

0.1775∗∗∗ (0.001) −0.0041 (0.165) 0.0028 (0.237) −0.0106 (0.388)

0.0041 (0.420) 0.0001 (0.942) −0.0080 (0.147) −0.0038 (0.418) 0.0538 (0.443) 0.0003 (0.174) 0.0647 (0.158) −39.997 1.12 0.247 319

−0.0083∗∗ (0.045) 0.0036 (0.386) −0.0001 (0.911) −0.0066 (0.128) −0.0031 (0.397) 0.0349 (0.535) 0.0002 (0.162) 0.0522 (0.167) −38.815 1.09 0.269 319

involved in plant closures post-1992. During this period, most closures were carried out by Ibstock, the firm at the heart of the two largest post-1992 acquisitions, while firms not involved in the acquisitions stood back.

DISCUSSION AND CONCLUSIONS The major issue addressed in this study is whether rationalization is best left to the market or whether M&A-led concentration is necessary before major rationalization can be achieved. The British brick industry provided an ideal testing ground following the collapse in the construction boom in the late 1980s. The initial response of firms was to close very small plants or those producing poor quality bricks. However, the greatest strides toward removing excess capacity were made during a sixmonth period a full three years after the collapse Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

42

A. Wood

in demand. The event that triggered this breakthrough was the acquisition of one large firm by another (Steetley Brick by Redland) in 1992. Two later acquisitions also involved subsequent closure programs. This link between acquisitions and closures is not of itself evidence of Bowers’ (1986) contention that concentration is necessary to overcome barriers to exit. A market-based explanation can also be derived from the size-exit models of Reynolds (1988) and Ghemawat and Nalebuff (1990). According to these models, firms may have an incentive to acquire rivals prior to closing capacity to internalize the gains from plant closures. Acquisitions merely facilitate the market mechanism. However, our failure to find a link between market share and the probability of closure suggests that the need to internalize the gains from incremental closures is not a determining factor in influencing the nature of rationalizations. Instead, our evidence suggests that nonmarket barriers to exit prevented a speedy reduction in excess capacity and that the acquisitions enabled the industry to overcome the resultant exit logjam. For instance, our interviews revealed a wish by key decision makers to be loyal to long-standing employees and a commitment to the plants that they themselves had developed. Takeovers allowed acquiring firms to sidestep these barriers by providing the opportunity to close mainly newly acquired plants. Our regression results indicate that newly acquired plants were significantly more likely to be closed than plants previously owned by the acquiring firm. Our results suggest that Bower’s (1986) concentration hypothesis can be generalized to small competitive industries where national political influence is a minor consideration. They reveal a twostage approach to rationalization on the basis of the sharp contrast between those plants that were closed prior to and after the commencement of concentration. Although the number and size of plants closed prior to the first acquisition was completely inadequate given the scale of excess capacity, those closed were the least efficient within the industry. The same is true of those plants that were closed around the time of the first acquisition, with the caveat that newly acquired plants were more likely to be closed irrespective of their relative efficiency. But as the dust settled, firms that were not involved in the acquisitions stood back and left the acquiring firm to shoulder the responsibility of Copyright  2008 John Wiley & Sons, Ltd.

removing the remaining excess capacity. Interestingly, this involved closing plants that were not necessarily the least efficient within the industry contrary to the neoclassical hypothesis. The initial market-based stage of rationalization is that firms cooperate to engage in the closure of small marginal plants. However this Nash-type solution does not resolve the fundamental problem of excess capacity in the face of a sharp drop in demand. Of all the barriers to exit, the most frequently cited during interviews was the concern of managers to maintain their respective market shares at all costs. This gives rise to a coordination and free rider problem that is best understood within the framework of prospect theory and is a major obstacle to rationalization. The second stage sees the managers of large or dominant firms taking the lead through a strategy of large M&As with resultant sizeable cuts in capacity mainly in the acquired plants. Such closures are easier for managers to accept because they form part of a net increase in the new firm’s capacity and market share. This solution has two main features. On the one hand, the net effect is that these dominant firms increase their overall market share, and is therefore consistent with their managers maximizing growth. On the other, the overall reduction in industry capacity frees the exit logjam and enables the remaining firms also to make closures while their managers can maintain their market share intact. In conclusion, our findings generalize the Bower (1986) hypothesis that concentration through acquisition is a necessary prerequisite for fundamental restructuring to the case of a relatively small competitive industry. Acquisitions enable acquiring and other firms to overcome many exit barriers and the exit logjam. We refine the Bower hypothesis by showing that the resulting rationalization is likely to be skewed with a disproportionate burden being shouldered by the predator firm, which favors the closure of newly acquired over previously owned plants. It would be interesting to see if our new results apply to other industries. This will be the subject of future research.

ACKNOWLEDGEMENTS I am grateful to Ciaran Driver, Jerry Coakley, William Dixon, Associate Editor Karel Cool, and Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Capacity Rationalization and Exit Strategies two anonymous referees whose detailed and helpful comments have substantially improved the exposition of the article.

REFERENCES Baden-Fuller CWF. 1989. Exit from declining industries and the case of steel. The Economic Journal 99(398): 949–961. Bower JL. 1986. When Markets Quake. Harvard University Press: Cambridge, MA. Building. 1989. Drop in demand forces London Brick layoffs. 26 May: 8. Deily ME. 1988. Investment activity and the exit decision. The Review of Economics and Statistics 70(4): 595–602. Deily ME. 1991. Exit strategies and plant closing decisions: the case of steel. RAND Journal of Economics 22(2): 250–263. Ghemawat P, Nalebuff B. 1985. Exit. RAND Journal of Economics 16(2): 184–194. Ghemawat P, Nalebuff B. 1990. The devolution of declining industries. The Quarterly Journal of Economics 105(1): 167–186. Gibson JK, Harris RID. 1996. Trade liberalisation and plant exit in New Zealand manufacturing. The Review of Economics and Statistics 78(3): 521–529.

Copyright  2008 John Wiley & Sons, Ltd.

43

Harrigan J. 1980. Strategies for Declining Businesses. Heath and Co: Lexington, MA. Jensen MC. 1993. The modern industrial revolution: exit and the failure of internal control systems. Journal of Finance 48(3): 831–880. Kahneman D. 1992. Reference points, anchors, norms, and mixed feelings. Organizational Behavior and Human Decision Processes 51(2): 296–312. Kahneman D, Tversky A. 1979. Prospect theory: an analysis of decision under risk. Econometrica 47(2): 263–291. Lieberman MB. 1990. Exit from declining industries: ‘shakeout’ or ‘stakeout’? RAND Journal of Economics 21(4): 538–554. Reynolds SS. 1988. Plant closing and exit behavior in declining industries. Economica 55(220): 493–503. Ridgway JM. 1982. Common Clay and Shale. Mineral Resources Consultative Committee, Mineral Dossier no. 22, HMSO: London. Scheibl F, Wood A. 2005. Investment sequencing in the brick industry: an application of grounded theory. Cambridge Journal of Economics 29(2): 223–247. Whinston MD. 1988. Exit with multiplants. RAND Journal of Economics 19(4): 568–588. Wood A. 2005. Investment interdependence and the coordination of lumpy investments: evidence from the British brick industry. Applied Economics 37(1): 37–49.

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj

Copyright  2008 John Wiley & Sons, Ltd.

0.05 0.06 −0.05 0.12 0.02 0.14 0.06 −0.03 0.04 −0.07 −0.08 0.00 0.00 −0.09 −0.01

23

0.13 −0.02 0.10 0.07 0.06 0.12 0.33 −0.08 0.13 0.00 0.05 0.17 −0.21 −0.11 0.19 0.11 0.10 0.06 0.01 −0.01 0.01 −0.02 0.02 0.12 −0.17 −0.17 −0.15 0.03 0.14 0.03 −0.03 −0.01 0.02 −0.05 −0.21 0.16 0.24 0.12 0.00 0.09 −0.07 −0.07 0.01 0.10 −0.18 0.07 −0.06 −0.06 0.01 −0.02 −0.02 0.00 0.00 −0.23 0.06 0.03 0.06 −0.02 −0.02 0.00 0.09 −0.21 0.05

To preserve the panel structure correlations are derived from the slope coefficient from two-way regressions.

0.52 0.05 −0.12 0.42 0.19 0.09 0.01 0.33 −0.05 −0.06 −0.02 0.03 0.01 −0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.01 0.06 0.11 −0.03 0.02 −0.02

−0.07 0.01 0.01

−0.01 −0.01

−0.54

CAP ASTH- ASTH- REALFIRM LOW HIGH PRICE

0.64 0.03 0.03 −0.04 −0.03 −0.03 0.01 0.03 −0.03

CLOSE SHARE PLANTS RIVAL- RIVAL- RIVAL ACQUI- ACQUI- RURAL FRIEND CAP IND INC DEC RED RER

Correlation matrix23

SHARE PLANTS RIVALINC RIVALDEC RIVAL ACQUIRED ACQUIRER RURAL FRIEND CAP IND CAP FIRM ASTHLOW ASTHHIGH REALPRICE STARTS

Appendix

44 A. Wood

Strat. Mgmt. J., 30: 25–44 (2009) DOI: 10.1002/smj