Small Business Economics

, Volume 39, Issue 1, pp 207–229

Consolidation by merger: the UK beer market

Authors

Article

DOI: 10.1007/s11187-010-9295-2

Cite this article as:
Esteve-Pérez, S. Small Bus Econ (2012) 39: 207. doi:10.1007/s11187-010-9295-2
  • 620 Views

Abstract

This paper examines the determinants of market structure in the UK brewing industry over 1949–1969. Sutton (Sunk costs and market structure: price competition, advertising, and the evolution of concentration, 1991) points to technology and advertising races as two key drivers of market concentration. This study uses an own-built longitudinal data set of the population of firms and breweries, and reveals the importance of institutional factors in explaining the dynamics of market structure. The practice of tying outlets to brewers and legal restrictions on opening retail outlets, together with a permissive policy towards mergers, made acquisition of medium-sized firms and brewery closure (shortly after acquisition) the main driving mechanism towards industry consolidation. During this period, the number of firms and plants fell sharply (by 74% and 60%, respectively) and production rose (by about 25%), with no firm entry and just one new brewery opening. As a result, concentration increased and the market transformed from a highly fragmented one into a stable oligopoly.

JEL Classifications

C41D21L22L26

Keywords

Concentrating industryExitAcquisitionLiquidationSurvival analysis

1 Introduction

The evolution of market structure has long been studied by economic and business researchers. Geroski and Mazzucato (2001) point out that a market adapts to changing environment through changes in the population of firms (“selection”) and through changes in the features and competencies of firms in the market at any time (“adaptation”). Firms’ entry, growth and exit constitute dynamic processes of selection which affect efficiency, toughness of competition and market structure over time. Firms’ efficiency and its evolution over time have been commonly highlighted as the main driving force (for instance, in the “entry and post-entry performance” literature: Jovanovic 1982; Mata and Portugal 1994 and the Special Issue of the International Journal of Industrial Organization 1995 on the Post Entry Performance of Firms; and the “shake-out” literature: Klepper 1996; Klepper and Simons 2000, K&S henceforth). Sutton (1991) challenges the dominant paradigm and develops a theoretical framework that points out the importance of changes in minimum efficient scale (exogenous set-up costs) and increasing competition in advertising (endogenous sunk costs) in order to explain the evolution of market structure. This author further emphasizes the importance of institutional factors and the particular history of a sector to explain market structure. Finally, other authors (Ghemawat and Nalebuff 1985, 1990, G&N hereafter; Agarwal and Gort 1996) underline the importance of strategic factors in declining phases of a market.

This paper contributes to this literature by examining a particular industry in which manufacturers are vertically integrated. The primary goal is to analyse the determinants of the process that transformed the UK brewing industry into an oligopoly with a fringe of very small competitors. The shift from a highly fragmented industry, comprising a large number of small (geographically dispersed) producers supplying local markets through the ownership or control of tied houses,1 to a rather (horizontally) concentrated market took place mainly over 1949–1969.2 The selection process was extremely fierce, with only 26% of the firms (and 40% of the plants) in the market in 1949 surviving up to 1969, while beer production rose by 25%, together with no firm entry, one new brewery opening and the birth of the National Brewers. Industry concentration rose sharply, as is shown by the remarkable increase of the five largest firms’ share of output (employment) from 18% (19%) to 62% (61%) during 1954–1968 (Mark 1974). Thereafter, market structure remained fairly stable until the 1990s.

On theoretical grounds the evolution of an industry towards consolidation can take place through two main routes: (1) internal growth of entrants and incumbents, and exit of non-successful firms (mainly through liquidation) or (2) external growth through mergers and acquisitions with capacity rationalisation. In some markets, a fringe of small firms may also survive, supplying particular market niches, leading to a dual structure (Sutton 1991).

In examining this industry, the paper makes the following contributions. First, it provides evidence of an industry that represents an exceptional case of industry consolidation driven by firm exit and capacity reduction with practically no entry. Secondly, the distinctive institutional features of this industry in the UK (when compared with that in other countries, such as the USA), due to the extent of vertical integration, the restrictive licensing laws, the stance of competition policy towards mergers, and consumers’ tastes, remarkably shaped the consolidation process. Thirdly, in contrast to most previous studies, this paper focusses on firms’ exit, distinguishing among two different forms of exit: liquidation and acquisition. The decision to close down a brewery is also examined.

The empirical work is carried out using a comprehensive own-built data set, which comprises the population of actively brewing firms (brewers) and plants (breweries) in the UK over 1949–1990. The analysis is based on 1949–1969, but in some cases the discussion is extended up to 1990 in order to highlight the intensity of the consolidation process during the first period. To proceed with the empirical work, I use survival methods, including the estimation of Cox proportional hazards models (Cox 1972, 1975). Survival methods exploit the longitudinal aspects of the data, i.e. they account for both the occurrence of an event (whether a firm and/or brewery exits, that is, it fails to survive) and the timing of the event (when that occurred), distinguishing between periods of high risk or low risk of exit. They also control for the presence of censored observations (i.e. a number of firms/breweries had not completed their life-time spells by the end of the sample period), and allow easy handling of time-varying explanatory variables. The latter is desirable since a firm’s ability to survive will probably change over time. A competing-risks model is further used to explicitly address the joint determination of the probability of exit and the form of exit: liquidation and acquisition.

The results of this paper suggest that the particular features of the industry in the UK shaped the process of industry consolidation. The steady decline in the number of independent brewers and plants, and the rise in beer production, point out the importance of economies of large-scale production, pushing the least efficient firms out of the market. However, the tied house system and the restrictions on new licences to sell beer, together with the lenient competition policy in respect of mergers, consumers’ tastes and product characteristics shaped both the form and the extent of the consolidation process. Despite the common tendency towards the rise in concentration across countries, this paper highlights that both the driving forces of the process and the resulting market structure in the UK differed from those in other countries. In the UK, those factors facilitating acquisition raised the firms’ risk of exit, so that acquisition became the primary route for elimination of inefficient breweries: acquiring companies grabbed their acquired rivals’ market share (through their tied retail outlets) and rationalised capacity shortly after acquisition. The sector consolidated a dual structure over time, with a large fringe of small craft brewers supplying their local market and a few large companies pursuing a national (and international) strategy.

In next section a brief revision of the related theory and some relevant features of the UK beer market are presented. Section 3 describes the data set, and the empirical methods are outlined in Sect. 4. The empirical analysis comes in Sect. 5, and finally Sect. 6 concludes.

2 Theory and industry background

The primary goal of this paper is to explain how an industry evolves over time, trying to shed light on the main determinants of this process. To this end, this paper relies on some theories of market structure that are applied to a particular market. This section briefly outlines the relevant theory for the empirical work and some features of the market that are relevant for the analysis.

Geroski and Mazzucato (2001) argue that markets adapt to changes in environment through changes in the population of firms (“selection”) and through changes in the characteristics and competencies of the firms that populate the market at any time (“adaptation”). In principle, there are two major routes for business growth and industry consolidation: (1) internal growth, and exit of non-successful firms (mainly through liquidation) or (2) external growth through mergers and acquisitions with capacity rationalisation.

Researchers have long and extensively studied the determinants of the processes that shape the evolution of market structure. The traditional approach relied on the structure–conduct–performance paradigm (Bain 1956), which predicts that industry concentration is driven by exogenous structural factors. Most theoretical and empirical studies point towards technology as the main factor driving market structure. Jovanovic’s (1982) model of passive learning emphasizes the role of exogenous sunk costs of entry and firm heterogeneity in their efficiency levels (unknown before entry) as a main factor explaining industry dynamics. Firms learn about their true efficiency as time goes by: the most productive ones survive and grow over time, whereas the least productive firms exit the market. Klepper (1996) suggests that dynamic scale economies in process innovation confer an advantage for large firms, leading to industry consolidation in mature phases of an industry life cycle. These technology-based models of industry dynamics make no direct predictions on the consolidation route. The related empirical work does not take account of the different exit routes: liquidation and acquisition.

Sutton (1991) develops a framework that challenges the structure–conduct–performance paradigm. According to this author, the combination of three main factors shapes the evolution of market structure over time: (1) changes in minimum efficient scale (due to changes in set-up exogenous sunk costs), (2) advertising races (endogenous sunk costs) and (3) institutional factors in the particular industry. He also provides a number of empirical investigations on the evolution of a number of markets, including the US beer market. The evidence points to the role of increases in minimum efficient scale due to technological innovation and competition in advertising outlays, in relation to market size, as the two key factors in explaining the evolution of market structure. His empirical work further points out the importance of two key institutional factors that have an influence on how an industry responds to the previous concentrating forces: on one hand, the stance of the competition policy authorities in respect of merger and acquisition activity; on the other hand, the practice of tying retail outlets to particular manufacturing companies. Hence, insofar as mergers are not discouraged by competition authorities, they offer a possible route toward accommodating changes in minimum efficient scale (and advertising races). Furthermore, if retail outlets are tied to manufacturers, the effectiveness of advertising in order to increase market share will be diminished.

2.1 Industry background

The number of independent brewers and brewing plants (breweries) has declined steadily since World War II, while production and concentration have increased. Despite the important changes in consumption, production and market structure during the 1990s (Pinkse and Slade 2004; Slade 1998, 2004), the process of consolidation of the industry took place mainly over 1949–1969. Table 1 illustrates the decline: the number of firms (breweries) fell from 353 (433) to 91 (175) between 1949 and 1970. Moreover, during 1900–1970 there was no firm entry and only one new brewery was opened, in 1969. Since 1970, the number of brewers and breweries fell to 64 and 105 in 1990, respectively.
Table 1

The number of beer producers and breweries in the UK (1949–1990)

Year

Number of firms

Number of plants (breweries)

Six largest firms by number of breweries

Number of firms brewing at three or more sites

Number of firms brewing at four or more sites

Total breweries

% Breweries

1949

353

433

28

6.5

16

5

1950

347

426

29

6.8

16

5

1955

280

364

30

8.2

18

7

1960

192

296

54

18.2

19

12

1965

129

230

75

32.6

15

10

1970

91

175

74

42.3

10

7

1975

80

146

59

40.4

10

7

1980

77

137

53

38.7

9

7

1985

70

123

41

33.3

11

8

1990

64

105

39

37.1

7

5

Source: Author’s estimations. See Sect. 3 for details

Notes: Arthur Guinness Son Co. Ltd. is not included as one of the “largest” since it had only one brewery actively brewing in the UK

Furthermore, during 1949–1969, the National Brewers emerged, and horizontal concentration sharply increased. The five-firm concentration ratio rose from 18% to 62% over 1954–1968 (Mark 1974), whereas it grew from 24.9% to 47.4% in the USA (Greer 1971). Beer sales and production rose by 25%. Foreign competition has always been very limited (imports usually exceeding exports, but seldom over 5% of UK production). Moreover, the seven largest brewing companies accounted for 73% of total UK beer production in 1967 (UK Monopolies Commission 1969, p. 5). After 1970, market concentration remained fairly stable, so that the six largest brewer companies (the “Six National Brewers”) supplied 75% of UK beer production in 1989 (UK Monopolies and Mergers Commission 1989, p. 2). Columns 4–7 of Table 1 reveal that brewery concentration mirrors the increase in output concentration over this period.3 The proportion of breweries owned by the six largest firms increased from 6.5% in 1949 to 42.3% in 1970.

The fall in the number of firms (by 74%) and plants (by 60%), together with the rise in beer consumption and production (by about 25%) point to the existence of economies of large-scale production during the concentrating phase of the industry’s life cycle. Greater economies of scale brought about by technological innovations in packaging, automation, water treatment and transportation likely played a role in explaining concentration. According to Sutton (1991), another important driver of concentration is the escalation in advertising expenditures by the largest firms. Greer (1971) attributes more importance to advertising than to scale economies as a driver of concentration in US brewing. In order to assess the role of advertising as well as to attain a better understanding of the concentrating process, some crucial features of the industry must be considered. They are outlined in turn.

First, historically, vertical relations in all phases of production and distribution in the UK brewing industry have been significant. The extent of them ranged from vertical integration (managed and tenanted houses) to control of independent retailers through loan ties (that is, independent retailers that receive loans at below-market rates in exchange for exclusive dealing arrangements). The brewers’ control on retailing (tied house system) was the result of restrictive licensing practices from the end of the 19th century together with no restrictions on the power of brewers to acquire premises to sell beer until the 1989 Monopolies and Mergers Commission report. Public policy imposed restrictions on both the number and the quality of premises licensed to sell beer. The Licensing Acts of 1869 and 1904 increased these restrictions, which were only slightly alleviated by the Licensing Acts of 1961 and 1964. Hence, brewers had control on a high percentage of pubs, where the availability of their competitors’ beers was limited to reciprocal agreements on very few national beers (especially bottled Guinness stout). Thus, in 1967, about 90% of the beer sold by the brewers was the sellers’ own brew. Secondly, competition policy towards mergers and acquisitions was rather lenient until 1969. Thirdly, on-premises consumption of draft beer made up a rather unusually high proportion of total sales (when compared with that in other countries, e.g. the USA). Moreover, the strong regional variation in beer preferences led to the existence of a large variety of beers, with very few truly national draft beers. In addition, lager’s market share was rather low until the 1970s, from under 1% in 1953 to 2% in 1965. As a result, national advertising played a less important role than in other countries (e.g. the USA).

The institutional set-up of the industry shaped the process of industry consolidation. It implied little scope for raising market share through price competition, given the existence of captive consumers. The existence of captive markets given the importance of on-premise consumption, the tied house system and the restrictions on new licences to sell beer also reduced the effectiveness of advertising in order to explain the evolution of market structure. During the 1960s, as a national market for beer was developing, promotional expenditures became more important, which gave large companies an additional advantage given the size of funds required. Moreover, advertising became more important in the lager segment as this product raised its importance after 1970.

Furthermore, the presence of high barriers to entry arising from the brewers’ tight control on retail trade, the high incidence of beer consumption at the pub and the legal restrictions on opening new outlets (and the large investment required to open and run them) effectively deterred entry from 1900 up to 1970. In addition, the existence of captive markets prompted by the tied house system let many inefficient producers survive longer than their inherent competitive ability would have suggested, and restrained (voluntary) production capacity rationalisation, despite the latent situation of excess capacity in the industry (Hawkins and Pass 1979, pp. 41–43).

These features made external growth a more likely route towards industry consolidation than internal growth. Expanding brewers had an incentive to acquire other brewers, and so obtaining access to their tied outlets in order to reach new consumers and enlarge their market share.4 Besides, acquisition also allowed attaining well-known brands and/or reaching particular geographic locations. Accordingly, I expect a higher incidence of firm exit through mergers and acquisitions than through liquidation. Acquirers could grab their acquired rival’s market share and also reduce costs of production and marketing. A second prediction relates to the way in which capacity reduction took place in the industry. Soon after acquisition, inefficient brewing capacity could be removed with no restriction, so that the scale of production of more efficient plants could be increased. Hence, I expect acquisition and brewery closure to be the main route for consolidation, rather than liquidation of inefficient brewers. The former was probably enhanced by the improvement of transportation conditions within the UK and the introduction of technological refinements that increased durability of the product after WWII. They probably brought about further incentives to concentrate capacity and take advantage of scale economies in production and distribution.5 Furthermore, the particular features of the market probably allowed a fringe of small, craft producers to survive, supplying their local captive markets.

3 Data

The study relies on an own-built longitudinal data set of the population of firms and plants (breweries) in the UK brewing industry over 1949–1990. The data have been gathered from several sources. The data comprise annual information on the number of UK brewing firms and breweries actively brewing,6 their entry and exit or closure dates (from 1949) and their location (both city and county according to the Economic Planning Regions reorganisation of regions in 1974). Therefore, the analysis is conditional on firms (and their breweries) having survived up to 1949. All firms alive in 1949 had entered the market before 1900. In order to obtain more accurate information on active breweries and firms and to isolate actual capacity reduction and firm exit the information on wholesale breweries provided by the Brewers’ Society (those registered with H.M. Customs and Excise) has been compared with that from Barber (1996), Richmond and Turton (1990) and other secondary sources (Gourvish and Wilson 1994, among others). The figures from the Brewers’ Society may exaggerate the actual number of active breweries, as some of them remained in the trade without brewing. Finally, other sources have been used to account for heterogeneity among firms leading to different fates.7

The choice of the period of analysis deserves some attention. First, the consolidation of the industry took mainly place over the 1950s and 1960s. It went through a severe restructuring process that transformed a rather disperse industry into a stable oligopoly. Secondly, data availability and reliability also led to selection of 1949 as the initial year for statistical analysis. For instance, information used to proxy firm size and profitability are obtained from the Cambridge/DTI Databank of Annual Company Accounts, which provides information starting in 1949. Thirdly, in 1989 the Monopolies and Mergers Commission published its second report on the sector so that its consequences/recommendations had not started to take place yet (Slade 1998).

A brewery is computed to be closed down (exit) in year t when this is the last year of brewing at that site (although that brewery could remain as a depot or to wholesale beer brewed by other breweries). The first year (after 1949) of production by a brewery or brewery company is recorded as the plant/firm entry date. Rebuilding work on a site is not considered to be closure of an old brewery and opening of a new one. A firm exits in year t if it is actively and independently brewing at t but no longer (or no longer independently) in the market in t + 1. That is, ceasing to brew independently is considered as firm exit. There are two main exit routes for firms: liquidation and acquisition. Therefore, the number of firms (breweries) in year t + 1 (Nt+1) consists of stayers (Nt) plus entrants in t + 1 (Et+1) minus exitors in t (Xt).

The treatment of mergers and acquisitions follows K&S (2000). Whenever two (or more) firms merge (sometimes bringing about a change in name), I do not record them as two companies exiting the market and one de novo entering, but the largest company is computed as continuing brewing and the other(s) as being acquired.8 When a company is jointly acquired by two firms, the brewer that eventually gains full control of the acquired company is regarded as the single acquirer. Furthermore, name changes are not considered as a change in the number of companies. Finally, Harp Lager Ltd. is not regarded as an independent brewery company, since it was a consortium of some other brewers to brew lager.

In what follows, I examine the main determinants of firm exit, distinguishing by form of exit (liquidation and acquisition), and brewery closure. The focus primarily lies on firm-level decisions, as the firm is the locus of decision-making that primarily influences the degree of competition in the industry.

4 Modelling the hazard of firm exit and brewery closure

Survival methods9 are well suited to analyse the determinants of the restructuring process that the UK brewing industry went through because of the following features.10 First, these methods take into account the evolution of the exit risk and its determinants over time, since they control for both the occurrence and the timing of exit. Secondly, they are appropriate to analyse the determinants of firm exit taking into account different exit routes. Thirdly, survival models allow easy introduction of time-varying explanatory variables in order to account for firms’ heterogeneity explaining their different fates, given that a firm’s ability to survive may vary over time. Furthermore, as Mata et al. (1995) point out, the introduction of time-varying covariates allows the effect of the explanatory variables to change over time, overcoming the potential limitation of considering pre-sample-period characteristics (or characteristics at time of entry) as the determinants of a firm’s survival probability. Fourthly, these methods also control for censored observations, i.e. they take into account the fact that some firms (breweries) may have not suffered the event (i.e., they survive) by the end of the sample period.

The focus of the empirical work lies on explaining how the conditional probability (i.e. the hazard rate) of firm exit, the hazard rate of liquidation (voluntary exit and bankruptcy) and the hazard of being acquired, as well as the hazard of brewery closure, depend on firm and brewery characteristics and how they changed over time. To this end, a multivariate analysis is undertaken estimating several semi-parametric survival models (Cox proportional hazards model, Cox 1972, 1975). Thus, the individual effect of each explanatory variable on the hazard rate, controlling simultaneously for the effect of the other variables, is assessed.

The econometric analysis is restricted to the period 1949–1969 for the following reasons: First, as Table 1 shows, the selection process in the industry was much tougher up to 1969, with no firm entry and just one brewery opened. Over 85% of the brewers exiting over 1949–1990 had already failed by 1970. Likewise, the transition from a highly fragmented industry to an oligopolistic industry mainly occurred during 1949–1969. In addition, the importance of free trade (i.e. non-tied trade) and lager consumption was still rather limited up to 1969. Secondly, by 1969 the large brewing groups had already consolidated their presence at a national level.11 Thirdly, some explanatory variables have been built using financial information of the companies included in the Cambridge/DTI Databank of Annual Company Accounts. This data set comprises all UK quoted companies (satisfying a size requirement since 1961), but it suffers a high rate of attrition after 1969 and becomes a stratified random sample after 1977.

Next, the estimation method and the explanatory variables are discussed in turn.

4.1 The Cox proportional hazards model

A key concept in survival analysis is the hazard function, that is, the probability of a firm exiting the market (or a brewery being closed down) at time t conditional upon having survived up to that time:
$$ h\left( t \right) = \mathop {\lim }\limits_{{{\text{d}}t \to 0}} {\frac{{\Pr \left( {t \le T < t + {\text{d}}t\left| {T \ge t} \right.} \right)}}{{{\text{d}}t}}}, $$
where T is a non-negative random variable (duration).
The estimations are carried out using the semi-parametric (continuous-time) Cox proportional hazards (CPHM) specifications, for the risk of firm exit and brewery closure. It is a reduced-form model that assumes that the hazard function h(t) for the exit time t of an individual i with a vector of covariates X takes the following form:
$$ h_{i} \left( t \right) = h_{0} \left( t \right) \cdot \exp \left( {X_{i}^{\prime } \left( t \right)\beta } \right) $$
(1)
where β is a vector of unknown regression coefficients and h0(t) is an unspecified baseline hazard function (for X = 0). The effect of a unit change in a covariate X is to produce a constant proportional change in the probability of suffering the event at time t conditional on having survived up to that period (hazard rate). The baseline hazard is left unspecified, i.e. no parametric relationship between time and covariates is imposed. Vector X may include both time-constant and time-varying explanatory variables. Hence, the firm’s characteristics at time of exit together with the embodied information on its survival up to that period are relevant in explaining a firm’s hazard rate.
The data consist of a list of n times (one per subject), m (m ≤ n) of which are exit times and n − m of which are censoring times. The censoring is assumed to be non-informative. Then the maximum partial likelihood estimator \( \mathop \beta \limits^{ \wedge } \) is the value that maximizes the partial likelihood function (Cox 1972, 1975):12
$$ L\left( \beta \right) = \mathop \Uppi \limits_{i = 1}^{m} \left[ {{\frac{{\exp \left\{ {\beta 'X_{i} \left( {T_{i} } \right)} \right\}}}{{\sum\nolimits_{{j \in R_{i} }} {\exp \left\{ {\beta 'X_{i} \left( {T_{i} } \right)} \right\}} }}}} \right], $$
where Ri is the set of labels attached to the individuals at risk at time Ti. Thus, it is only the ordering of events that matters for the estimation of the CPHM and not the actual time of the event. This is a desirable feature given that the analysis is based on calendar time.13

The lack of information on tied houses per firm may lead to the presence of unobserved heterogeneity, which biases the estimates towards negative duration dependence (\( {\frac{{{\text{d}}h}}{{{\text{d}}t}}} < 0 \)). Nonetheless, Han and Hausman (1990) show that this problem may be less important when non-parametric hazard specifications are used. Moreover, Dolton and van der Klauw (1995) find that neglecting or misspecifying unobserved heterogeneity has almost no consequences when using a flexible baseline specification. In fact, they obtain that unobserved heterogeneity considerably affects the estimates when an incorrect parametric form of the baseline hazard is imposed. These authors also conclude that semi-parametric estimation procedures prevent inconsistent estimation of the covariate coefficients due to a misspecified baseline hazard, which is especially severe with time-varying covariates, although these methods involve some loss of efficiency. In addition, Wheelock and Wilson (2000) state that, although the baseline function varies over time and not over individuals at a given time period, it is evaluated at different times for different individuals (those surviving at the time) and therefore captures individual heterogeneity among those individuals exiting at different times.

The determinants of firm exit may vary for different forms of exit. Hence, I also estimate a competing-risks survival model where both whether and when firm exit takes place, as well as the exit route (i.e. the ultimate state which the firm moves into), are jointly addressed. The ultimate states must be mutually exclusive and exhaustive, i.e. a firm will exit either through liquidation or by being acquired (or survive). Following Narendranathan and Stewart (1991), it is also assumed that these risks are independent. Thus, firm exit for reasons other than the single exit route being analysed at a time are statistically equivalent to censored observations. Hence, every actively brewing firm faces two possible risks: voluntary exit/bankruptcy and cessation of independent brewing after being acquired. Thus, I estimate the (continuous-time) proportional competing-risks hazard model:
$$ h_{ir} \left( t \right) = h_{0r} \left( t \right) \cdot \exp \left( {X_{ir} \left( t \right)\beta_{r} } \right)\;\;\;\;{\text{r}} = \left\{ {{\text{liquidation;}}\;{\text{acquisition}}} \right\}, $$
(2)
where hr(t) is the instantaneous probability of suffering risk r conditional on surviving up to time t (r = {voluntary exit/bankruptcy; being acquired}).

4.2 Explanatory variables

This section presents and discusses expected effect of the explanatory variables to be included in the X vector, which are described in Table 8 in the Appendix. All covariates in the firm-level and brewery-level analyses are time varying. That is, within an individual (either brewery company or brewery), the variable may take different values at different times. Given the definition of exit/closure (Sect. 3), explanatory variables in survival analysis are “predictable” (van der Berg 2001); i.e. observable characteristics of a firm/brewery that contribute to explain the exit (cessation of independent operation) in t + 1 are observable just before the event takes place, avoiding potential endogeneity problems. Most covariates are dummy variables with the aim of capturing possible non-linear effects of covariates on the hazard rate.

The literature on entry and exit suggests that hazard rates of exit decline with age and start-up (and current) size (Jovanovic 1982; Special Issue of the International Journal of Industrial Organization, 1995, on the Post Entry Performance of Firms; Klepper 1996; K&S 2000, among others). However, both the exit from declining industries literature (G&N 1985, 1990) and Agarwal and Gort (1996) argue that small firms may have a strategic advantage over their larger rivals in mature phases of an industry life cycle. Nevertheless, in the UK beer market, firms’ age and their size at entry do not appear to have been important drivers of the concentrating process, given that all the firms and breweries in 1949 had entered before 1900. That is, by 1949, firms had survived over 50 years, whereas the focus of the previous literature lies on survival after entry. In addition, during the sample period, production and sales first slightly declined (1949–1958) and then rose (1959–1969), leading to an overall 25% increase.

The effect of firm size on exit hazards is assessed using three dummy variables that split the population of firms into three groups each year (Size1, Size2, Size3). They are calculated using information from the Cambridge/DTI Databank of Annual Company Accounts that provides financial information on: (a) all UK quoted companies from 1948 to 1960, and (b) from 1961 to 1969, on UK quoted companies with net assets and/or income over £0.5 million and £50,000, respectively. The variable Size1 takes the value of 1 if the firm is excluded from the Cambridge/DTI Databank of Annual Company Accounts. Among the included brewery companies, Size2 is equal to 1 if the brewer’s net assets are equal to or smaller than the median of net assets, and Size3 takes the value of 1 if net assets are above the median.

Size1 includes non-quoted, most likely family-owned, companies probably facing a size disadvantage. Besides, this group probably contains small firms supplying captive local markets facing little competition, whose shares are difficult to attain and performance difficult to monitor. This might reduce their risk of exit, despite their size disadvantage. They are expected to suffer a disproportionately higher hazard of liquidation than of being acquired. A lower risk of brewery closure is also predicted, as it involves either exiting the market or a 50% capacity reduction.

On the other hand, along efficiency considerations the largest-quoted companies (Size3) are expected to survive longer. However, companies in this group are predicted to face a higher risk of being acquired than non-quoted firms (although more funds are required), as it is easier to gain access to their shares to enlarge market share, as long as financial size proxies the latter. In addition, some of the largest companies may have pursued a strategy of expansion through mergers with and acquisitions of other brewery companies (to broaden their consumer base and/or gain control of well-known brands) and subsequent removal of redundant capacity. Hence, a positive correlation between Size3 and the hazard of brewery closure is expected, although this could be better collected by other variables (in particular, Multipla—dummy variable equal to 1 in year t if the firm has two or more active breweries). Small quoted companies (medium-sized quoted companies since 1961), Size2, are expected to bear a relatively higher risk of exit, especially of acquisition as large firms grew and consolidated through acquisition of other brewers, which is facilitated when they are quoted.

The variable Multipla is also likely to affect the probability of firm survival and brewery closure. At any time, multi-plant brewers are expected to face a lower risk of exit, especially of liquidation, since it implies the closure of all their plants at once. However, they are more likely to shut down breweries, as they enjoy greater flexibility to re-allocate resources, thus facing lower costs of plant closure. In addition, and reinforcing this argument, the decision to shut down one plant by single-plant firms involves leaving the market entirely. Hence, after controlling for firm size, I anticipate multi-plant firms to be more likely to close down plants. In the UK brewing industry, it seems plausible to expect single-brewery firms to be those supplying their local “captive” market, which allows them to enjoy rather pleasant survival conditions. Taking these factors into account, the overall impact of Multipla on exit may be unclear, probably reducing the instantaneous risk of liquidation. On the other hand, as long as the number of a firm’s tied outlets (and consequently, its market share) is positively related with the number of breweries, multi-plant firms could be an attractive target for other companies willing to expand. Alternatively, multi-brewery firms may be expanding themselves, which would reduce the risk of being acquired.

The effect of the number of breweries operated by a brewery company at any time (Nplant) may be non-linear. Firms with one (or, at most two) breweries in operation will be mostly those small-sized, family-owned companies supplying their local market. Firms with an intermediate number of breweries may be both expanding companies in early stages of expansion and/or medium-sized companies, which may be the target of other companies. Firms with a relatively large number of breweries will probably be pursuing a national strategy. In line with the discussion in the previous paragraph, Nplant would be negatively related to firm exit and liquidation, but positively related (at least for intermediate values) to the risk of being taken over. In addition, I would expect a positive relationship between Nplant and the risk of closing down breweries, as firms with more brewing sites incur lower agency costs of reducing capacity.

As discussed in Sect. 2, the institutional set-up of the industry probably led growth through acquisition of brewery companies and their valuable assets to be a common route for firm expansion. Shortly after acquisition, redundant brewing capacity could be closed down. Acquisition might also have been a strategy to reduce the risk of being acquired by other firms. The variable Earlyacq aims to capture a brewer’s efforts to grow (taking advantage of scale economies in production) and survive independently. Thus, it is expected to be negatively related to the risk of each firm exit route. Besides, as long as brewers acquire other companies in order to reach consumers, a positive impact on the risk of brewery closure is anticipated.

Low-profit firms are expected to face higher exit rates: they probably both suffer a higher risk of being acquired as they involve “cheaper market share” and are more likely to go into liquidation. To assess the effect of this variable, three dummy variables that split the population of firms into three groups every period are used, from the information in the Cambridge/DTI Databank of Annual Company Accounts. Excluded companies are the first group (Size1), and included companies are split into two groups according to whether their ratio profits to net assets is above the median (Profit23) or not (Profit22). The use of this set of dummies introduces a shortcoming in the analysis, because the excluded category from this financial data source (Size1) does not necessarily include the least profitable firms.14 Quoted low-profit brewers (Profit22) will probably suffer a disproportionately higher hazard of acquisition given that their shares are easy to acquire and they are relatively cheap. This could overcome their possibly lower risk of liquidation (if they are not the least profitable firms) and lead to a higher risk of (overall) exit. As for Profit23, high-profit firms are expected to survive longer, although their high value may attract other companies prone to form large groups through mergers. These companies are more likely to be those pursuing a national strategy, which could raise the probability of brewery closures.

The effect of advertising in the UK brewing industry has always been limited by the presence of captive markets and the pattern of beer consumption (higher proportion of draft and ales versus packaged and lager beers, respectively). Earlyadv is a variable switching from 0 to 1 when a firm first incurs advertising expenditures over 1950–1960 and then remains equal to 1 thereafter. Early advertisers may enjoy a first-mover advantage given that this could have allowed them to build up a reputation for some brands which would have placed them in an advantageous position to compete in the national market, as free trade (beyond tied outlets) raised its importance. It could have also made competitors more willing to host well-known brands through reciprocal agreements, enlarging the brewer’s market share. The latter would have enhanced a brewer’s survival prospects, facing a lower risk of being taken over and a higher risk of closing down breweries. However, brewers selling well-known brands may have also attracted acquirers, which could have been important in the industry after the disastrous experience by some well-established companies that launched advertising campaigns of their own products (e.g. Red Barrel by Watney Mann in the mid-1950s). Therefore, the overall effect on takeover risk could be ambiguous.

If expanding brewers have to acquire inefficient breweries to gain access to consumers, many brewery closures would have typically taken place shortly after acquisition. In the brewery analysis, Feacq is a time-varying dummy covariate switching from 0 to 1 the first time a brewery is acquired by a different company after 1949, and then it remains equal to 1 thereafter.

5 Results

5.1 Preliminary evidence

This section provides preliminary evidence on the evolution of the UK beer market. The empirical analysis focusses on the period 1949–1969, although references up to 1990 are introduced given that they highlight that the selection process was much fiercer up to 1969.

Table 2 presents the declining number of firms and breweries over the sample period. Interestingly, 82.4% (74.2%) of the 353 brewers in 1949 had exited by 1990 (1970). As for breweries, 337 (260) breweries were closed down over 1949–1990 (1949–1969). The risk of firm exit and brewery closure was heavily concentrated over 1949–1969, with no firm entry and two brewery openings: one brewery resumed brewing (though not independently), and a new brewery opened in 1969.
Table 2

Number of firms and breweries. Entry and exit in the UK brewing industry (1949–1990)

Year

Brewing firms

Breweries

Total number (1)

Exit number (2)

Exit rate (%) (3) = (2)/(1)

Form of exit

Entry number (7)

Total number (8)

Brewery closure (9)

Closure rate (%) (10) = (9)/(8)

Brewery opening (11)

Voluntary + bankruptcy (4)

Being acquired (5)

Exit rate (%) (6) = (5)/(1)

1949

353

      

433

   

1949-58

 

134

38.0

22

112

31.7

0

 

119

27.5

0

1959

219

      

314

   

1959–1969

 

128

58.5

16

112

51.1

0

 

141

44.9

2

1970

91

      

175

   

1970–1979

 

14

15.4

4

10

11.0

0

 

42

24.0

4

1980

77

      

137

   

1980–1989

 

15

19.4

0

15

19.4

2

 

35

25.6

3

1990

64

      

105

   

1949–1989

 

291

82.4

42

249

70.5

2

 

337

77.8

9

1949–1969

 

262

74.2

38

224

63.5

0

 

260

60.1

2

Source: Author’s estimations. See Sect. 3 for details

Entry through acquisition of on-going firms by ten non-brewery firms after 1970 are not regarded as exit of an incumbent and entry by a new firm, since there is neither real change in industry capacity nor change in number of producers (following K&S 2000)

Entry by microbreweries since 1970 is not included. Their impact on trade is marginal. The two firms that entered over 1980–1989 are firms re-entering the market, which would be better classified as microbreweries. One single-brewery firm regained independence for a few months in 1960 but ceased brewing that year

Brewery opening includes both opening of a new brewery and re-opening of an old site

Overwhelmingly, brewery firms ceased brewing independently (i.e. exited) after being acquired rather than exiting voluntarily or going bankrupt: 85.6% (85.5%) of exiting firms (70.5% of initial population) were acquired by other brewers over 1949–1990 (1949–1969). These figures are remarkably high and exceed those of takeover activity in other UK manufacturing sectors. For instance, Dickerson et al. (1998) reported that, for the period 1948–1970, 40% of the companies included in the Cambridge/DTI Databank of Annual Company Accounts were taken over. It is remarkable that 72.4% of the brewery companies included in the same data set were acquired during this period. Moreover, 63.5% of all brewers in 1949 had been acquired by 1970. These figures sharply contrast with those for the US brewing industry, where less than 1% of all firms exited through merger or acquisition over 1800–1988 (Horvath et al. 2001).

The discussion above highlights the high mortality rates during the period of analysis and points to the existence of distinctive features of this sector when compared with both other UK manufacturing sectors and the US brewing industry, making one exit route (exit after being acquired) far more common. This result broadly confirms the arguments of Sect. 2.1.

Focussing on breweries, Table 3 reveals that brewery closure shortly after acquisition was rather common in the industry, as predicted in Sect. 2.1. Up to 35% of acquired breweries (brewery acquisition is almost perfectly correlated with firm acquisition) were closed down the same year of acquisition. The median survival time of breweries after being acquired is 3 years. Although this pattern slows down, as 35% of acquired breweries remained in operations after 7 years and 10% survived over 33 years, two considerations are in order: first, the breweries owned by large companies that merged over the 1950s and 1960s to form the National Brewing Groups (Bass Charrington, Scottish & Newcastle, Allied Breweries, Watney Mann-Grand Metropolitan, Courage and Whitbread) are included; second, estimations in the bottom panel of Table 3 overestimate duration because they are restricted to the brewery’s first loss of independence, and at that time firm amalgamations took place in several stages, involving brewery acquisition of previously acquired breweries.
Table 3

Brewery survival in the UK, 1949–1990

Number of breweries: 442

169 (31.2%) never acquired

69 remained up to 1990

100 shut down before 1990

273 (61.8%) acquired

36 remained up to 1990

237 shut down before 1990

Brewery duration after being acquired by a different company (years)a

Number of breweries: 273

Percentilesb

Years

35

1

44

2

50

3

65

7

75

13

85

21

90

33

Source: Author’s estimations. See Sect. 3 for details

aBrewery duration after first loss of independence, independently of calendar year of acquisition

bPercentiles have been obtained from the product limit estimate of the survivor function, corrected for censoring by means of the actuarial adjustment method

Finally, Table 4 presents the distribution of breweries per firm over 1949–1990. Reading across the first row, over 79% of the firms in the market always brewed in just one site. Column 4 shows that, in 1989, only 13 out of the 64 firms (actively brewing) operated more than one brewery. More importantly, of the 51 single-brewery companies in 1989, 36 had just one plant brewing during the period 1949–1989, which gives an idea of their relatively local and stable scope over time. This table also allows rough tracing of the formation of the Six National Brewers over the 1950s and 1960s. By 1969, these companies were brewing at a large number of sites given the rather local scope of beer trade. Afterwards, as the national market for beer consolidated, proximity to consumers progressively depreciated as an invaluable asset in the industry, and production became more concentrated at fewer, large sites. This is fully confirmed by the geographic pattern of expansion of the major brewers through the location of their active and inactive (that is, acquired and then closed down) breweries since 1949.
Table 4

Distribution of brewery ownership per firm

Number of breweries

1949

1969

1989

Freq.

%

Freq.

%

Freq.

%

1

299

84.7

78

80.4

51

79.7

2

38

10.8

9

9.3

6

9.4

3

11

3.1

3

3.1

2

3.1

4

2

0.6

1

1.0

1

1.6

5

1

0.3

    

6

2

0.6

  

2

3.1

7

  

1

1.0

  

8

    

1

1.6

9

  

1

1.0

  

11

  

2

2.1

  

12

    

1

1.6

20

  

1

1.0

  

24

  

1

1.0

  

Total firms

353

100

97

100

64

100

(Total breweries)

(433)

 

(191)

 

(105)

 

Source: Author’s estimations

Microbreweries entering after 1970 are not included

So far, the preliminary evidence suggests that: (a) the UK brewing industry suffered a severe restructuring process, especially over 1949–1969; (b) the intensity of the process changed over time, being concentrated over 1959–1969; (c) this process involved no firm entry and very few brewery openings; and (d) the consolidation of the sector involved brewers acquiring other firms, gaining control of their tied outlets in order to reach new consumers. Soon after acquisition, redundant brewing capacity was reduced.

Therefore, the effect of acquisitions on market structure was probably twofold: on the one hand, they broadened the acquiring company’s consumer base (by enlarging their market share and/or accessing new markets through the control of tied houses); and, on the other hand, they facilitated the adjustment of capacity in an industry with problems of excess capacity.

This process led to the consolidation of a stable oligopoly, with a small group of large producers and a large fringe of small producers that managed to survive producing at a local scope.

5.2 Estimation results

Let us start with some basic data, ignoring covariates. Two longitudinal samples are used in the analysis: the population of both firms and breweries actively brewing over 1949–1969. Figure 1 illustrates non-parametric estimates of the hazard rate15 of each one of the risks examined. It shows: (a) the very low incidence of voluntary exit and bankruptcy; (b) that the risk of firm exit and brewery closure changed sharply over time, which suggests that survival methods are appropriate to carry out the empirical work; and (c) that the risk of firm exit was generally higher (except for 1968 and 1969) and more concentrated at the beginning of the period than that of brewery closure. The former reached its peak in 1959, whereas the latter in 1969. Moreover, 50% of brewers in 1949 had failed by 1960, whereas the same attrition rate is reached in 1966 for breweries.
https://static-content.springer.com/image/art%3A10.1007%2Fs11187-010-9295-2/MediaObjects/11187_2010_9295_Fig1_HTML.gif
Fig. 1

Non-parametric estimates of hazard functions

In order to evaluate the effect of firm and brewery characteristics on the exit hazard controlling simultaneously for the effect of the other covariates, I inspect the regression results from several reduced-form CPHM given by (1) and (2). All estimations are carried out using robust standard errors.16 At each exit time, correlation across covariates is not high.

The estimation results are reported in Tables 5, 6 and 7.17 They are contingent upon the firms (breweries) having survived up to 1949. Hazard ratios, their p values and chi-square for each regression are presented. A hazard ratio lower than 1 implies that the hazard rate decreases and the probability of survival rises with increases in the value of the explanatory variable; a value greater than 1 indicates that an increase in the variable raises the hazard rate. In the competing-risks model, the results for the risk of firm liquidation (voluntary exit, bankruptcy) are rather weak, as only 10.5% of the firms had suffered this event by the end of the sample period. However, the competing-risks model allows the effect of the explanatory variables on the risk of acquisition to be examined separately from that on the pooled risk of exit. In the latter case, the effect of some covariates cancels out. A test of whether exits to different states are behaviourally distinct rather than simply incidental (Narendranathan and Stewart 1991) cannot be carried out since covariates in each regression are different. However, visual inspection of non-parametric estimates of hazard functions (Fig. 1) suggests that the two exit routes are rather different.
Table 5

Risk of firm exit (1949–1969). CPHM estimates. Dependent variable: exit risk

Variable

(1)

(2)

1949–1969

1949–1958

1959–1969

1949–1969

1949–1958

1959–1969

Earlyadv

0.9038

(0.484)

0.5891**

(0.029)

1.1473

(0.541)

0.9072

(0.498)

0.6181**

(0.048)

1.1274

(0.583)

Size2

1.1030

(0.509)

0.8548

(0.460)

1.5383**

(0.047)

   

Size3

1.0123

(0.954)

0.9177

(0.763)

1.2304

(0.589)

   

Profit22

   

1.4232**

(0.020)

1.2557

(0.263)

1.8547***

(0.006)

Profit23

   

0.7333*

(0.090)

0.4841**

(0.014)

1.0950

(0.731)

Nplant

0.8907

(0.312)

0.7364

(0.599)

0.8358

(0.149)

0.8633

(0.240)

0.6446

(0.481)

0.8151

(0.115)

Multipla

0.9994

(0.998)

0.5423

(0.478)

1.9007*

(0.054)

0.9858

(0.959)

0.6352

(0.620)

1.8006*

(0.071)

Earlyacq

1.0742

(0.674)

0.7318

(0.411)

1.0558

(0.844)

1.1202

(0.489)

0.7994

(0.554)

1.0360

(0.889)

Log likelihood

−1390.282

−1371.513

 

−1384.189

−1364.697

χ2 (df)

3.85 (6)

21.24 (12)

 

16.95 (6)

44.83 (12)

p Value

(0.697)

(0.0018)

 

(0.0095)

(0.0000)

No. of observations

4555

 

No. of subjects

352

 

No. of events (exit)

261

 

The analysis is conditional on firms surviving up to 1949. The coefficients indicate the effect on the hazard for one standard increase in a continuous variable or a shift from 0 to 1 for a dummy variable. Estimation carried using Efron method for handling ties, robust standard errors

p Values correspond to two-sided significant tests, versus the null hypothesis that each true multiplier is equal to 1, in brackets

The discrepancy with Table 2 arises given that one single-brewery firm that exited the market in 1952 is excluded due to the lack of information on the location of its brewery

* Significant at 10% level; ** significant at 5% level; *** significant at 1% level

In addition, estimations allowing for a change in the effect of explanatory variables from 1959 were also carried out. This choice is based on a number of factors: (a) most non-parametric estimates of survivor and hazard functions for different values of a covariate (not reported) show a significant change by 1959. For instance, the risk of firm exit is clearly higher for non-quoted companies (i.e. Size1 = 1) before 1959, but smaller during the 1960s; (b) exit rates are significantly higher over 1959–1969, as reported by column (3) of Table 2; (c) the trend in beer sales shifted from slightly declining over 1949–1958 to continuously rising for 1959–1969. The estimation equation takes the following form:
$$ h\left( {t;X} \right) = h_{0} \left( t \right) \cdot \exp \left( {X\beta_{1} + X\left( {{\text{year}} > 1958} \right) \cdot \beta_{2} } \right), $$
(3)
so that estimated hazard rates are \( \text{e}^{{\mathop {\beta_{1} }\limits^{ \wedge } }} \) up to 1958 and \( \text{e}^{{\mathop {\beta_{1} }\limits^{ \wedge } + \mathop {\beta_{2} }\limits^{ \wedge } }} \) from 1959. This specification is the preferred one.

Furthermore, specification error tests and the proportional hazards assumption tests (both globally and for individual variables) proposed by Grambsch and Therneau (1994) have been carried out. In general, the joint significance of estimated coefficients cannot be rejected, and the proportional hazards assumption cannot be rejected, though results clearly improve when a break in 1959 is allowed. The regression results for the risk of firm exit, including a competing-risks analysis, and the hazard of brewery is discussed below.

5.2.1 Hazard of firm exit

Table 5 sets out the results from the CPHM for the (overall) risk of firm exit. Column (1) shows that, after controlling for other variables, the instantaneous probability of firm exit is not significantly higher for non-quoted and smaller firms (Size1, omitted category) than for the largest (and quoted) ones. During the 1960s, medium-sized quoted brewers face a significantly higher risk of exit than their smaller and larger competitors.

Column (2) suggests that low-profit (quoted) brewers face a higher exit hazard (especially over 1959–1969), whereas high-profit (quoted) companies endure a lower hazard (especially over 1949–1959).

The remaining covariates are included in both specifications (1) and (2). Thus, multiple breweries (Multipla) has a positive effect on the hazard of firm exit in the 1960s. Being an early advertiser (Earlyadv) is negatively, and significantly, related to exit only over 1949–1958. The remaining covariates do not significantly affect the survival prospects of firms.

The regression results from the two competing-risks of firm exit are presented in Table 6. Columns (1) and (3) and columns (2) and (4) are equivalent specifications for the risk of being acquired and liquidation. However, not a single early advertiser (Earlyadv), quoted and large (Size3), quoted and high-profit (Profit23), early acquirer (Earlyacq) brewer exited through liquidation. As for size, columns (1) and (3) suggest that small quoted brewers suffer a higher risk of being acquired (especially over 1959–1969), but a lower hazard of liquidation. The two risks cancel out when the overall risk of exit is evaluated [column (1), Table 5].
Table 6

Competing-risks estimation of firm exit, 1949–1969. CPHM estimates. Dependent variable: exit risk

Variables

Risk of acquisition

Risk of liquidationa

(1)

(2)

(3)

(4)

1949–1969

1949–1958

1959–1969

1949–1969

1949–1958

1959–1969

1949–1969

1949–1969

Earlyadv

0.9970

(0.984)

0.6353*

(0.068)

1.3192

(0.234)

0.9992

(0.996)

0.6664

(0.105)

1.2902

(0.260)

  

Size2

1.4173**

(0.028)

1.1108

(0.638)

1.9607***

(0.004)

   

0.0823**

(0.015)

 

Size3

1.2670

(0.295)

1.1615

(0.617)

1.5011

(0.310)

     

Profit22

   

1.8156***

(0.000)

1.6211**

(0.024)

2.3425***

(0.000)

 

0.1198**

(0.031)

Profit23

   

0.9386

(0.741)

0.6204

(0.116)

1.3936

(0.240)

  

Nplant

0.8893

(0.314)

0.7010

(0.550)

0.8397

(0.158)

0.8617

(0.243)

0.6097

(0.437)

0.8188

(0.124)

0.3120**

(0.043)

0.4174

(0.104)

Multipla

0.9368

(0.813)

0.5976

(0.561)

1.6873

(0.117)

0.9109

(0.749)

0.6987

(0.704)

1.5610

(0.178)

  

Earlyacq

1.1611

(0.394)

0.7666

(0.491)

1.1932

(0.540)

1.2082

(0.258)

0.8393

(0.649)

1.1589

(0.575)

  

Log likelihood

−1192.74

−1175.96

−1186.81

−1169.25

−186.94

−189.66

χ2 (df)

7.71 (6)

34.19 (12)

21.42 (6)

48.98 (12)

6.17 (2)

4.87 (2)

p Value

(0.260)

(0.0006)

(0.0015)

(0.000)

(0.046)

(0.088)

No. of observations

4555

4555

No. of subjects

352

352

No of events (exit)

224

37

The analysis is conditional on firms surviving up to 1949. The coefficients indicate the effect on the hazard for one standard increase in a continuous variable or a shift from 0 to 1 for a dummy variable. Estimation carried using Efron method for handling ties, robust standard errors

p Values correspond to two-sided significant tests, versus the null hypothesis that each true multiplier is equal to 1, in brackets

The discrepancy with Table 2 arises given that one single-brewery firm that exited the market in 1952 is excluded due to the lack of information on the location of its brewery

* Significant at 10% level; ** significant at 5% level; *** significant at 1% level

aCaution must be taken when interpreting these results, as only 10.51% of firms suffered the hazard. Not a single firm with Earlyadv = 1, Size3 = 1, Profit23 = 1, Earlyacq = 1 exited the industry through this route.Multipla and Nplant are perfectly collinear in this case

Low-profit (quoted) firms bear a much higher risk of being acquired, the effect being higher in the 1960s, and a lower risk of liquidation than non-quoted firms. The effect of being an early advertiser on the hazard of being acquired is rather weak: only over 1949–1958 is this risk reduced, at 6.8% [column (1)] and 10.5% [column (2)] significance level, respectively. Early acquirers did not change their risk of being acquired. Column (3) suggests that, as the number of breweries raises, the risk of liquidation declines. The remaining covariates do not significantly affect the hazards examined.

5.2.2 Hazard of brewery closure

Table 7 shows the estimation results from the CPHM on the risk of brewery closure by firms. After controlling for other covariates, the risk of brewery closure declines with the brewer’s size and profitability. Moreover, the effect seems to have intensified over the 1960s. The number of breweries and whether the brewer operates more than one brewery increase the hazard. Early acquiring companies (Earlyacq) are more likely to close down breweries, especially over 1949–1958, which seems reasonable given the construction of this variable. The risk of closure by a brewery sharply rises after the first time it is acquired (Feacq). The effect of Earlyadv is non-significant.
Table 7

Risk of brewery closure, 1949–1969. CPHM estimates. Dependent variable: closure risk

Variable

(1)

(2)

(3)

1949–1969

1949–1958

1959–1969

1949–1969

1949–1958

1959–1969

1949–1969

1949–1958

1959–1969

Earlyadv

0.8368

(0.401)

0.7844

(0.336)

0.8179

(0.529)

0.8918

(0.583)

0.8814

(0.601)

0.9305

(0.828)

0.7683

(0.203)

0.7450

(0.236)

0.7267

(0.275)

Size2

0.4888***

(0.006)

0.5276**

(0.045)

0.4797*

(0.088)

0.5108**

(0.012)

0.5445*

(0.070)

0.4471**

(0.035)

   

Size3

0.3453***

(0.000)

0.4137***

(0.002)

0.2846***

(0.008)

0.3602***

(0.000)

0.4520***

(0.003)

0.1910***

(0.002)

   

Profit22

      

0.4191***

(0.000)

0.4735***

(0.010)

0.3976**

(0.027)

Profit23

      

0.3818***

(0.000)

0.4436***

(0.003)

0.3558***

(0.008)

Nplant

1.0407***

(0.003)

1.0806

(0.429)

1.0488***

(0.001)

0.9959

(0.784)

1.0883

(0.372)

1.0074

(0.647)

1.0375***

(0.008)

1.0636

(0.532)

1.0452***

(0.003)

Multipla

8.5112***

(0.000)

7.2900***

(0.000)

11.0353***

(0.000)

5.7958***

(0.000)

5.3779***

(0.000)

5.3908***

(0.000)

8.3863***

(0.000)

7.3075***

(0.000)

10.9979***

(0.000)

Earlyacq

2.4833***

(0.000)

3.0247***

(0.000)

1.6405

(0.223)

0.6269

(0.108)

0.6063

(0.223)

0.6915

(0.456)

2.3026***

(0.000)

2.9335***

(0.000)

1.3359

(0.445)

Feacq

   

15.6854***

(0.000)

18.2799***

(0.000)

14.0076***

(0.000)

   

Log likelihood

−1351.624

–1348.715

−1254.463

−1249.012

−1352.326

−1349.425

χ2 (df)

188.63 (6)

197.97 (12)

355.19 (7)

403.56 (14)

184.06 (6)

193.52 (12)

p Value

(0.0000)

(0.0000)

(0.0000)

(0.0000)

(0.0000)

(0.0000)

No. of observations

6522

 

No. of subjects

433

 

No of events (exit)

259

 

The analysis is conditional on firms surviving up to 1949. The coefficients indicate the effect on the hazard for one standard increase in a continuous variable or a shift from 0 to 1 for a dummy variable. Estimation carried using Efron method for handling ties, robust standard errors

p Values correspond to two-sided significant tests, versus the null hypothesis that each true multiplier is equal to 1, in brackets

The discrepancy with Table 2 arises given that one single-brewery firm that exited the market in 1952 is excluded due to the lack of information on the location of its brewery

* Significant at 10% level; ** significant at 5% level; *** significant at 1% level

5.2.3 Discussion

Small firms did not face a significantly higher risk of (overall) exit. This result contrasts with the findings by the “entry and post-entry” and “shake-out” literature.18 It is tempting to argue that these results could be in accordance with the “stake-out” literature (G&N 1985, 1990). However, small firms (Size1) appear to have borne a disproportionately higher hazard of leaving the market through liquidation by closing down their plant(s), whereas small and medium-sized, quoted brewers (Size2) were more likely to be acquired by other brewers. Furthermore, at the time of exit through liquidation no firm was large and quoted and only one firm was small-quoted,19 which may indicate that the latter two groups were the target of important bids to form large groups. Moreover, once I control for other factors, medium-sized and large firms had a lower risk of closing down breweries, against G&N’s (1990) prediction. Therefore, rather than strategic factors in a declining market, the distinctive features of this industry and the nature of the variables included in the estimation specification make other explanations more appealing.

In an industry with restrictions on opening retail outlets and retailers tied to brewers, quoted companies were an attractive target for expanding brewers. The larger their size, the larger the market share attained through a merger or acquisition, although more funds would be required to acquire them.20 The results in the paper are consistent with Dunne and Hughes’ (1994) finding that the smallest and largest companies suffer a lower risk of being acquired than medium-sized companies. They are also in line with Dean’s (1997) findings for the English brewing industry for the period 1945–1960. It may be more difficult for family-owned firms (which are probably over-represented in Size1) to be acquired, given the scarce availability of shares and the greater difficulties in monitoring their performance (when compared with public quoted companies). Small-quoted companies are easier to acquire than large-quoted ones. The latter group, given the large amount of funds required to acquire them, are more likely to be involved in mergers. In fact, the analysis differentiating between periods (1949–1958 and 1959–1969) suggests the existence of a sharp difference between the two periods. During the 1950s, there is no significant difference in the hazard of being acquired between small-quoted and large-quoted firms. During the 1960s, however, the National Brewers consolidated and expanded by acquiring quoted, medium-sized firms. An alternative explanation for this unusual size–survival relationship is that small firms might have not been significantly more inefficient than their rivals and/or that small brewers, by focussing on their local captive markets, brewing particular varieties and using traditional brewing methods, managed to survive for a long time. Furthermore, the relatively low risk of exit of non-quoted companies is in line with the findings by empirical studies of resource partitioning that concentration rises coincide with increasing survival chances of specialist firms relative to generalist firms (Carroll and Swaminathan 1992). Thus, by finding a niche market, firms increase their opportunities to survive in a concentrating industry.

The National Brewers formed primarily through successive mergers between medium-sized and large companies over the 1950s. In the 1960s, they further expanded by acquiring other brewers, then removing redundant brewing capacity as a national market for beer started to arise. Improvements in transportation means and technological breakthroughs further enhanced this process. The existence of captive consumers made quoted medium-sized firms an ideal target, given that these brewers had an acceptable market share and were easily acquired.

As expected, profitability is negatively related to the exit hazard. However, low-profit (quoted) brewery companies faced a significantly higher hazard of being acquired and a lower hazard of liquidation. Moreover, not a single highly profitable (quoted) company exited through liquidation. Acquisition of low-profit (quoted) firms probably involved a cheap means of enlarging expanding brewers’ market share. This result may be also in accordance with the result in Dickerson et al. (1998) that takeovers discipline poorly performing companies. In addition, the effect is stronger during the 1960s, supporting the discussion in the previous paragraph. As for brewery closure, the most profitable brewery companies were less likely to close down breweries, probably given their better prospects. At first glance, this may be unexpected in the UK brewing industry if a firm’s profitability is positively correlated with pursuing a “national strategy”. However, in the multivariate analysis, Multipla may better capture this effect.21

The effect of advertising has always been limited given the distinctive features of the UK beer market. Early advertising brewers faced a lower exit hazard during the 1950s, turning non-significant over the 1960s. This result arises from their lower risk of being acquired over 1949–1969. This may reflect the existence of an early-mover advantage in building up a reputation by brewers pursuing a national strategy. However, brewers brewing well-known brands when the “national market” for beer was consolidating could have attracted other companies, probably pursuing a national strategy. These firms probably acknowledged the fact that acquiring these brands was a better strategy than launching campaigns on their relatively less popular products (i.e. brand reputation as a valuable asset). The two effects are consistent with Sutton’s (1991) argument that advertising increases a product’s perceived quality, thus enhancing the consumers’ willingness to pay for it, so that they may have offset in the regressions.

Early acquiring brewers did not go into liquidation. However, that did not prevent them from being acquired or merging with other brewers. This result is reasonable since, by construction, this variable switches from 0 to 1 the first time a company acquires another brewer during the 1950s. For this reason, it is positively related to brewery closure over the 1950s. Brewery firms acquired other brewers (and pubs), and soon after acquisition closed down their breweries. Related to that, multi-brewery firms (Multipla) were more likely to close down breweries, especially over the 1960s. This may reinforce the argument that firms acquired other companies as a means to reach new consumers, and also to gain control of successful brands, to later rationalise capacity. The previous result is further supported by the strong positive effect on the instantaneous hazard of brewery closure of the variable Feacq, which controls for those breweries acquired by another brewer after 1949. The latter is consistent with the finding that 50% of acquired breweries did not survive longer than 3 years (Table 3), whereas 50% of the population of breweries in 1949 stayed in operation at least 18 years.

Multi-brewery firms bore a higher risk of exit during the 1960s as a result of their higher (at 11.7% level of statistical significance) risk of being acquired over the 1960s. Furthermore, they faced a lower risk of liquidation, although they were more likely to close down breweries. In UK brewing the number of breweries was directly related with a firm’s number of tied outlets, and therefore with its market share. Multi-brewery brewers were probably expanding but were also attractive to buyers. Moreover, this result seems to be consistent with previous findings that multi-plant firms close down plants more easily than single-plant firms, but they do not exit (Mata and Portugal 1994).

The role of a firm’s number of breweries (Nplant) in shaping the evolution of the different risks deserves some attention. The risk of brewery closure rises with the number of breweries operated by the firm. However, the magnitude of this effect is small and restricted to 1959–1969. This is probably related to the fact that, during the 1960s as a national market for beer was developing, firms began to concentrate production at fewer sites. So far, I have argued that some firms acquired others then reduced capacity, but did not (voluntarily or going bankrupt) leave the industry. However, the number of breweries is not significant in explaining firms’ exit (or the risk of being acquired). It is tempting to argue that, as suggested by G&N (1990), multi-plant firms started to close down plants to become single-plant firms and later exit the market. Nevertheless, the direction of this dynamic effect cannot be accurately captured by the CPHM specification, since this method evaluates time-varying covariates just at their value at each specific exit time. However, the following evidence reinforces my argument in contrast to that in G&N (1990). Thirty-five out of 37 firms exiting voluntarily or going bankrupt over 1949–1969 did not change their number of breweries over time. One firm reduced its number of breweries, whereas another company increased them from one to two before exiting the market.22

6 Concluding remarks

The UK beer market over 1949–1969 represents an exceptional case study on the determinants of market structure. The market transformed from a highly fragmented one into a stable oligopoly in a process characterised by a sharp reduction in the number of firms and breweries, together with neither firm entry nor brewery openings, while beer consumption increased.

The market structure literature points to a combination of technical change and advertising outlays as two key drivers of the process of industry consolidation (Sutton 1991). This paper reveals the importance of institutional factors of an industry in explaining the dynamics of market structure. The practice of tying outlets to brewers, the legal restrictions on opening retail outlets and the high proportion of consumption of draft beer at the pub, together with a permissive policy towards mergers, made acquisition of medium-sized firms and brewery closure (shortly after acquisition) the main route towards market consolidation (rather than internal growth and liquidation of inefficient firms). Besides, these factors effectively deterred entry from 1900 up to 1970, and limited the extent of exit of inefficient brewers.

The empirical evidence broadly supports this argument, given the much higher incidence of brewers’ exit after being acquired, even when compared with this industry in other countries (e.g. the USA) and with other manufacturing industries in the UK. In a sector in which brewers enjoyed captive markets encouraged by the tied house system, once technological refinements (e.g. keg beer) made the product last longer and transportation costs declined (making national brands of draft beer feasible), expanding firms had to acquire their rivals throughout the country and their tied outlets in order to enlarge their market share. Shortly after acquisition, inefficient brewing capacity could be removed with no restriction and production could be concentrated at fewer sites to take advantage of scale economies in production.

Accordingly, medium-sized and quoted companies suffered a proportionately higher risk of being taken over than non-quoted companies, which endured a relatively higher hazard of liquidation. Therefore, factors facilitating acquisition and the availability of a certain market share enhanced the risk of being acquired, especially during the 1960s. Likewise, low-profit and quoted firms, probably cheaper to acquire, had a shorter expected duration due to their higher risk of being acquired. Unlike in the US beer market, advertising does not appear to have played a crucial role (at least, in the early days) in the process of consolidation.

Therefore, the tied house system may actually have constrained both the growth of concentration of brewing capacity, by making it primarily feasible through acquisition of tied houses through mergers or takeovers, and the extent of capacity rationalisation, by allowing many inefficient firms to survive. It may also have restrained the effectiveness of other forms of competition, such as advertising and research and development competition.

Thus, the UK beer market evolved towards a stable dual structure (Sutton, 1991). On the one hand, a reduced number of large companies (with new large, diversifying, entrants from adjacent sectors during the 1970s) account for almost 80% of UK beer production and compete at the national/international (or, at least, regional) level. They consolidated their position by shaking out intermediate-sized firms by taking them over and gaining access to their tied outlets, especially over 1959–1969. On the other hand, a large number of very small brewers primarily left the industry either voluntarily or by going bankrupt, while a number of them managed to survive by supplying their local “secured” niche markets provided by their tied outlets and brewing, using traditional methods, their own varieties of the product. Since 1970, they further benefited from campaigns stressing the taste for traditionally brewed beer, taste for variety. From that date, the segment has been characterised by a high rate of entry and exit.

The findings of this paper have some policy implications. Despite the importance of technology and advertising competition in explaining the evolution of market structure, institutional factors are fundamental drivers of the extent, nature and speed of consolidation of industries. Therefore, it is important to include the insights of institutional economics when investigating the evolution of market structure. In particular, the stance of regulators towards mergers and acquisitions affects how market selection proceeds, which has important welfare implications. For instance, it affects the number of competitors and their relative size as well as the extent of price competition, and thus concentration and market power. On the other hand, these factors also have an effect on the number of varieties available for consumption. A sound policy towards mergers should adopt a dynamic perspective by taking these factors into account.

Footnotes
1

Licensed houses (pubs) managed or franchised (tenanted) by brewers are often called tied houses.

 
2

However, concentration was lower (and the number of product varieties much larger) than in other countries. For instance, in 1988, the four-firm concentration ratio was 58% in the UK, 82% in the USA (in 1985), 90% in Denmark and about 95% in France and Canada.

 
3

Two considerations are in order in relation to Table 1. First, these figures would more accurately reveal output concentration if all breweries were of equal size and equally efficient. This is not a serious limitation up to the 1970s, when the largest firms built up sizeable brewing sites (the so-called macrobreweries) to concentrate production and take further advantage of scale economies. Second, Guinness is excluded from the group of largest firms as it only brewed at one site in the UK, despite its large market share.

 
4

See Cosh and Hughes (2008) for an extensive revision of takeover activity in the UK. The authors assess the relative characteristics of acquiring and acquired companies and their effects on performance using financial information.

 
5

However, the high proportion of taxes on final price limited the scope for cost reduction from scale economies in production. For instance, in 1964/65, excise duty accounted for 61.4% of a brewing firm’s overall costs, production costs 31.5% and distribution costs 7.1% (Gourvish and Wilson, 1994, p. 510).

 
6

Companies and breweries in the Channel Islands, Northern Ireland and the Isle of Man are not included. In addition, the opening of the Carlsberg brewery in Northampton in 1973 is not included since it was dominated by a foreign company.

 
7

The main limitation of the data set is the lack of information on both the number of tied outlets owned by each brewery company and either products or output/sales per firm and brewery.

 
8

The size qualification I apply is based on firms’ total net assets in the 4 years previous to the merger.

 
9

See Kiefer (1988) for an overview of these methods.

 
10

In contrast to traditional cross-section regression methods: ordinary least squares, logit and probit.

 
11

The exceptions to this pattern are Scottish & Newcastle and Courage, mainly based in the north and the south, respectively, which attained a national scope through their merger in the early 1990s.

 
12

Cox (1975) shows that this partial likelihood can be treated as an ordinary likelihood to derive valid (partial) maximum-likelihood estimates of β.

 
13

The method proposed by Efron (1977) for handling ties has been used. It assumes that all firm exits taking place at the same time have the same probability of exiting in a certain order, and corrects the risk set at any exit time with ties accordingly.

 
14

The regression analysis has also been run (not reported herein) using a dummy variable for the excluded firms and continuous size and profit variables for the included firms. The results are in line with those reported using sets of dummy variables.

 
15

This is estimated as the hazard contribution to the Nelson–Aalen cumulative hazard function between two exit times. This hazard is recorded at all periods at which exit occurs and is obtained as \( \widehat{h}\left( t \right) = {\frac{{d_{j} }}{{n_{j} }}} \), where dj is the number of events (exit) at time j (j = 1949,…,1969) and nj is the number of firms or breweries at risk at this time, just before the occurrence of the event.

 
16

Standard errors are computed by the robust method with clustering by firm (brewery), given that the data comprise multiple records for each individual, thus correcting the standard errors for possible within-firm (brewery) serial correlation and heteroscedasticity in random survival outcomes.

 
17

In order to control for the presence of unobserved individual heterogeneity (that is, unobservable—to the researcher—firm-level characteristics, such as unobserved firm organisational capabilities, access to specific assets, etc., which may affect the firms’ survival chances), I have estimated frailty survival models, which are generalisations of the proportional hazards models given by (1) and (2). The null hypothesis of no unobserved individual heterogeneity cannot be rejected at a 1% significance level. Hence, the non-frailty models [expressions (1) and (2) in Sect. 4.1] are the appropriate models to estimate.

 
18

Firm size (and age) are taken to represent the efficiency disparities arising from differences in experience, managerial abilities (Mata and Portugal 1994), scale of production and firm organisation (Dunne et al. 1988, 1989), due to cumulative scale economies in process innovation (Klepper 1996).

 
19
Distribution of firms by size group at time of exit, distinguishing by route of firm exit
 

Size1

Size2

Size3

 

Voluntary exit/bankruptcy

36

1

0

37

Acquired

106

66

52

224.

 
20

Dickerson et al. (1998) also find that large firms are usually less likely to be taken over, which may be related to the existence of capital market imperfections preventing buyers from raising funds to make an acceptable bid. They may be more prone to be involved in mergers.

 
21

Indeed, I ran two univariate CPHM regressions (not reported) with Profit22 and Profit23, respectively, and a significant positive effect on brewery closure was obtained. The effect turned negative when I added Multipla to these regressions.

 
22

In addition, less than 5% of the brewers that were acquired did not reduce their number of breweries before being acquired.

 

Acknowledgements

I would like to thank the associate editor E. Santarelli and two anonymous referees for their helpful comments and suggestions. Financial support from the Spanish Ministry of Science and Technology (project number ECO2008-04059/ECON) and from Generalitat Valenciana (project reference GVPROMETEO2009-098) is gratefully acknowledged. The usual disclaimer applies.

Copyright information

© Springer Science+Business Media, LLC. 2010