Death on the stock exchange: The fate of the 1948 population of large UK quoted companies, 1948–2018

Abstract This article provides a long-term demographic analysis of the principal members of the population of companies quoted on the UK stock exchanges in 1948. Motivated by theories of natural selection and of corporate governance, it traces the survival records of the 1948 population over the biblical ‘threescore years and ten’ to 2018. Of the population of 1513 companies in 1948, only 19 survived for the full seventy years. The survival rate, as captured by the half-life of companies in the population, is remarkably stable over the period. As well as charting death rates over time, the study explores the causes of death – takeover, failure, etc. Also it analyses the relationship between survival and the characteristics of the firm and its environment.

and the heightened threat to weak or self-serving managers from disciplinary takeover when shares are publicly traded and widely dispersed (Manne, 1965;Shleifer & Vishny, 1997). On the other hand, attrition by take-over is necessary for this process of economic natural selection to operate (Alchian, 1950;Meeks, 1992;Nelson & Winter, 1982;Winter, 1964); although Matthews' (1984) discussion of Darwinism draws attention to the loss of organisational capital -the experience, information, contacts, etc. that enable people to cooperate effectively as a group (see also Arrow, 1984) -that can accompany a company's disappearance. If long-run survival rates for the company population are low, this calls into question the dependability of long-term contracts that companies might sign -for example, to pay a defined benefit pension (Meeks, 2017), or clean up a nuclear power station site (Taylor, 2007) many years later. In relation to industrial organisation, Tepper and Hearn (2019) cite evidence for the US that 'between 1996 and 2016, the number of stocks…fell by roughly 50% from more than 7300 to fewer than 3600' . 3 And, as predicted by the Structure-Conduct-Performance paradigm in the industrial organisation literature, the resulting increase in concentration is associated with 'higher profits, lower wages, and less competition' . 4 Wu (2018) explores wider implications for political economy: he links the attrition of the company population to 'enfeebled' anti-trust policy, leading to concentration of economic and political power and 'the reemergence of an outrageous divide between the rich and the poor' (p. 20).

The 1948 population
To qualify for inclusion in the NIESR's 1948 population, companies must have been: i. admitted to the Official List of a UK stock exchange 5 ii. independent companies (i.e. not majority owned by another company) iii. operating primarily in the United Kingdom iv. engaged principally in manufacturing, distribution, construction or transport, and certain other services (companies whose main activities were in agriculture, mining, shipping, insurance, property, banking, and finance were excluded). 6 The main part of our analysis focuses on the 1513 larger members of this population. 7 To permit consistent inter-year comparisons of death rates, a subset of the NIESR 1948 population was excluded from our analysis. These were the very smallest companies -including moribund and shell companies -for whom the government analysts ceased collecting data during the period. Although numerous, their collective contribution to the population's economic activity was very small: their net assets summed to less than those of just three of the companies at the opposite end of our population's size distribution. 8 Of course, if we have in mind the lifetime analogy of people or -as in Marshall (1920) and Hannah (1999) of trees, or in Alchian (1950), or Winter (1964) of species generally -our 1948 starting date does not correspond with birth or fertilisation. Many of the companies in our population were 'born' (started trading) before their listing on the Stock Exchange. Some were born long before; and for some, but by no means all, records have survived of their earlier development from, say, sole trader to partnership to company, to listed company. For example, Johnson Matthey, one of our seventy-year survivors, dates its founding as 1817 (Johnson Matthey 2017). And the first listing on the Stock Exchange long predates 1948 for many of our population. Equally, 'death' by merger or acquisition can lead to a long after-life (for some of the assets, under different control). Moreover, a complete analysis of the evolution of the population would involve the study of births, new entries to the population, as well as deaths. The study of the survival of legal listed entities does nevertheless have important implications for those who contract with them and other stakeholders.

Data definitions and sources
For years up to 1990 we recovered our data on the death of companies within this population 9 from the working papers of the NIESR and then of government departments, successively the Board of Trade, the Department of Trade and Industry (DTI), and the Business Statistics Office. These organisations were collecting company accounts in order to construct aggregate financial accounting data for the quoted company sector and its constituent industries. In order to ensure that the aggregate analyses of UK quoted companies were complete, and distortions in year-to-year comparisons were not created by omitting live companies, those maintaining the population investigated every case where a quoted company failed to file an annual report, and so traced the timing and cause of any disappearance from the 1948 population. The date and cause of death was established from newspaper reports and from enquiries of individual companies. We were allowed to consult these handwritten records of death before they were destroyed, and we digitised the core information. In some cases, where they were incomplete, we have augmented the data; and we have continued to hand collect them for years after 1990 (when the government discontinued their analysis). For these further investigations of the fate of individual companies, and for the data that we relate to the demographic information (e.g. size, growth and industry) our sources have included the Stock Exchange Official yearbook, the London Business School Share Price Database, the financial press, companies' annual reports, and (in the most recent years) internet searches.
We are then defining the death of a company as its disappearance from the population of larger independent UK companies listed on the London Stock Exchange. This definition does have some advantages: it is relatively well-defined and recorded; it corresponds to important economic events -a change of management, governance and regulatory framework, loss of financing facilities for the business and of access to a public market for existing equity holders, and consequential changes in the position of employees, pensioners, suppliers and other stakeholders. It will therefore capture many of the cases which would qualify as death on any definition. For example, where a company is liquidated, its exit is more closely comparable with the death of a person or a tree, the standard comparators in the economic natural selection literature.
But the definition is open to misunderstanding: a company which has been taken over (the most frequent cause of death in our study) does see a change of control, etc.; but its assets, tangible and intangible, might continue in use; and the 'death' might have only a limited impact on customers or employees. For example, the Dutch firm Akzo Nobel which acquired the remnants of ICI (the second largest member of our population in 1948) still sells paint under ICI's Dulux brand and produces it in a UK factory. Hence there is scope for further research into the subsequent 'after-lives' of taken-over companies to discover whether, and to what extent, they survive and flourish under their new ownership; although often the activities and assets of the target will have been assimilated into the acquirer and may be unidentifiable, and intensive analysis of individual cases would be required for such research.
Another possible confusion arises with companies which have 'merged' -implying that both parties have been assimilated into a new entity, in contrast with an 'acquisition' , where one party (dying) is assimilated into another (surviving). In practice, transactions recorded as mergers are sometimes in substance acquisitions, but adopting the legal form of merger yields benefits. For example, merger/pooling accounting, rather than acquisition/purchase accounting procedures, allowed the entity to avoid recognising purchased goodwill in the balance sheet and its amortisation in the profit and loss account, resulting in higher reported earnings after the amalgamation. The difficulties of distinguishing between mergers and acquisitions, combined with the widely divergent accounting methods applied to them, has led the International Accounting Standards Board to ban merger ('pooling of interests') accounting. 10 The convention adopted by the analysts whose working papers we used up to 1990 was to label both parties in a 'merger' as deaths, and the new entity as a birth, whereas an acquisition is death of the acquired company only. We have tried to maintain this convention in the subsequent years but are aware that, in practice, very close study of the terms of the deal and of post-merger control would sometimes show that the economic substance of the transaction does not correspond with its legal form. Typically, this would involve an acquisition being wrongly characterised as a merger, because of the accounting advantages of the latter treatment. In the case of the (very small) number of these (sometimes giant) transactions wrongly classified as mergers, we regard both of the merging companies as dying. If some of these are in substance acquisitions, with one firm actually surviving, our tally of deaths is biased somewhat upwards.
A further ambiguity with our definition of death arises when the share is delisted. Often this is associated with an upheaval in the business, such as financial distress subsequently ending in liquidation. But occasionally it is just that a flourishing listed business finds that the costs of maintaining a listing exceed the benefits. For example, Everards Brewery repurchased its remaining preference shares in 1997, leaving the stock market. But the business continued its long life as a private company, its board chaired by the fifth generation of the Everard family -longevity indeed. In some other cases, especially for earlier decades, we know of no extant evidence on the reasons for delisting.
Assigning a date to a company's death is also less straightforward than in the human counterpart. For example, a company in financial distress might first have its share suspended; then, after a lag, have an administrator (or receiver) appointed; and then, after a further lag, go into liquidation. We have generally defined the death year as that following the last accounting period in which the company submitted independent financial accounts. As well as those intrinsic difficulties of dating some deaths, this dating convention introduces some imprecision in aggregate time series: the accounting period does not reliably correspond to the calendar year. Only a minority of our companies have accounting year-ends in December (although it is the most popular choice), so the dating procedure leads to some 'spillage' of deaths to adjacent calendar years; and year to year movements reported below should not be interpreted too precisely.
The data used below on initial industry membership are those prepared by NIESR for 1948 (see Meeks et al., 1998 for full details). Where companies belong to more than one SIC 2-digit industry, their principal activity is chosen, even though, in a diversified company, that may account only for a minority of its business. This problem has become more serious over the seventy years of our analysis because of trends to diversify (Goudie & Meeks, 1982); and in some cases, companies' principal industry will be quite different in 2018 from 1948. For example, Whitbread, one of the few survivors in our population, was a brewer in 1948, but became a hotelier and café operator and then decided in 2018 (in the face of shareholder pressure) to divest its coffee shop division. This illustrates the degree of change that may be necessary for 'survival' on our definition: in such cases, the activities of the business may have changed as much as if they had been acquired and restructured by another company, which would count as 'death' on our definition. We discuss such migration further below.
The data on initial size are obtained from companies' balance sheets in the databank we have constructed . The corresponding data for 2018 are extracted from companies' annual report and accounts. Net assets are used as the size measure: alternatives such as sales or employment were not required to be reported in 1948.

Seventy year survival rates, the population half-life, and fluctuations in the death rate
We see from Figure 1 the rapid and drastic attrition of the population. On our definitions just 19 of the initial population of 1513 companies survived to 2018 as independent listed companies, 1.26% of the initial members.
Few previous studies have analysed survival rates over such long periods. One exception is Hannah (1998), which focusses on the top 100 industrial firms in the world in 1912. 14.5 11 UK companies belonged to this population. Their chances of survival and, even more stringent, of still enjoying membership of the global top 100 83 years later, in 1995, were as high as 47% -many times higher than even our survival rate for the shorter period of 70 years. 12  Sources: Data compiled by the authors and stored in the Cambridge/Dti Databank of company accounts , the stock exchange official yearbook, the London Business school share Price Database, the financial press, companies' annual reports, and internet searches.
However, this was a select group of companies -those UK firms which were giant by world standards. And Hannah conjectured that for smaller firms by contrast 'the probability of survival over the century approximates to zero' . Our attrition rate is consistent with that conjecture. Previous work has found an inverse relationship between size and death rates (e.g. Singh, 1975;Whittington, 1972); and we explore the relationship between size and survival further below. The same patterns were previously less pronounced in the UK. Going back much further in time, and to a broader size distribution, including unlisted companies, Shannon (1954) traced the demographics of limited companies formed soon after the English Joint Stock Companies Act of 1856. He traced their fate through to 1929, a similar length of time to ours. Despite being typically much smaller than those in our population, these companies suffered less severe attrition (defined as failure to survive as registered companies, rather than, as in our case, failure to retain a stock exchange listing) than ours: some 10% were still alive in 1929. A similar attrition rate to Shannon's was found by Payne (1980) for the 2936 companies formed in Scotland between 1856 and mid-1895: 331 were still in existence in 1960. In an international context, Napolitano et al. (2015) conclude that 'the typical features for extraordinarily long-living companies throughout history are relatively small or medium size and family ownership' . The studies cited above are not confined to listed companies, and the comparison between these studies and our results based entirely on listed companies suggests that a stock exchange listing is not a passport to survival. Obtaining a market listing may increase access to capital, which may assist survival of the business activity, but it may increase vulnerability to take-over and therefore reduce the chances of survival as an independent entity.
The half-life of our 1948 population (the number of years it took for half the population to disappear) was 17 years: by the end of 1965, half the original 1513 companies had already ceased to exist as independent listed companies. The largest declines in absolute numbers came in the takeover waves of the late 1950s and late 1960s (Kynaston, 1994;Meeks, 1977), a process we pursue below. However, a 17 year half life does not reflect the typical experience over the period. Table 1 reports the percentage of those surviving at the beginning of each decade which disappeared in the subsequent 10 years. It shows that the first decade, following the Second World War, had unusually low attrition rates, probably because the possibilities for financial restructuring and take-overs were restricted by post-war controls. The half-lives of the survivors in subsequent decades (1959-68 onwards) were typically around 10 years (a 10-year survival rate of 50 percent). From 1959 onwards, the average death rate in the population did not vary so greatly over the decades. Thus, apart from the first decade, soon after World War 2, when restructuring of the private sector began only slowly, around half of those 'alive' at the start of each decade survived till the next decade. This relatively steady rate of attrition is illustrated in Figure 1: on a log scale, after the first decade, the slope -reflecting the rate of changedoes not vary greatly. Table 2 provides an analysis of the cause of death for members of the population. Overwhelmingly the most important cause of death is takeover/merger (1255 companies, some 83% of the initial population): a familiar result for the UK (Meeks, 1996), 13 as it is for the US (Carlton & Perloff, 2000), though not one that occurred in this period for corresponding populations in Continental Europe or Japan 14 (Eckbo, 1992). Eighty-nine of the remaining deaths were due to outright failure (liquidation/receivership/administration). The resulting annual failure rate (0.08% p.a.) is close to that recorded by Mueller (1986) for the top 1000 corporations in the US: 0.09% p.a. over the period 1950-72; and it is not greatly different from the rate of around 0.16% p.a. recorded by Harhoff et al. (1998) for German joint stock companies (AG) in 1989-94. 15 This definition of failure is narrower than ones often used in the bankruptcy literature (Goudie & Meeks, 1991); however, other categories in Table 2 (especially 'ceased quote') are likely to consist mainly of failing companies. And, more important, legal form may differ from economic substance: to avoid bankruptcy costs, takeover is sometimes used as an alternative to receivership/administration for a company that is, or is about to become, insolvent (Peel, 1990). Hence the boundary between our takeover and failure categories is likely to be somewhat indistinct. includes companies withdrawn from the Dti population because accounts were inadequate or unavailable even though no record of death could be found, companies which died but were subsequently re-born, often after reorganisation, and ones for whom a cause of disappearance from the population could not be established from the available sources. Sources: Data compiled by the authors and stored in the Cambridge/Dti Databank of company accounts , the stock exchange official yearbook, the London Business school share Price Database, the financial press, companies' annual reports, and internet searches.  , the stock exchange official yearbook, the London Business school share Price Database, the financial press, companies' annual reports, and internet searches.

Characteristics contributing to longevity
i) Initial size, growth, and long-term survival rates Previous studies of company populations have reliably found that the probability of survival is higher for bigger firms (e.g. Dunne & Hughes, 1994;Singh, 1975;Troske, 1996;Whittington, 1972). There is no shortage of explanations for this relationship. Greater size often brings greater market power (Cowling et al., 1980) and/or greater diversification (Goudie & Meeks, 1982), both of which may afford higher protection from exogenous life-threatening shocks. It typically also brings access to cheaper finance (Dimson, 1988;Prais, 1976), as well as to scale economies in production, marketing, R&D, etc. (Pratten, 1971;Scherer et al., 1975). And, since it limits the range of potential bidders, it has been shown to provide greater insulation from takeover (Chatterjee, 1994;Singh, 1975).
The data in Table 3 for our population provide some confirmation of this relationship. In constructing the Table, the 1948 population has first been ranked by size (net assets, at book value). Then the ranking has been split into deciles. Then each of the 2018 survivors has been listed in its respective size decile in 1948. Only 2 of the 19 survivors began in the six smallest deciles; and 15 of the survivors began in the two largest deciles. The smallest survivor was Tesco, a minnow in 1948. The probability of survival within any decile is extremely low (no size provides absolute immunity from takeover or failure); but the very largest listed companies did enjoy somewhat better survival prospects. Napolitano et al. (2015) report that, at the opposite end of the size distribution, when unlisted companies are included (not the case in our analysis), very long-living companies tended to be small and family-owned. Table 3 also shows a stronger measure of survival: membership of the FTSE 100 index in 2018. This indicates that the company had a very high stock market capitalisation at the end of the period. Of the 19 survivors, eight were members of the FTSE 100 index and, of these, six started in the top two size deciles in 1948. Thus, successful survival was associated with large initial size, although Tesco was a striking exception. Table 4 shows that by 2018 Tesco had grown from a minnow to one of the largest fish in the pool. In constructing Table 4, the net assets of each survivor in 2018 have been divided by the respective net assets of the business in 1948. The companies are then ranked by this quotient, to reflect their growth over the seven decades. The range of growth experience is startling. In 2018 Tesco heads the ranking: 64 thousand times its 1948 size. At the opposite end of the Table is Cohen/600 Group, which ended the period 22 times bigger than it started: its book value of net assets did not keep pace with the Retail Price Index, whose 2018 level was 36 times that of 1948. The median performer on this growth ranking was Unilever, with a quotient of 268.
By 2018, UK GDP at current prices was 175 times its level in 1948 (https://www.ons.gov. uk/economy/grossdomesticproductgdp/timeseries/ybha/ukea). This increase comprised the rise in prices and the real growth in output. The growth represented by the quotient for companies comprises the rise in prices, and the firm's 'organic' growth, and the 'cannibalistic' expansion by acquiring other companies. Eight of our nineteen surviving companies did not keep pace with the growth of the economy. Hence, growth does not appear to be a necessary condition for survival; but, unsurprisingly, all eight of the survivors which were in the FTSE100 index in 2018 were among the 10 most rapidly growing of the 19 survivors.  , the stock exchange official yearbook, the London Business school share Price Database, the financial press, companies' annual reports, and internet searches. yardsticks: Consumer price index and GDP: https://www.ons.gov.uk/economy.  , the stock exchange official yearbook, the London Business school share Price Database, the financial press, companies' annual reports, and internet searches.
ii) Industry and long-term survival rates Although Riviezzo et al. (2015) found that the business history literature on longevity has devoted limited attention to environmental characteristics, relative to organisational characteristics, links between the industry in which a company operates and its survival prospects are not difficult to rationalise. Industry will affect the scale of the company's potential market, the opportunities offered by technology to secure economies of scale and scope, the market's growth or decline, and the competitive conditions, including entry barriers and exposure to foreign competition and to exchange rate fluctuations. Chandler (1990) explores these links with a historical and international perspective. Barker and Peterson (1987) demonstrate the diversity of industrial experience and behaviour in the UK. Goudie and Meeks (1986) translate this into the long-term experience of some companies within the population: for example, the capital employed of continuing companies in the chemicals industry rose 20 fold over the period 1948-77 (at current prices), 12 fold in textiles, but only 5 fold in shipbuilding. McGahan (2008) confirms the association between industry and measures of company performance for U.S. corporations. Table 5 shows the survival rate by S.I.C. 2-digit industry. None of the 1948 population survived the seventy years as independent listed companies in 11 of the 22 industries. In some of these cases, the result is not surprising -the attrition of British companies is only too well-known in industries such as shipbuilding and vehicles. Table 6 explores the migration of companies across these 2-digit industrial boundaries. It shows the industrial classification of each survivor in 1948, and the principal industry in which it operated in 2018. At first sight, the industry membership may seem surprisingly stable: only 5 of the 19 survivors migrated. However, the two-digit classification is a broad one, setting a high barrier for the recognition of industrial migration, and even then, in some cases, migration is perhaps a misleading term. For example, Unilever was in 1948 already represented in both food manufacturing (industry 21) and personal/home care products (industry 26, to which it was allocated). During the period, the balance of its activities shifted somewhat so that its principal industry changed. On the other hand, Low and Bonar appears to have transformed itself but remained allocated to its initial industry. 16 In 1948 it was allocated to industry 41, textiles -an industry which has lost so many members as many forms of textile production have moved offshore. By 2018, the company was still producing fabrics, and so allocated to industry 41; but in specialist areas, for the auto and construction industries and coastal defence -markets and applications which did not exist in 1948. Again, Marks and Spencer remained in the retail distribution industry (82) throughout, but at the beginning it sold everyday clothing; whilst at the end, 60 per cent of its turnover was derived from selling food -mostly in prepared form, a market which scarcely existed in 1948. Marks and Spencer's ability to adapt, develop its brand, and build its sales and profits was reflected in Little's (1962, p. 412) comment that 'statistically speaking, Marks and Spencer does not exist' , meaning that its performance was outside the normal range of experience. But even it came close to death in 2004, at the hands of a hostile bidder, Philip Green, and it has recently (in 2018) announced a store closure programme as part of a restructuring plan to adapt to recent structural changes in retail markets. Some of the migrations were associated with a major change in business model, and might (as with Low and Bonar, and Marks and Spencer) be characterised as a shift to markets characterised by higher income elasticity of demand than for their original activities. For example, Whitbread moved out of brewing (industry 23) into hospitality (hotels, restaurants and coffee shops, part of service industry 88); Marstons continued to brew, but by 2018 derived most of their income from the same service activities as Whitbread. Smith and Nephew, in the unpromising textile industry (41) in 1948, derived its income in 2018 from the health sector -replacement joints, implants, etc (industry 49, other manufacturing).
At one time, the DTI, who were then assigning industry classifications, created separate categories for conglomerate firms which could not be assigned to any single principal industry. But none of the famous conglomerates spanning a wide range of disparate industries -such as Hanson and BTR-continued in our population for these seven decades. Perhaps their modern counterparts are private equity businesses, which were particularly active in the final decades of our period, and often aimed to hold a company for only a few years while taking steps to increase its market value. Lord Hanson had famously said that all his businesses were for sale all of the time, should an attractive offer be forthcoming. Similarly, one of the major companies in our population which almost survived our period, GKN, fell prey in 2018 to a young turnaround company, Melrose, whose published plan was to dismember GKN and sell parts off within a few years.

Conclusions and implications
The most striking 'stylised fact' to emerge from this study is that typically about one half of our population of large listed UK companies disappeared from the population within the next decade. This high rate of attrition meant that only 1.3% of the 1948 population (19 out of 1513) survived until 2018 (Table 2). Moreover, the rate of attrition, as captured by the 10-year half-life of the surviving population, was remarkably stable, the only exception being the slower rate of attrition in the first decade, 17 which was the period of post-war recovery. 18 If, following Coase, a firm is seen as a 'nexus of contracts' , our survival rate prompts questions about the credibility of some of the long-term contracts to which these companies commit. One such common contract in this period has been the promise by companies such as British Home Stores (deceased), or British Steel (deceased) to a 20-year old recruit of a defined benefit pension seventy or more years later (Meeks, 2017); another has been the promise by British Energy (deceased) to decommission nuclear power plants many decades after their construction (Taylor, 2007); and, more trivial, the promise by the owners of Doc Martens to guarantee their boots for the wearer's lifetime. 19 A second, almost equally striking result is that the vast majority of these disappearances (83%) were due to take-overs rather than insolvency (Table 2), thus raising the expectation that the business will continue under new or reformed management. This draws attention to the critical role played, in this period, by the take-over process, which is often described as 'the market for corporate control' . There can be little doubt that take-overs led to a drastic reallocation of assets to new management teams or changed the policies of existing management teams. Moreover, take-over bids must have had a powerful effect on management even when they did not succeed (as in the celebrated ICI bid for Courtaulds in 1962) and even when the threat did not materialise. It is notable that the companies which survived over the period showed the ability to adapt their business activities, in some cases radically. Thus, the take-over mechanism may be credited with introducing a high degree of mobility and dynamism into the quoted company sector. On the other hand, it is not clear that this was necessarily a benign process. Questions have been raised (supported by empirical evidence 20 ) about whether take-overs were successful, for example in terms of improved profitability, and whose interests they served (management, shareholders of the acquirer or the acquired company, customers, employees, or creditors such as pension funds). Potential conflicts of interest in relation to takeover are illustrated by the finding of Harford and Li (2007) that 'even in mergers where bidding shareholders are worse off [as a result of an acquisition], bidding CEOs are better off three-quarters of the time' . A further concern has been short-termism: the maximisation of short-term financial performance (often assisted by creative accounting methods) to the detriment of longterm investment. This was most clearly demonstrated by asset-stripping take-overs, such as were carried out by some conglomerate companies during the period of our study. However, our concern here is to record the extent of attrition in our population rather than to evaluate whether it had benign effects on the economic performance of the corporate sector.
The high attrition rate in our population, with only 19 survivors in 2018 from an initial population of 1513 in 1948, suggests that the survivors are worth further study in order to establish whether there are distinctive characteristics that aid survival. 21 An obvious initial characteristic to investigate is opening size, whose influence is well-established in take-over studies: larger companies are less likely to be taken over because there are fewer potential bidders with the necessary financial and managerial resources. We do indeed find that the majority of our survivors were in the top two deciles of the initial (1948) size distribution (Table 3). However, some of the survivors were small in 1948, including Tesco, which grew to be a giant during the period. Moreover, most of the 1948 companies in all size groups 'died' (on our definition of losing their independent stock market listing) during the period. Thus, large initial size is neither necessary nor sufficient for survival, although it may help a little.
Another obvious candidate, which could complement initial size, is subsequent growth. This might preserve or enhance initial size and therefore reinforce the relative immunity to take-over that size might bring. However, although there is some indication that survivors tended to grow rapidly (Table 4), the evidence is not strong. Eight of the nineteen survivors grew (in terms of net assets) at a lower rate than nominal (current price) GNP, although the very smallest (in 1948) survivor did grow at the fastest rate of all.
A final characteristic of the survivors that was investigated was industrial classification. Clearly, operating in a declining industry seems likely to increase the probability of death, and this is borne out by Table 5, which classifies our 19 survivors by their 2-digit standard industrial classification. This shows that 12 out of 21 industries had no survivors, and these industries were mainly ones such as shipbuilding and vehicle manufacture which suffered significant decline in the period. Hence, initial industry seems to be a relevant factor in survival, although with only 19 survivors spread over 21 industries, no industry had a high survival rate, the highest being 3.1 per cent. Thus, as with the other factors considered, industrial location does not provide an elixir of life.
A more significant indicator of survival capacity may be that given by Table 6, which compares the opening industrial classification of each survivor with the closing classification. Over the 70-year period, 5 of our 19 survivors changed their 2 digit standard industrial classification. Admittedly, 70 years is a long time, but the industrial classification is a broad one, so that changing it implies drastic change. Moreover, we have documented significant changes of activity amongst those that did not change classification, such as Low and Bonar's move into emerging specialist markets within the ailing textile industry.
Thus, the lesson provided by our survivors is that adaptation to a changing environment is essential for survival, although by no means does it guarantee survival. The vast majority of firms perished by take-over in this period. Many of these were forcibly changed by the take-over, and those that survived may have avoided take-over by anticipating the actions that an acquirer would take. An example of this is Whitbread's recent decision (2018) to divest its Costa Coffee arm, in order to fend off a bidder who proposed to do precisely this. Another current example is that of GKN, which offered, in its response to the Melrose bid, to divest some of its interests. In this case, the proposal came too late to satisfy shareholders, who accepted the Melrose bid.
In view of the dominance of the take-over process in re-shaping the quoted company population, an implication of our study is that it is important to understand the nature of this process. We have already alluded to the apparent poor long-term shareholder returns following take-over and the possibly malign consequences for other parties with a long-term interest in the taken-over business, such as (topically) members of defined benefit pension schemes. There is, however, a need for deeper research into what happens after take-over: the extent to which the acquired business survives and prospers under its new ownership and contractual arrangements made under the previous ownership continue to be honoured. We have also noted that the take-over mechanism is less important in some economies that are not obviously less successful than those where it is dominant. Furthermore, we have noted that there has been a recent trend in the UK towards delisting of businesses (or not listing in the first place) in favour of ownership by private equity. Notably, this has been the case for several of the UK's privatised utilities. Thus, in the future, the listed company sector may decline in importance as other corporate governance models emerge which are less reliant on the market for corporate control and better adapted to the long-run interests of stakeholders.
Finally, it is hoped that our limited empirical study stimulates further research on the questions which it raises. We have already alluded to the problem of identifying the survival of the business, as opposed to the legal listed entity. Apart from the identification of 'life after death' , there is a question of 'life before birth': how long did companies or businesses exist before obtaining a stock exchange listing? Another important question is the pattern of births in our population. We have focussed entirely on deaths, but a balanced understanding of the evolution of the population of listed companies (as opposed to those existing in 1948) requires an understanding of the birth process, which has the potential to replace the deaths and thus sustain the size of the population whilst allowing its composition to change, in the manner of biological populations.  Grullon et al. (2017). 5. In 1948, in addition to the London Stock Exchange, there were several provincial exchanges.

Notes
These gradually became more integrated with the London exchange and were finally incorporated in it in 1973. Subsequently, the London exchange promoted the Alternative Investment Market (AIM, see footnote 8), for the shares of smaller companies. One of our 'survivors' in 2018 was listed on AIM and another on NEX, an independent exchange for emerging companies. Note that these data do not include all securities listed on the UK stock exchanges in 1948. Our main analysis includes 1513 companies; and our subsidiary analysis of small companies reported in footnote 8 analysed a further 1264 companies. The remainder of 1948 listings on the Exchange (totalling 9821 in April 1948) included (still separately quoted) subsidiaries of companies which are already included in our analysis (to avoid double-counting this is confined to consolidated companies); banks and other financial organisations, whose accounts cannot readily be compared with those of the sectors examined here; and companies operating mainly overseas (oil companies are significant examples), because the founders of the dataset were focusing on the British economy. Then the listings also include government securities. We are grateful to the editor for prompting this explanation. The editor also suggested giving some indication of the relative size of sectors excluded and included. NIESR (1956, p. 7) provided some helpful indicators of the significance of this population of listed companies in Britain's company population. Their data were for 1953, and comprised 3025 quoted companies, having increased by some 250 since our population was formed in 1948. They wrote: 'The 3,000 or so quoted public companies included…form part of the total of 270,269 companies with share capital, 11,444 of them public companies, which were on the company registers in London and Edinburgh at the end of 1953. In terms of paid-up capital the tabulated companies accounted for …43% of the total for all companies, and 67% of the total …for all public companies. ' In terms of profits, 'the quoted company sector represented in 1949-53 …56% of the total corporate sector, the corresponding figure for manufacturing only being 71%' . 6. For further details of the population see Meeks et al. (1998). For business history of this population see, e.g. Higgins and Toms (2000). 7. A company is excluded if the book value of its net assets was less than 2 million pounds in 1968, or its equivalent of 289 thousand pounds in 1948. See footnote 8 for more detail. 8. In most cases, such small companies would probably not have been members of the main market by 2018. In later years an alternative market was developed for them (first the Unlisted Securities Market (USM)), then the Alternative Investment Market (AIM)): between 1980 and 1990, 817 smaller companies were listed on the USM (Michie, 1999). In 2015, 20 years after AIM was established, 1,074 businesses, with a combined market capitalization of £75 billion, were listed on AIM (London Stock Exchange Group, AIM 20 Key Statistics, available on the London Stock Exchange Group web site, https//www.lsseg.com/markets-products-and-services). The government statisticians responsible in later years for collecting data truncated the population on four occasions by removing a total of 770 very small companies in the years 1960, 1964, 1968, and 1973. This creates a lack of consistency over time if we wish to follow the fate of a fixed population, defined in 1948. We have therefore excluded those companies throughout. This causes a potential downward bias in our observed survival rates because, by definition, the excluded companies had all survived from 1948 to the date of their first exclusion, so that the excluded group are all survivors, at least up to the date of the first cull. In order to eliminate this problem and to achieve a comparable population of large listed companies in 1948, we have, additionally, excluded from our opening popula-tion those small companies which were not subsequently culled but would have been culled in 1948 had a size qualification been in force then. These are small companies that did not survive long enough to be culled or those that were small in 1948 but had grown rapidly to exceed the size threshold by the time of a subsequent cull. These number 494 companies. Their experience was not greatly different from that of the larger companies reported below in this paper. They experienced a half-life of just 10 years, compared with slightly more for the bigger firms. And while takeover was still the dominant cause of death (68% vs 83%), a rather higher percentage (13% vs, 6%) than we found for the larger companies ended in receivership/administration/liquidation. The statisticians who were collecting the data in order to construct aggregate annual financial accounts for industries ceased monitoring the very smallest companies because of their limited economic significance for this purpose. Our estimate of the net assets in 1948 of the largest of the excluded companies is £289k. If all the 1264 excluded companies were this maximum size, their combined net assets would have been £365m. In our retained population, the combined net assets of just three larger companies, Unilever, ICI and Imperial Tobacco, were £408m. 9. We refer to the companies studied as a population rather than a sample because they are a complete set of companies falling within the definition. They are not a sample of a larger population, drawn with the intention of making inferences about that population. 10. For example, the 'merger' of BP (not in our population) with Amoco in 1998 was accounted for in this way. Merger (better described as 'pooling of interests') accounting was subsequently banned by the new International Accounting Standard, IFRS 4 (2004). 11. A half company arises in a country's tally where a company has dual nationality (e.g. Unilever). 12. Napolitano et al. (2015) cite evidence for the Fortune 500 for the period 1970 to 1983 that 'the life expectancy of big companies was between 40 and 50 years'; and that 'of the largest British, French and German companies in 1907,…the majority had vanished by the end of the 1980's' . 13. But it is vastly more important for this population and period than it was for Shannon's (1954) study of UK limited companies in the seven decades up to 1929, when just one-sixth of the companies were 'sold, amalgamated or reconstructed' (p. 383). The Shannon population is dominated by unlisted companies (but also influenced by late Victorian speculative capitalism; Shannon is critical of the apparent gullibility of investors). Unlisted companies may be relatively immune to hostile takeovers because a potential bidder cannot build up a shareholding in the target company through routine market purchases. Liquidations were the predominant cause of death in Shannon's population, and half of these were compulsory, suggesting business failure.  Napolitano et al. (2015) note that the oldest continuously operating company in the world was Kongo Gumi in Japan, at 1428 years: and we know that Japanese M&A is exceptionally low by international standards (Mauboussin & Callahan, 2015). 15. These are simple arithmetic averages, not compound rates. 16. Law & Bonar has been delisted since 2018: it is now part of the German Freudenberg group.
We have checked that, at the time of revision, no further 'survivors' have disappeared in the meantime. 17. Kynaston (1994) dates 1958-9 as the turning-point in the M&A market. He refers (p. 239) to the conjunction of easier money and a bull market with the 'Aluminium War' -the takeover battle for British Aluminium: 'the government had changed for ever the market for corporate control by siding with the aggressor' . 18. It would be possible to refine the analysis further by examining the attrition rate over somewhat shorter periods and attempting to relate it to changes in the state of the economy and institutional changes. However, apart from the first decade the attrition rate appears to be remarkably stable over the longer term despite the many institutional changes in these seven decades, so this would not necessarily be a fruitful exercise. At first sight, the stability of this measure of the death rate in different decades is surprising. The economic conditions faced by this population of companies have changed significantly from decade to decade. For example, the UK economy was stronger in the 1960s than in the 1970s, and in the decade up to 2008 than in the one after. It could be that there are offsetting mechanisms: recession brings more business failures (Goudie & Meeks, 1991), and 'scavenger' takeovers of moribund firms; while stock market booms are associated with opportunistic takeovers in which bidders exchange their highly priced shares for targets (Botsari & Meeks, 2018;Shleifer & Vishny, 2003). This would be a promising area for further research. Another possible development would be to analyse death rates over very short periods -years or even quarters, linking them to fluctuations in, for example, interest, exchange and tax rates. This would require a substantial macroeconomic modelling exercise (as in, for example, Goudie & Meeks, 1991, well beyond the scope of this paper. 19. A promise no longer offered. However, at the time of writing, Darn Tough offer such a promise for their socks! 20. E.g. Cuypers et al., 2017;Gregory, 1997;McKinsey & Co, 2010;Meeks, 1977;Moeller et al., 2005;Ravenscraft & Sherer, 1987. 21. By concentrating on survival factors, we do not intend to imply that survival is necessarily a virtue. As our discussion of the take-over mechanism indicates, there is some evidence that this, the principal cause of death, is not an efficient process, but equally it can be interpreted as a necessary means of adaptation to a changing environment, as in biological populations. We are grateful to an anonymous reviewer for suggesting this clarification.