Professional Documents
Culture Documents
Fabio Braggion and Steven Ongena 2017
Fabio Braggion and Steven Ongena 2017
Accepted Article
BANKING SECTOR DEREGULATION, BANK–FIRM RELATIONSHIPS
We study the effects of the 1971 deregulation of UK banking on firms’ financial and
investment policies. The deregulation was a turning point in the evolution of firm–bank
relationships during the twentieth century. Indeed, for more than 80 years prior to
deregulation most firms had had a relationship with only one bank: this was no longer the case
from 1971 on. Deregulation and intensifying competition in the banking sector spurred firms—
in local markets with many banks already active—to increase leverage and to invest more in
research and development. Bank debt similarly expanded while trade credit contracted.
∗ Corresponding author. Department of Banking and Finance, University of Zurich (UZH), Plattenstrasse 32, CH-8032 Zurich,
Switzerland; e-mail: steven.ongena@bf.uzh.ch.
We would like to thank two anonymous referees and an Associate Editor, along with Liam Brunt, Fabio Castiglionesi, Bob
DeYoung, Leslie Hannah, Phil Molyneux, Lyndon Moore, Daniel Paravisini, Morten O. Ravn (editor), Peter P. Robejsek,
Anthony Saunders, Klaus Schaeck, John Turner, participants at the following—the AEA Meetings (San Diego), the ACPR-BdF
Conference (Paris), the CESifo-Bundesbank Conference (Munich), the ESRC Seminar (Bangor), the FDIC-CFR Workshop
th
(Washington, DC), the SUERF-BoF Conference (Helsinki), the MFS Symposium (Larnaca), the 6 Summer Macro-Finance
st
Workshop (Paris), the 1 EABH Summer School in Financial and Monetary History (Madrid), SFI Research Days (Gerzensee),
rd
the 3 Financial History Workshop (Tilburg), the XX Finance Forum (Oviedo) and the European Banking History Conference
(Rotterdam), and seminar participants at the Catholic University of Milan, Erasmus University (Rotterdam), Frankfurt SFM,
the Free University of Bolzano, NUS, ULB, the universities of Bath, Maastricht, and Navarra (Pamplona), and Aarhus,
Bocconi, Oxford, Paris West and Tilburg universities for helpful comments. We thank Massimo Giuliodori, Andy Haldane
and Peter Richardson for sharing data and Dave Brooks (the English Language-Coaching Service, www.elcs.ch) for excellent
editorial assistance, and we acknowledge financial support from CAREFIN, EBC, ERC (ADG 2016-GA740272 lending,
Ongena), the FDIC-CFR, NWO (VIDI, Braggion) and TCF.
This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1111/ecoj.12569
To give a sense of the significance of the reform, we first document that for more than 80 years
(since reliable data has been available) firms established a business relationship with only one bank.
However, starting in 1971 a remarkable shift took place from bilateral to multilateral relationship
banking. Indeed, about 85 per cent of UK companies in our sample were involved in a single bank
relationship between 1906 and 1970. This figure considerably declines to 72 per cent in 1976 and to
64 per cent in 1986.
In order to establish a causal relationship between the 1971 banking sector deregulation and firms’
financing, we use a difference-in-difference analysis and treat the 1971 reform as a natural
experiment. We compare the pre- and post-reform financial policies of firms located in cities where
many banks already had multiple branches with those of firms located in cities where only a few
banks had some degree of branch presence. In the former type of city, banks could start competing
immediately after the reform was introduced, while in the latter type of city this was not possible
without setting up new branches. It is important to note that, due to the long history of cartelisation
in the UK banking sector, local banking market conditions were historically inherited and did not
depend on the current level of economic activity. As a result, we are more likely to identify the true
effect of the reform on firms’ financial and investment policies.
Our analysis finds that following deregulation, firms in less concentrated local markets increased
their number of bank relationships, as well as bank debt over total debt, and leverage. They
decreased their trade credit considerably more than did firms in more concentrated local markets,
while at the same time increasing their net investment in intangible assets. A one to zero change, for
example, in the concentration index increases the number of bank relationships by one tenth, bank
debt over total debt by about ten percentage points, and debt over total assets by six percentage
points.
The effects on the number of bank relationships and the use of bank debt are stronger for firms
below median age and firms especially engaged in R & D. This suggests that the lack of competition
especially reduced access to finance for young and more innovative firms.
Our paper makes two contributions to the extant literature. First, our work contributes to the
growing literature that studies the evolution of financial phenomena during very long periods of
time.1 We place deregulation in a larger historical context and show that deregulation broke (as far
as we can look back) a secular trend in the number of firm–bank relationships.
1
E.g. Rajan and Zingales (2003), Chambers and Dimson (2009), Frydman and Saks (2010), Jordà et al. (2013),
Schularick and Taylor (2012), Philippon and Reshef (2012) and Knoll et al. (2017). Recent work by Graham et al.
Unlike US banking reforms that were introduced stepwise over many years, the UK introduced most
elements of reform in a single month—that is to say, in October 1971. Both Zarutskie (2006) and
Rice and Strahan (2010), for example, study a later step in US reform: the deregulation of interstate
branching.2 While Rice and Strahan (2010) find that interstate branching reduced loan pricing but did
not increase firms’ leverage, Zarutskie (2006) finds that, following reform, young firms were less
likely to use debt. However, these results do not exclude the possibility that positive effects on loan
quantities could have been observed in the earlier stages of deregulation, when local banking
markets were even less competitive.
And unlike the 1985 French deregulation considered in Bertrand et al. (2007), the UK reform does
not involve the elimination of industrial subsidies provided by the government via the banking
sector. As a result, we can isolate the effects of bank competition on firms’ financing from the
possible effects related to the role of the government in providing credit to the economy. Bertrand
et al. (2007) find that following deregulation firms’ use of bank debt and leverage decreased, a result
that supports the notion that with the elimination of government subsidies banks became more
careful with regard to screening borrowers and granting loans.
As our main empirical strategy consists of a difference-in-difference analysis based on the banking
sector deregulation of 1971 and on the local degree of banking concentration prior to deregulation,
two additional salient points need to be clarified. First, banking deregulation led to changes in the
supply of capital conditions. The historical evidence in fact suggests that deregulation was driven by
the need of the Bank of England to have a more effective way of conducting monetary policy, rather
than by any effort to accommodate the specific needs of the corporate sector. Second, deregulation
arguably intensified banking competition most immediately in those local markets where many
banks already had a ‘bricks-and-mortar’ presence. We base this conjecture on Stigler (1964) who
states that in an area with a large number of producers (in our case, a large number of banks and
branches) cartel agreements are more difficult to enforce. The larger the number of banks in a city,
the less visible are the actions of each bank, and the lower is the probability of detecting a bank that
breaks any given cartel agreement.3
(2015) shows that US corporate leverage doubled from the 1920s to the 1970s. The authors explain this
phenomenon not only by looking at the changes that took place in taxation, but also by linking the doubling
with fundamental developments in industry composition, firms’ characteristics, assets and investments, and—
important for our paper—credit supply conditions.
2
See also e.g. Black and Strahan (2002), Stiroh and Strahan (2003), Huang (2008) and—recently—Hoffmann
and Stewen (2015) and Berger et al. (2015) on the relevance of banking deregulation for credit and real
growth, and Leary (2009), Sufi (2009), Lemmon and Roberts (2010) and Banerjee and Duflo (2014) on the
relaxation of credit supply constraints and corporate borrowing.
3
Also important for our identification strategy is the fact that the distribution of bank branches in the UK in
the late 1960s was historically determined and unlikely to depend on then current economic conditions. In
1948, the Bank of England and the Treasury promoted the introduction of cartel agreements among British
Section 1 provides some institutional background and introduces the data. Section 2 demonstrates
that deregulation was indeed a unique turning point during the twentieth century with regard to the
way in which corporations engaged with the banking sector. Section 3 analyses the impact of
deregulation on the number of firm–bank relationships, and on firm financing, across local banking
markets according to the concentration of bank branch presence. Section 4 analyses how the reform
led to new banks penetrating more concentrated banking markets. Section 5 concludes.
In the late 1960s, the government and the Bank of England recognised the inadequacy of this
arrangement. In October 1971, the cartel was dismantled and, via a set of reforms known as
‘Competition and Credit Control’, the UK authorities promoted greater competition between
financial institutions. In particular, both ceilings on interest rates and direct controls on credit were
lifted; reserve requirements were also relaxed. Such a change in policy generated strong
competition between banks and other financial intermediaries on both the deposit and loan
markets. The available historical evidence suggests that the lifting of the provisions related to the
cartel was independent from possible changes in any given firm’s demand for loans, which would
imply a form of reverse causality: deregulation happened because firms wanted to borrow more.
Reverse causality could in principle inflate the impact we estimate; however, it appears that the
cartel was lifted because:
(1) Of the inefficiencies related to the cartel itself. Like every monopoly, it led banks to waste
resources, suffer agency problems and lack dynamism and entrepreneurship.
banks. The cartel appears to have frozen local market conditions at the 1948 levels. Our own data, as well as
the history of the period, reveal that local market conditions changed little between 1948 and the late 1960s
(Pressnell (1970), p. 385; Carnevali and Hannah (1995), p. 74; Ackrill and Hannah (2001), p. 182).
Based on this historical record, we analyse the ‘Competition and Credit Control’ reform in 1971 in
three steps. First, we assess how important a turning point this deregulation was for firm–bank
relationships, then we analyse the impact of deregulation on bank relationships and firms’ financing,
and—finally, in a third step—we consequently investigate the importance of the transition to
multiple banking for corporate financing.
For each company, the Yearbook provides the name of the company and its location, the names of
its directors, the total amount of nominal share capital issued by the company and information
related to the company’s corporate governance arrangements, such as its voting rights. In some
cases dividend payments information is also provided. After 1948, the Yearbook also provides a
4
For the years 1896 to 1966 we retrieve our data from eight issues, in particular from those of 1896, 1906,
1916, 1920, 1924, 1934, 1938, and 1948. Starting in 1956 and until 1966 and between 1976 and 1986 we
access ten issues on a biannual basis, and between 1966 and 1976 eleven issues on an annual basis. Our
sample period ends in 1986 because by then the transition to multiple relationships (we detail later) appears
to be concluded (see the UK numbers in Ongena and Smith (2000) for example). We also do not study the so-
called ‘Big Bang’—the effects of the liberalisation of the London Stock Exchange undertaken in October 1986.
In principle, we consider this to be outside the scope of our study as it concerned principally the structure of
trading and the ownership of the London Stock Exchange rather than the bilateral relationships between firms
and banks.
5
This information can be directly retrieved from the Yearbook.
6
Before 1905, the Yearbook provides sufficient information only for a selection of firms. Usually, the largest
and the most traded firms are included. As a result, we suspect that our 1896 sample is biased towards large
and liquid companies.
We also collect data on the location of bank branches, this from the Bankers’ Almanac. Each year,
the Bankers’ Almanac lists the locations and addresses of the branches of each bank located in the
UK. We collect this information for 1948, and at biannual bases between 1966 and 1986. Data on
population and employment in the various historical British counties are taken from Lee (1979), who
reports this information for every ten years starting in 1841. We use the values indicated for 1961.
To provide more detail on the transition from bilateral to multilateral banking, Table 1 Panel B
reports the number of firm–bank relationships annually for the period 1964‒76 for 664 UK firms that
7
To be sure that the name and the number of banks indicated in the Yearbook corresponds to the actual bank
relationships maintained by the firm, we compare the information in the Yearbook with the information
provided in a random sample of firms’ balance sheets selected throughout our time period. In all cases, the
information reported by the Yearbook corresponds with the information reported by the firm in its balance
sheet.
8
These results are similar to the numbers presented in Ongena and Smith (2000) for the year 1996. The
authors investigate the number of bank relationships by sampling 138 large companies in the UK, and find that
only 23 per cent of firms in their sample maintained only a single bank relationship.
The results basically corroborate those shown in Panel A. The average number of relationships
increases in twelve years from 1.2 to 1.4. Between 1964 and 1970, the number of bank relationships
oscillates between 1.24 and 1.27. In 1971 the number is 1.30; in 1973, 1.34; and even higher values
are registered in the following years.
We also investigate which type of bank is added to firms’ firm–bank relationships. We distinguish
between clearing banks (mostly large, London-based banks), other British banks, and foreign banks.
Table 2 lists the number and percentage of relationship – year observations between 1966 and 1986
by the type both of relationship bank and added bank. Clearing banks account for 96 per cent of all
relationship bank – year observations (85 per cent are headquartered in London, 8 per cent in
Scotland, and 3 per cent in Ireland), while other British and foreign banks account for only 2 per cent
each.
Surprisingly given these proportions as presented in Table 2, many firms add another clearing bank
as a second bank, resulting in 62 per cent of the added bank observations. Other (secondary) British
banks account for 13 per cent; foreign banks for more than 25 per cent (of which 8 per cent
corresponds to commonwealth banks and the remainder to other foreign banks). These percentages
suggest that while many firms simply engaged another clearing bank so as to increase access to
credit, other firms ‘traded down’ to a (secondary) British bank so as to obtain a better size fit, or
engaged with a foreign bank in order to obtain better trade-related financial services.
In sum, it appears there was a fundamental shift from bilateral to multilateral banking that started in
1971, marking the deregulation that occurred then as a watershed event that we now aim to study
further.
Second, the degree of banking competition and concentration has to be exogenous to the then
present economic conditions. This was indeed the case in the UK in the late 1960s. The introduction
of the banking cartel after 1948 appears to have mostly frozen local market conditions at 1948
levels. The available historical evidence suggests that the opening of new branches between 1948
and 1971 was a rather wasteful activity and seemingly did not follow sound economic criteria
(Pressnell (1970), p. 385; Ackrill and Hannah (2001), p. 182). It appears as though banks were
opening new branches for self-serving motives (e.g. empire building) rather than in response to
changes in local economic conditions. Put differently, ‘[…] in expanding their branches, [banks were]
growth maximizers; they were not in general profit-maximizers’ (Carnevali and Hannah (1995), p.
74). This fact helps our identification strategy because, if anything, the opening of new branches
between 1948 and 1971 made our local measures of competition/concentration particularly
unrelated to then present economic conditions.
To give a sense of the persistence of local market competition before and after the reform, we
compute at the town level the Herfindahl–Hirschman Index (HHI) of banking concentration using
branch data for each UK town in 1948, 1966 and 1984. We find that in 1966, 83 per cent of UK cities
were in the same quartile of banking concentration as in 1948. This figure reveals a quite high
degree of persistence in local banking concentration in the years preceding the reform. In 1984, 60
per cent were in the same quartile as in 1966: a drop of 23 percentage points in respect of the pre-
reform period.9
We inspect the validity of our identification strategy by constructing graphs that display the annual
averages of firms’ numbers of bank relationships and leverage between 1968 and 1975.10 In Figure 1,
we differentiate between firms located in cities (excluding London) with high and low banking
concentration using the HHI measured in 1966 as well as firms located in London.11 We consider
London separately as it housed the London Stock Exchange and the Eurodollar market, and the many
foreign banks located in London were providing a large number of financing alternatives to
9
The 1984 Bankers’ Almanac does not cover as many towns as the 1948 and 1966 editions. Smaller towns with
a relatively high HHI in 1966, for example, are missing. To the extent that, after deregulation, more branches
were opened in these towns, we are underestimating the degree of persistence between 1966 and 1984.
10
Bank debt over total debt displays a similar behavior. To reduce the possibility that the result is driven by
issues related to sample composition, we consider firms that appear for at least two years both before and
after the reform.
11
We define as cities with high banking concentration those cities that have an HHI above the sample median,
and as cities with low concentration those with an HHI below the sample median.
3.2. Specifications
The equation we wish to estimate is
, = + + ∗ , + ∗ , + , (1)
with the firms’ fixed effects, the year fixed effects and the error term. Equation (1) captures
the idea that the effect of the 1971 deregulation should be stronger in markets that have lower
banking concentration at the beginning of the period. yit indicates (consecutively) the firm i’s
number of bank relationships, bank to total debt, leverage ratio and trade credit total debt. These
variables are intended to capture the main financial policies of a firm—that is to say, the importance
of its reliance on debt and the relative importance of bank debt vis-à-vis other forms of credit. Post is
a dummy variable that equals one for the years following 1971 and equals zero otherwise.
HHIi,1966 is defined as
ℎ (2)
, = .
∑ ℎ
Equation (2) represents the HHI of the local banking market in which firm i has its headquarters and
is calculated on the basis of the number of bank branches of all banks k locally present in 1966 (with
N the total number of banks in the country).
Controls (defined in Table 3) include, at the county level, County Population, which is the number of
people resident in the county in which the firm has its headquarters, and Employed over Total
Population of the County, which is the number of people employed divided by the total number of
people resident in the county in which the firm has its headquarters, and at the firm level, Firm Has
HQ in London (0/1), which equals one if the firm has its headquarters in London and equals zero
otherwise, Officially Listed (0/1), which equals one if the company had any class of its outstanding
shares officially listed in London and traded on the floor and equals zero otherwise, One Share - One
Vote (0/1), which equals one if the company applies the one share–one vote principal and equals
zero otherwise, Arm's Length Debt (0/1), which equals one if the company has outstanding public
debt securities and equals zero otherwise, Book Value of Assets, Age (of the company) in the sample
year, Board Size, which is the number of members of the administration board, Subsidiary (0/1),
which equals one if the company is controlled by another company and equals zero otherwise, and
Return on Equity, which is total profits divided by total capital and reserves.
In order to correct the estimates for potential serial correlation, we follow Bertrand et al. (2004)—
that is to say, we average the dependent variable before and after deregulation, and estimate
Equation (1) in the following form:12
, − , = + , + , + . (1’)
The dependent variables are the changes between the 1968 to 1971 (pre) and 1972 to 1974 (post)
period averages of the dependent variables we have described above.13 Detailed definitions and
summary statistics for all variables appear in Table 3.14
In cities where market concentration is low, we expect banks to immediately compete vigorously,
and consequently that firms will increase their number of relationships and use more bank debt and
increase leverage. Hence, in those specifications we expect a negative sign on this main variable of
interest (i.e. we expected β to be negative). We cluster the standard errors at the city level. In
Appendix Table A.1 we also assess if the even more predetermined variable 1948 HHI of the Local
Banking Market Where Firm HQ Is Located yields comparable results. It does.
12
At this stage, it is worth observing that we also revisit all exercises (reported below) starting from the
following alternative specification:
, = + + ∗ , + , + , (1 Alt.)
, − , = + , + ( , − , )+ . (1’ Alt.)
But the estimates from this alternative specification (which is not so conservative) are similar to those we
report below and are therefore not discussed further.
13
The parameter represents the time fixed effect in the first differenced version of Equation (1).
14
We restrict the analysis to firms that have at least one observation in both the pre- and the post- period. To
maintain consistency across specifications we also restrict the analysis on the number of bank relationships to
firms that are included in the Cambridge DTI Databank. Lifting this restriction makes the results somewhat
stronger.
We find that the local degree of concentration is negatively related to number of bank relationships,
bank debt over total debt, and leverage: all this is consistent with the notion that after the 1971
reform, in areas with more banks, firms increased their reliance on bank debt and leverage.
In all cases, the estimated coefficients are not only statistically significant but also economically
relevant. Take the number of firm–bank relationships. The estimated coefficient on the 1966 HHI is
mostly unaffected by the inclusion of a varied set of firm characteristics and the 1966 HHI of the
Neighbouring Banking Markets (in Model II),15 and is estimated to equal -0.098 (in I) and -0.071 (in
II). An increase in the HHI from zero (unconcentrated) to one (monopoly) amounts to a decrease of
about 7 to 9 per cent in the probability of adding an extra bank relationship.
The same is true for the change in bank debt over total debt and for the change in total debt over
total assets. In both cases the estimated coefficients on the HHI are consistently negative and
statistically significant and amount to an effect that—in terms of economic relevancy (for HHI from
zero to one)—is equal to a decline of about 10 percentage points (45 per cent of its pre-reform
average) of the reliance on bank debt and a decline of about 6 percentage points (that is, 15 per cent
of its pre-reform average) of leverage. These figures are large and meaningful, and strongly suggest
that deregulation affected firms’ financing, especially in those locations where we would a priori
expect this to be the case—that is to say, where branches of many different banks were already
present and could start competing immediately.
For the dependent variable Change in Trade Credit over Total Debt we find that the estimated
coefficients on the HHI are positive, though not statistically significant, and amount to an impact (for
a zero to one HHI increase) corresponding to 8 to 9 per cent of its pre-shock average.
Overall, these estimates strongly suggest that deregulation immediately affected firms’ financing, as
well as the local presence of banks and the likely intensity of competition that resulted from that
presence.16
3.4. Extensions
In Table 5 we add interactions (to Model II from Table 4) of the 1966 HHI of the local banking market
with three salient firm characteristics that capture opacity—that is, Above Median Patents and
Goodwill (0/1), Below Median Book Value of Assets (0/1) and Below Median Age (0/1). While Table 5
only presents the interaction term estimates (for reasons of parsimony), the estimated equations do
15
This variable is defined as the Herfindahl–Hirschman Index of the neighbouring banking markets in which the
firm has its headquarters calculated on the basis of the number of bank branches present there in 1966. Recall
that the definitions of all variables used in the analysis appear in Table 3.
16
We also control for variables that describe the capitalisation and liquidity of the bank that originally traded
with the firm. Results remain unaltered (Appendix Table A.2).
Our results show that especially firms with above median patents and goodwill or younger firms,
when located in less concentrated markets, were more likely to add a bank and experienced more
significant changes in financial policy. Firms below the median age when moving from monopoly to
perfect competition for example would double their reliance on bank debt and increase their
leverage by 28 per cent. These results are in line with the ‘traditional’ view that banks with more
market power limit access to finance for young firms and firms engaged in R & D.
In Table 6 we assess whether deregulation not only affected both the number of banks and
corporate financial policies, but also resulted in immediate changes in firms’ investments. If
deregulation allows firms to acquire additional debt financing (as suggested by the results in Table
4), we may expect these extra resources to be used for additional investment. We check this
possibility by studying the total investment made by firms, in particular analysing two of its
components: Change in Net Investment in Tangible Assets and Change in Net Investment in
Intangible Assets. All models we estimate in Table 6 contain similar controls and fixed effects to the
similarly numbered models in Table 4.
The estimated coefficients on the HHI variables suggest that there is a consistent positive effect on
investment—that is to say, in markets that prior to deregulation had lower concentration
investment increases more. The effect is statistically significant and economically very large for
investment in intangible assets, which grows fivefold, moving from monopoly to perfect
competition.17 These results strongly suggest that deregulation helped firms to increase their
investments in goodwill and research and development.
In sum, not only corporate financial policies and the number of bank relationships but also corporate
investment is affected by deregulation. The effect is distributed along banking market concentration
in a way that is consistent with the resulting immediate intensification in the degree of competition
present.
3.5. Robustness
We run a number of robustness tests (in addition to the various model specifications and main
variable definitions that were already deployed). First, we construct a falsification test to see
whether the parallel trends assumption is satisfied. Any difference-in-difference analysis rests on the
assumption that the dependent variables of each group (in our case, firms located in different cities)
display the same trend in the periods preceding the reform.
17
Notice that the pre-shock average of net investment in intangible assets over total assets is very small:
around 0.069%. All models in Table 6 contain similar controls and fixed effects as in the similarly numbered
models in Table 4.
Our identification strategy is also designed to minimise the effects that changes in reserve
requirements (which were included in the 1971 reform) may have on our results. In principle,
variations in reserve requirements alone should not have any differential effects between
concentrated and not concentrated markets if banks are not in a position to compete. We rerun
similar analyses for two different estimation windows around 1957 and 1963, years when the UK
banking market was still operating under the cartel agreement but the reserve requirement was
raised from a minimum of 25 per cent to 30 per cent (in 1957) and then lowered to 28 per cent (in
1963). We find no effects of these changes in reserve requirements on the various debt ratios
(Appendix Table A.3).
We also assess our main findings for wider time windows—that is to say, 1966‒71 (pre) to 1972‒76
(post) and 1956‒71 (pre) to 1972‒86 (post). In the latter case, we also employ an HHI based on 1948
branch information. Albeit slightly weaker in terms of statistical significance, findings remain
qualitatively similar and economically very relevant (Appendix Table A.4).18
For each year and each city, we first compute the average interest rate on lending by dividing the
total interest received by the total loans outstanding. Then we calculate the interest rate spread,
subtracting from the average interest rate the corresponding Bank of England rate. With this data in
18
We also pursue a general method of moments (GMM) Arellano and Bond (1991) estimation with two- and
three-year lags of the dependent variable as instruments for its one-year lag, which is added to Equation (1).
Our estimate on the interaction of the 1966 HHI of the Local Banking Market Where Firm HQ Is Located and a
dummy that takes the value of one after 1971 and equals zero otherwise is similar to the estimates we have
obtained so far with a quite a different econometric approach (Appendix Table A.5). And, following Angrist and
Krueger (1999) and Lemmon and Roberts (2010), we adopt a difference-in-differences analysis coupled with
propensity score matching to understand whether adding a bank to the existing set of relationship banks has
an effect on firms’ financial policies. Results are consistent with those reported here (Appendix B and Table A.6
and A.7).
19
The cities are Bristol, Exeter, Maidstone, Southampton, Leeds, Newcastle, Sheffield, York, Liverpool,
Manchester, Preston, Birmingham, Leicester, Norwich and London.
A possible interpretation of this result is that, in the immediate aftermath of the reform, Midland
branches located in low banking concentration cities were the first to lower interest rates. By 1974,
other Midland branches followed. This interpretation appears plausible when we consider that the
interest rate sample covers a selection of medium/large sized cities where in principle banks were
already facing some form of competition and where we would expect the adjustment of interest
rates to occur more quickly. Interestingly, the cross-sectional variance of the interest rate spread
increases after deregulation. We find that the cross-sectional variance equals 0.26 per cent before
deregulation, whereas the variance increases to 0.39 per cent in 1972 and to 0.50 per cent between
1972 and 1974. This increase in dispersion is consistent with the idea that deregulation—by
removing interest rate controls—allowed banks to freely set their interest rates.
where ∆ ,( , ) denotes the change in the total number of banks or branches in city i between
year t and year t+j, , indicates the HHI of industry concentration as defined in Equation (2) and
, is a dummy variable that takes the value one if the city is London and zero otherwise.20
In the analysis, we distinguish whether the relationship between local competition and entry is
different in the pre- and post-reform periods. We report the results in Table 8 (in Appendix Table A.8
we provide summary statistics). Models I and II deal with the period before the reform and show a
positive correlation between banking concentration in 1948 (1958) and the opening of new branches
20
Unfortunately, we do not have any data on city populations or economic activity that would allow us to
include additional controls.
Regarding the opening of new branches, Models IV and V show that cities with high banking
concentration in 1948 (1958) experienced a decline in their numbers of bank branches, statistically
significant at the 1 per cent level. In particular, a move from monopoly to perfect competition in
1948 would have explained around 50 per cent of the standard deviation of the dependent variable.
While we find no opening of new branches in the pre-reform period, the decline in the numbers of
branches in cities with high HHI is consistent with the notion that existing banks, more than anything
else, moved branches between cities.21 Model VI reveals a positive correlation between cities’ HHI in
1966 and the opening of new branches between 1966 and 1984: moving from monopoly to perfect
competition explains around 20 per cent of the standard deviation of the dependent variable. All in
all, after the 1971 reform, cities with high banking concentration experienced an increased presence
of new banks and new branches. This result is consistent with the notions that banks also responded
to the new competitive environment by opening new offices and by enlarging their geographical
presence.
5. Conclusion
In this paper, we document that banking deregulation and the consequent intensifying competition
may be one of the factors that explains the increase in corporate leverage observed in the UK and,
potentially, throughout the rest of the world during the twentieth century. Based on how the
relationships between firms and banks evolved during the twentieth century in Britain, we first show
that the 1971 deregulation indeed entailed a remarkable shift from bilateral to multilateral
relationship banking during this period. Then we document that bank and total debt increased more
after banking deregulation for firms located in banking markets with more intense pre-existing
competition, and for firms that added a banking relationship compared to those that did not.
21
Despite this observed reduction, there exists a high intertemporal correlation between all pre-1971 HHIs.
References
Ackrill, M. and L. Hannah (2001). Barclays: The Business of Banking, 1690-1996, Cambridge:
Cambridge University Press.
Arellano, M. and S. Bond (1991). 'Some Tests of Specification for Panel Data: Monte Carlo Evidence
and an Application to Employment Equations', Review of Economic Studies, vol. 58(2), pp.
277-297.
Banerjee, A. and E. Duflo (2014). 'Do Firms Want to Borrow More: Testing Credit Constraints Using a
Targeted Lending Program', Review of Economic Studies, vol. 81(2), pp. 572-607.
Berger, E. A., A. W. Butler, E. Hu and M. Zekhnini (2015) Credit Be Dammed: The Impact of Banking
Deregulation on Economic Growth, Houston TX, Rice University.
Bertrand, M., E. Duflo and S. Mullainathan (2004). 'How Much Should We Trust Differences-in-
Differences Estimates?', Quarterly Journal of Economics, vol. 119(1), pp. 249-275.
Bertrand, M., A. Schoar and D. Thesmar (2007). 'Banking Deregulation and Industry Structure:
Evidence from the French Banking Reforms of 1985', Journal of Finance, vol. 62, pp. 597-628.
Black, S. E. and P. E. Strahan (2002). 'Entrepreneurship and Bank Credit Availability', Journal of
Finance, vol. 57(6), pp. 2807-2834.
Calomiris, C. W. and S. H. Haber (2014). Fragile by Design: The Political Origins of Banking Crises and
Scarce Credit, Princeton NJ: Princeton University Press.
Carnevali, F. and L. Hannah (1995). 'The Effects of Banking Cartels and Credit Rationing on UK
Industrial Structure and Economic Performance since World War Two', in (Bordo, M. D. and
R. Sylla eds.), Anglo-American Financial Systems. Institutions and Markets in the Twentieth
Century, pp. 65-88, New York: New York University Salomon Center.
Chambers, D. and E. Dimson (2009). 'IPO Underpricing Over the Very Long-Run', Journal of Finance,
vol. 64(3), pp. 1407-1443.
Cheffins, B. R. (2008). Corporate Ownership and Control: British Business Transformed, Oxford:
Oxford University Press.
Frydman, C. and R. E. Saks (2010). 'Executive Compensation: A New View from a Long-Term
Perspective', Review of Financial Studies, vol. 23(5), pp. 2099-2138.
Graham, J. R., M. Leary and M. Roberts (2015). 'A Century of Capital Structure: The Leveraging of
Corporate America', Journal of Financial Economics, vol. 118(3), pp. 658-683.
Hoffmann, M. and I. Stewen (2015) Holes in the Dike: The Global Savings Glut, U.S. House Prices and
the Long Shadow of Banking Deregulation, 5332, Munchen, CESifo.
Knoll, K., M. Schularick and T. M. Steger (2017). 'No Price Like Home: Global House Prices, 1870-
2012', American Economic Review, vol. 107(2), pp. 331-353.
Leary, M. T. (2009). 'Bank Loan Supply, Lender Choice, and Corporate Capital Structure', Journal of
Finance, vol. 64(3), pp. 1143-1185.
Lemmon, M. and M. R. Roberts (2010). 'The Response of Corporate Financing and Investment to
Changes in the Supply of Credit', Journal of Financial and Quantitative Analysis, vol. 45(3),
pp. 555–587.
Ongena, S. and D. C. Smith (2000). 'What Determines the Number of Bank Relationships? Cross-
Country Evidence', Journal of Financial Intermediation, vol. 9(1), pp. 26-56.
Philippon, T. and A. Reshef (2012). 'Wages and Human Capital in the U.S. Finance Industry: 1909–
2006', Quarterly Journal of Economics, vol. 127(4), pp. 1551-1609.
Pressnell, L. S. (1970). 'Cartels and Competition in British banking: A Background Study', Banco
Nazionale del Lavoro Quarterly Review, vol. 95, pp. 373-406.
Rajan, R. G. and L. Zingales (2003). 'The Great Reversals: The Politics of Financial Development in the
20th Century', Journal of Financial Economics, vol. 69, pp. 5-50.
Rice, T. and P. E. Strahan (2010). 'Does Credit Competition Affect Small-Firm Finance?', Journal of
Finance, vol. 65(3), pp. 861-889.
Schularick, M. and A. M. Taylor (2012). 'Credit Booms Gone Bust: Monetary Policy, Leverage Cycles,
and Financial Crises, 1870–2008', American Economic Review, vol. 102 (2), pp. 1029-1061.
Stigler, G. J. (1964). 'A Theory of Oligopoly', Journal of Political Economy, vol. 72(1), pp. 44-61.
Stiroh, K. and P. Strahan (2003). 'Competitive Dynamics of Deregulation: Evidence from U.S.
Banking', Journal of Money, Credit, and Banking, vol. 35(5), pp. 801-828.
Sufi, A. (2009). 'The Real Effects of Debt Certification: Evidence from the Introduction of Bank Loan
Ratings', Review of Financial Studies, vol. 22(4), pp. 1659-1691.
Zarutskie, R. (2006). 'Evidence on the Effects of Bank Competition on Firm Borrowing and
Investment', Journal of Financial Economics, vol. 81, pp. 503-537.
Net Investment in Intangible Assets - Average net investment in intangible assets divided by the book value of assets at the beginning of
the year between 1972 and 1974, minus average net investment in intangible assets divided by the 5,035 0.001 - 0.008
book value of assets at the beginning of the year between 1968 and 1971
Independent Variables
1966 HHI of the Local Banking Market Where Firm HQ Is Located 0 to 1 The Herfindahl-Hirschman Index of the local banking market where the firm has its headquarters 1,203 0.186 0.167 0.129
calculated on the basis of the number of bank branches present there in 1966
1948 HHI of the Local Banking Market Where Firm HQ Is Located 0 to 1 The Herfindahl-Hirschman Index of the local banking market where the firm has its headquarters 1,160 0.193 0.167 0.105
calculated on the basis of the number of bank branches present there in 1948
1966 HHI of Neighbouring Banking Markets Where Firm HQ Is Located 0 to 1 The Herfindahl-Hirschman Index of the neighbouring banking markets where the firm has its 1,192 0.188 0.198 0.041
headquarters calculated on the basis of the number of bank branches present there in 1966
County Population - The number of people present in the county where the firm has its headquarters 8,823 2,120,449 1,674,406 1,303,708
Employed over Total Population in the County - The number of employed divided by the total number of people present in the county where the firm 8,823 0.628 0.477 0.282
has its headquarters
Firm Has HQ in London (0/1) 0/1 =1 if the firm has its headquarter in London, =0 otherwise 8,823 0.280 - 0.449
Officially Listed (0/1) 0/1 =1 if the company had any class of its outstanding shares officially listed in London and traded on 5,065 0.841 1.000 0.366
the floor, =0 otherwise
One Share - One Vote (0/1) 0/1 =1 if the company applies the one share - one vote principal, =0 otherwise 4,819 0.463 - 0.499
Arm's Length Debt (0/1) 0/1 =1 if the company has outstanding public debt securities, =0 otherwise 5,059 0.513 1.000 0.500
Book Value of Assets ('000 of £) Firm book value of assets 5,211 26,126 6,379 79,805
Age years Age of the company in the sample year 8,823 68.258 69.000 33.324
Board Size - Number of members in the administration board 5,008 6.412 6.000 2.524
Subsidiary (0/1) 0/1 =1 if the company is controlled by another company, =0 otherwise 8,791 0.088 - 0.283
Return on Equity - Total profits divided by total capital and reserves 5,211 0.796 0.640 0.818
Notes . The table reports the definitions and descriptive statistics of the dependent and independent variables employed in the ensuing analysis is based. All variables are collected from the Cambridge DTI database. We compute the descriptive statistics on the
full sample of firms between 1968 and 1974 before taking the averages of the variables before and after the 1971 reform.
Change in the Number of Firm- Change in Bank Debt over Change in Total Debt over Change in Trade Credit over
Bank Relationships Total Debt Total Assets Total Debt
I II I II I II I II
1966 HHI of the Local Banking Market Where Firm HQ Is Located -0.100*** -0.075* -0.098** -0.096** -0.054** -0.064** 0.059 0.061
(0.029) (0.044) (0.047) (0.046) (0.023) (0.026) (0.036) (0.040)
1966 HHI of Neighbouring Banking Markets Where Firm HQ Is Located 0.346 0.276 0.178* -0.306*
(0.477) (0.187) (0.103) (0.167)
County/Location Controls Yes Yes Yes Yes Yes Yes Yes Yes
Firm Controls No Yes No Yes No Yes No Yes
Industry Fixed Effects No Yes No Yes No Yes No Yes
R-squared 0.002 0.042 0.007 0.062 0.013 0.116 0.012 0.092
Number of Observations 1,312 970 684 648 684 648 684 648
Economic Relevancy -9.1% -6.8% -47.1% -46.1% -13.7% -16.2% 8.9% 9.2%
Notes . The estimates in this table come from ordinary least squares models. All variables are defined in Table 3. The estimated coefficients are listed in the first row, standard errors that are clustered at the town level are
reported in the second row between brackets, and the corresponding significance levels are in the first row adjacent to the estimated coefficients. County/Location Controls include ln(County Population), Employed Over
Total Population in the County, and Firm Has HQ in London (0/1), while Firm Controls include Officially Listed (0/1), One Share - One Vote (0/1), Arm's Length Debt (0/1), ln(Book Value of Assets), ln(1 + Age), ln(1 +
Board Size), Subsidiary (0/1), and Return on Equity. The Economic Relevancy is calculated for a change in the 1966 HHI of the Local Banking Market Where Firm HQ Is Located from zero to one to be divided by the
mean of the level of the dependent variable. *** Significant at 1%, ** significant at 5%, * significant at 10%.
Change in the Number of Change in Bank Debt over Change in Total Debt over Change in Trade Credit
Firm-Bank Relationships Total Debt Total Assets over Total Debt
Model I II I II I II I II
1966 HHI of the Local Banking Market Where Firm HQ Is Located -0.045 0.001 0.017 0.007 -0.002 -0.007 -0.007 -0.006
(0.063) (0.034) (0.024) (0.027) (0.017) (0.018) (0.037) (0.039)
All Controls of Equivalent Model in Table 3 Yes Yes Yes Yes Yes Yes Yes Yes
Industry Fixed Effects No Yes No Yes No Yes No Yes
R-squared 0.007 0.054 0.006 0.055 0.001 0.070 0.009 0.055
Number of Observations 1,684 1,254 1,215 1,169 1,215 1,169 1,215 1,169
Notes . The estimates in this table come from ordinary least squares models that are equivalent to Models I and II in Table 4. All variables are defined in Table 3. The estimated coefficients are
listed in the first row, standard errors that are clustered at the town level are reported in the second row between brackets, and the corresponding significance levels are in the first row adjacent to
the estimated coefficients. *** Significant at 1%, ** significant at 5%, * significant at 10%.
1948 HHI of the Local Banking Market Where Firm HQ Is Located 0.032 -2.804***
(0.035) (0.111)
1958 HHI of the Local Banking Market Where Firm HQ Is Located 0.036 -0.756***
(0.040) (0.072)
1966 HHI of the Local Banking Market Where Firm HQ Is Located 1.214*** 0.279***
(0.067) (0.097)
R-squared 0.038 0.125 0.942 0.211 0.051 0.014
Number of Observations 3,429 3,232 2,522 3,429 3,232 2,522
Economic Relevancy 5.2% 4.8% 28.1% -50.1% -17.1% 18.5%
Notes. The estimates in this table come from ordinary least squares models. The dependent variables are the change in the number of banks or branches in the local banking market where the firm
headquarters is located. All specifications include Firm Has HQ in London (0/1) a dummy variable that equals one if the firm has its headquarters in London, and equals zero otherwise. All other
variables are defined in Table 3. The estimated coefficients are listed in the first row, standard errors are reported in the second row between brackets, and the corresponding significance levels are in the
first row adjacent to the estimated coefficients. *** Significant at 1%, ** significant at 5%, * significant at 10%.
Fig. 1
Accepted Article Assessing the Parallel Trends Assumption
Leverage
.5
.45
.4
.35
.3
Notes. This figure plots the average number of firm-bank relationships and total debt over total assets
for firms located in cities with a low versus high HHI, and in London. Cities with low HHI are
defined as cities where the Herfindhal-Hirschman Index of banking concentration measured in 1966 is
below the sample median. Cities with High HHI are cities where the Herfindhal-Hirschman index of
banking concentration measured in 1966 is equal or above the sample median.
Panel B 1968-1974
0
-.005
-.01
-.015
-.02
-.025
Notes. Each point on the figure represents the interest rate spread after the reform minus the interest
spread before the reform for each city, represented on the y-axis with its own 1966-Herfindhal-
Hirschman index (represented on the x-axis), for the periods 1968-1972 and 1968-1974, respectively.