What Determines M&A Legal and Financial Advisors’ Competitiveness in an International Financial Centre: Using China’s Going Out Policy as a Natural Experiment

 

 

Bryane Michael, Dariusz Wojcik, Douglas W. Arner,

Chen Lin, Wilson Tong and Simon Zhao

 

 

 

Abstract

 

Roughly 60% of all publically announced advisors to China’s “Going Out” M&A transactions from 2000 to 2014 were from international financial centres (representing over 70% of deal value). Why did advisors, located so far away from both acquirer and target, manage to dominate the M&A advisory market in the early stages of the “Going Out” policy? What can we learn from the smaller advisors located outside of these financial centres who managed to capture a growing share of this business in “Going Out’s” more recent stages? In this paper, we hypothesize the existence of a “legal complexity externality” that had the effect of increasing a financial centre’s ability to attract international business. We look at the way Going Out advisors have responded to advisory opportunities using what management theorists call “blue ocean strategy.” We show that relationships across geography changed, as large global advisors lost their share of advisory business to advisors outside of international financial centres due to the interplay of these legal complexity externalities and blue ocean strategies. As cities helps foster changes in the law governing Going Out transactions – and as financial and legal advisors adapted their strategies to compete – cities gained or lost Going Out business. We provide 5 recommendations to existing and aspiring international financial centres looking to capture a larger share of global M&A and other investment advisory business.

 

 

 

Keywords: Going Out, international law firms, global investment banks, blue ocean strategy
JEL Codes: K40, G34, N35, O53

 

 

 

The authors gratefully acknowledge the Hong Kong Research Grants Council’s Theme-based Research Scheme Project “Enhancing Hong Kong’s Future as a Leading International Financial Centre” for providing funding for this research. Nothing in this paper represents or necessarily reflects the views and opinions of funders or the institutions to which the authors are affiliated.


 

 

Table of Contents

 

Introduction. 4

The Mystery of China’s Missing Advisors. 5

Why do Chinese firms choose advisors in third-party and IFC jurisdictions?. 5

The role of specialization and an international division of advisory labour 13

Dearth of advisors in Chinese cities?. 16

What Do We Know About Going Out Companies’ Advisors?. 17

A snapshot of advisors in China’s Going Out 17

The move over time away from elite advisors. 20

How have going out advisors seized their opportunity over time?. 23

The need for new theories of advisor choice. 27

Legal Complexity and Blue Ocean Strategy. 29

All roads lead to lawyers. 29

The data on legal systems, complexity and the “legal complexity externality”. 32

Going Out’s Blue Ocean strategists. 33

Model and Econometric Analysis. 36

A model of advisory geography. 36

Legal complexity determines whether a target chooses a local advisor 40

Law schools help make advisors and financial centres. 40

Does financial theorizing help advisors compete?. 42

Differentiation helps advisors compete. 43

Policy Implications. 44

a) Focus on London and New York. 44

b) Especially for civil law jurisdictions, allow for rapid dissemination of foreign and domestic judgments into regulation  45

c) Encourage investment banks and law firms to fund local law school research, policy work and teaching. 45

d) Emerging market law firms to differentiate themselves through size and marketing. 46

Conclusion. 46

References. 47

 

Appendix I: Model 54

Overview.. 54

A model of going out firms’ choice of geographical distribution of advisory mandates. 54

Building indifference curves for Chinese clients. 58

Understanding advisors’ production possibilities. 61

Equilibrium between clients and their advisors. 63

Driving real-world like results using this simple model 67

Model conclusions. 68

Appendix II: Major Areas of Harmonization for Aspiring International Financial Centres. 69

Appendix III: Background to Statistical Analysis. 71

Description of data. 71

Dealing with missing data. 75

Regression analysis. 76


What Determines M&A Legal and Financial Advisors’ Competitiveness in an International Financial Centre: Using China’s Going Out Policy as a Natural Experiment

 

Bryane Michael, Dariusz Wojcik, Douglas W. Arner,

Chen Lin, Wilson Tong and Simon Zhao

 

Introduction

 

In 2008, the Industrial and Commercial Bank of China (ICBC) sought to buy a large stake in Standard Bank Group )(South Africa). Deutsche Bank (Germany), Simonis Storm Securities (Namibia), and Bowman Gilfillan (South Africa) acted as advisors for Standard Bank. For ICBC were Goldman Sachs (USA), ICEA Capital (Hong Kong), Linklaters (UK), King & Wood (China), and Webber Wentzel (South Africa). The consummation of the $3.3 billion deal would thus involve advisors headquartered in New York, London, Johannesburg, Beijing, Hong Kong, Frankfurt, and Windhoek. How did advisors from New York, Hong Kong, Frankfurt and Windhoek become involved in a deal involving firms headquartered thousands of miles away? What does their involvement mean for the development of the financial centres in which these advisors operate from? 

 

In this paper we look at the factors affecting legal and financial advisors to international business transactions and the cities they are headquartered. We use China’s Going Out Policy as a natural experiment. “Going Out” represents a large-scale investment flow which legal and financial advisors responded to as they sought international business. The magnitude of these flows clearly impacted significantly on advisory business of the cities involved – allowing us to detect changes in advisors’ business acquisition strategies as well as see the influence of the financial centres in which they operate. We find the operation of two factors over the 14 years we consider. A “legal complexity externality” helped advisors in places like New York and London offer complex legal and financial solutions to Chinese acquirer and foreign firm alike. To compete with the inherent advantages conferred by these externalities (and the other typical advantages already addressed elsewhere in the literature), advisors in emerging markets had to develop new and different ways of competing.[1]

 

We have organized our this working paper in six sections. The first section describes the mystery of China’s missing advisors. We document the tendency for acquirer and target companies to relatively rarely choose advisors from their own cities – instead seeking out advisors from cities which are major financial centres. We show that the usual explanations for M&A activity and advisor hiring practices cannot explain the data we observe. Such a puzzle leads us to look at various advisor characteristics. In the second section, we present data about Going Out advisors. We show how Going Out firms increasingly chose smaller advisors outside of the traditional international financial centres over time. These observations set the stage for hypothesizing a theory of “legal complexity externality” and the “blue ocean strategies” that advisors have used to compete with the large advisory firms based in New York and London.[2] The third section presents some evidence showing how such a legal complexity externality and blue ocean strategy might determine which cities Going Out companies look to in choosing advisors for financial and legal advice. We show some of the trends in simple graphs or readers can interpret the more complex econometric analysis we conduct later in the paper. The fourth section presents the model we develop, the econometric methods we use and their results. We find that financial and legal advisors perform better in cities with more complex financial law. We also find that a shift in advisors’ strategies has led to increasing advisory business going to non elite advisory firms outside of top 10 international financial centres. The fifth section provides recommendations for policymakers in large and aspiring financial centres based on our findings. In general, these lower ranked financial centres should find ways to make their financial law more robust and encourage their advisors to diversify their marketing strategies. The final section concludes, while three appendices provide technical details for the results in our paper.  

 

The Mystery of China’s Missing Advisors

 

Why do Chinese firms choose advisors in third-party and IFC jurisdictions?

 

China’s Going Out policy has led to substantial investment flows from China to other parts of the world, bringing with it substantial changes in demand across geographical regions for financial and legal advice in support of the underlying transactions.[3] Figure 1a shows the global geography of all Chinese “investments and contracts worldwide (excluding bonds)....failed and successful – valued at more than $100 million in all industries.”[4] Estimates of the amount of these investments vary depending on the definitions used and the source, and the 2014 figures probably grossly under-estimate the true value of such investment.[5] Yet, most analysts agree that these flows exceeded $70 billion in 2014. Likewise, most estimates suggest that investment going out from China will exceed investment going into China for the first time in 2015. Regardless of the amount, these flows have impacted the size and distribution of related advisory services. As shown in Figure 1b, the location of financial and legal advisors involved in Chinese companies’ Going Out M&A activity corresponds little with the actual location of such investment.[6] Both maps show the US as the largest single destination of such investment.  Interestingly, fees paid for advisory services in support of such M&A investment to financial and legal advisors corresponds to neither general investment flows nor to M&A flows. Financial flows to advisors and the geographical location they work in represents a separate trail – partially separated from investment in the “real” economy.

 

 

A number of commentators have attempted to explain the geographical separation of companies from their advisors in the Chinese context. Sutherland et al (2012) show that many Chinese companies use off-shore jurisdictions such as the British Virgin Islands (BVI) and the Cayman Islands to export capital and access finance unavailable to them in Mainland China. Given China’s restrictive lending practices (often favouring specific companies), Chinese companies will seek M&A activity in international financial and off-shore centres to obtain loans, share capital and/or the ability to buy assets like patents in other jurisdictions more easily. In such a case, the choice of advisors to such deals would not depend on their location – such M&A represents primarily a paper transaction. Others have noted that tax considerations, the ease of doing business and even investment “roundtripping” has helped make these international financial centres a focus for Chinese companies Going Out (Vlcek, 2008).[7] Most persuasively, authors like Sharman (2012) have argued that international financial centres provide “surrogate” institutions, allowing Chinese companies to avail themselves of institutions unavailable at home. In this view, Chinese companies register in New York and use New York advisors to overcome regulatory and other barriers in China.[8] In all these cases, Chinese firms require the advantages that the target companies’ legal and financial systems offer far more than the markets, technology, know-how or other resources the target country offers.[9]

 

Network analysis of advisors to China’s Going Out M&A activity shows a similar trend. Figure 2 shows a network map of the relationships between advisors to both sides of a Going Out acquisition. As shown, London and New York serve as joint hubs of financial and legal advice as funds stream from China abroad.[10] Chinese financial centres like Beijing and Shanghai serve as very small nodes in the network. The location and relationships between financial and legal advisors do not correspond to the location and relationships between acquirer and target firms. More importantly, Going Out has both solidified and created new relationships between advisors in these far-flung geographic locations. The magnitude of Going Out flows allows us to observe how investment flows affect the market for cross-border legal and financial advisory services – a perfect “natural experiment” for thinking about the theory of advisor selection and competition.[11]

 

 

 

 

 

 

 

Figure 2: Going Out Probably Affected Geographical Relationships Between Advisors and the Companies they Advised

 

the overall network with New York centric network in blue and London in purple

Hong Kong’s city partners revolve around major money centres

Cayman Islands’ partners are other major money centres

Toronto-based advisors  partner in North America and Australia (for some reason)

 

The figure shows the network connections between the cities from which China’s Going Out financial and legal advisors (or advisors to their targets) came from. We show the acquirer advisors’ city as the source and the target advisors’ cities as the target (or sink). We do not provide detailed analysis as we use these networks to illustrate how financial flows lead to a geographic dispersion of advisory relationships.

 

Existing theory fails to provide adequate explanations for the geographical dispersion of Going Out financial and legal advisors. According to existing theory, professional service firms congregate around their clients (or around each other) to take advantage of network and agglomeration externalities (Dorry, 2014). In such a scenario, we would expect to see far more clumping around established financial centres and/or acquirer/target firms than we do. New York, London and Hong Kong remain large financial and professional service centres – but far less than a gravity model might predict.[12] Hyun and Kim (2009) summarize a branch of the literature finding that firms from similar institutional backgrounds tend to “gravitate” (in terms of engaging in M&A activity) toward each other. Yet, as we will show further in this paper, no “gravity” (if we can use the term to explain flows of M&A money and funds to advisors on both sides of the transaction) appears to exist for Going Out firms. We would particularly expect to these centres grow relative to others if their policymakers become “captured”, as Sikka (2003) and others have claimed. Yet, the stable existence of a wide and far network of advisory relationships in Canada and Australia (as we will show in the next section) defy traditional explanation. 

 

What do these relationships look like? A look at the origin of many of these M&A deals explains the reliance on international financial centres. Figure 3 shows the reported value of M&A deals coming from various Chinese cities.[13] Interestingly, Chinese companies registered in BVI represent the highest average value of M&A deals from 2000 to 2014. Naturally, Beijing-based companies play in important role in these transactions. Yet, taking these data at face value, Beijing ranks 6th among Chinese cities in terms of creating high-value cross-border M&A deals. Advisors in locations like Canada and Sydney can compete with their legal and banking institutions which resemble those in the US and UK. And their specialization in natural resources may make their advice better.

 

Figure 3: China's Large Cities Conspicuously Absent as Homes to

High Value Going Out Deals

The figure shows the average value of Going Out M&A transactions for Chinese companies registered in the cities shown in the figure from 2000-2014. As shown, M&A deals are widely dispersed across China (and in several off-shore financial centres). The off-shore centres refer to Chinese companies incorporated in BVI and the Cayman Islands.

Source: Zephyr (2015).

 

Yet, even after accounting for the role that international financial centres play in China’s Going Out story and the possible “surrogate” effect of Anglo-Saxon and other countries institutions, a geographical disconnect still exists between companies and their advisors. Figure 4 shows the major advisors’ location relative to acquirer and target Going Out companies. As shown, financial centres often do not provide advice to clients located in the same metropolitan area. Target companies in the top 3 financial centres (New York, London and Hong Kong) work primarily with advisors in the same location. However, target companies in other metropolitan centres generally work with advisors from other countries. Similarly – as shown in Figure 5 – acquirers tend to work with advisors located in major financial centres. In both figures, we show only jurisdictions with a relatively large proportion of advisors – omitting to name the numerous other jurisdictions which won only one or two mandates. 

 

Figure 4: Most Target Companies Do Not Deal with Advisors

from their Same City or Country

(same city/country shown in bold)

 

Target Company’s City

City of major advisors to either acquirer or target company

London

London (46%), New York (23%)

New York

New York (80%)

Hong Kong

New York (30%), New York (20%), London (20%) Hong Kong (10%), Paris (12%), Zurich (10%),

Singapore

New York (30%), London (30%), Hong Kong (15%)

Sydney

Sydney (50%), London (41%)

Calgary

Toronto (40%)

Georgetown*

New York (20%), Hong Kong (15%), London (15%)

Hamilton*

New York (30%), London (12%), Hong Kong (12%)

Brussels

New York (65%), London (15%), Paris (15%)

Melbourne

Sydney (22%)

Kazincbarcika (HU)

London (50%), New York (30%)

Vancouver

Toronto (61%)

Perth

Sydney (70%)

Washington DC

New York (50%), London (16%)

Paris

London (37%), New York (25%)

San Francisco

New York (67%)

Milan

Milan (30%)

West Perth

London (30%), Sydney (30%)

We highlight in bold cities where acquiring companies’ or target companies’ advisors’ city matches the target companies’ city. We show the proportion of advisors to target companies in the cities shown on the left-hand side of the figure in brackets. For example, London-based advisors served as same side or opposite side advisors in transactions where the target company sits in London.  We show only “major” cities – namely cities with the largest proportions of advisors. We do not show cities with small proportions of advisors.

* represents financial centre where company registered. New York, London and Hong Kong clearly have in-depth competency in dealing with these offshore centres.

 

 

 

Figure 5: Most Acquiring Companies Worked with Advisors in the International Financial Centres Rather than Those at Home 

 

Acquirer’s City

City of major advisors to either acquirer or target company

Beijing

New York (74), London (54), Hamilton (3), Toronto (7), Zurich (20), Sydney (11), Brisbane (3), Tokyo (1), Paris (8), Frankfurt (5), Hong Kong (9), San Fran (3), Charlotte (2), Milan (2), Manchester (1) 

Guangzhou

New York (2), Sydney (2), Paris (1), San Fran (1)

Shanghai

Hamilton (1), New York (12), London (13), Zurich (5), Sydney (1), Tokyo (1), Paris (4), Frankfurt (1), San Fran (3), Charlotte (1), Milan (1)

Shenzhen

New York (14), London (5), Zurich (6), Tokyo (1), Frankfurt (1)

The figure shows the location of advisors to acquiring companies in the cities shown on the left-hand side of the figure (with the number of those advisors in brackets). For example, 54 London-based advisors served as same side or opposite side advisors in transactions where the acquiring company sits in Beijing.  We show only “major” cities – namely cities with the largest proportions of advisors. We do not show cities with small proportions of advisors.

 

These data provide three patterns which existing theories fail to explain. First, acquirers and targets rely on international financial centres far more than expected by theories suggesting advisors gravitate toward their clients, but far less than theories where advisors gravitate toward each other. Di Giovanni (2005) finds that market capitalization best explains Going Out firms’ location decisions. Yet, even a cursory glance at the data show their theory cannot be right (neither for targets nor their advisors). If Chinese companies really do look to “roundtrip” investment (using foreign acquisitions to raise money in foreign capital markets), why choose relatively low-cap Brussels or Calgary as target destinations? Why choose Paris-based advisors at all when these firms have access to New York and London-based advisors (and their capital markets)?

 

Second, they rely far more on peripheral advisors than any theory might predict. Figure 6 shows the more interesting geographical relationships emerging from our analysis. Why would London-based target companies (with their choice of an excellent array of local advisors) choose advisors in Brisbane or Oxford? Why would a Laverton-based company choose an advisor in Edinburgh? Authors like Chen and Wang (2012) note that M&A advisors play a key role in helping to bridge differences (such as in resources and operational methods) between acquirer and target. The extent to which advisors can help arbitrate these differences would determine their success in attracting Going Out clients. Yet, again, prima facie, nothing about New York or London based financial and legal advisors would suggest any special insight into energy markets, IT or the other domains that Going Out firms work tend to focus their investment activities. Certainly, a Singaporean target may not need an Amsterdam-based company to arbitrate differences between it and a Chinese company. Bala Ramasamy et al. find that Going Out firms’ ownership explains their motivations. State-owned (controlled) companies seek natural resources – often in countries with high political risks. Yet, nothing about New York or London based firms suggest they have special insight into the law and economics of these frontier markets.

 

 

 

 

Figure 6: More Unusual Advisor Locations From Target Companies

(domestic advisors listed to illustrate that depth of local advisory markets)

 

Target Company’s City

(International)

City of major advisors to either acquirer or target company

Target Company’s City

(Domestic)

Advisors’ city

London

Charlotte, Oxford, Brisbane,

Qingdao

Hangzhou

Georgetown (KY)

St. Louis, Los Angeles, Hamilton

Chengdu

Harbin

Vancouver

Sydney, Charlotte

Jining

Hefei

East Tamaki

New York,

Ningbo

Kunming

Bad Staffelstein

San Francisco

Rizhou

Ningbo

Laverton

Edinburgh

Huzhou

Qingdao

Greenback

Atlanta

Zhuhai

Wuhan

The figure shows a sample of the lower frequency matches between target companies’ cities and the cities where they advisors come from. While the odd personal relationship between company management and advisors may explain a few such unusual city matchings, the breadth and on-going new creation of such relationships suggests something structural about the market for advice sets the “long tail” relationships. 

 

Third, the extent to which advisors do work in the same city as their client suggests that the usual explanations of advisor’s role in these M&As leaves much to be desired. Sun et al. (2012) argue that China’s Going Out firms choose locations based on firms’ operations, location and internalization (the classical Dunning’s OLI model). In their view, Going Out firms should choose advisors which help them obtain factor endowments and natural resources, learn new things over time, put the firms together in a new and better way, help design their value chains and help navigate local particularities. In this way, firms need management consultants such as McKinsey more than financial or legal advisors (though we do not discuss such consulting firms in this paper and our dataset does not cover these consultants). An Amsterdam-based advisor may help a Chinese acquirer understand a Helmond-based target. But little about a Dusseldorf-based advisor suggests they can teach their Chinese counterparts about working with a Cologne-based company. We must look elsewhere for explanations.  

 

Baoteng et al. (2008) reiterate a common theme in the literature. Chinese Going Out firms – in their view – focus on increasing market share. They seek faster expansion into new markets, strive for diversification across industries and even within them, and look to obtain foreign technologies and other resources. As such, international financial centres either possess these advantages, or can connect Going Out firms to such advantages. Yet, the data do not seem to support such a conclusion. Figure 7 shows the extent to which IFC-based advisors worked on different types of transactions. The first bar on the left hand side of the figure shows non-IFC based advisors had roughly 2% more transactions involving an increase in the target firms’ capitalization from slightly above 0% to 10%. The second bar shows that IFC-based advisors handled roughly 4% more transactions than non-IFC based advisors which involved an increase in the target firms’ capitalization of 11% to 50%. The other bars report similar results. As shown by the longest downward pointing black bar, non-IFC based advisors helped acquirers buy 50% of the target firms’ shares roughly 4% of the time more often than IFC-based advisors.[14] Otherwise, these data show few differences.

 

 

The role of specialization and an international division of advisory labour

 

A fair amount of theoretical and empirical evidence suggests that advisors specialize in order to capture advisory opportunities. Graham et al. (2012) find evidence that M&A advisors which specialize in a particular industry or activity tend to generate higher returns for acquirers, close deals more quickly, charge lower fees, and offer better advice. Zhang (2012) offers a contrary perspective: she finds that apparent specialization does not provide specific advantages. Chen and Young’s (2010) research implies that advisors to Going Out firms would necessarily specialize, given the specific focus of their clients. In their view, government interests – rather than economic ones – drove Chinese M&A activity abroad. These interests focused around the strategic acquisition of raw materials and other strategic resources. If the Chen and Young view held true, we would expect to see specialization by advisors which reflect the specialization of the firms they advise.

 

Yet, the data fail to show any specific specialization. Figure 8a shows the share of M&A Going Out deals in various industries.[15] Contrary to received wisdom – at least in M&A related investment – no specialization appears. The top 5 advisors to China’s Going Out companies all worked on financial sector-related M&A – with JP Morgan leading in proportional terms. Credit Suisse and UBS had a relatively high proportional participation in materials-related M&A. Morgan Stanley and Citi had relatively high participation in industrial-related M&A. Yet, these proportions hardly point to specialization. Similarly, a clear division of labour fails to appear among the legal advisors working on the largest Going Out transactions. All legal firms advised on finance-related transactions. Linklaters and Clifford Chance had proportionally larger participation in materials M&As. Clifford Chance and White & Case had significant participation in industrial deals. Yet, no specialization appears. 

 

 

What about specialization across types of advisors? Do elite advisors (which we call Bulge Bracket banks and Big Law) have a particular hold over M&A activity in any particular industry?[16] Figure 9 shows the proportion of M&A deals in each industry handled by elite advisors (Bulge Bracket banks or Big Law as we define them in this paper). Among Bulge Bracket financial advisors, the highest proportion of deals centred on health care and utilities. The lowest proportions focused on industrials and materials. As for Big Law legal advisors (as defined in this paper), consumer staples and telecoms represented the highest proportion of their work. These data show no obvious specialization.

 

 

International financial centres also do not seem to specialize in particular industries – further eroding the specialization hypothesis we described earlier. Figure 10 shows the proportion of Going Out M&A transactions handled by financial and legal advisors located in leading IFCs (as we define them in this paper).[17] Advisors in IFCs handled about 80% of industrial-firm acquisitions for acquirers and almost half that proportion in consumer staples. For target firms, they handled about 60% of finance-related Going Out transactions and only about 40% of materials-related ones. Geographic specialization around certain industrial sectors clearly does not occur.

 

 

Dearth of advisors in Chinese cities?

 

Why do Chinese advisory firms fail to represent a larger proportion of these Going Out firms? Numerous scholars have hypothesized – either directly or by implication – that Chinese financial advisory firms and law firms have not reached a critical size or experience level to engage in such representation. Wang and Xie (2011) for example argue that the choice of advisor depends on the advisor’s size and experience in a sector. Yet, judging by domestic M&A activity, many of these advisors have significant experience. Figure 11 shows the location of China’s financial and legal advisors to domestic M&A over the Going Out period. As shown, the largest Chinese advisors (both financial and legal) handled very significant deal values. Most of the value of these transactions concentrated in Beijing or Shanghai. Moreover, advisors serve many of China’s second-tier or third-tier cities. As such, places such as Urumqi or Tianjin did not suffer from a lack of advisors. Interestingly, New York and Hong Kong based financial advisors and New York and London based legal advisors continued to play an important role in China’s domestic M&A. Yet, judging by the depth of China’s advisory markets, Chinese advisors could well have played a greater role in Going Out.  To put these numbers into perspective, Hong Kong advisors handled 1% of the roughly $400 billion in transactions value handled by these Beijing advisors. 

 

 

Why did not Chinese advisors simply copy Western practices – and why did not Western companies simply offer their services from outpost offices on the Mainland? The evidence shows that both strategies met with mixed success. As Li and Liu (2012) note, Chinese firms have made significant headway adapting foreign law firm management practices to the Chinese context. Indeed, the 2008 global financial crisis provided Chinese firms (and probably firms in other emerging markets) with the opportunity to seize market share from entrenched elite advisory firms. Stern and Li (2015) find that foreign law firms keep an office in China, but generally do not succeed there. Instead, firms compete as much as jurisdictions do – as an advisor’s jurisdiction and competitive strategy provides it with ineffable competitive advantages which it cannot replicate abroad.

 

A jurisdiction’s law probably affects it ability to compete in China (and elsewhere). Bird (2011) reviews the literature on legal advisor choice – finding that a country’s legal framework determines the competitiveness of its legal advisors. Continuing with the logic inherent in Evans and Gabel (2014), a country’s financial law provides its advisors with the “legal flexibilities” needed to adjust to complex M&A agreements (39). Michael et al. (2015b) find that the quality of legal institutions (the scope, complexity and sheer size of financial regulation) statistically significantly explains financial institutions’ competitiveness. A complex corpus of financial law provides the fodder which advisors and their clients can use to their advantage during investment contract negotiations, renegotiations and during disputes. Legal complexity of the advisor’s jurisdiction helps determine its competitive advantage internationally as much as its strategy adaptation to avoid competition (which we will describe in greater detail in the next section).  To the extent that financial law from top tier financial centres influences China, we should expect to see its advisors play a more important role in Going Out transactions.[18]

 

What Do We Know About Going Out Companies’ Advisors?

 

A snapshot of advisors in China’s Going Out

 

Going Out represents such a large opportunity for these advisors because Going Out deals represent the largest deals since 2000. Figure 12 shows deal sizes for the three largest M&A jurisdictions in the 2000-2014 period. As shown, the average value of deals managed by advisors to Chinese cross-border deals exceeded those involved in US or Japanese cross-border deals. Moreover, firms in IFCs handled transactions of one order of magnitude larger than those outside of IFCs – with IFC-based firms handling €22.3b in pre-deal value and those outside of IFCs handling €2.9b. Advisors in IFCs handled on average deals worth €920 million while those outside of IFCs €667 million.[19] Modelled fee income for advisors in IFCs equalled €4.7 million compared with a statistically significant difference of €3.6 million for firms outside IFCs. International financial centres possess something worth about €1 million more.

 

 

Which types of advisors seized the opportunities availed to them by China’s Going Out Policy? Figure 13 shows the extent to which elite advisors in international financial centres handled Going Out transactions. As shown, throughout the 2000-2014 period, IFC-based Bulge Bracket banks managed to represent Going Out firms (either acquirers or targets) most frequently. As for legal advisors, non-Big Law firms not based in international financial centres won the most mandates to represent clients on either side of the M&A transaction. As we describe later though, these data underestimate the influence of Big Law in these transactions. Many Bulge Bracket banks hired and worked with Big Law firms in international financial centres – increasing their influence and representation in these transactions. Yet, even accounting for their dominance over the M&A advisory business, local law firms used something to compete more effectively with their larger rivals.

 

Figure 13: IFC-based Bulge Bracket Banks and Non-IFC, Non-Big Law Firms

Advised Going Out Firms

The figure shows the headcount proportions (not weighted by transaction size) of advisors to China’s Going Out companies for the period 2000-2014. We show the extent to which those advisors represent Bulge Bracket banks or Big Law firms (as we define them in the paper) and whether these firms come from top 10 international financial centres (also as we define them in this paper).

Source: Zephyr (2015).

 

Which advisors specifically benefitted the most? Figure 14a shows the added revenue for financial advisors from large Chinese cross-border M&A activity from 2000 to 2014. As shown, Morgan Stanley has profited the most – dealing with almost €50 billion worth of deals. Of the 20 out of the 32 transactions we could obtain data for, the firm earned more than €111 million in modelled advisory fees during the period.[20] Yet, if the company earned the traditional 1%-2% of deal value, then such earnings (before expenses) could come to €500 million to €1 billion. Yet, the geography of these mandates shows the importance of international financial centres in China’s Going Out Policy. As shown in Figure 14a, roughly half of Morgan Stanley’s mandates centres around international financial centres. Following the usual power law distribution, we see that advisors in the bottom are involved with deal sizes one order of magnitude less than those at the top.

 

 

 

The distribution of deal sizes among these top 10 firms raises the same questions posed earlier. Why do we observe so much clustering of activity among firms based in international financial centres? Why do certain firms not capture more work than others? Even without quantitative analysis, we can surmise that financial law in certain jurisdictions clearly provides part of the explanation. We can also surmise that no firm completely dominates the Going Out advisory trade because of competitive reactions/responses by other advisors.[21] The distribution of deal sizes across advisor geographies already tells us much about the contribution of financial law and rivalry among advisors. Even though Going Out firms provide completely new markets for these advisors, these companies demonstrated inherent abilities to differentiate, alter strategies in the face of competition and use their jurisdictions’ law to their advantage. As we will see, other, smaller firms (outside the Top 10) adapted in Going Out’s later periods.

 

The move over time away from elite advisors 

 

Competition among advisors during the Going Out period (so far) has shown a shift away from elite advisors based in IFCs, and toward other advisors. Figure 15 shows this shift in four key statistics. First, the percent of advisors in the target companies’ own city rose more during the later part of Going Out than in the earlier part. In the early part, the target firms worked with advisors based in their city about 25% of the time. In the later part of Going Out, this percent increased to about 40%. Second, Going Out firms on both sides of the deal (acquirers and targets) chose advisors in IFCs almost three-quarters of the time. By the late Going Out period, this proportion fell to about 55%. Going Out companies chose elite advisors about half the time during the early Going Out period, and about 30% of the time during the later Going Out period. Advisors’ client mix also shifted, with private clients rising from about 40% to 60% over these two periods. The client mix also shifted. In the early days of Going Out, these non-elite advisors only represented about 30% of targets. By the later period, they represented 44% of targets. Going Out still mainly consisted of SOE acquirers with a preference for IFC-based elite advisors. But the market for advisory services has slowed shifted. Taken together, these data suggest that the market for M&A advisory shifted to smaller advisors outside of the major IFCs.

 

 

A year-by-year analysis shows these trends in greater detail. Figure 16 shows the trends described in the previous figure in more detail. In general, acquirers have increasingly sought to acquire firms in their same industry (up from about 25% of all transactions to around 65%). Advisors to these transactions based in IFCs fell from about 75% of all transactions over the period to about 50%. Advisors from the target companies’ own cities have managed to obtain mandates slightly more frequently over the period. The percent of SOE clients has fallen slightly. All together, these data suggest other advisory centres managed to bolster their competitiveness and adapt to the exigencies of winning Going Out mandates. 

 

 

The picture of such an evolution depends on whether one looks at markets for financial advisor or legal advice. In general, Bulge Bracket banks marginally tightened their grip on clients of all kinds, whereas smaller law firms significantly attracted both SOE and private clients away from Big Law. Figure 17 shows the change in market share of Bulge Bracket banks and Big Law, broken down over the 2000 to 2014 period by SOE and private company clients. Bulge Bracket banks managed to increase their market share of SOE clients, while increasing their share of private clients by 9%. In contrast, non Big-Law law firms increased their share of SOE clients by 22% and of private clients by 41%. Such a picture confirms the previous finding that Bulge Bracket banks have found ways to increase their competitiveness over this time as have smaller law firms. As we show later, these data support the hypothesis of rapid legal change (and advisor competitiveness) in most jurisdictions.

 

 

Maybe lower fees have driven these changes? If advisors responded by changing fees, then fee changes (and not deep-seated competitive factors) would determine which advisors won Going Out mandates. Yet, as shown in Figure 18, in our sample, Bulge Bracket banks earned an average of €6.4 million.[22] In contrast, their colleagues earned only a statistically significantly different €3.8 million. Big Law earned €4.4 million in modelled fees as opposed to a non-statistically significantly difference of €3.9 million for their colleagues. Bulge Bracket banks worked on €1.3b in deals, as opposed to their colleagues of a statistically significant €640 million. Big Law worked on deals worth €835 million as opposed to a non-statistically significantly different €690 million. Neither type of advisor dealt with pre-merger total assets values of any statistically significant difference.

 

 

How have Going Out advisors increased their competitiveness over time?

 

The top Going Out advisors have managed to grow their deal sizes more slowly than their colleagues working on other deals. Figure 19 shows the way that the top financial advisors’ deal sizes have changed with changes in deal sizes for the top 10 advisors in general. Simply put, the figure measures the extent to which the top financial advisors’ deal sizes have grown in line with their peers. The top financial advisors’ global deal sizes outside China grew faster than those working on China-related deals in the later Going Out period. Since 2005, the top 10 advisors to Going Out M&A deals did not grow deal sizes in line with all advisors (as shown by elasticities less than 1). In other words, smaller advisors grew their deal sizes more than the top 10 advisors did – suggesting that something about their financial law and competitive strategy made them more competitive.       

 

 

Even among top 10 advisors, we observe significant deconcentration of deal values – as the most successful advisors managed to grow deal sizes less than their colleagues. Figure 20 shows the proportion of Going Out deals won by the leading financial advisor and legal advisor each year – compared with colleagues working on all deals globally. Top 10 financial advisors on Going Out deals managed to capture 4% less of the total proportion of deal values than their peers working on global deals. Legal advisors on Going Out deals captured 8% less of the total market than their peers working on global deals. These reflect broader trends whereby the proportion of the top Going Out financial advisor was close to 80 times the deal values of the tenth place advisor, whereas the top Going Out legal advisor’s deal sizes exceed 110 times the value of the 10th place advisor. These proportions did not come close to global top financial advisors’ deal sizes of almost 140 times the 10th place advisor and the top global legal advisor’s deal sizes coming in at only about 40 times the 10th place advisor. Such data suggest that top advisors have monopolized less cross-border investment than their peers working on other deals. Smaller firms have acted more competitively.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 20: The Top Financial and Legal Advisors to China’s Going Out Companies Handled Less than Their Fair Share of Deal Values over the Years

The figure shows the deal values handled by top 10 financial and legal advisors on Going Out deals and all deals world-wide. We show how the proportion of deal values won by top 10 advisors relative to the total deal values won by all advisors has changed over time. These proportions saw a spike in 2012.  

Source: Zephyr (2015).

 

 

Patterns of mandate acquisition among the top 10 advisors paint a similar picture of the way that advisors’ home city legal systems and their own competitiveness have driven competition. Figure 21a shows the elasticity of the first place advisor on Going Out deals (the advisor winning the largest value of deals) compared with total deal values world-wide from 2000 to 2014. As shown by the left-most bars, deal sizes for the leading financial advisor grew roughly 4 times faster than those of the entire top 10 group of advisors. The 10th place financial advisor to Going Out deals did not manage to keep up with its counterpart working with all deals globally. Yet, this 10th place financial advisor managed to increase deal sizes faster than the top 10 in general. Going Out legal advisors (both in first place and tenth place) did not manage to keep up with expanding deal values for the top 10 overall. Figure 21b shows the way that top first place and tenth place financial and legal advisors’ deal sizes changed over time. Financial advisors’ at the top and bottom of the list tended to grow their deal sizes less quickly than the entire top 10. First place legal advisors’ deal sizes also grew less quickly than the top 10. Yet, the tenth place legal advisors’ deal sizes grew faster than the total for the top 10. Legal advisors all the way down the list tended to become more competitive over time. Top tier financial advisors managed to stay competitive.

 

 

What do these data mean for the financial centres which host these advisors? Figure 22 shows the ranking of financial centres in our sample. Rankings among financial advisors change far more fluidly than for legal advisors. London-based advisors may rank from 5th place in 2004 (by deal size) to 10th place in 2006 – and all the way to 1st place in 2014. Zurich on the other hand, shows a steady deterioration in rank (likely the result of Credit Suisse and UBS shrinking their investment banking operations after the 2008 global financial crisis). Nevertheless, the constant appearance of these cities in the top 10 show that something about these jurisdictions – besides simple strategic actions of its financial advisors – account for their high ranking. We argue later that London’s, New York’s, Zurich’s and Paris’ evolving financial law helped them in the competition for Going Out mandates.

 

The figure shows the change in city rankings for “Going Out” financial advisors and legal advisors from 2000-2014. We ranked advisors from these cities by deal value and averaged these ranks when advisors from the same city appeared in the top 10 more than once. The lack of cities for legal advisors (we only list 2 cities) shows how completely legal advisors in London, New York, and legal advisors without a main headquarters dominate the rankings. “No centre” means that advisors had no identifiable headquarters city – both according to our dataset and our own research into the advisors’ corporate structure.

Source: based on Zephyr (2015).

 

The stability of London and New York as top legal advisory hubs show how completely the complexity of their financial law and their own competitive actions have helped them to dominate the Going Out market. For both financial and legal advisory, “no centre” represents an important jurisdiction. “No centre” financial firms and law firms have organized themselves in a way which defies any attempt to categorize them as having a headquarters in any one city (or group of cities). Globally headquarter-less advisory firms have the perfect choice of financial law and can apply laws and strategies as needed to win mandates.

 

The need for new theories of advisor choice

 

Even a cursory look at existing theories of advisor choice show that these theories prove inadequate in explaining about Going Out investment. For authors like Agrawal et al. (2014), choice of advisor reflects “reputation effects.” In their view, parties to an M&A transaction often – and ill-advisedly – chose advisors based on market share or reputation rather than past performance. Yet, as we have shown in Figure 9 and Figure 10, the volatility of these advisors’ shares suggests that simple attraction based on market size or reputation cannot explain the competitiveness of Going Out firms’ advisors. Even theories offered in studies very close to ours fail to gain any traction. De Jong et al. (2010) look at almost exactly the same market we do. Using a database similar to our own, they look at Going Out firms’ choice of various kinds of financial and legal advisors and the returns the merger generated. They argue that the advisor’s global experience and its experience in target markets makes up for the lack of the acquirer’s own experience. They also find that advisors from the target’s or acquirer’s country create far more value than advisors from third-party countries. Yet, we have already illustrated the complete lack of relationship between Going Out firms’ location, the location of their advisors and any supposed competencies and specializations these advisors should have. Their client-centric theory of advisor value leaves something out.

 

Indeed, all the theories of advisor choice fail to explain the structure of the Going Out advisory market we have illustrated previously. The evidence even fails to explain the most obvious decision criteria – that advisor choice depends on actual or expected profitability to the acquirer and/or target companies.[23] Figure 23 provides an overview of the theories we have discussed thus far, either directly or indirectly. Certain advisors may have better completion rates – but what explains the probability of completion? Why do Credit Suisse and Clifford Chance complete more deals than local firms? We already showed that expertise probably does not matter – as rivals can easily copy such expertise and a base in New York hardly explains expertise in Mongolian mining (or other areas). Nothing about these data suggest a home bias or the importance of a fixed distance from the client for transaction success. As for the other theories, little explains why certain firms may become socially embedded with each other or their clients.  

 

Figure 23: Strands of Literature Looking at the Way M&A Activity Impacts on Advisory Demand

 

What determines choice of advisor?

Description

Completion rate

Krishnan and Masulis (2013). De Mong et al. (2011).

Top firms have better completion rates and have higher engagement rates for complex cases. Yet, advisors with industry-specific knowledge may reduce such completion rates.

Expertise matters

Klasa et al. (2013), Chang et al. (2015), Wang et al. (2014).

Expertise increases an advisor’s chances of working on an engagement. However, concerns about information leakage may encourage firms to seek advisors outside the sector. Experience adds value by supplanting firms’ lack of expertise in an area.

Specialization

Graham et al. (2012),

Separate – though closely correlated with expertise – specialization allows advisors a breadth of relationships and scale to complete transactions. Often couched in market share terms.

Advisors’ effect on profits 

Bao and Edmans (2011), Hunter and Jagtiani (2002).

Many authors find the lack of correlation between choice of advisor and profits. Switching advisors will often have a positive effect on profits, suggesting a balancing relationship effect and a competition effect.

Social embeddedness and home market bias

Chang et al. (2012), Francis et al. (2012a; 2012b), Sibilkov and McConnell (2014).

Firms may (or may not) hire based on own previous deals, firms may hire advisor to counterpart’s previous deals. Allocation of mandates to advisors depends on overall embeddedness in Boards and other relations. Francis and colleagues find a home-market effect.

Reputation effects

Agrawal et al. (2014)

Parties to an M&A transaction may hire advisors to capture a greater share of the returns to union. They often – and stupidly – chose advisors based on market share rather than past performance.

Deal size and complexity

Kurkela (2014)

Larger and more complex transactions require more specialised advisors. Yet, Kurkela (2014) finds that “boutique” advisors less likely for large, complex acquisitions.

Synergy effect

Laux (2001), Flood (2009)

Firms engage elite legal advisors to off-set value destroying aspects of elite financial advisors’ advice.

Distance

Karsten et al. (2014), Wang and Xie (2011).

Firms do not choose based on location. Expertise often relates to distance. A trade-off exists between local knowledge and global scale (Wang and Xie).

 

The previous studies ignore some deeper factor that drives much of the theoretical evolution so for. Interestingly, De Jong et al. (2010) pick up these deep factors peripherally. They find that attributes of the financial and legal systems in the countries from which the target and acquirer come (like financial depth, extent of investor protection and other factors) affect their choice of advisor. Once we take the usefulness (or complexity) of advisors’ legal systems and advisor strategic differentiation into account, the other theories make more sense.  Companies would choose certain companies based on deal size – if these advisors’ financial law helped them to deal with such sizes better than their rivals. Differences in law and differentiation can drive social embeddedness and investments in certain kinds of industry expertise. To understand the literature on advisor choice (and the way those advisors structure their relationships), we must understand those deeper factors.

 

Legal Complexity and Blue Ocean Strategy

 

All roads lead to lawyers

 

Legal complexity plays a far more important role in international and Going Out M&A than apparent at first glance. The pivotal role of legal advisors in the M&A industry structure shows why. Even small growth in international M&A requires exponential growth in legal advice – and lawyers. Consider our sample of Going Out firms. For the 578 transactions conducted from 2000 to 2014, each transaction would have required at least six different lawyers with a different client. The buyer, seller, and vendor (the organization selling part or all of a company) each would retain their own law firms. Yet, the financial advisors on these transactions will also usually retain their own counsel. As such, a transaction may involve lawyers for the 6 parties – buyer, seller, vendor, buyer’s financial advisor, seller’s financial advisor, and vendor’s financial advisor. Increased demands placed on in-house counsel to the three parties may also incrementally add demand for lawyers. Figure 24 shows the counsel retained by major financial advisors to Going Out transactions. 

 

Figure 24: Advisors to the Advisors Based in the Big Three

International Financial Centres

(advisor in red rows and advisors to the advisors in black columns, with the city where the advisors’ advisor works in gray columns)

 

Citigroup Inc.

AZB & Partners

N/A

Shearman & Sterling LLP

International

Skadden Arps Slate Meagher & Flom LLP

London

Credit Suisse

AZB & Partners

N/A

Freshfields Bruckhaus Deringer

International

Deutsche Bank AG

Clifford Chance LLP

New York

Linklaters

N/A

Skadden Arps Slate Meagher & Flom LLP

London

Slaughter and May

Hong Kong

JP Morgan (includes JPM Cazenove)

Appleby Global

Cayman Islands

Paul Hastings Janofsky & Walker LLP

Hong Kong

Simmons & Simmons Ltd

International

Slaughter and May

Hong Kong

Skadden Arps Slate Meagher & Flom LLP

London

Morgan Stanley

Freshfields Bruckhaus Deringer

International

Jones Day

International

Shearman & Sterling LLP

Palo Alto

Slaughter and May

Hong Kong

The figure shows examples “advisors of advisors” for Going Out transactions. We chose the financial and legal advisors with the highest numbers of advisors for illustrative purposes.  For example, Morgan Stanley retained Freshfields, Jones Day, Shearman & Sterling, and Slaughter and May as advisors to them directly (rather than Morgan Stanley’s client firm). 

 

The demand for so many lawyers demonstrates an obvious revealed preference for legal complexity.[24] Consider the case of 6 legal advisors on a transaction, the firms and other advisors involved would engage so many separate law firms unless they saw a value in bringing advisors who could advise on the law for at least 4 jurisdictions and a range of specialist law.[25] Clearly, demand for lawyers increases exponentially. For a 100 case increase, the number of law firm engagements could engage by 600 or more.[26] In 2010, the number of advisors retained in top 10 international financial centres (as we have previously defined them) doubled those outside these centres. By 2014, the proportion of legal advisors hired outside IFCs roughly equalled those located inside them. For this reason, the number of legal advisors retained each year has steadily increased – and will increase faster than financial advisors. As more and more lawyers create cases, provide advice on banking and securities regulation and are involved in the creation of ever more related law and regulation, the ever “complexifying” legal centres create ever more complex financial law.  

 

Other theorists have not failed to observe the pivotal role financial law plays in explaining the stock/flow of investment as well as the stock/flow of advisors. Bottazzi et al. (2009) for example find that “investor’s legal system is more important than that of the company in determining investor behavior.” Van Criekingen et al. (2005) find that law firms provide a “glue” which binds global cities together. Flood (2007) provides an overview of the reasons why such “glue” exists. Figure 25 summarizes the reasons why international financial centres like New York and London have housed large global law firms – and other jurisdictions’ reliance on their law. Large New York and London based law firms have typically served to make the use of law as important a part of the deal as the use of financial and business analysis.[27] Clearly, these law firms generate strategic value for firms – otherwise why pay enormous fees to such a large number of separate firms?[28]

 

Figure 25: Reasons Why Big Law Has Traditionally Advised on Many Large Investment (M&A) Deals and Why That May Change

 

Factor

Description

Effects Pulling for Big Law

“Big Law” for Complex Financial Legislation

Large law firms have organizational advantages in digesting a body of voluminous and complex law. As such, the size and complexity of large themselves comes to reflect the size and complexity of the law they work with.

“Big Law effect”

Large international law firms get business because they are large international firms. They provide stability and depth for risk-averse managers.

Tie in with Bulge Bracket banks

Stable relationships with the large international banks make finding clients far easier.

Dominance of Anglo-Saxon Law

Anglo-Saxon law dominates contractual terms, arbitration and other aspects of a transaction. Many contractual elements (such as interest rates and commodity prices) set in Anglo-Saxon countries.

Mini-UN

effect

Big Law attracts talent from a range of countries and lawyers will circulate across jurisdictions. Given access to Bulgarian or Indonesian talent in the law firms’ IFC-based headquarters, clients have little need to shop for lawyers in Sofia or Jakarta.

Document standardisation allows for economies of scale

Transactions documentation often follows similar patterns. As such, law firms which pay the “fix costs” during a few large transactions can take on more clients with relatively low marginal costs.

Competencies in dealing with cases hard to replicate

International law firms create internal cultures and procedures for dealing with large cross-border transactions which smaller firms find difficult to replicate.

Factors shifting away from Big Law

Robo-Counsel and Deprofess-ionalisation

Greater use of IT and computer programmes to complete low-value paperwork reduces Big Law’s cost advantages. Many jurisdictions relaxing rules on allowing non-lawyers to do non-legal tasks (like prepare paperwork and conduct financial analysis).

Capital market finance

Unlike in the US and UK, law firms in many jurisdictions can raise money from sources other than partners and banks (such as stock markets).

Global legislative harmonisation without wage harmonization

Many countries have adopted many of the provisions found in Anglo-Saxon law (see Appendix II for major areas). Lawyers working with these provisions may hail from urban areas with significantly lower wages/costs.

New Clientele

Large BRICS and other companies may have less awe for big Bulge Bracket banks and the law firms they bring in-tow.

Increasing professional regulation of lawyers

Many jurisdictions “imported” the voluminous self-regulation imposed of lawyers when they hire lawyers from IFCs. By developing their own lengthy standards, emerging markets make IFC lawyer’s professional self-regulation less applicable.

Sources: Flood (2007) and Ribstein (2010) with reinterpretations to fit the framework of our paper.

 

The data on legal systems, complexity and the “legal complexity externality”

 

Clearly, the complex nature of financial law in these IFCs helps increase demand for advisors – both financial and legal. As Dubai, Shanghai and other centres’ policymakers seek to bolster demand for professional services, they have passed voluminous swathes of new financial law (Arner, 2009). Such additions merge complex judge-made and regulator-made law which has filtered into legislative instruments over decades in the US and UK (Barton, 2011; Beaverstock et al., 1999).  The data also suggest (though no more than suggest) that international financial centres have more complex financial law than other places. Participants in M&As and other transactions avail themselves of such law as “standardized contractual terms” – as a way to reduce transactions costs involved in negotiating contracts and paying for unexpected outcomes (Druzin, 2009). The comparative institutionalists such as Beck et al. (2003) showed that Anglo-Saxon financial law adjusts more quickly to changing exigencies of international finance. Thus, international market actors will choose the law that provides them with the greatest amenity. The international financial centres provide such law.[29]

 

More complex financial law generally corresponds with more attractive financial centres. Figure 26 shows two measures of financial law. The TMF legal complexity measure relies on surveys of company securities in different countries and other officers/experts (TMF Group, 2014). The Michael et al. (2014a) measure uses a complex statistical procedure known as multidimensional scaling for various provisions in countries’ banking and securities law to look for similarities and differences. In theory, as countries start to lie toward 1 on the measure, they have increasingly complex financial law. Clearly a pattern exists between groups of countries. The jurisdictions attracting significant proportions of Going Out deals have significantly more complex financial law than the jurisdiction where target companies often locate.

 

 

Could a positive “legal complexity externality” exist for advisors based in international financial centres? Such a positive externality – if it exists – would provide advisors in a jurisdiction with a competitive advantage.[30] Legal and financial advisors could reliably use such law to generate producer and consumer surplus for their clients. Competitors in other jurisdictions would need to find ways of competing – even if by copying and amending existing “best practice” financial laws.[31] A statistically significantly positive relationship between a jurisdiction’s legal complexity and advisors’ ability in that jurisdiction to compete for deals would represent the first step toward thinking through such an externality.

 

Going Out’s Blue Ocean strategists

 

How do financial and legal advisors adjust their marketing and other service offering strategies to their rivals? When Clifford Chance (for example) wins several large Going Out deals, how do rival law firms in China and Germany react? What differentiates Clifford Chance from Bonelli Erede Pappalardo? Taxonomies like Mayson’s (2007) for law and Smith and Walter (2003) for investment banks prove abstract, subjective and difficult to apply across groups of companies. Quantitative methods of identifying advisors’ strategy look like the best way of isolating differences in competitive strategy of advisors competing for Going Out business – even if we do not know what those differences are.[32]

 

Elite advisors in international financial centres react to Going Out opportunities differently than smaller advisors outside of the centres. Figure 27 shows the classification of Going Out advisors – based on the deal sizes these advisors won, the cities they work from, the number of other advisors they work with, the industries they work in, and other variables.[33] Advisors differ according to two constructed variables which we label as “competitive factor x” and “competitive factor y.” We see that non-IFC-based advisors differ along the x-variable. We do not know exactly how without in-depth qualitative analysis. We only know that we can quantify differences in these advisors’ reactions to Going Out transactions. Figure 28 shows the identities and scores of several of the smaller advisors shown on the right-hand side of Figure 27. We speculate that these advisors choose different strategies because they do not come from IFCs. Following Mauborgne and Kim’s (2005) notion of Blue Ocean Strategy – these advisors seek to compete in areas without intense competition.[34]

 

 

 

 

 

 

 

 

 

 

 

 

Figure 28: Examples of Blue Ocean Advisors to Going Out Companies

 

Advisor name

Type

Scaling

 

 Advisor name

Type

 Scaling

YES Bank Ltd

FA

5.77

 

Shin & Kim

LA

2.57

Lefosse Advogados

LA

5.30

 

Four Partners Advisory SIM SpA

LA

2.56

Mattos Filho Veiga Filho Marrey Jr e Quiroga Advogados SC

LA

5.29

 

Bonelli Erede Pappalardo

LA

2.52

Ping An Securities Co., Ltd

FA

4.63

 

Negri Clementi Studio Legale Associaton

LA

2.49

Standard Bank Group Ltd

FA

4.63

 

Bonelli Erede Pappalardo

LA

2.42

Qinisele Resources (Pty) Ltd

FA

4.59

 

NCTM Studio Legale Associato

LA

2.39

Guosen Securities Co., Ltd

FA

4.57

 

King & Wood

LA

2.31

Bowman Gilfillan

LA

4.55

 

Haitong Securities Co., Ltd

FA

2.22

Deneys Reitz Inc.

LA

4.46

 

Guotai Junan Securities Co., Ltd

FA

2.12

Deutsche Securities (SA) (Pty) Ltd

FA

4.38

 

King & Wood

LA

1.91

KPMG Corporate Finance

FA

4.25

 

Gide Loyrette Nouel AARPI

LA

1.85

UBS

FA

2.80

 

ABN AMRO Bank NV

FA

1.78

WOLF THEISS

FA

2.78

 

Flichy Grange Avocats

LA

1.78

Credit Suisse

 

2.71

 

China International Capital Corporation Ltd

FA

1.73

Unicredit SpA

FA

2.59

 

BNP Paribas SA

FA

1.72

 

Analysis of advisors’ jurisdictions also shows clustering in the way that IFC-based advisors and non-IFC-based advisors behave. Figure 29 shows the same factors shown in Figure 27, but we list the advisors’ home jurisdiction, instead of name (and we flip the factors as x and y have little meaning in this context). As shown, advisors working from different cities react in different ways.

 

 

Our analysis though disabuses the notion that advisors act in similar ways – namely that IFC-based advisors may act differently than non-IFC ones. We certainly observe generalized clustering within these groups. However, we observe numerous occasions where statistical analysis identifies an elite advisor acting in particular years like non-elite advisors. Credit Suisse and Norton Rose have years where their actions are statistically indistinguishable from those of non-IFC based advisors. New York and London appear in some years lumped together with smaller jurisdictions. Grouping advisors and financial centres into neat, tidy clusters provides a useful heuristic for theorizing. Yet, the real-world data show that the two groups likely compete in each others’ spaces.

 

Our analysis also expands upon the idea of the “geography of finance.” In the classical view, natural geography and the diversity of institutions (such as law and regulation), determines how money and advice flows as measured in latitude and longitude (Clark and Wojcik, 2007). We can measure the geography of Going Out advisory in “strategic space.” Every point in such a strategic space corresponds to real advisors in real locations. Yet, using “maps” of legal complexity and/or advisors’ strategies can help us observe more clearly how Going Out (and other) deal sizes change by advisor and advisors’ location.

 

Model and Econometric Analysis

 

A model of advisory geography

 

How do changes in advisor’s competitiveness affect the distribution of Going Out mandates across financial and legal advisors? Figure 30 shows a representation of the probability of receiving a Going Out mandate across geographical space. New York and London based advisors have the highest probability of receiving a Going Out mandate.[35] The other locations indicated have lower probabilities – with areas in very dark gray having the lowest (but still positive) probability. The darkest gray areas also represent the areas which would benefit most from changing factors which affect their likelihood of winning Going Out mandates. In this view of geography, only advisors in the places indicated on the map have a chance to change their luck at attracting Going Out mandates. The marked jurisdictions serve as “ruts” or fixed points – and the “geography of finance” consists of understanding the factors driving the importance of each area.[36] We refer to our model as a discrete probability model because in-between places attract no probability of scoring advisory work.

 

 

As discussed in the paper, several factors determine the probability (frequency) of advisors from a certain jurisdiction to receive Going Out mandates. To illustrate our model’s logic, we show in Figure 31 the equation-of-best-fit between the variables we discussed in the paper and the probabilities we just described. Two points stand out. First, we can describe each probability as a function of the factors we discussed in our paper. Second, at least two of these factors maximize these probabilities at a certain point. In the case of legal complexity, when legal complexity for a legal jurisdiction falls below or exceeds 0.83, we can improve these probabilities by moving closer to 0.83.[37] As we show in the equations below the figure, we can describe these probabilities as a combination of these factors.

 


 

Figure 31: Probability of Attracting Going Out Mandates Depends on the Value of Factors We Described in this Paper

 

The figure shows the lines-of-best-fit for the probabilities of various cities’ advisors receiving Going Out mandates. We show in the table the bivariate relationships (taking only one factor into account). The equations below the table show the “equation” for such a probability, taking into account the “breaks” in 2 of the relationships we explored. The final statement with the arrow describes the way that these probabilities must be translated into each city’s advisors. We take 4-th power roots of these equations because we show each equation individually results in the city’s probability of snagging a Going Out mandate. If we multiply all four factors together, we need to “un-multiply” them to get back to the original final probability of a city’s advisors winning a Going Out mandate.

 

Our model of competition between jurisdictions in which Going Out companies’ financial and legal advisors work matches the two major factors we have described in this paper. Figure 32 shows the way these two factors interact.[38] Advisors, particularly in well-established IFCs benefit from competitive advantages gained through the legal complexity of their headquarters jurisdiction (and to a lesser extent complexity from other jurisdictions they work out of). For jurisdictions without this legal complexity, their financial and legal advisors must differentiate. We show New York and Zurich (Switzerland) as two examples. Cities have a production possibilities frontier (which we show as cities’ PPF). Chinese Going Out companies assign implicit value to each jurisdiction’s legal complexity and its advisors’ differentiation – as shown by “indifference curves.” These curves tell the trade-off between legal complexity and differentiation needed to achieve a particular probability of winning Going Out mandates. The “higher” the indifference curve, the higher the probability of winning mandates. Advisors in cities like Hamilton or St. Helier need to modify their jurisdictions’ legal complexity, their own differentiation (and as we will shortly show, their law schools’ quality) to maximize their probability of winning mandates. Obviously, points in “legal complexity-differentiation space” shown in Figure 32 correspond to actual city locations. The real-world geography of Going Out advisors’ mandate volumes corresponds to the mathematical geography of competitiveness in Cartesian space.  

 

 

So how do the variables we discussed result in a jurisdiction’s competitiveness (probability of winning mandates)? Figure 32 above shows a system-wide view of such competitiveness. Figure 33 shows the way advisors in particular jurisdictions can enhance their competitiveness (probability of attracting Going Out mandates).  A jurisdiction’s probability of winning mandates rises with its legal complexity (up to a point). Advisors in a jurisdiction differentiate because they have to in order to win mandates. Improvements in law school quality (or the quality of specific law faculties, as ranked by QS World University rankings) increases the probability of winning mandates – as lawyers and financial advisors use local law better. Such a change corresponds with less needed differentiation. Shrinking mandate levels correspond with lower probabilities of winning a mandate and lower levels of level complexity needed to deal with a liquid market in mandates. 

 

 

 

Assuming our model accurately reflects reality, changes in our variables of interest will “redistribute” advisors’ probability of winning Going Out mandates across cities. Figure 34 shows the new distribution in probabilities assuming that currently low probability jurisdictions engage in a 1 standard deviation change in their law school scores, legal complexity and differentiation. London and New York lose mandates very quickly because of the backward bending nature of the probabilities we discussed earlier and because of the finite amount of probability of winning mandates which these cities share. The lowest probability jurisdictions’ probabilities of winning mandates rise quickly – though they remain low relative to New York and London’s previous probabilities. Any one city that can improve the competitiveness of its advisors will experience a very sudden rise in demand for its legal and financial advisors.

 

 

 

Legal complexity determines whether a target chooses a local advisor

 

Legal complexity statistically significantly correlates with the probably that a target company will choose an advisor in its own city. Figure 35 shows the range of parameter estimates from the regression analysis we discuss in the Appendix. As shown, an increase of a 1 unit change in the TML legal complexity score (of .1) increases the probability of choosing the target firm’s own city by 5.7%.[39] Cities with more complexity take business away from these cities. As firms choose advisors in jurisdictions with more legal complexity, changes in these leading jurisdictions’ legal complexity might result in a 6% decrease in the chance of local advisors getting the M&A mandate. As cities go up in Y/Zen rating scores, the probability of using a local advisor decreases 3%. Taken together, these results show the role that both legal complexity and IFC “desirability” play in attracting business to advisors.

 

 

These results also highlight other aspects of competition among elite advisors. Companies which expand in their own industry show no statistically significant preference for advisors. Deal values also play little role in the decision to choose a local or foreign advisor. As we shall see in a moment, the quality of the city’s law school(s) may determine the extent to which its advisors of both types (legal and financial) attract clients.

 

Law schools help enhance the competitiveness of  advisors and financial centres

 

What determines the probability of advisory work going to international financial centres? Figure 36 shows the impact of several variables on Going Out firms’ decision to hire elite advisors. As we showed previously, Going Out firms choose elite advisors (Bulge Bracket banks and Big Law firms) located in international financial centres. Country size has a negative influence – likely reflecting the advisor’s costs and other opportunities.[40]  The quality of the law school where the advisor sits also has a statistically significant relationship on whether their clients choose an elite advisory firm. As the quality of other centre’s law schools increases, the probability that companies will choose an advisor based outside a top IFC increases. In other words, if smaller jurisdictions’ advisors want to attract customers like Going Out firms, they need to work with government to improve the quality of local law firms.[41] Unsurprisingly, larger deals tend to go to IFCs. Interestingly, the quality of a jurisdiction’s finance school (finance or accounting department in the local university) does not have a statistically significant relationship with advisor choice.

 

 

Similar patterns unsurprisingly exist in the way that Going Out firms engage advisors headquartered in top 10 international financial centres (IFCs). Figure 37 shows several of the variables we analysed when looking at Going Out firms’ choice to engage advisors in top tier IFCs. As shown, they choose these advisors when the IFC also serves as the home of the target company. Like the previous regression, companies choose advisors in a top tier IFC when the law school quality of their advisors falls – again suggesting that raising law school scores serves as a way of increasing demand for that school’s local legal and financial advisors. Complex financial regulations also draw clients to IFC-based advisors. As the GDP of an advisor falls, the chance of customers choosing IFC-based advisors rises. 

 

Both regressions confirm anecdotal and other evidence suggesting that the quality of a jurisdiction’s economics and finance teaching/research does not spill-over into demand for advisors from that jurisdiction (Kim et al. 2009; Lewis, 2010). Economics faculties appear to create little usable benefit which advisors (either legal or financial) can use to attract Going Out clients. This analysis also confounds another piece of folk wisdom – that elite advisors in top IFCs handle larger deals. Nothing in this analysis suggests that elite advisors in IFCs have any kind of competitive advantage in winning larger deals.

 

Does financial theorizing help advisors compete?

 

What value does financial theorizing have – if not to help advisors win clients? The data suggests that financial theorizing by academics in financial centres may help advisors form the coalitions needed to successfully complete larger deals. Figure 38 shows the relationship between the number of reported advisors to various Going Out deals, and the variables we analysed for this study.[42] As shown, three variables proved significant in our analysis of the number of advisors working on Going Out deals. As deal values increased, so did the number of advisors on both sides of the deal. Such a relationship might suggest that larger deals needed more advisors to complete. As the score of the finance department of the university of the city in which advisors worked rises, the number of advisors also rises. Such a relationship suggests that finance research and teaching helps advisors coordinate and/or cooperate somehow.[43] Advisors from lower scoring IFCs (as ranked by Z/Yen) tend to work with more advisors. Such a finding suggests (and only suggests) that advisors from higher ranked jurisdictions need less help than those from lower ranked ones.

 

 

 

 

Differentiation helps advisors compete

 

Advisors from jurisdictions with more complex financial law seem to choose less differentiated strategies when competing for Going Out mandates. Figure 39 shows the (negative) relationship between the complexity of advisors’ jurisdictions’ financial law and our constructed measure of strategic differentiation. Just plotting the two variables together gives a slightly negative “apparent” relationship. After controlling for other factors such as the size of the advisors’ jurisdictions’ GDP and the quality of its legal and finance education, the relationship looks far more negative. Advisors seem to choose to work on different kinds of Going Out mandates from elite advisors when their financial law is less complex.    

 

 

A detailed look at the data reveals a more complex pattern than first suspected. Figure 40 shows the results from regressing advisors’ differentiation on various attributes of their jurisdictions’ legal system and professional education. As we suspected, advisors from low ranking IFCs (as scored by Z/Yen) engage in significant differentiation (choosing client types very different from those chosen by elite advisors). Scores of advisors’ jurisdictions’ law schools and finance departments positively correlate with such differentiation. Better law and financial departments thus clearly provide some grounding for advisors looking to enhance competitiveness. Our two measures of the complexity of a jurisdiction’s financial law show markedly different results.

 

 

Our proxy for financial law complexity focusing on subjective perceptions correlates positively with advisor differentiation, while our objective measure which focuses on similarities across countries correlates negatively with such differentiation. Taking each variable at face value, advisors differentiate as financial law becomes more complex, but differentiate less between jurisdictions with similarly complex financial law. Frankly speaking, we trust the second variable (the measure of legal similarity between jurisdictions) as our measure of a jurisdiction’s legal complexity more. That measure relies on objective data. Yet, the difference between these indicators may represent an area for future research. 

 

Policy Implications

 

What should international financial centres do to rise in the ranks?

 

a) Focus on London and New York

 

Lawmakers and regulators in jurisdictions whose advisors failed to garner large and frequent Going Out mandates should focus on New York and London’s financial law as a model. If our results our right, by copying their financial law, they might be able to increase their competitiveness vis-à-vis the large financial centres, encourage top tier advisors to locate there in greater quantities, and generate demand of their own for advisory services. We include a list of key areas which arise repeatedly and which bear specific attention in considering these issues in Appendix II.

b) Especially for civil law jurisdictions, allow for rapid dissemination of foreign and domestic judgments into regulation

 

Numerous scholars and practitioners alike have argued that the flexibility of London and New York’s law and practice encourages the use of their advisors (Haselman et al., 2010). Our study adds another stone in the edifice of the theory of evolutionary legal change and financial centre growth. We show that jurisdictions with “richer” (read more complex) financial law tend to capture more Going Out mandates. When they have vibrant law schools and financial faculties able to create that law, use it, comment on it, and teach it to students and working professionals, their advisors compete more successfully for mandates. In emerging markets, legal change usually happens too slowly – as Ministries of Justice prepare legal changes which take years to prepare, adopt and then filter down to regulators.

 

Introducing provisions at the legislative level which provide regulators with the authority to adopt changes quickly can help add much-needed flexibility into emerging markets’ financial law. Regulators in New York, London, Hong Kong and Singapore have the authority – as described in legislation – to adopt changes to various rulebooks and regulatory instruments quickly and easily.[44] Adding provisions in banking, securities and insurance laws which give regulators the authority to adopt rules based on international practice, foreign judgments or regulations can make these legal systems more flexible. Legislators and the relevant interest groups would define the limits of such regulatory action.

 

c) Encourage investment banks and law firms to fund local law school research, policy work and teaching

 

According to our research, law schools play a pivotal role in developing the competitiveness of a jurisdiction’s law firms and financial advisors. Law school faculty teach future legal advisors (and sometimes financial advisors as well). They also work on projects which comment on domestic financial sector legislation and regulation. The positive externalities generated by high ranked law schools speak for themselves in our research. Yet, as noted by several commentators, the economics of law schools favours production of lawyers instead of conducting excellent research and engaging with policymakers (Kissam, 1999). Government budgets do not provide the same kind of indirect and direct budgetary assistance to legal research and engagement that they do for other social sciences. Law firms can provide funding for the research and legal changes which will eventually benefit them in the long-run (subject to conflict-of-interest monitoring). Unsurprisingly, we have previously found that demand for compliance advisors and other types of financial advisors depends on their productivity. Engaging in research can only boost this productivity in the long-run.  

 

Local law firms and investment banks can provide grants and other funding on a competitive basis just like governments do. They can release requests-for-proposals for grants aimed at study local financial law (and producing concrete recommendations for reforming it). They can also provide researchers with access to their relationships and other assets they possess. The results of this research can be branded with the university’s logo and the funders’ logos as stipulated in their contracts and any rules specific to each organisation governing such PR activity.

 

d) Emerging market law firms to differentiate themselves through size and marketing

 

How can financial and legal advisors in emerging markets differentiate themselves from their larger and better funded rivals in New York and London? The large advisors (both financial and legal) had access to pools of capital in the past which allowed them to achieve their size and international scale. We showed that advisors do not seem to specialize along sectoral or geographic lines. How can advisors from places like New Zealand or the Middle East hope to compete with their larger rivals.

 

These jurisdictions can adopt more relaxed regulations governing bankers and lawyers than those in the US and UK. For lawyers, these rules can relax strict requirements that only qualified lawyers handle even mundane tasks – such as preparing certain kinds of standardized documents or giving standardized advice. These rules can also allow law firms dealing exclusively in cross-border investment and other activities to list on stock exchanges. As for bankers, regulators can ease restrictions on bankers from soliciting business across borders (and loosen requirements for being solicited).[45] 

 

Conclusion

 

In this paper, we have analysed the way that financial and legal advisors have responded to China’s Going Out policy. The policy represented a natural experiment, in that we can track the way these large flows of funds have affected advisory relationships – both between advisors and their clients as well as between advisors themselves. We found that over the Going Out period covered (from 2000 to 2014), large elite advisors based in top tier international financial centres dominated the advisory business across all types of transactions. Yet, their influence decreased over time (particularly for lawyers) as smaller firms found ways to compete. What accounts for the way these advisors compete for Going Out mandates? What does Going Out teach us about such competition in general?

 

We found three factors not discussed in previous studies account for these advisors’ competitiveness. First, the complexity of the advisors’ home jurisdictions’ financial law gives these advisors the base material from which to fashion solutions to all kinds of current and future problems. Going Out firms favoured advisors (both large and small, though mostly big) proficient with more complex financial law. Second, the quality of law schools in the advisors’ jurisdictions correlates with advisor choice. Law schools affect the quality of local financial law and train the lawyers needed to use that law. Such law also governs what investment bankers can do. Thus, legal complexity and law school quality unsurprisingly affect demand for financial advisors’ services just as much as legal advisors. Third, advisors which can not draw on rich local financial laws or leading law schools can still differentiate themselves. They can go after clients looking for advice in specific areas, at specific times and grab markets away from elite advisors in other ways. Using clustering analysis, we found that smaller firms outside large international financial centres used these strategies – even if we cannot tell exactly what those strategies are. This is an area for future research.

 

Like all studies, these results provide tentative explanations – as all data and theory should bear constant scrutiny. Yet, we hope to leave four long-lasting contributions to this area of study. First, we hope to raise interest in the study of advisors to cross-border investment activity – an area which remains understudied. Second, we hope to provide a more “structuralist” view of financial and legal advisors. Much theorizing and media analysis of these advisors – particularly elite advisors – treats their success as the result of their own decisions. We show that the environment in which they originate and operate plays a key role. Third, this study is a precursor analysis to a further study examining the ways in which advisory networks formed and changed over the Going Out period. This current paper is foundational for the subsequent study it that it analyses the deeper factors driving such structure. Fourth, we seek to highlight the value of maps in “strategic space” in helping to describe the geography of finance – pointing the way to changes in advisory demand in terrestrial space. The geography of advisory services related to cross-border investment depends on geographical/topological features of a space described in terms of legal complexity, service differentiation and law/finance school quality.

 

References

 

Agrawal, Anup, Tommy Cooper, Qin Lian, and Qiming Wang. (2014). Does Hiring M&A Advisers Matter for Private Sellers? Available online.

 

Alford, Duncan. (2007). The Influence of Hong Kong Banking Law on Banking Reform in the People's Republic of China, Journal of International Banking and Regulation 22(266).

 

Arner, Douglas. (2009). The Competition of International Financial Centres and the Role of Law. In Karl Matthias Meessen, Marc Bungenberg, Adelheid Puttler,

Economic Law as an Economic Good: Its Rule Function and Its Tool Function in the Competition of Systems. Munich: Sellier European Law Publishers.

 

Bao, Jack and Alex Edmans. (2007). How Should Acquirers Select Advisors? Persistence in Investment Bank Performance. Available online.

 

Bao, Jack and Alex Edmans. (2011). Do Investment Banks Matter for M&A Returns?

Review of Financial Studies 24(7): 2286-2315.

 

Barton, Benjamin. (2010). Judges, Lawyers, and a Predictive Theory of Legal Complexity. College of Law Faculty Scholarship. Available at: http://trace.tennessee.edu/utk_lawpubl/38.

 

Beaverstock, J., P. Taylor and R. Smith. (1999). The long arm of the law: London's law firms in a globalising world economy. Environment and Planning 31(10): 1857 – 1876.

 

Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine. (2003). Law and finance: why does legal origin matter? Journal of Comparative Economics 31(4): 653–675.

 

Bird, Robert. (2011). Law, Strategy, and Competitive Advantage. Connecticut Law Review 44(1): 63-94.

 

Boateng, Agyenim, Wang Qian and Yang Tianle. (2008). Cross-border M&As by Chinese firms: An analysis of strategic motives and performance. Thunderbird International Business Review 50(4): 259-270.

 

Bottazzi, Laura, Marco Da Rin, and Thomas Hellmann. (2009). What is the role of legal systems in financial intermediation? Theory and evidence. Journal of Financial Intermediation 18(4):559–598.

 

Chang, Xin, Chander Shekhar, Lewis Tam, and Jiaquan Yao. (2015). Industry Expertise, Information Leakage, and the Choice of M&A Advisors. 24th Australasian Finance and Banking Conference 2011 Paper.

 

Chaffee, Eric. (2010). Internationalization of securities Regulation: The United States Government's Role in Regulating the Global Capital Markets. Journal of Business and Technology 5(187). 

Chen, Feiqiong and Yin Wang. (2012). Integration risk in cross-border M&A based on internal and external resource: empirical evidence from China. Quality & Quantity 48(1): 281-295.

 

Chen, Yuan and Michael Young (2010). Cross-border mergers and acquisitions by Chinese listed companies: A principal–principal perspective. Asia Pacific Journal of Management 27(3): 523-539.

 

Chey, Hyoung-kyu. (2014). International Harmonization of Financial Regulation? London: Routledge.  

 

Clark, Gordon and Dariusz Wojcik. (2007). The Geography of Finance: Corporate Governance in the Global Marketplace. Oxford: Oxford University Press

 

Cole, Rebel, Kenneth Ferris, Arie Melnik. (2010). The Cost of Advice in Merger & Acquisition Transactions. Available online.

 

Crotty, James. (2008). If Financial Market Competition is Intense, Why are Financial Firm Profits so High? Reflections on the Current 'Golden Age' of Finance. Competition & Change 12(2): 167-183.

 

Di Giovanni, Julian. (2005). What drives capital flows? The case of cross-border M&A activity and financial deepening. Journal of International Economics 65(1): 127-149.

 

De Jong, Abe, Steven Ongena and Marieke Van der Poel. (2010). The International Diversification of Banks and the Value of Their Cross-Border M&A Advice. CEPR Discussion Paper No. DP7735. Available at: https://www.academia.edu/1079543/_Thinking_about_Going_Abroad_The_Choice_of_Advisor_in_Cross-Border_M_and_A_Deals

 

De Mong, Richard, Ira Harris and Susan Williams. (2011). Financial and Legal Advisors in Merger and Acquisition Transactions. International Journal of Business, Humanities and Technology 1(3). Available online.

 

Dorry, Sabine. (2014).Strategic nodes in investment fund global production networks: The example of the financial centre Luxembourg. Journal of Economic Geography 14(3).

 

Druzin, Bryan. (2009). Buying Commercial Law: Choice of Forum, Choice of Law, and Network Effect. Tulane Journal of International & Comparative Law 18, 2009

 

Evans, Justin and Anthony Gabel. (2014). Legal Competitive Advantage and Legal Entrepreneurship: A Preliminary International Framework. North Carolina Journal of International Law and Commercial Regulation 39(2): 333-358.

 

Flood, John. (2009). Ambiguous Allegiances in the Lawyer-Client Relationship: The Case of Bankers and Lawyers. Available online.

 

Flood, John. (2011). The re-landscaping of the legal profession: Large law firms and professional re-regulation. Current Sociology 59(4) 507–529

 

Flood, John. (2007). Lawyers as Sanctifiers: The Role of Elite Law Firms in International Business Transactions. Indiana Journal of Global Legal Studies 17(1): 35-66.

 

Francis, Bill, Iftekhar Hasan, and Xian Sun. (2012a). Home preference at selecting financial advisors in cross-border M&As. Bank of Finland Research Discussion Paper 31. Available online.

 

Francis, Bill, Iftekhar Hasan, and Xian Sun. (2012a).Does relationship matter? The choice of financial advisors. Bank of Finland Research Discussion Paper 28. Available online.

 

Gordon, Richard. (2010). On the Use and Abuse of Standard for Law: Global Governance and Offshore Financial Centers. North Carolina Law Review  88.

 

Graham, Michael, Terry Walter, Alfred Yawson, and Huizhong Zhang. (2012).

Comparative Advantage, Industry Specialization, and the Role of Investment Banks in M&As. Available online.

 

Hansanti, Songporn, (2008). Recent Issues in International Finance: A Literature Review. In Hansanti, Songporn, Sardar Islam and Peter Sheehan. International Finance in Emerging Markets Issues, Welfare Economics Analyses and Policy Implications.

 

Haselmann, Rainer, Katharina Pistor, and Vikrant Vig. (2010). How Law Affects Lending. Review of Financial Studies 23(2): 549-580.

 

Heremans, Dirk and Katrien Bosquet. (2011). The Future of Law and Finance after the Financial Crisis: New  Perspectives on Regulation and Corporate Governance for Banks.

University of Illinois Law Review 1551.

 

Hunter, William and Julapa Jagtiani (2003). An analysis of advisor choice, fees, and effort in mergers and acquisitions. Review of Financial Economics  12(1): 65 – 81.

 

Hyun, Hea-Jung and Hyuk Hwang Kim. (2010). The Determinants of Cross-border M&As: The Role of Institutions and Financial Development in the Gravity Model. The World Economy 33(2): 292–310.

 

Kaplow, Louis. (1995). A Model of the Optimal Complexity of Legal Rules. Journal of Law, Economics, & Organization 11(1): 150-163.

 

Karsten, Christel GJ, Ulrike Malmendier, and Zacharias Sautner. M&A negotiations and lawyer expertise. Working paper, University of Amsterdam, 2013. Available online.

 

Kim, Chan and Renee Mauborgne. (2005). Blue Ocean Strategy: How to Create Uncontested Market Space and Make Competition Irrelevant. Cambridge MA: Harvard Business School Publishing.

 

Kim, Han, Adair Morse, and Luigi Zingales. (2009). Are elite universities losing their competitive edge? Journal of Financial Economics 93(3): 353–381.


Kissam, Philip. (1999). Lurching Towards the Millennium: The Law School, the Research University, and the Professional Reforms of Legal Education. Ohio State Law Journal  60(1965).

 

Krishnan, C. and Ronald Masulis. (2013). Law Firm Expertise and Merger and Acquisition Outcomes. The Journal of Law & Economics 56(1). Available online.

 

Krishnan, C. and Paul Laux. (2007). Legal Advisors: Popularity Versus Economic Performance in Acquisitions. Available online.

 

Kurkela, Mikko. (2014). "Boutique" financial advisors vs. full-service investment

banks: An analysis of advisor choice and deal outcomes in mergers and acquisitions. Available online.

 

Klasa, Sandy, Lubomir Litov, Jordan Neyland and Simone Sepe. (2013). Does Legal Counsel Expertise Add Value? Evidence from Mergers and Acquisitions. Wharton Financial Institutions Centre Working Paper 13-27. Available online.

 

Laux, Paul (2001). Legal Intermediaries in Mergers and Acquisitions: Roles and Effects.

Available online.

 

Levine, Ross. (2010). The governance of financial regulation: reform lessons from the recent crisis.  BIS Working Papers No 329.

 

Li, Xue-yao and Sida Liu. (2012). The Learning Process of Globalization: How Chinese Law Firms Survived the Financial Crisis. Fordham Law Review 80(6):

 

Lewis, Michael. (2010). Liar’s Poker. New York: Norton.

 

Michael, Bryane, Joseph Falzon and Ajay Shamdasani. (2015a). A Theory of Financial Services Competition, Compliance and Regulation. Available online.

 

Michael, Bryane, Christopher Hartwell, and Gary Ho. (2015b). Does Financial Market Development Explain (or at Least Predict) the Demand for Wealth Management and Private Banking Services in Developing Markets? Journal of Investment Consulting 15(2).

 

Michael, Bryane. (2014). Does Objectives-Based Financial Regulation Imply a Rethink of Legislatively Mandated Economic Regulation? The Case of Hong Kong and Twin Peaks Financial Regulation. University of Hong Kong Faculty of Law Working Paper 2014/5.

 

Mayson, Stephen. (2007). Law Firm Strategy: Competitive Advantage and Valuation.

 

Prakash, Loungani, Ashoka Mody, and Assaf Razin, The Global Disconnect: The Role of Transactional Distance and Scale Economies in Gravity Equations, 49 Scot. J. of Pol. Econ. 5, 2002.    

 

Prentice, Robert. (2005). Regulatory Competition in Securities Law: A Dream (That Should Be) Deferred. Ohio State Law Journal 66(1155).

 

Ramasamy, Bala, Matthew Yeung, and Sylvie Laforet, . (2012). China's outward foreign direct investment: Location choice and firm ownership. Journal of World Business 47(1): 17-25.

 

Salidjanova, Nargiza. (2011). Going Out: An Overview of China’s Outward Foreign Direct Investment, USCC Staff Research Report.

 

Segal-Horn, Susan and Alison Dean. (2011). The rise of super-elite law firms: towards global strategies. Service Industries Journal 31(2): 195-213.

 

Sharman, Jason. (2012). Chinese capital flows and offshore financial centers. The Pacific Review 25(3): 317-337.

 

Sibilkov, Valeriy and John McConnell. (2014). Client Performance, Choice of Investment Bank Advisors in Corporate Takeovers, and Investment Bank Market Share. Review of Financial Studies 27(8): 2474-2503.

 

Sikka, Prem. (2003). The Role of Offshore Financial Centres in Globalization. Accounting Forum 27(4): 365-399.

 

Smith, Roy and Ingo Walter. (2003). Global Banking. Oxford: Oxford University Press.

 

Stern, Rachel and Su Li. (2015). The Outpost Office: How International Law Firms Approach the China Market. Law and Social Inquiry.

 

Sun, Sunny Mike Peng, Bing Ren, and Daying Yan. (2012). A comparative ownership advantage framework for cross-border M&As: The rise of Chinese and Indian MNEs.

Journal of World Business 47(1): 4–16.

 

Sutherland, Dylan, Ben Matthews, and Ahmad El-Gohari. (2012). An exploration of how Chinese companies use tax havens and offshore financial centres: roundtripping or capital-augmenting OFDI? In Fu, Xiaolan. China's Role in Global Economic Recovery. London: Routledge.

 

TMF Group. (2014). Global Benchmark Complexity Index 2014. Available online.

 

Van Criekingen, Mathieu, Jean-Michel Decroly, Moritz Lennert, Pierre Cornut and Christian Vandermotten. (2005). Local Geographies of Global Players:

International Law Firms in Brussels. Journal of Contemporary European Studies 13(2): 173–187.

 

Vlcek, William. (2008). From Road Town (BVI) to Shanghai: FDI, IBCs and global capital flows. Third Biennial Oceanic Conference on International Studies (OCIS)

School of Political Science and International Studies, University of Queensland

2 - 4 July.

 

Wang, Cong, Fei Xie, and Kuo Zhang. (2014). Expert Advice: Industry Expertise of Financial Advisors in Mergers and Acquisitions. Available online.

 

Wang, Cong and Fei Xie. (2011). The Choice of Financial Advisors in Cross-border Acquisitions: Determinants and Consequences. Available online.

 

Wenbin, Huang and Andreas Wilkes. (2011). Analysis of China’s overseas investment policies. CIFOR Working Paper 79.

 

 

 

 

 

 

 

 

 


 

Appendix I: Model

 

Overview

 

Our model of financial and legal advisor competition for Going Out mandates focuses on advisors’ probabilities of winning these mandates. We follow the Krugman’s Razor approach to modelling, specifying a simple rule for Chinese Going Out firms – and looking at how advisors might react to that rule.[46]  We take the current centres of Going Out advice (and the frequency by which they offer that advice) as given – and look to see what might affect the probabilities of each centre receiving a Going Out mandate. We show that these centres will maximize their probability of receiving such mandates, subject to a resource constraint of changing their financial law’s complexity, the quality of their law schools and the differentiation of their advisors. Advisor selection across geography in our model is completely divorced from the geography of acquirers and target firms.  

 

Our model provides 9 observations on the state of supply and demand for Going Out advisors.  We observe that a geometric model of supply and demand for Chinese investment adequately explains the Going Out experience – with a jurisdiction’s legal complexity, law school quality and advisors’ own strategic differentiation serving as new and interesting explanations for such investment.[47] We show that financial centres looking to gain more advisory business need to aim for the optimal levels of the variables we describe in this paper.

 

A Model of Going Out Firms’ Choice of Geographical Distribution of Advisory Mandates

 

We start our model with the stylized facts pertaining to Going Out. We assume – particularly in the short-run – a fixed number of Going Out transaction and their value. To keep the math simple, we assume one large mandate of value Q value – which different advisors “share” through their pi probability of winning the mandate.[48] We define p as the number of mandates won by an advisor in a jurisdiction divided by the total offered.[49] Multiple advisors from a jurisdiction just add up their probabilities to give each jurisdiction a probability of attracting Going Out advisory business. We do not differentiate between financial and legal advisors – as no obvious pattern emerged in the data need explanation. We impose two restrictions on such a set-up. First, we require a fixed number of advisors i set at the jurisdictions receiving at least one mandate from 2000 to 2014 – giving a fixed advisor set A={New York, London...Auckland}. Second, we can rank the set of geographical areas (cities or financial centres) by their probabilities of winning Going Out mandates.

 

Going Out firms express preferences for advisors with particular backgrounds – and their preferences determine the geographical allocation of Going Out mandates. In our model, we assume that Going Out firms act together, taking decisions about allocating Going Out mandates as one body.[50] Figure A1 shows the variables which impact on their decision to choose an advisor. Equation (1) below that shows the way these variables combine into Going Out firms’ client decisions about their advisors. We choose a very simple combination of these variables, as a simple union of probabilities – as clearly the level of level complexity affects the extent to which differentiation strategies “pass through” into final probabilities – and so forth.[51] We assume all advisors receive the same pay and advisors “commodities” for model purposes. We mention the effect of wages in equation (1) and Figure A1, but drop any assumption they play a serious role in advisor selection in equation (2).[52] Equation (3) shows the restrictions we described above. The last restriction in the equation requires that each probability “map” (or place) each jurisdiction in a particular ranking.[53]

 

 

 

 

 

 

 

Figure A1: Variables and Parameters Related to Going Out Advisor Choice

 

Greek/ English

Letter

Variable

or parameter

Description

C

legal complexity

A proxy describing the complexity of an advisor’s financial law

d

size effect of legal complexity

The way that the “volume” or size of financial law affects the way that legal complexity affects Going Out firms’ choice of an advisor

q

legal complexity transaction factor

“Translates” the way the number of pages of financial law (or a proxy for it) feeds into the probability that firms choose a particular advisor.

D

extent of advisor differentiation

Some measure of the extent to which advisors differentiate – by choosing different markets, types of clients, etc.

y

differentiation translation factor

Converts units of differentiation into units of probability

z

extent of differentiation

Describes the way that more differentiation affects the way that  differentiation affects the probability of winning Going Out mandates

c

pass-through of law school effect

Converts the way that law school quality and the “size” of that quality impacts on the probability local advisors get Going Out mandates.

X

proxy for law school quality

A measure of law school quality, as “quality-adjusted” journal papers, students test scores or other factors scaled into one simple measure.

w

“size” effect of quality

Same as above – takes into account that very good law schools’ improvement will affect probabilities of advisors getting Going Out deals differently than if a “bad” school improves

w

wage levels

Average wage for advisors

L

amount of advisors used

A simple measure of staffing as a headcount of advisors from a particular jurisdiction

b

effect of employment size

Measures any agglomeration or size effects to staffing in a financial centre.

 

                                                                                          (1)

                                                                                   (2)

such that     and       and       (3)

 

Our model of advisor demand implies several things about the market for M&A advisors (and investment advisors more generally). First, as long as jurisdictions and advisors in A (the set of jurisdictions with advisors to Going Out companies) keep their legal complexity, differentiation and law school policies within certain bounds, they will always have at least some Chinese demand for their services. Second, one advisors’ loss is another advisor’s (or advisors’) gain. Because probability always adds to unity, advisors in places like Texas or Tel Aviv can only score more mandates by snagging them from other financial centres. Third, diminishing returns to scale and other factors likely imply a specialisation in the way advisors’ jurisdictions raise their probability of getting Going Out work. If New York (for example) focuses on legal complexity as a way of maximising its probability pNY, then Madrid could likely focus on differentiating from the New York advisors as a way of raising its probability pMadrid.

 

For readers less comfortable with abstract math, we can illustrate the way that these probabilities might correspond with our model’s variables. Figure A2 shows the bi-variate relationships between the probability of obtaining a Going Out mandate and several of the variables we talked about in the paper. As shown, the probability of obtaining a mandate “splits” for certain levels of our variables of interest. For example, for systems displaying very high legal complexity, they show the same low probability of obtaining a mandate and those with very low levels. The “optimal” level (before taking other factors into account) hovers at around 0.83. As for deal size, the “optimal” deal size (before taking other factors into account) comes in at around €1 million. “Better” law schools (as defined by global rankings) always correlates with more Going Out mandates. Moreover, differentiation matters...up to a point. After that point, differentiation does not increase the probability of scoring Going Out mandates. At this point, readers only need to know that relationships certainly exist between probabilities of getting Going Out work and the model variables.

 

Figure A2: Relationships Between Probability of Winning a Mandate

and Key IFC Traits

 

Factor

Equation

Range

Size matters?

rank

1<Rank<30

No

legal complexity (C)

for C<0.83

for C>0.83

No

deal size (Q)

for Q<1074

for Q<1075

Yes

law school quality (X)

for 0<X<1

Yes

differentiation (D)

for 0<D<5

No

 

In theory, a piece-wise relationship best describes the way that the probability of winning Going Out mandates relates to the model variables shown in Figure A2. The function p takes on different values depending where a particular jurisdiction’s C, D, and X sit. We know for these equations that a jurisdiction reaches it maximum probability around C=0.83, Q=1074, X=1, and D=5. 

 

Figure A3: The Piecewise Relationship Between p and the Other Variables

 

for C<0.83 and Q<1074

 

for C>0.83 and Q<1074

 

for C<0.83 and Q>1074

 

for C>0.83 and Q>1074

 

p

 
     

 

The figure shows the multiplication of the lines-of-best-fit for the probabilities of various cities’ advisors receiving Going Out mandates. We simply shoved the equations describing the bivariate relationships (taking only one factor into account) into equation (2). We take 4-th power roots of these equations because we show each equation individually results in the city’s probability of snagging a Going Out mandate. If we multiply all four factors together, we need to “un-multiply” them to get back to the original final probability of a city’s advisors winning a Going Out mandate.

 

Building indifference curves for Chinese clients

 

Observation 1: Going Out clients looking urgently for advisors will look for advisors choosing differentiation levels at

 

In the main body of our paper, we speculated in Figure 32 that advisors’ probabilities depend on the extent to which they can match the preferred attributes of their clients. To what extent do Chinese firms want to use advisors’ jurisdictions’ complex financial law – and how complex do they like it? Equation (4) describes jurisdiction certain to get these tenders. If we abstract away from law schools for the moment, equation (5) tells us that the level of legal complexity and differentiation equals the reciprocal of the “pass-through” effects we described earlier. Such a result would certainly exceed 1 and thus fall into the range for C and D we describe in the next appendix. Equation (6) shows the solution for C as a function of D. The level of C for “sure thing” advisors (with a probability close to 1) gets bigger as differentiation decreases and as the size effect of such complexity decreases. Taking the functional inverse of equation (6) provides the equation shown in Observation 1.

 

                                                                                     (4)

                                                                                                               (5)

                                                                                                            (6)

 

Observation 2: Going Out firms’ desire for strategic differentiation increases as their interest in using more sophisticated law expands.

 

How do legal complexity and strategic differentiation change otherwise? Equation (7) shows this relationship using the implicit differentiation rule. By testing different values for this differential, we can get some idea about how the red curve shown in Figure 32 labelled “Going Out companies’ indifference curve” behaves. After some math, equation (8) shows how this differential acts when changing at the constant rate of 1. That equation says that the difference between the way the probability of a Going Out engagement changes as companies differentiate versus if their regulators made local financial law more complex equals the size effect on legal complexity grows. As we promised in Observation 2, if our maths and analysis are correct, for particular probability of giving Going Out mandates, Chinese companies’ desire for differentiation expands with their desire for more complex law.

 

                                                            (7)

                                                   (7a)

                                                                                 (7b)

                                                                                                 (8)

 

A second way to demonstrate this relies on elasticity of these curves. Equation (12) shows the elasticity of changes in advisor differentiation with respect to changes in legal complexity. As shown, such elasticity equals the probability of advisors getting a mandate minus the way that probability changes for rises in legal complexity (scaled by the level of such complexity) divided by the way differentiation changes such probability in relation to advisors’ probability of getting mandates. Equation (14) shows the constant elasticity case (with constant c). When such elasticity is constant, the probability of obtain Going Out mandates equals the way that probability changes with differentiation (scaled by the level of such differentiation) plus the way that legal complexity adds to such probability (again scaled by the level of such complexity). The important thing to note is that the probability of winning mandates simply equals the sum of the way changing the variables of interest affect such probability. As we describe a bit later, a additive model of such probability used for regression helps us understand the way geometric probabilities come together to decide a jurisdiction’s success at winning Going Out mandates.

 

                                                                              (12)

                                                                                                  (13)

                                                                                               (14)

 

Another way to see this is to ask under what conditions do these indifference curves slope down? Recalling from equation (6), that C= f(D-z/q). C will slope down when -1<–z/q<0. Taking each part separately, we know that z>q or (1/q)<0.

 

Observation 3: Going Out companies’ seek advisors with a mix of differentiation, access to legal complexity and legal knowledge.

 

How else does legal complexity act compares with advisor differentiation? Equation (9) shows when equation (7) approaches a very small number close to zero which we label as D. Equation (10) shows the final result – where the change in probability as legal complexity changes equals legal complexity’s size effect expressed as a proportion of the level of such legal complexity C. Equation (11) shows equation (7) set to infinity. The ratio of Cd-1 to Dz approaches infinity only as -D approaches zero (and thus conversely approaches negative infinity when D approaches zero). In all these cases, we expect the nice convex shape shown in Figure 32. Legal complexity grows infinity as differentiation goes to zero and goes to zero as differentiation approaches infinity (as shown in equation 10a). These examples show us that the probability of winning mandates decreases at the size of each variable increases. These examples also show us that “corner solutions” (or solutions where Chinese firms want only access to legal complexity for example) probably don’t exist.

 

                                                                (9)

                                                                                                   (9a)

                                                                                                              (10)

                                                                                                             (10a)

                                                                                                           (11)

 

We can see this as we reintroduce law school quality into Going Out firms’ advisory choice. To what extent do the indifference curves we painted shift around? Returning to equation (4) for simplicity, we can solve for C as a function of D – as shown in equation (12). Equation (13) shows the way that C needs to change to maintain a fixed level of probability. Equation (14) shows change in terms of the original C. Equation (15) rearranges and equation (16) shows the change in C as a proportion of the level of C. As we see, as we increase X, C rises as does the requirement to increase D. Unsurprisingly, our explanatory variables are linked – like in our analysis.

 

                                                                                                                  (12)

                                                                                                (13)

                                                                                    (14)

                                                                           (15)

                                                                                          (16)

 

Understanding advisors’ production possibilities

 

Under what conditions will legal and financial advisors bit on Going Out mandates? Equation (17) shows that advisors maximise profits when they maximise the probability of their winning a mandate (as a function of C, D and X) while minimising any costs they incur as a result of their government’s or own decision to provide these public goods. Equation (18) just plops specific relationships on each of the function forms from the previous equation. Figure A4 describes the parameters we use to model advisors’ decisions. We assume that advisors and government officials work together, thus making policy questions like the complexity of financial law and law school development issues that these advisors can affect.

 

                                                                                                            (17)     

                                                                               (18)

 

Figure A4: Variables and Parameters Related to Advisors’ Competition for Going Out Mandates

 

Greek/ English

Letter

Variable

or parameter

Description

D

legal complexity’s help in winning mandates

Describes the way that the size of a country’s level complexity can help its advisors increase their chances of winning Going Out work.

z

differentiation’s help in winning mandates

Describes the way that advisors’ differentiation can help increase their chances of winning Going Out work.

Q

law school quality’s help in winning mandates

Describes the way that advisors’ jurisdictions’ law schools help them win Going Out work.

tc

“tax” used to pay for legal complexity

Represents the direct payment or indirect costs that advisors must bear to pay lawmakers, regulators and third-parties for making, commenting on and otherwise changing financial law’s complexity.

tD

“tax” used to pay for differentiation

Represents the direct payment or indirect costs that advisors must bear to differentiate themselves from their competition as they struggle to attract Going Out companies.

tX

“tax” used to pay for law school development

Represents the direct payment or indirect costs that advisors must bear to pay to law schools, officials who set their policies and otherwise concern themselves with law schools’ development.

 

Observation 4: Provision of legal complexity, differentiation and law school quality unsurprisingly depend on the costs of making them.

 

Like in the previous example, we can abstract away from law school quality for the moment to figure out how the investments that various advisors in various jurisdictions make raise their probabilities of obtaining Going Out work. Differentiating equation (18) by C and D, setting these derivatives equal to zero (their maximum) and putting the resulting equations in a system gives equation (19).[54] If  and , then equation (20) shows the solutions for investment in legal complexity C and advisor differentiation D.

 

                                                                       (19)

 and                                                      (20)

 

Observation 5: While clients see the variables in our model as complements, advisors see them as substitutes when faced with costs of “investing” in them.

 

What trade-offs need to be made as advisors decide to fund legal complexity or differentiation? Returning to equation (18) and removing law school quality for the moment, we see from equation (21) that – at any particular level of profit  -- we can not solve C as a tidy function of D.  Instead, we can see how C varies with D by using the implicit differentiation function as shown equation (22). In equation (23), we require that ∂C/D<0, or that C varies negatively with D – reflecting the scarcity of resources. As a result, the interaction between C and D must be less than the cost of differentiating advisors (tD) and any effect on profits of such differentiation (as shown in equation 24). By taking the logs of both sides and solving for D, we see that the level of differentiation must be less than the cost of developing such differentiation (in terms of the opportunity cost of C), scaled by the way D impacts on profits. As shown in equation 25, clearly (just by looking at the reciprocal relationship between C and D), we see that substitutability of these factors – with the cost of develop them a key consideration.

 

                                                                                                  (21)

                                                                                            (22)

                                        (23)

                                                                                                        (24)

                                                                                                                 (25)

 

The relationship between C and D illustrates the way that advisors face a type of “production possibilities frontier” when developing their competitive position in bidding for Going Out work. Figure A5 shows the way we might interpret real-world data, in light of the findings from our model. On the figure, we drew several “frontiers” between C and D – corresponding with higher investment levels. As these frontiers move toward the northeast part of the figure (as we showed in the main body of our paper), advisors’ probability of winning Going Out work increases. We do not show Chinese firms’ preferred trade-off between C and D in this figure.

 

 

Equilibrium between clients and their advisors

 

Observation 6: Advisors’ may “lump” together or specialise in “strategy space” depending on advisory tastes and technologies

 

What is the equilibrium level of legal complexity and differentiation? Differentiating both Chinese clients’ preference function and advisors’ profit function by C and D gives equations (26). Setting the difference between these rates of change equal to zero (such that they change at the same rate) gives equations (27). At their optimum (as shown in equations 28), the level of legal complexity equals a function of D and X – and the level of strategic differentiation represents a function of C and X. Equations (29) simply reorganise the variables (using the quadratic equations formula), to put all the Cs and Ds on one side of the equation.

 

                                                   (26)

                                   (27)

                                                     (28)

                      (29)

 

Nothing forbids jurisdictions from choosing the nearly the same levels of C, D, and X – and thus have the same probabilities of receiving Going Out mandates. On the demand side of the Going Out relationship, nothing in our model prevents advisors from using high levels of X,C, or D or achieve a particular probability of going a mandate. On the supply side, nothing prevents jurisdictions from having similar levels of legal complexity or similar law schools. The data clearly show such clumping – and differentiation. Figure A6 shows the relationship between Going Out mandates and law school qualities (just as an example of one of the variables we analyse in this study).We see clumping for law school scores at about 70-80, at about 50 and limited clustering around 20. We also see differences in law school scores – reflecting differences in the ways some universities succeed or fail to raise their law school’s quality.

 

 

Observation 7: Linear rather than geographic models may provide better explanations for the distribution across advisors of Going Out mandates

 

Recalling equations (29), we do not observe many interactions between C, D and X in the solutions we derived. The solution for C for example, as only one term with a possible interaction  . That term (if we assume that Xw=XQ) probably represents a very small part of the overall effect on C. Similarly, D contains terms completely separable from any interactions. Equation (31) shows equation (29) written in a way we might see in a linear regression. The beta coefficients – in an idea world – would simply reflect the parameters we show in the equation. C can either vary by 0 or 1 with D. A non-linear function of X must be used to pick up b2 in this expression (which likely represents a small part of the effect and can probably be ignored). But otherwise the term remains separate from C. Yet, these functions look surprisingly linear.

 

                                 (31)

 

Indeed, non-linear models perform significantly worse. Figure A7 shows the regression results for a regression attempting to explain whether the variables we discussed in our model help explain the probability IFC-based advisor won a Going Out mandate. Such a regression explains about 3% of all Chinese firms’ choices to use IFC-based advisors. Looking further down the regression, we see the residuals for such a regression look pretty bad. Such a model would over-predict the use of IFC-based advisors for low values of our model variables – and over-predict for high levels. These results though don’t suggest a problem with our multiplicative model of probability.[55] Because the method of least squares “projects” variance onto the additive terms in our regression model, our linear model works better than logit or geometric models.[56]

 

Figure A7: Do Not Use a Geometric Regression Model to Test a Geometric Theoretical Model

 

Regression Summary for Dependent Variable: Probability that

Advisor Comes from an Log IFC

R= .21 R²= .045 Adjusted R²= .03 F(6,353)=2.8022

 

Beta

Beta

Std.Err.

B

B

Std.Err.

t(353)

p-level

Intercept

 

 

0.23

0.19

1.19

0.23

Log of Deal Value

0.38

0.13

0.03

0.01

2.94

0.003

Log of Law School Proxy

0.006

0.24

0.0006

0.02

0.02

0.97

Log of  Finance School Proxy

0.57

0.38

0.054

0.03

1.51

0.12

Log of Legal Complexity

0.44

0.54

0.040

0.04

0.83

0.40

Log of Legal Similarity

0.88

1.04

0.08

0.09

0.85

0.39

Log of Z/Yen IFC score

-2.31

1.19

-0.21

0.10

-1.96

0.051

 

Observation 8: Our model requires no modification for understanding differences in supply/demand for legal advisors as opposed to financial advisors

 

As a general model, our model can apply to legal advisor and financial advisor sub-sectors just as well to the general class of all advisors. Equation (32) shows how we might relabel equation (20) – to take differences in the type of advisor into account. If FA subscripts stand for financial advisors and LA subscripts stand for legal advisors, then equations (25) clearly show how we might adjust our analysis for each advisor segment. Moreover, the resulting probabilities could correspond to different geographical places.

 

                                    (32)

 

Figure A8 shows how different model parameters likely led to different probabilities of Going Out work for different types of advisors in different locations. New York represented the highest probability jurisdiction for financial advisors – scoring about 13% of Going Out mandates. On the other hand, London scored 27% of them – a far larger overall proportion. Zurich came in at third place in terms of financial advisory mandates, but Sydney came in third place for legal advisory mandates. Estimating our model for different kinds of advisors simply involves different parameters for each group of advisors.

 

Figure A9: Geography of Financial Advisors Probably Results from Different Model Parameters than Legal Advisors

(percentages of headcount mandates for 2000 to 2014)

 

Financial Advisors

 

Legal Advisors*

New York

13%

 

London

27%

London

8%

 

New York

22%

Zurich

5%

 

Sydney

10%

Hong Kong

2%

 

Beijing

7%

Paris

2%

 

Toronto

5%

Frankfurt

2%

 

San Francisco

4%

* excludes advisors-to-advisors.

 

Driving real-world like results using our simple model

 

Can our model reproduce the results we observed in the real-world? Figure A9 shows the distribution of mandates across geographical space. The result looks like the result of multiple geometric curves multiplied together.[57] Imagine that New York had a 0.6 probability of getting mandates from each of our three variables. Then multiplying the three together gives about 0.22. Figure A10 shows the way that changes in a jurisdiction’s legal complexity scores affect its advisors’ probability of winning Going Out work. As legal complexity scores move away from the obvious optimal, advisors’ chances of winning such work clearly fall exceedingly quickly. We do not need to build a full model to see that the way advisors’ X, C and D “match” Chinese companies’ preferences probably influences their chances of winning Going Out mandates.

 

 

Observation 9: Small changes in model variables around the optimum can lead to large changes in emerging markets probabilities of winning Going Out work

 

Figures A9 and A10 illustrate the principle easily. Imagine that New York (for example) loses 5% of its probability in attracting 100 Going Out projects. As shown in Figure A6, for its 25 projects, it would lose roughly 2-3 projects.  A redistribution of those projects to Amsterdam (with 1 project), would result in 200%-300% proportional increase in the number of Amsterdam’s projects. Any investment in X, D, and C which brings Amsterdam closer to New York might redistribute these projects. Figure A7 shows, if Amsterdam’s legal complexity comes in at 0.6, a small change in such complexity makes little difference to its advisors. However, for complexity measure values closer to 0.79, a small change makes a large difference. The decision to undertake such a change clearly depends on the cost (tC). Clearly, some countries can compete, while others can not.

 

Model Conclusions

 

In this appendix, we have tried to predict how Chinese companies’ might “match” with advisors (based on supply-like and demand-like functions) to give Going Out mandates. We showed that demand for advisors’ services increases as they advisors have access to a rich, complex local financial law, increasingly differentiated strategy and local law schools enriching the whole local and global financial system. Advisors respond by developing these attributes of the local competitive competencies – but subject to their ability to develop them... and at a cost.

 

In our model, advisors match their advisory competencies with Going Out companies’ preferences. Jurisdictions whose legal complexity, law school quality scores and whose advisors’ differentiation lies far away from Chinese companies’ preferred levels attract low levels of probability for winning Going Out mandates. Advisors with levels closer to New York’s and London’s might steal some business (or “probability” in our model’s terms) from these gigantic international financial centres. Our model implies that financial centres – including New York and London – can earn more Going Out mandates by closer matching their advisory environments with Chinese Companies’ preferences. 

 


Appendix II: Major Areas of Harmonisation for Aspiring International Financial Centres

 

In our analysis of Going Out transaction, we saw points of law arise on a regular basis. For regulators from financial centres aspiring to catch up with London and New York (in terms of deal numbers and value), we highlight the following areas of financial law for focus.

 

International term loan agreements

 

To what extent do requirements on form and content of international (cross-border) loan agreements provide debtor and creditor with clarity and recourse to dispute resolution? What enables the use of standard form agreements (like those from the Loan Market Association)? What differences exist between New York and London versus our jurisdiction in terms of precedent conditions, representations/warranties, default provisions, structure/style of commitment letters and term sheets.

 

Rescheduling, restructuring and standstill agreements

 

What about the drafting of these agreements makes US/UK law so preferable? What standardized conditions can we introduce at the regulatory level to encourage the use of these agreements outside of our jurisdiction (but using modifications to US/UK law we introduce in our domestic regulation)?

 

Primary and secondary syndicated loan agreements and transfers

 

How are our provisions on the form and content of international syndicated loan agreements more complex/ costly than those from UK-US?  What about obligations and rights of syndicated lenders and intermediaries have academics and professionals recently criticised? If you wanted to transfer a $1 billion book of loans, what about domestic law would you prefer—after mapping out requirements under US/UK law?

 

Project finance

 

What documentation requirements most affects preferences for and against your jurisdiction vis-a-vis London and New York? Are certain financing structures generally prohibited which financiers can “get around” in New York and London?

 

Securitization and structured finance

 

How does your jurisdiction treat “true sale” differently than New York and London? How are securitisations structured in your jurisdiction (if at all)? What parts of UK-US regulation in this area might tempt financiers to look to your jurisdiction to base their transactions (like the 5% retention requirement, etc.)

 

International bond issues

 

This component will examine the process of issuing an international bond; the parties to such issues and the fundamental terms typically incorporated. The legal nature of international bonds will also be examined as will the manner in which they are traded. The role and duties of the bond trustee will also be examined in detail.

 

Alternative Investment Funds

 

Do what extent does private equity and hedge fund finance provide a viable method of “crowding in” other types of investment using your jurisdiction as the place of record? Esplain the differences in PE/VC regulation (provision by provision) between your jurisdiction and UK/US. If you could change US-UK regulations in order to get more foreign businesses to use these regulations to “piggy-back” other kinds of investment, what would you change? 

 

Legal opinions

 

What about local regulations makes the production of lawyers’ opinions more reliable/useful in UK-US than here? If possible, obtain several opinions on large transactions involving London and New York based advisors – and local advisors. What are the obvious differences – both in form and substance? How can they be “copied”? (Or what policy activities can result in the production of skills needed for local counsel to produce the same kinds of analyses?)

 

Conflict of laws in international finance transactions

 

What about London and New York use of UK and US law (respectively) allows them to use provisions aimed at conflict of laws more efficiently than at home? What are those provisions (provision-by-provision comparison)? What does your analysis of court cases and arbitration reveal?

 

 

 

Sources: Compiled by authors used Penn  and Singh to structure the presentation of our analysis.

 


 

Appendix III: Background to Statistical Analysis

 

Description of data

 

Our sample of Going Out firms consists of all cross-border M&A transactions listed in the Zephyr database from 1 January 2000 to the 31 December 2014. Our search resulted in 578 transactions, which we parsed to 372 transactions which listed at least one paired advisor and city. The original data source sometimes listed multiple advisors for a single transaction – though much more frequently failed to identify either legal or financial advisors to the target or acquiring company. After listing each advisor’s role in these transactions separately, we obtained 810 separate advisor transactions. We also produced a separate file of 1,536 matched pairs of relationships across these transactions, which we will use in a future paper to map out advisory relationships between these companies.[58]

 

We identified several parts of these transactions for our analysis. Figure B1 shows the variables we used in our analysis of M&A advisors and the cities they work in. We identified the major attributes of each transaction – the industry in each acquirer and target work, ownership of the acquirer (SOE or private), the types of deals they engaged in, and attributes of the deal like deal size, fees likely paid and the target company’s size before the transaction. We used about 372 observations – capable of providing enough discriminating variability for us to run even sensitive tests.

 

Figure B1: List of Variables Used in the Study

 

Variable Name

Description

No. valid

Name of acquirer and target companies

Shows name of Chinese acquirer and its cross-borders target. 

372

Type of Ownership

Classifies companies as either SOEs or private companies. We used main shareholder identities and their holdings to establish if a state-owned or controlled entity owns or runs the company. When we could not make a determination using this data, we used popular references in the financial press to determine membership in each group.

372

Target name

Shows name of target company outside of China

372

Name of acquirers’ and targets’ advisors

Shows names of financial and legal advisors on a transaction.

839

Advisors’ cities

Shows the city of each advisors’ headquarters. We omit for advisors without a single city serving as headquarters.

792

Acquirer and target city and country

Shows the city and country for acquirer and target firms. Acquirer firms in China only.

371

Industry codes

We converted US SIC codes for each company into Global Industry Classification Standard (GICS) sector codes. These categorize companies as energy, materials, industrials, consumer discretionary, consumer staples, health care, financials, information technology, telecoms and utilities.

115

Number of cities and advisors

We added the number of advisors on each side of the M&A transaction and the number of cities they come from.

see above

Deal type

The original source lists the nature of the M&A activity, such as increasing share holdings or buying share holdings. We binned these types in increases in shareholdings (0%-20%, 20%-50%, 50% or more), and share purchases (1%-20% of shares, 21%-50%, 50% or more, or a full buy-out).

368

Deal value

Shows the value of the transaction – in millions of euros.

317

Transaction year

We show the year in which the transaction closed (was completed). 

333

Modelled fee income

Shows estimated advisors’ revenues for each transaction.

183

Pre-deal target asset

Shows target company’s total assets before the transaction.

195

Sources: Zephyr (2015).

 

Figure B2 provides background statistics on our variables. We report most of the important statistics in the body of our paper. Our sample contained roughly 2/3 private companies and 1/3 SOEs (not weighted by market capitalisation).  We record a separate advisor transaction when an advisor participates in some way in a transaction. Throughout the entire period, financial advisors have represented target firms 64% of the time and legal advisors advised targets 57% of the time.

 

Figure B2: Descriptive Statistics for M&A Advisors Dataset

(weighted by number of advisors)*

 

Variable

Valid N

Mean

Sum

Minimum

Maximum

Std.Dev.

Quantitative Variables

 

 

 

 

 

 

Probility Acquirer Private

810

0.58

 

0

1

.49402

Deal value th EUR

317

494,721

156.m

12.000

7306271

923954

Year

333

 

 

 

 

 

Modeled Fee Income th EUR

183

3469

634k

0.744

15197

3611

Probability of Elite Advisor

810

.36

 

0

1

.48

Probability of IFC-based Advisor

645

.609

 

0

1

.48829

Pre-deal target total assets th EUR Last avail. yr

195

12339341

240m

18.029

870223000

84740191

Deal value to assets

173

61

1.06k

0.000

9000

690

Fee to deal value

183

0

3.65

0.005

0

0

Qualitative variables

 

 

 

 

 

 

Number Acquirer cities

371

1

48.2

0.000

6

1

Number Target advisors

371

1

35.7

0.000

8

1

Number Acquirer advisors

371

1

48.0

0.000

7

1

Number Target cities

371

1

31

0.000

7

1

* weighting by the number of advisors gives indicators like whether the target company’s city and its advisor’s city matches a higher weight for transactions which involved numerous advisors. Specifically, when creating the spreadsheet on each advisor’s transactions, we would repeat the same transaction several times when several advisors worked on the same transaction. Simple means from such a spreadsheet would obviously count transaction with multiple advisors more than those with only one advisor,

 

Figure B2 shows the statistics we constructed in order to find correlations in the data. For example, do advisors in top 10 international financial centres specialise in M&A activity in a particular industry (or when the acquirer seeks to acquire a firm in its same industry)? In our sample, financial advisors constituted 415 of our observations and 362 of them (with 35 missing). Similarly, our data set contained 444 advisors to acquiring companies and 333 advisors to target companies (again with 35 missing). We used two major categorisations of advisors in our study – the extent to which they belonged to elite advisors (Bulge Bracket banks or Big Law) and whether they operated from a top 10 international financial centre.[59] Members of Bulge Bracket banks included Morgan Stanley, Credit Suisse, Citigroup, JP Morgan, UBS, Deutsche Bank, Lazard, Goldman Sachs, Bank of America, and Rothschild. Members of Big Law firms included Linklaters, Clifford Chance, Freshfields, White & Case, Skadden Arps, Shearman, Baker & McKenzie, Allen & Overy, O’Melvery, and Jones Day. The top 10 international financial centres for our study consisted of New York, London, Hong Kong, Singapore, Zurich, Frankfurt, Luxembourg, Paris, Tokyo, and Seoul.

 

Figure B3: Constructed Statistics Used for Analysis

 

Variable Name

Description

No. valid

Industry match?

A dummy variable which results in a 1 if the acquirer’s GIC sector code equals the target company’s code – 0 otherwise.

643*

Type of advisor

We categorise advisors as legal or financial.

777

Client side

Shows whether the advisor worked for the acquirer or target company.

777

Target-Advisor City Match?

A dummy variable equalling 1 if the advisor is based in the same city as the target company.

 

Deal value to size

Shows the ratio of the deal size to the target’s pre-deal asset value.

173

Fee to deal value

Shows the ratio of deal fees to deal value

183

Elite Advisor status

We classified financial advisors as “Bulge Bracket” (to borrow a US euphemism) or simple “Bank” depending on whether the financial advisor ranked in the top 10 for deal values during the entire 2000-2014 period. Similarly, we ranked law firms as “Big Law” or “Law Firm” depending on whether the advisor ranked in the top 10 by deal size for the whole period.

810

IFC membership

An advisor’s city qualifies as a top 10 International Financial Centre if the city scored in the top 10 on the Y/Zen IFC Index. We selected only the top city in a jurisdiction if the country has more than one city in the top 10 list.

645

Natural log panel

We constructed variables based on the natural log of deal size, law school score, finance school score, regulatory complexity score, Michael et al. score, and scale IFC ranking. As most variables range from 0 to 1, the natural log values range from 0 to -36.

-

* refers to advisor weighted numbers as discussed in the previous figure.

 

In order to find potential reasons for the development of world-class Going Out advisors in each jurisdiction, we compiled data on the variables we described in this paper. Figure B4 shows these variables, their description and the sources we used. We used law score and finance department scores from QS World University Rankings – accepting the obvious limitations and drawbacks that come with using this kind of data. As we discussed in the paper, we played with two measures of the complexity of financial law – one compiled by the company TMF and the other from a statistical analysis of World Bank data. We naturally used log values of GDP (to deal with the fact that some countries’ GDP can weigh it at hundreds of times those of smaller countries).[60] Our differentiation measure comes from our other data – as a measure of everything except the way a jurisdiction impacts on advisors’ ability to win Going Out projects.

 

Figure B4: Other Statistics Used

 

Variable Name

Description

No. valid*

Law School Scoring

We used the QS World University Rankings by for Law in 2014 to score cities according to the “best” law school in their city. When a city has more than 1 law school score, we chose the highest. 

643

Finance School Scoring

We used the QS World University Rankings by Finance programmes in 2014 to score cities according to the “best” finance programme in their city. When a city has more than 1 university with a scored finance department, we chose the highest. 

643

TMF Measure of Legal Complexity

Shows the ranking of various cities based on their financial law’s legal complexity. We rescaled the scores to a 0-1 scale.

600

Michael et al. financial law index

Shows the 0-1 score for each jurisdiction based on quantified similarities in hundreds of provisions of these countries’ financial law. The paper uses clustering/multi-dimensional scaling to find similarities statistically.

614

GDP

We used World Bank data reported GDP at PPP for 2013 at constant 2009 prices.

641

IFC Rank Score

We used the numerical score assigned to each jurisdiction in Z/Yen’s ranking of international financial centres.

557

Differentiation measure

We used multi-dimension scaling of our dataset (excluding all the factors shown in the current Figure) to arrive at similarities and differences in the ways advisors worked on Going Out deals.

537

* refers to advisor weighted numbers as discussed in the previous figure.

 

Figure B5 describes the jurisdictional “environmental” variables we used in our study. As shown, financial faculties had higher average quality scores than law faculties. With a mean of 0.94 (out of a possible 1), most of the cities we looked at scored pretty high on Z/Yen’s index anyway. Figure B6 continues the description – by showing simple correlations between these data with each other. None of these correlations comes in about 0.50 – with the exception of finance department quality scores and IFC ranking (a relationship we did not use in our analysis anyway). As such, multi-collinearity probability represented a very minor problem – if a problem at all. Our two measures of financial law complexity correlated by -0.17 – meaning that overall when one measure goes up, the other one goes down very slightly. For practical purposes, one could say that these variables had little relationship with each other – making the search of a better proxy for the complexity of financial law a necessity.[61]

 

 

 

Figure B5: Overview of IFC Characteristic Variables

 

Indicator

Valid N

Mean

Minimum

Maximum

Std.Dev.

Law School Quality Proxy

643

72.95

1.0

94.00

28.5

Finance School Quality Proxy

643

77.90

1.0

97.50

24.9

TML Measure of Legal Complexity

600

0.61

0.10

1.00

0.25

Michael et al. Measure of Legal Complexity

614

0.82

0.07

1.00

0.19

Differentiation Measure

537

2.02

0.85

5.93

1.43

ln GDP

641

15.10

6.92

16.63

1.49

IFC Rank Score

557

0.94

0.69

1.00

0.06

 

Figure B6: Correlations Between Various IFC Attribute Variables

(all correlations significant at the 5% level)

 

Variable

Law School

Finance School

TMF Measure

Michael Measure

IFC Rank Score

Law School

 

0.66

0.44

0.50

0.54

Finance School

 

 

0.34

0.48

0.75

TMF Measure

 

 

 

-0.17

0.43

Michael Measure

 

 

 

 

0.35

IFC Rank Score

 

 

 

 

 

Number of observations totalled 507.

 

Dealing with missing data

 

At the risk of seemingly overly cautious, we wanted to flag the issue of data ommission and potential bias for the reader’s consideration. We used a database (Zephyr) which obviously omitted aspects of the deals concerned (and possibly a range of deals themselves). We have no guarantee that our data serves as a random sample of some deeper population – and many times advisors’ cities were omitted. Many transactions reported no advisors at all.[62] We think that the database covered the major transactions and many minor ones of significance.  Imagining our data availability for a moment like a pyramid, the top (representing the largest deals, with the larger advisors in a range of sectors) probably pretty accurately reflects our target population. Looking down the pyramid, smaller deals and smaller advisors could likely have been omitted. Yet, our statistical analysis basically looks at differences while “sliding” down the pyramid. We don’t try to guess at the extent the bottom of the pyramid (where many deals and advisors might lie). We look at subtle changes in the top the pyramid to tell us something about the complexity of financial law, advisors’ differentiation strategies and law school qualities. We don’t care about painting a portrait of the entire Going Out advisory market. We only care about the part that deals with our research question.

 

We think even missing transactions would not affect our results for three reasons. First, our regression diagnostics – particularly residual analysis -- looked very good. If we had large numbers or types of omitted transactions affecting the quality of our data, we would expect to see the problems we encounter in noisy datasets. Second, unlike in most jurisdictions, the Chinese keep very good records about their companies’ investment activity abroad. After consulting the other resources we reference in our paper, we could find no obvious areas of omitted reporting. We do not think that a large gaggle of Mexican advisors (for example) were systematically omitted because they are not concerned with league table scores. Third, none of our analysis produced borderline results which the discovery of a new data trove might overturn. When relationships were statistically significantly, they were significant at the 1% level of less. Our discussions with advisors themselves tend to confirm the trends we describe.

 

Regression analysis

 

We used exceedingly simple regression methods because our analysis seems to suffer from none of the problems found in the other studies. We did not need to use instrumental variables or other fancy methods to deal with endogenously bias. Our dependent variables of advisor selection clearly did not influence structural issues in the jurisdictions we studied – like law school quality or the complexity of financial law. Our proxy for advisor differentiation naturally (and hopefully) highly correlated with attributes like whether these advisors worked in an IFC – as we constructed this proxy using such information. However, we did not use our differentiation proxy in any regressions – outside looking at the way it adjusted to jurisdictions’ structural attributes. Residuals were generally small and normally distributed. All in all, this project required rudimentary, pedestrian econometric methods.  

 

Our regression-based modelling work tested four alternate views of competition among advisors. As shown in Figure B7, in the first view, such advisor competition depends on attributes of the industry itself. In this view, differences in acquiring or target companies and their strategies drive advisor selection. In the second view, competition depends on advisors’ attributes – their size, their location and so forth. The third view might argue that competition depends on the attributes of the financial centres that advisors come from. Their legal complexity, law school scores, and similar variables make up this third view. The final view takes attributes of all these models into account. At first glance, the other models may look like strawman descriptions of advisor competition. Yet, recalling that regression analysis “penalises” the use of alot of random variables, models which mix variables from these various views of advisor competition might perform worse than than “purer” models.

 

 

Figure B7: Overview of Regression Models Tested

 

Model

Description and variables

Industry Characteristics Model

Tests the extent to which characteristics of the acquirer and target company determine attributes about their advisors and the financial centres they operate in/from.

Variables used: type of ownership, acquirer’s and target’s country, size and industry, whether acquirer buys a company from the same industry,  deal type (increase capitalisation, joint venture, etc), deal value and value of target company.

Advisor Characteristics Model

Tests whether attributes of the advisors play a role in firms’ decision to engage them.

Variables used: advisors’ type, city, transaction side, whether they sit in an IFC, status as an elite advisor.

IFC Characteristics Model

Tests whether attributes connected to the IFC which advisors work in determines advisor choice and/or choice of city.

Variables used: proxy for local law school quality, quality of finance department in the local university, extent of legal complexity.

Total Model

Tests the extent to which a combination of the previous models (as discussed in the Model section of our paper) explains advisor and IFC choice.

Variables used: A mix of the above.

The figure shows the philosophies giving the selection of different groups of variables as we tested the extent to which various factors determine advisor selection in Going Out deals. Econometricians can not just plunk variables into a regression at random and see which does do best. The world views shown in the figure provide differing logics for including and excluding variables in test regressions.

 

So what does the regression analysis say? The following represents the raw regression results for the regressions reported on in our study. We do not report on the various diagnostic tests and statistics used to assess the validity of these regressions. For some models, the regression obviously performs worse. We note these models with relatively low R-squared statistics – leaving out any discussion of these models’ defects. We do not show the range of models tested for our differentiation proxy as the dependent variable – as we could not test the range of models. We used the other variables to construct the differentiation proxy – thus using it again in regression analysis would generate spuriously good results.

 

Figure B8: Variables Affecting the Probability that the Target Companies’ Advisors Work in the Same City as Them

(standard deviations of b-value estimates shown in grey)

 

 Variable

 

 

Industry Traits Model

Advisor Traits Model

IFC Traits Model

Total  Model

Intercept

0.37

1.21

 

2.40

 

0.08

0.46

 

0.50

Acquirer Company Dummy variable

-0.02

 

 

 

 

0.06

 

 

 

Acquirer-Target Industries Match Dummy

0.03

 

0.01

0.00

 

0.07

 

0.04

0.00

Deal value

0.00

 

0.00

0.00

 

0.00

 

0.00

0.00

Pre-deal target company’s total assets

0.00

 

 

 

 

0.00

 

 

 

Difference in Acquirer’s city*

-0.06

 

 

 

 

0.03

 

 

 

Year

0.01

0.10

0.01

 

 

0.01

0.01

0.01

 

Advisor Type and Client

 

0.09

 

 

 

 

0.02

 

 

Probability of Elite Advisor

 

-0.19

-0.06

 

 

 

0.04

0.05

 

Probability of Advisor in IFC

 

-0.04

-0.12

 

 

 

0.08

0.11

 

GDP (in natural log terms)

 

-0.02

0.05

0.05

 

 

0.02

0.04

0.04

IFC Rank Score

 

-0.77

-2.07

-2.26

 

 

0.63

0.99

0.56

Number advisors

 

0.01

 

 

 

 

0.01

 

 

Law School

 

 

0.00

0.00

 

 

 

0.00

0.00

Finance School

 

 

0.00

 

 

 

 

0.00

 

TML Measure

 

 

0.57

0.58

 

 

 

0.16

0.16

Michael Measure

 

 

-0.60

-0.62

 

 

 

0.19

0.19

Adjusted R2

 

 

0.03

0.2

0.2

0.2

The figure shows the b-values and their standard deviations for the variables shown as regressed on the dependent variable show in the title. We have coded these variables as described in the methodology section of this paper. Adjusted R2 refers to the proportion of variance explained by the variables listed. The low variance explained probably stems from our omission of typical variables used (like M&A premia), noisiness in the original data and likely biases in the data which we discuss in this paper’s appendix.

 

Figure B9: Variables Explaining the Probability that the Chosen Advisor Comes from an Top 10 International Financial Centre

(standard deviations of b-value estimates shown in grey)

 

 

 Variable

 

 

Industry Traits Model

Advisor Traits Model

IFC Traits Model

Total  Model

Intercept

51.73

0.03

-5.49

-4.75

 

26.14

0.47

0.30

0.30

Prob Private

0.03

 

 

 

 

0.06

 

 

 

Industry Match Dummy

0.18

 

 

 

 

0.08

 

 

 

Deal value (thous EUR)

0.00

 

 

0.00

 

0.00

 

 

0.00

Year

-0.03

0.00

0.00

0.00

 

0.01

0.00

0.00

0.00

Pre-deal target company’s total assets

0.00

 

 

 

 

0.00

 

 

 

Acquirer city number

0.02

 

 

 

 

0.03

 

 

 

type and client

 

-0.01

 

 

 

 

0.02

 

 

Prob of Big

 

0.40

 

0.12

 

 

0.04

 

0.02

Prob of Target City Match

 

-0.19

 

 

 

 

0.04

 

 

Number advisors

 

0.00

 

 

 

 

0.01

 

 

Law School

 

 

-0.01

0.00

 

 

 

0.00

0.00

Finance School

 

 

0.00

0.00

 

 

 

0.00

0.00

TML Measure

 

 

0.07

-0.04

 

 

 

0.07

0.07

Michael et al. Measure

 

 

0.37

0.46

 

 

 

0.09

0.09

ln GDP

 

 

-0.08

-0.13

 

 

 

0.02

0.02

IFC Rank Score

 

 

8.06

7.78

 

 

 

0.25

0.27

Number advisors

 

 

 

0.00

 

 

 

 

0.00

 Adjusted R2

0.41

0.25

0.84

0.86

The figure shows the b-values and their standard deviations for the variables shown as regressed on the dependent variable show in the title. We have coded these variables as described in the methodology section of this paper. Adjusted R2 refers to the proportion of variance explained by the variables listed. The low variance explained probably stems from our omission of typical variables used (like M&A premia), noisiness in the original data and likely biases in the data which we discuss in this paper’s appendix.

 

 

Figure B10: Probability of Choosing an Elite Advisor

(standard deviations of b-value estimates shown in grey)

 

 Variable

 

 

Industry Traits Model

Advisor Traits Model

IFC Traits Model

Total  Model

Intercept

39.88

0.19

-4.58

18.73

 

 

 

24.84

0.05

0.69

20.39

Prob Private

 

0.00

 

 

-0.10

 

 

 

0.06

 

 

0.06

Industry Match Dummy

0.05

 

 

-0.10

 

 

 

0.07

 

 

0.06

Deal value (thous EUR)

0.00

 

 

 

 

 

 

0.00

 

 

 

Year

-0.02

 

0.00

-0.01

 

 

 

0.01

 

0.00

0.01

Pre-deal target company’s total assets

0.00

 

 

0.00

 

 

 

0.00

 

 

0.00

Acquirer city number

0.04

 

 

 

 

 

 

0.03

 

 

 

type and client

 

-0.02

 

-0.03

 

 

 

 

0.02

 

0.03

Prob of IFC

 

0.41

 

 

 

 

 

 

0.04

 

 

Prob of Target City Match

 

-0.18

 

 

 

 

 

 

0.04

 

 

Number advisors

 

0.01

0.01

0.00

 

 

 

 

0.01

0.01

0.01

Law School

 

 

-0.01

-0.01

 

 

 

 

 

0.00

0.00

Finance School

 

 

0.00

0.00

 

 

 

 

 

0.00

0.00

TML Measure 

 

 

-0.04

0.00

 

 

 

 

 

0.17

0.21

Bryane Measure

 

 

-0.02

-0.16

 

 

 

 

 

0.22

0.30

ln GDP

 

 

0.09

0.17

 

 

 

 

 

0.05

0.06

IFC Rank Score

 

 

4.51

4.26

 

 

 

 

 

0.59

0.74

Adjusted R2 

0.05

0.26

0.28

0,36

The figure shows the b-values and their standard deviations for the variables shown as regressed on the dependent variable show in the title. We have coded these variables as described in the methodology section of this paper. Adjusted R2 refers to the proportion of variance explained by the variables listed. The low variance explained probably stems from our omission of typical variables used (like M&A premia), noisiness in the original data and likely biases in the data which we discuss in this paper’s appendix.

 

 

 

 

 

 

 

 

 

 



[1] The literature on international financial centres has bulked out over the years – with authors finding evidence of agglomeration externalities, quality of life externalities (for attracting good advisors), historical accident, and good policies (Hansanti, 2008).

[2] We describe these in greater detail later. In brief, a legal complexity externality relates to the benefits that a jurisdiction’s complex financial law may endow on advisors from that jurisdiction. Authors like Kaplow (1995) theorised about such benefits without giving them a name. Blue ocean strategy relates to an advisory firm’s decision to compete in different markets or services than other advisors. Kim and Mauborgne (2005) provide the canonical description of the strategy.

[3] The Going Out Policy consists of a series of high-profile speeches and subsequent administrative decisions encouraging Chinese companies to engage in foreign direct investment. We do not have space to describe the policy or summarise its well-known effects. Salidjanova (2011) and Wenbin and Wilkes (2011) provide competent overviews of the Policy and the subsequent investment flows. 

[4] The Heritage Foundation database provides a broader view of Chinese cross-border investment than we use in our study of cross-border M&A activity only. One reason for showing this figure centres on highlighting the geographical discrepancy between these investments (many centred on resource-rich, frontier markets) and M&A activity. See original source for the definition of “contract.”  See Heritage Foundation, China Global Investment Tracker, available online.

[5] Until recently, the Ministry of Commerce had tried to record all Chinese investment abroad. Recent regulatory streamlining has since eliminated this requirement. See Ministry of Commerce, Circular of the General Office of the Ministry of Commerce on Setting up An Information Database of Foreign Investment Intention of Enterprises, MOC Circular 39/2003, 2003, available online. Wenbin and Wilkes (2003) provide an overview of these regulations over time.

[6] Figure 1a shows the flows of investments in general from the Heritage Foundation, while Figure 1b shows M&A-related investments recorded in the database we used for our own analysis. The figures do not exactly compare the same data. Yet, they illustrate the broader trend in the data – that the markets for advice are divorced from the flows of these investments themselves.

[7] In such round-tripping transactions, Chinese companies invest money in a foreign company with the express purpose of using that foreign company to raise money for activity back on the Mainland. The Chinese-owned foreign company sends the money back to the Mainland, where the company invests it and uses the relationship with the foreign entity it owns to pay back foreign investors more easily than China’s restrictive foreign exchange regulations currently allow. 

[8] The data seem to support such a hypothesis. Many countries hosting international financial centres possess very good World Bank Doing Business cores (with Singapore in first place in 2014, Hong Kong in third, the US in seventh place, the UK in eighth place and China in 90th place).

[9] The evidence for such “institutional arbitrage” remains far from convincing. Michael et al. (2015) for example find few differences between China’s banking regulation and regulation in several OECD countries housing top ranked international financial centres.

[10] We do not provide a fuller analysis of Figure 2. We only want to illustrate – using real data – how China’s cross-border M&A activity has led to a constellation of advisory relationships between international financial centres themselves as much as between the companies they advice. We describe and analyse these relationships more fully in a forthcoming paper. 

[11] Economists often refer specifically to heightened M&A activity as a natural experiment as such activity substantially alters variables of interest in an experiment in the same way a scientist would if he or she had the power to alter conditions in the real-world.

[12] Even after taking into account factors like language, gravity models still break down. Loungani and co-authors (2002) provide a solution similar to the one we propose in this paper – by postulating the importance of “transactional distance” (and scale economies). We describe these transactional distances in the empirical section of our paper.

[13] We refer to reported values as the actual value may differ significantly from reported values. We use one database of public M&A deals (Zephyr) which may or may not omit data. While we have no strong pre-existing reason to suspect that the database has significant holes, we urge the reader to exercise caution when reading and interpreting data we present in this paper.

[14] IFC-based advisors refer to advisors based in the largest international financial centres in a country – namely New York, London, Hong Kong, Singapore, Tokyo, Zurich, Seoul, Toronto, and Frankfurt. See the Appendix for a further discussion of our choice of these international financial centres.

[15] We categorized deals into 7 potential categories (materials, energy, consumer discretionary, consumer stapes, information technology, communications, and utilities). “Other” in the figure means one of these industries did not comprise a large proportion of its deals.

[16] We classify Bulge Bracket banks as those top 10 banks as ranked by the total value of Going Out M&A deals from 2000 to 2014. Similarly, we classify membership in Big Law by a top 10 ranking in the value of these deals. We show the members of these groups in Figures 14a and 14b.  

[17] We used the Z/Yen ranking of international financial centres for our top 10. We only selected one city from each jurisdiction (in cases like the US where a jurisdiction had several top ranking cities). The top 10 includes New York, London, Hong Kong, Singapore, Tokyo, Zurich, Seoul, Toronto, Luxembourg, and Frankfurt.  

[18] Many commentators have already noticed that importing Hong Kong banking and securities law to the Mainland has led to more competitive financial and legal advisors (Alford, 2007)

[19] The difference is statistically significantly different at the 5% with a p-value of 0.02.

[20] Modelled fee income refers to estimated advisory fees earned by the advisor, as reported by the Zephyr database. 

[21] Segal-Horn and Dean (2011) describe the nature of such global competition among elite law firms while Crotty (2008) describes such competition for global investment banks. The firms which dominate Going Out deals also dominate deals from other regions. Yet, the size and nature of Chinese deals represents a new market for the entire industry. Thus, observing advisors’ responses to Going Out provides a view of such competition unhindered by the legacies of prior years’ competition. 

[22] We can not ensure that the database contains a complete set of Going Out transactions – and many transactions omitted numbers for such modelled fee income. As such, we present these data mostly for the sake of completeness (offering the data we have available to the reader) rather than to provide a lynchpin for our argument.

[23] The evidence suggests that Big Law does not produce superiour returns than other law firms. Krishnan and Laux (2007) find no relationship between long-run M&A returns and choice of legal advisor. Yet, the same author with another author Krishnan and Masulis (2013)) find a positive relationship. .

[24] Such “complexity” refers simply to the number of provisions in banking and securities law – as a proxy for the types of contingencies covered by the law, types of transactions and so forth. As we do not discuss the specific content of this law, we do not cite the standard references in the law and finance literature or talk about the way this literature has evolved. See Heremans and Bosquet (2011) for a fascinating discussion of the way this literature has evolved.

[25] The jurisdictions in question include the acquirer’s home country, the target’s country and the law of advisors on each side of the transaction.

[26] Some transactions show more than one advisor, with a large number also showing none. Given the secretive nature of international business, we think the reported number of advisors-of-record understates the actual employment of these advisors.  

[27] With the exception of Klasa et al. (2013), few scholars have actually researched the role that law firms play in raising the profitability of an investment transaction or a client.

[28] Cole et al. (2010) for example find that

[29] Offshore financial centres seem at first glance an exception to this rule. The Cayman Islands, British Virgin Islands or Bermuda do not create large tombs of financial law. They have reputations for having little regulation. Instead, they use law in other jurisdictions – acting only as a conduit (Gordon, 2010). 

[30] The stock of financial law clearly represents a public good – a non-excludable and non-rivalrous good for use by financial service firms in ALL jurisdictions (Seigel, 2005). The creation and use of banking and securities law in London for example clearly benefits third parties in developing countries who use that law. “Better” transactions (lower risk, higher return) which result from using such law clearly benefit contracting parties as well as consumers who use the goods and services of transacting companies, their suppliers, bankers, and other third-parties in a position to benefit from the increased producer welfare enjoyed by transacting parties.  

[31] Governments themselves clearly provide the diffusion, copying, and adjusting of laws to more closely match US and UK law (Chaffee, 2010). Policy entrepreneurs in aspiring financial centres may also seek to confer the same positive externality public goods on their financial service firms which those in New York and London enjoy (Prentice, 2005).

[32] The benefit of “statistical discrimination” consists of finding differences using variances in variables, even without knowing why those differences occur. As we describe in the Appendix, we do not try to interpret the dimensions of difference found by our statistical analysis – in order to keep from over-interpreting our data.

[33] We describe in the Appendix the multidimensional scaling procedure we used to assign the variance in a range of variables into two constructed variables.

[34] The authors contrast “blue ocean” competition (an area of competition relatively free from rivalry) with “red ocean” competition – the bloodied area of rivalry where intense rivalry leads to low profits.

[35] These probabilities represent the number of mandates won by advisors in these jurisdictions divided by the total number of mandates. If we had to guess what the distributions of mandates for the next year would be, these relative frequencies provide the best guesses (before taking into account factors that might affect these probabilities).

[36] In this model, places like Mexico City have no chance of attracting Going Out mandates – at least in the short run. The reader may naturally critique such a view as too deterministic – forbidding the chance for new urban centres to join this “geography of finance.” We accept this critique – and warn readers that our model can not predict new centres of Going Out advisory work.

[37] In probability terms, these distributions resemble Fatigue or Beta distributions (which show very pointy modes). In regression terms, these distributions might be described as piecewise continuous geometric relationships. We want to keep our discussion simple enough for a non-economist to understand, so we do not characterize these equations more here.

[38] A simple glance at the shape of the data in Figures 26 and 27 show the resemblance between our model and the data. We do not overlay the data with the Figure in order to keep an already cluttered figure simple.

[39] We discuss in other sections the probability of a city’s advisors winning these mandates – regardless of whether the city houses the target company’s headquarters or not.

[40] As the advisor’s country grows, the advisor will have other deals to chase besides Going Out deals. Moreover, the most competitive IFCs sit in small countries (on a GDP basis). Thus large countries like Germany, Italy and even Japan clearly compete at a disadvantage relative to the Cayman Islands and British Virgin Islands.

[41] A negative relationship between law school quality and the probability of choosing an elite advisor means that as a jurisdiction’s law school quality scores rise, Going Out firms choose advisors from that jurisdiction, rather than from large IFCs. These companies thus appear to choose elite advisors when they can not be sure that advisors in a jurisdiction have a superlative legal education. As we point out, the same relationship does not apply for a finance-related education.

[42] As a reminder to the reader, the dataset we used may have omitted data about Going Out transactions – including data about the number, type, and location of advisors to these transactions. 

[43] Such a result contradicts research showing that economics education makes individuals more selfish. We do not try to explain the underlying reasons for our discovery.

[44] Michael (2014) reviews the way such “objectives-based legislation” delegates authority to regulators and reviews the performance of such regulations on financial stability.

[45] Anti-solicitation rules work for both the solicitor and solicitee sides of the attempted transaction. If rules in the US (for example) prevent US broker-deals from soliciting Chinese clients, then Chinese rules prevent US banks from soliciting and operating in China without the appropriate authorisations. Both sides of the relationship have relied on work-around “safe harbour” clauses to get around a system basically targeted against them. Easing such safe harbour rules or even redesigning rules completely to allow for better regulation of cross-border advice could help deepen markets for all advisors. 

[46] Krugman’s Razor follows from a well-read paper Paul Krugman (1995) wrote about the role of modelling in economic and development theorising. Models use the least amount of complexity to sufficiently adequately explain the real-world and generate predictions. In line with this approach, we choose the least complex empirical methods as well – using simpler methods to illustrate results we may have checked using more sophisticated methods. Krugman also persistently argued for the use of plain, simple English when making economic arguments. As such, we try to write our paper in a non-technical way which any educated and interested reader could follow.

[47] We do not deny the role of more obvious variables that other authors have already studied – like the role of institutions, past profitability of advisors’ mandates, probabilities of success in closing a transaction, and other variables. We only wish to step back, and look at some of the deeper drivers which explain why these variables have their explanatory power.

[48] We could model n mandates out of a possible total of N, of equal value q -- such that all qs add up to Q. Clearly, if a firm wins n/Q mandates of q value, the value would equal the probability p of winning one Q.

[49] We could have used the proportion of each advisors mandate value to the total value offered. However, the “spirit” of the analysis in our paper clearly looks at winning mandates, regardless of their value. Thus, we focus on the way advisors win work – without regard to the size of such work. Anyway, we show that no relationship exists between deal sizes and types of advisors. Thus, providing a model explaining geographical differences in deal sizes would not help us when the real-world shows no such differences exist.

[50] Such an assumption simplifies the model – and may not represent a wild departure from reality. Chinese firms have a higher degree of coordination than we are used to in the West – with extensive coordination through professional bodies and State control over finance, foreign exchange and strategic sectors for foreign investment. Even if firms don’t coordinate directly, government control over various aspects of company behaviour may have the result of coordinating their foreign investment and advisor choice.

[51] In other words, “interaction effects” exist between the variables in our study. Again, following the Krugman’s Razor principle, we do not discuss these interactions at any length – so this would make our already 100 page paper even longer. Future researchers can analyse first-order interaction effects between these variables if our work garners any interest.

[52] In theory, the amount of money Going Out companies need to pay for advice decreases the probability that advisors from a particular jurisdiction might attract increases in their probability for being hired. As we discuss in the paper, differences in advisory fees seem statistically insignificantly distinguishable from jurisdiction to jurisdiction. Moreover, in a supposedly global market for advisory talent, these differences should remain small on a quality-adjusted basis.

[53] The function “Rank” refers to the operation we describe in the text of mapping each city’s advisors’ probability of receiving Going Out deals into a discrete, ordered-list ranking of these jurisdictions – and not to the “rank” of a matrix.

[54] Advisors’ profits change with legal complexity as and with their own differentiation as . In order to provide some of the intermediate steps that other researchers need to reproduce our results, using Cramer’s Rule, we find  and .

[55] In theory, by taking logs of our variables, we can se linear regression to test a form of equation (20) as , where b1 from the standard regression equation equals (D-d) and so forth.                          

 

[56] A logit model assumes that the probabilities we work with follow a logistic or normal probability. We do not need to impose a structure the relationship between our variables (like a logit model does).

[57] The actual distribution of best fit – a fatigue distribution – basically does the same thing. The fatigue equation “penalises” observations far away from its very leftward skewed mean.

[58] For example, if the acquirer reports 2 advisors and the target reports 3 advisors, we would record 5 advisors tied to that one transaction. The 2 acquirer advisors have 6 relationships with target company’s advisors. The 3 advisors have 6 relationships with the acquirer 2 advisors. In total, these 5 advisors have 12 relationships. 

[59] In the following lists, we use abbreviated rather than full names. Law firm names especially can be lengthy, and we use only enough of the name to clearly identify which organisation we are talking about.

[60] In fact, we use log values to deal with the geometric relationship inherent in much of our data. As we suggest in our model, the size of these variables affects the way they impact on the probabilities of firms obtaining Going Out work. Taking logs of data represents one way to isolate the role that the size of a country’s financial law or other variables play in determining the way a change in that size affects the probability of obtaining Going Out work.

[61] As we describe in the main paper, the Michael et al. measure uses objective measures of law, rather than TML’s subjective measures. As such, we relied more heavily on that measure.

[62] Lack of advisors could just mean reporting companies handled their own financial and legal work. Indeed, de Jong and co-authors (2010) treat lack of advisors as a situation where companies chose no advisor! Given our knowledge about the extent to which companies protect their privacy and secrecy, we tend to be more chary about assuming that advisors did not participate in particular transactions. Yet, as we explain in this Appendix, we don’t think that mis-reporting produced significant bias in our dataset.