Bank Affiliation and Co-Investment in the Financial Crisis – “Negative Certification” Effects for Private Equity?

1

Master in International Finance

Master Thesis

Bank Affiliation and Co-Investment in the Financial Crisis – “Negative

Certification” Effects for Private Equity?

Jayme Rafael Bossi de Castro

Thesis Supervisor: Prof. Stefan Arping

2

Contents

Abstract ... 3 Introduction ... 3 Literature Review and Hypothesis ... 6

Captivity in Private Equity Firms 6

Bank Affiliation background 8

Financial Crisis and Future 11

Hypotheses Development 13

Sample Description ... 16

Criteria and Capture 16

Brief Analysis 17 Methodology ... 18 Results ... 22 Robustness Check 25 Conclusion ... 27 References ... 29 Appendix ... 31

Bank Affiliation per Country 31

Bank Affiliation per Industry 32

Variables Description 33

Summary Statistics 34

Dependent Variables – Distribution 35

Panel Regressions – Individual Effects 36

3

Abstract

Banks play fundamental roles in the Private Equity industry, be it by providing funding (mainly to LBO’s) or services (e.g.: underwriting, consulting, etc). It is therefore presumable that their involvement in the direct ownership of Private Equity firms should lead to exceptional performance as result of synergies, certification properties, networking and other strategic advantages.

By analysis three metrics of financial soundness for 715 private companies in Europe between the years of 2005 and 2013, this study aims to verify whether such advantages persisted through the financial crisis or if the association to banks amidst bailouts and scandals had a negative impact on the companies managed by bank-affiliated PE-firms and their partners in co-investment.

Despite of finding significant influence of the ownership type over the solvency of companies in the crisis period, we find no significant relation between bank affiliation and RoA or Financial P&L, thus contradicting some of the assumptions traditionally made by PE professionals and researchers.

Introduction

The quality and the strength of relationships play a central role amongst the sources of value creation and improved performance traditionally attributed to Private Equity firms (Bloom, Sadun, & Reenen, 2008). The influence of the connections maintained by the general partners can be observed in virtually every stage of the investment cycle as we know it - from the identification of opportunities for

4

investment, to the options for disinvestment, passing by the ability to raise funds and to establish operational synergies, unlocking value in the portfolio companies. Such relationships present themselves in multiple forms, such as commercial agreements with strategic suppliers and customers, partnership in co-investment and participation in club deals, or even the direct affiliation to another corporation, government or financial institution.

Theory suggests that in the particular case of affiliated or captive firms, the characteristics of the major shareholder to whom the Private Equity (or “PE”, as we’ll be referring to it along this paper) firm is related can largely determine, among other things, its long term strategy and competitive advantages, as well as the style, capacity and industries of choice for its funds, given the influence that different owners can have over factors such as the volume of resources made available to the firm, the goals of the ultimate owners and the incentives to the asset managers.

Although the potential conflicts of interest resulting from this close alignment of the PE firm and the paternal institution can make the participation in co-investments or direct investment (when possible) in funds managed by captive firms unattractive for many investors, the possibility to “free-ride” on the reputation, expertise and network of the paternal company can, in theory, outweigh such issues and lead to improved returns.

This dilemma is further aggravated in the case of PE Firms affiliated to Banks when taken into account the pros of supposed facilitated access to funding at lower costs, and the cons of the eventual secondary interests of a bank in selling financial services to the portfolio companies at the expense of third parties involved in the deals.

Added to that, theory suggests that the access to non-public information by banks, through the offering of financial products to private companies, provides them with strategic informational advantages, resulting in more accurate valuations of the companies purchased and consequently better investment

5

returns (Chen & Martin, 2011), besides certifying their quality towards the market (Duarte-Silva, 2010), thus leading to higher valuations when exiting deals.

Although apparently intuitive in a first moment, these concepts have been confronted by empirical studies, such as Fang, Ivashina, & Lerner (2013), who after analyzing U.S. transactions between 1983 and 2009, found support neither to the alleged better investment abilities, nor to the certification benefits of bank affiliation, still reporting bank-affiliated deals to show slightly higher probabilities of bankruptcy and downgrade.

However, as ponders Tereza Tykvová (2006) and other authors, to assume that the characteristics and relations observed in the U.S. market are necessarily representative of the entire Private Equity industry across the globe, might lead to simplistic conclusions that overestimate the homogeneity of the globalized markets, not taking into account the particularities of different countries and the still substantial differences between institutional environments in economies such as the U.S. and the European Union (Cumming, Fleming, & Schwienbacher, 2006). Such differences are likely to have been intensified even further in the recent years, driven by the global financial crisis of 2008 and European crisis of 2010, and their subsequent effects over regulations, banks’ ownership and macro-economic scenarios, suggesting that a more complete and specific understanding of the European post-crisis Private Equity industry still requires construction.

Is it possible that in Europe, factors such as the loss of credibility of banks, combined with credit downgrades, bonuses cuts, and overall more risk-averse times might have led the “bank ownership” effect (if any), to turn from positive to negative?

The announcement by the European Parliament in January of 2014 of its intention – still subject to approval in higher instances - to impose tighter restrictions to the involvement of large European Banks in Private Equity, in an “European version” of the American Volcker Law, increases the demand for

6

academic production on the topic in order to enrich the debate over the consequences that the participation of banks in proprietary investment activities in PE brings to the economy and the society.

Taking into consideration the questions introduced above, this study aims to contribute to the previously existing literature by analyzing the effects that the exposure to the global financial crisis brought to companies managed by bank-affiliated Private Equity firms during the crisis and the years immediately before and after it (assuming its main phase is finished).

Literature Review and Hypothesis

Captivity in Private Equity Firms

Private Equity funds can be sub-categorized in many different ways given the generalist nature of the term. Besides the possibility of classifying them according to their industry of specialization - if any-, or their investment style (e.g.: Venture Capital, Angel Investment, Leverage Buyouts, Mezzanine Debt..), a secondary but not necessarily less important aspect used for this purpose involves the characteristics of the PE firm who runs them, namely whether its management is captive or independent. Below we formalize the two concepts as proposed by previous literature:

Captive Firms: Firms controlled by a major institutional investor, such as corporation, a government, a university endowment, or a bank. These firms and its funds are fully or mostly dedicated to the management of investments of a major shareholders’ own resources.

Independent Firms: Firms dedicated to the management of third parties resources, at the cost of commissions or fees as payment for their investment services, dedicating only limited proprietary resources to the deals. The majority of the existing Private Equity firms.

7

The importance attributed by investors to such factor is naturally connected to their individual goals, mandates, and profiles. Whether and how captive and non-captive investments differ in terms of performance, investment style and other characteristics is still objective of extensive debate.

In a qualitative research interviewing 15 Venture Capital fund managers, Osnabrugge & Orbinson (2001) reported independent VC firms to have stricter investment processes than captive ones, attributing the difference to the supposedly looser agency pressure to which captive funds are subject. Dufour, Nasica, & Torre (2013) find that independent VC Private Equity firms participate in more profitable syndicated projects than captive ones, while Fang, Ivashina, & Lerner (2013) find no material differences between target industry and investment’s evaluation processes between different independent and bank-affiliated firms after interviewing a number of asset managers in the U.S.

Added to the level of independence of a firm, there is strong evidence that the individual characteristics of the major sponsors- such as governments, endowments, banks, or strategic corporative investors - are to a lower or higher extension inherited down the ownership chain, influencing characteristics of the funds and portfolio companies invested.

Tzikvová (2006) reports that in the German market, corporate private equity firms tend to have a more pronounced role in corporate governance and monitoring of the portfolio companies, besides directly investing in larger portions of equity of the underlying companies, whereas bank-affiliated and governmental PE firms tend to act only as bridge investors. Davis, Grob, & Haan (2007) also observe influence of the sponsor’s characteristics in the pension fund industry, after analyzing funds captive to 100 Dutch sponsoring firms and reporting a relation between the level of leveraging of the sponsor and the cover ratio of its funds.

Theory suggests that this influence of the main investor characteristics takes place through mechanisms such as the volume of resources made available to the PE firm (Chen & Martin, 2011), the goals of the

8

ultimate owners (Hellmann, Lindsey, & Puri, 2008), or the incentive structures to the managers (Lerner et al. 2005). However, in the very particular case of bank-affiliated firms, the influence of the major shareholder over the PE firm strategy and portfolio company’s operations can take place in different and combined manners, especially in the cases where the paternal bank accumulates other functions such as the one of debtor or service provider to the deals entered by the PE firm, many times in syndication with other banks, resulting in even more complex agency conflicts

Such a fact led Fang, Ivashina, & Lerner (2013) to propose the distinction of bank-backed deals between simple “bank-affiliation” and “bank paternal sponsoring”. In the first case, the involvement of the bank is limited to the regular exposure obtained through the participation in the deal’s equity, while the second involves also the participation of the paternal bank as the funding provider to the deal syndication, in the case of Leveraged Buy Outs. Among the findings of their research, it is reported that such paternal-sponsored deals enjoy better financing terms, while regular bank-affiliated deals seem to face the same, or slightly worse, terms than the ones backed by independent PE firms.

For the purposes of this paper and for the sake of simplicity we do not make distinctions between bank-affiliated and parent financed deals in this study, defining as “bank-bank-affiliated” any companies who had bank-affiliated PE Firms among their investors as classified by Orbis database. Nevertheless, in order to better understand the implications and the issues that the different types of bank involvement can have for the multiple stakeholders, it is important to understand the incentives that lead to bank affiliation in Private Equity, as covered in the next section.

Bank Affiliation background

Although the supposed reasons for the intense participation of banks in Private Equity (about 30% of all deals, according to Lopez-de-Silanes, Phalippou, & Gottschalg, (2011)) are many, existing literature has

9

traditionally focused on 3 fundamental factors to which the occurrence of bank affiliation in Private Equity is attributed:

Leveraging Capability: Firms affiliated to banks in theory enjoy facilitated access to funding and thus the possibility to optimize the leveraging level of their individual portfolio companies at lower costs and under looser covenants (Hellmann, Lindsey, & Puri, 2008; Fang, Ivashina, & Lerner, 2013). This privilege of immense strategic importance – especially in times of reduced availability of funding in the regular market – benefits the sponsor and the PE firms mutually. On the one hand, the enhanced credit capacity enjoyed enables the PE firm to aim for larger-scale investments in terms of size and leveraging - such as Leveraged Buy-Outs - with lower dependency on pooled loans. On the other, the paternal bank benefits from providing finance to syndicated deals structured by the PE firm, acting as an anchor investor and then profiting from lending to the whole deal syndicate, thus obtaining exposure to the companies’ equity while transferring great portion of the risks and debt burden to the co-investors (Fang, Ivashina, & Lerner, 2013). Affiliated companies also tend to use parent debt as substitute to external loans in countries where access to external debt is limited or expensive, as in many developing countries (Desai et al. 2004)

Cross Selling and Relationship Management: The involvement of a bank with Private Equity investment can be also influenced by the importance of the maintenance of relationships with corporations, clients, and other banks for its core lending activities. By retaining a portion of ownership or providing funding to a PE deal, the bank increases its chances to sell other services and products to the company or the other owners in the future (Duarte-Silva, 2010; Hellmann, et al. 2008). Just as in the case of the funding aspect, the sale of these additional financial services increments the total result for the bank, while partially transferring costs to the co-investors. At the same time, the PE firms and portfolio companies also benefit from the improved access to otherwise costly financial services, such as underwriting and

10

other M&A related services (Drucke & Puri, 2005), advisory and consulting (Hellmann, et al., 2008) and lower costs when going public, and lower underpricing of their IPO’s when going public (Chahine & Filatotchev, 2008; Erhemjamts & Raman, 2012).

Still in the dimension of costs-saving, beneficial synergies can potentially emerge from the overlap or integration of vital functions such as risk management, reporting, middle-office, and IT infrastructure, enhancing the ability of a PE firm to meet the increasingly stricter regulatory requirements of monitoring and accountability in the funds industry.

Information Asymmetry & Certification: A third and perhaps most important hypothetical reason for bank sponsoring in PE resides in the opportunity enjoyed by both the portfolio companies and the PE firm, to benefit from the brand and reputation of the paternal institution, associating their images, the quality of their services/products and their credit profiles to those of the bank backing their operations, even if in association with other investors.

The precise mechanisms through which this reputation pegging occurs are not always clear and seem to vary across countries and markets (Cumming, Fleming, & Schwienbacher, 2006), however the most well accepted view, and in line with information asymmetry theory, is that the certification effects are closely linked to the well-known limited availability of information on Private companies. The pre-existing relationships of banks with private companies, through the offering of loans and other financial services, grants them access to confidential information about the financial condition and performance of private companies (Duarte-Silva, 2010; Kracaw, 1980; Fama, 1985; Diamond, 1991), this enabling them to take well-informed investment decisions based on more accurate valuations (Chen and Martin, 2009). Although the transference of confidential information about clients across areas within a bank is, at least in theory, forbidden and some studies indeed report not to find confirmation of a superior investment ability of bank-affiliated firms (Fang et. Al, 2013), the participation of a bank (a better

11

informed participant) in a deal theoretically signalizes positively to external parties, influencing the quality of relationships maintained by the portfolio companies.

The hypothetical positive or negative effects of bank-affiliation certification are most clearly observable at the moment of the exit of an investment through an IPO or a public sale to a strategic buyer or another fund. Chahine & Filatotchev (2008) observed that bank affiliation to the IPO underwriter leaded to lower underpricing levels of French IPO’s compared to the public offers ran by non-bank-affiliated firms. On the other hand, Osnabrugge & Orbinson(2001), who analyzed the impact of the underwritting bank’s reputation over the prices of the shares of private placements, and how pre-existing investment relationships (such as affiliation) between the issuer and underwritter influence the price discount and gross spreads, found an inverse relation between bank-relationship and the fees paid for underwritting services, suggesting that banks may take advantage of privileged information and a supposed dependence of theses firms with whom they have a pre-existing relationship to impose higher placement agent fees.

Financial Crisis and Future

The reverberations of the Global Financial Crisis of 2007-2009 and the European Credit Crisis of 2009-2010 brought substantial changes to the defense of interests and relations between banks, investors, government and the society as a whole. If in one hand the Private Equity industry in 2014 indicates a strong recovery of fundraising, with capital overhang estimated in more than $ 1 trillion by the end of 2013, on the other, much of the privileged negotiation power enjoyed by PE managers back in the years when many of the studies referenced by this thesis were written is gone.

The greater risk aversion and scrutiny from policymakers and investors has increased the focus on costs reduction and more reasonable compensation arrangements, leading to an observed flexibilization of

12

the traditional “2 and 20” agreements (2% of administration fee and 20% of carried interest), popularization of deal-by-deal compensation, and more investor-friendly preferred returns terms (McCahery & Vermeulen, 2013), in a trend that might reduce – or aggravate - the differences between incentives in bank-affiliated and independent PE firms, and consequently their riskiness and management style. At the same time, the advance of deal syndication as instrument for risk mitigation must further increase the importance of strong institutional partnerships for PE investment.

Notwithstanding, no post-crisis change has brought – and will bring - more impacts and repercussion for the bank-affiliated category than the new regulatory scenario for banks. If the passage of the Gramm-Leach-Bliley Act in 1999, which suspended the prohibitions of the transference of information between lending and overwriting activities (Gummi, 2013) imposed by the Glass-Steagall Act, is deemed by many as a milestone among the events that led to the crisis, the proposal of the Volcker Rule in 2010 as part of the Dodd Frank framework, restricting the engagement of banks in proprietary trading and Private Equity investment is likely to bring irreversible structural changes to the Private Equity industry.

By limiting the interest of a bank in a fund to a maximum of 3% of ownership, and not exceeding 3% of the bank’s Tier 1 capital, the Volcker Rule nearly forbids the direct participation of banks in PE deals, redefining – if not extinguishing – the concept of bank-affiliated PE firms in the U.S. Aiming to avoid the circumvention of the subjective concept of “ownership” and “proprietary trading”, the law requires also the divestment of CLO’s (collateralized loan obligations), frequently used in parent bank-sponsored deals, and one of the main pillars of Leverage Buy-Out structures. After a series of revisions of the deadlines, banks are currently expected to align to such requirements before July of 2017.

While on one side, supporters of tougher regulation, such as policy makers, individuals and many academics argue that such restrictions are a necessary reform and still a very small price to be paid for institutions who enjoy the “safety net” of the too-big-to-fail status (Volcker, 2010), on the other, many

13

market participants, bankers and academics criticize, among other aspects, the economic consequences that the intervention might bring, such as loss of competitiveness, inability to manage risk and tightening of credit (Gummi, 2013), but also questioning the role that the Private Equity industry and its leveraged loans had in aggravating the crisis (Brissette, 2011).

In the beginning of 2014, the European Commission announced its proposal of similar legislation, with expectation to go effective also by 2017, and to be applicable to “the largest and most complex” European banks involved in trading activities. Although less specific than the American Volcker rule, focusing on concepts rather than specifications, the law signalizes a response to the public insatisfaction with the benefits previously enjoyed by banks - ultimately exemplified by the bailouts though the LTRO program, which consequently resulted in many European banks becoming government-owned. Furthermore, it represents an attempt to prevent that the involvement of banks in supposedly risky and non-innate activities, such as Private Equity sponsoring, may bring undesirable risks to the society and eventually lead to a new crisis. Such facts leads the author to hypothesize that both the ability of banks to certificate Private Equity - and their preferences when investing - must have been affected by the after crisis scenario.

Hypotheses Development

By assuming, based on the work of previous authors, that bank-affiliation certification has significantly positive correlation with many of the aspects that affect the performance of a private company; it might be plausible to assume that the financial condition and public image of portfolio companies is pegged to those of their sponsoring banks also in the pessimistic scenario of a credit crisis.

14

In these circumstances, besides the credit deterioration and damaged public image, it is presumable that the lower risk tolerance, as result of tougher regulation and scrutiny from the public, should lead to the choice of less volatile – and theoretically, with lower returns – companies for bank’s portfolios.

Finally, the legal and moral restrictions to the payment of extraordinary bonuses to bank executives and front-office, might have affected to some extent their PE firm affiliates, thus compromising the incentives structure and results of their portfolio companies.

The consideration of the possibility that the financial crisis led the association to banks to have a negative effect on the performance of portfolio companies, even if temporarily, is the basis for the first hypothesis to be tested in this work:

H1: The ownership by Bank-Affiliated PE Firms presented lower or negative influence over the results of PE companies after the beginning of the Global Financial Crisis.

Second, and taking into account the previously discussed growing importance of co-investment in the post-crisis Private Equity industry, we test whether a mere “second degree” relation with bank-affiliated PE firms may be sufficient to influence the returns of Private companies. For such purpose we formulate the following hypothesis:

H2: The ownership by PE Firms who had Bank-Affiliated PE firms among their main co-investors presented lower or negative influence over the results of PE companies after the beginning of the Global Financial Crisis.

Next to that, and assuming that a reduced willingness and/or ability of the Bank-Sponsors to lend could have directly impacted the funding waterfall, impacting the Solvency of the portfolio companies, we

15

compare the situation of their solvency ratios before and after the crisis through the following two hypothesis:

H3: The ownership by Bank-Affiliated PE Firms presented lower or negative influence over the solvency ratios of companies PE companies after the beginning of the Global Financial Crisis.

H4: : The ownership by PE Firms who had Bank-Affiliated PE firms among their main co-investors

presented lower or negative influence over the solvency ratios of companies PE companies after the beginning of the Global Financial Crisis

Last, assuming that the limitation to borrow from their parental bank sponsors might have led these portfolio companies to make use of more expensive alternatives of funding, such as mezzanine or subordinated debt to externals, and end up with worse terms, thus increasing their expenses with interest, we build up our last two hypothesis:

H5: The ownership by Bank-Affiliated PE Firms presented lower or negative influence over Financial Result of companies PE companies after the beginning of the Global Financial Crisis

H6: The ownership by PE Firms who had Bank-Affiliated PE firms among their main co-investors presented lower or negative influence over Financial Result of companies PE companies after the beginning of the Global Financial Crisis

In the next section, we specify and provide details on the data used in our study, before discussing the statistical approach utilized.

16

Sample Description

Criteria and Capture

As starting point for the research we selected a sample of over 5000 Private Equity Firm-owned companies from Orbis, a database maintained by the information company Bureau Van Dijk containing information on 130 million of listed and non-listed private companies worldwide. Our selection considered only companies that were linked to PE firms as portfolio companies, this way ignoring other regular private companies, such as family businesses or subsidiaries. This classification is made by Orbis. Considered the objective to contribute to the debate on regulation for bank-affiliation in Europe and in order to avoid the contamination of the data by particularities of other PE markets, such as the U.S., we selected only companies in the region of the extended European Union (28 countries). This option also takes into account the incomparability between markets for this period given that American, European and Emerging markets were affected by the crisis in different way and timing.

Next to that, considering the focus on the specific period of the Financial Crisis and the accentuated structural changes that have occurred in the PE industry in the last decade, we defined the scope of the sample as the years between 2005 and 2013, this way also trying to avoid contamination by the early 2000’sinternet bubble.

We then further refined the selection based on the number of observations per company in order to avoid excessive intermittence of data and companies who were either added to the database too recently or who had an unrepresentatively short life. For such purpose, companies that reported no RoA for more than 3 times between 2007 and 2011 were discarded.

Also for the sake of comparability and due to availability of data we restricted our sample to mid-sized companies, using as criteria only the operating revenues dimension, frequently defined as the range between 10 and 50 million EUR of revenues per year, which roughly converted do USD resulted in our

17

lower and upper thresholds of respectively 12 and 60 million USD per year. The main fundament for the decision is that while very few small companies provide complete financial information, which could lead to an overweight and consequent selection bias towards large companies, companies that are too large may not be sensitive enough to the influences of the paternal figure, as they frequently engage in multiple businesses and have their ownership dispersed among multiple partners.

After obtaining our initial sample of companies, we matched them with their respective PE firms and their natures in order to identify Bank Affiliation and main co-investors. Since detailed PE firm information is not available in Orbis we obtained this information from Thomson One database of companies and PE firms. Where the link between the company and PE firm was not directly available, keys were created and utilized to link them based on the company name. For the PE firms where this information was available, a list of the five main co-investors was also generated in order to identify Bank-Affiliated partners. The result after the application of all criteria was a sample of 715 companies.

Brief Analysis

From the 715 companies covered in the research, 84 were identified as Bank-affiliated. Such proportion is in line with previous findings of Hal S. Scott (“nearly 12%”), Professor of Law Systems at Harvard Law School, suggesting good representativeness of the sample, however they are well under the 30% observed by Fang, Ivashina, & Lerner (2013) in the U.S. The percentage of companies ran by PE firms who had the bank-affiliated among their main co-investors was of approximately 18%.

Table 1 provides details on the distribution of Bank Affiliation (1.1) and Bank-Affiliated Co-investment (1.2) per country observed in the sample. Although the proportions in an individual basis may be distorted in the case of countries with too few observations (e.g. Lithuania), in other cases, it reveals interesting and meaningful information on country-specific characteristics (e.g. Belgium).

18

The substantially higher proportion of bank affiliation in countries such as the UK, France and Germany suggests that bank-involvement in Private Equity increases with institutional quality and financial market development, while the nearly absolute absence in Eastern Europe seems to contradict the theoretical importance attributed to certification in inefficient and emerging markets. On the other hand, the unpopularity of bank affiliation in Nordic Countries despite of the relatively large number of observations seems conclusive and coherent with the studies of La Porta (1998) about the influence of legality systems over ownership behavior: certification through affiliation to a major shareholder has little or no relevance in countries with strong investor and creditor protection mechanisms – besides being a leader in these two aspects, Scandinavia also presents the best (lowest) scores for risk of expropriation and corruption in the world.

Intriguingly, this pattern is not reproduced in the participation of bank-affiliated co-investors - 3 Scandinavian countries among the top 7 - which might be attributable to some concentration of deals around a same major non-bank player, or the difficulty of foreign banks to penetrate such market, requiring partnerships with local firms and limiting the role of banks to the background.

Table 2 provides details on the distribution of bank affiliation (2.1) and bank-affiliated co-investment (2.2) per industry as observed in the sample. We observe that the participation of bank-affiliated PE firms is particularly high in the industries of Metals, Hotels & Restaurants and Food, Beverage & Tobacco.

Methodology

The testing of the research questions presented in the previous sections is performed through the use of an OLS Random Effects regression on our panel data. Although the option for a fixed effects regression could have been interesting given our interest in comparing a “treatment group” (bank-affiliated

19

owners) to a “control group” (independent owners), we understood that the invariant nature of some of the variables made this option unfeasible, thus our decision for the first.

In order to verify hypothesis 1 and 2 – the influence of bank-affiliation of direct owners and co-investors over returns of portfolio companies – we utilize the Return on Assets (ROA) ratio as dependent variable and performance measure, this way also taking into account the use of particularly large portions of debt and the limited availability of information on capital structure of private companies, which could hamper the use of Equity-based ratios, for example.

The dependent variable is then regressed on three main explanatory variables that aim to attribute coefficients for size, efficiency and risk, these variables being respectively Size (SIZ), productivity (PRD) and gearing (GEA). Due to data availability limitations, we opt for using the Total Assets account as proxy for Size and Operating Revenues per number of employees as estimator for Productivity, while we use Gearing exactly as provided by Orbis.

Then, aiming to reduce the skewedness of our variables we utilize them in the form of natural logarithms. The specifications and sources for each of the variables are provided in Table 3 and their respective histograms in Table 5. The regressors chosen are to some extent in line with those utilized by Bloom, Sadun, & Reenen (2008) in their comparison between results of PE and Non-PE private companies, in which they utilize labour, capital and materials as three-fold explanatory variables.

Next, we include our two crucial variables, namely whether the PE firm owning the company is Bank-Affiliated or not (AFL) and whether it has Bank-Bank-Affiliated PE firms among its main co-investors (COI), together with an auxiliary dummy to capture the exposure to the financial crisis (CRS) from 2008 on. Finally, we add two multi-factorial variables controlling for the Industry (IND) and Country (CTY) to which the portfolio companies belong, this way getting to the following function for our model:

20

Then, aiming to properly address the structural changes occurred in the period and the interactional effect that the financial crisis had over Bank-Affiliation, possibly changing its coefficients we generate two factorial variables simulating this interaction between the Crisis factor and Bank-Affiliation factors, specifying the second as constant in the two-way interaction. The results are the new variables Crisis X Affiliation (CXA) and Crisis X Co-investor’s Affiliation (CXC).

In a second moment, we analyze the effects that the same variables might have had over the Solvency ratios of the companies in our sample. Here, our assumption is that in a scenario of limited ability to borrow from the paternal bank, PE managers of companies facing liquidity problems could be led to liquidate assets, issue securities or increase borrowing from external parties, impacting the solvency of the individual companies. Pierret (2014), finds that banks themselves have their access to short term funding reduced in crisis scenarios where the market expects insolvency from their side. By looking at the proportion of shareholder funds over the total assets of the company, we aim to examine whether the paternal banks’ liquidity necessities might have influenced the solvency management of portfolio companies. For such purpose we run regressions with the same specifications and same independent variables as before, replacing the dependent variable for Solvency (SLV) and again including the interaction factorial variables in a second run. The functions are given as follow:

21 And:

Finally, we analyze whether the increased costs for funding, as result of, among other things, the inability of the bank sponsors to lend to the portfolio companies; or the increased credit spreads caused by the deterioration of the paternal bank’s credit rating, led to an increase of the companies’ interest expenses relative to their interest income, this way impacting negatively the Financial Profit and Losses (FPL) of such companies. As Körnert (2006) ponders, raising loans and issuing securities is only optimal in an environment of rising interest rates, while crisis and recessions usually precede periods of decreasing interest rates as attempt to stimulate economy, resulting in losses for issuers. We do so, by replicating the function utilized for the two previous dependent variables, this time replacing it for the variable Financial P&L as follows:

And once more, after regressing the dependent variable on the independent factors considered individually, we include interactional variables to account for the influence of the crisis over our core regressors, getting to the following equation:

The summary statistics for the primary variables of our sample can be found at Table 4 and the results are discussed in detail in the next section

22

Results

Returns on Assets

Our analysis fails to reject the null hypothesis for H1 and H2, as we observe statistically significant relation neither for bank-affiliation nor for bank-affiliated co-investment and returns on assets (Table 6). The p-values reported for the two core regressors for the period were of 0.256 for Bank-Affiliation of the PE firms and 0.103 for Bank-Affiliation of the co-investor. After the inclusion of the interaction variables (Table 7) we still find no significant influence of Bank Affiliation of the PE firm or its co-investors over the RoA, observing p-values of 0.252 and 0792 for their respective interactions with the Crisis variable.

As expected, both estimations report strong correlation with the three main explanatory variables - productivity, size and gearing - at a 99% confidence level, however the signals of the coefficients observed, both before and after the beginning of the crisis, seem counter intuitive, as one would expect an increase in leveraging to result in higher risk and consequently, higher returns. At the same time, Size should result in, among other things, gains of scale and negotiation power, having positive influence over returns, while our result suggests a negative correlation. A possible explanation for such coefficients might be that substantially increased costs of borrowing may have undermined the returns of companies regardless of their increased risk, to the point of distorting the risk-return relation and leading highly leveraged deals to be less profitable. As for the size coefficient, we believe that the greater exposure of larger companies to the depressed global markets might have influenced their businesses more significantly than smaller ones. The strong influence of the Industry factor over the returns seems in line with the premise that crisis has impacted different industries with different intensities since it is well-known that sectors such as Textiles and Hotels & Restaurants tend to be more hardly affected by crisis, as consumers reduce their willingness to spend. The Crisis dummy variable

23

presented also no significant correlation with the returns, leading us to consider the revision of the model as this relation was a pillar to our assumptions.

Solvency

When analyzing Solvency we find the only statistically meaningful relation with Bank-Affiliation of PE firms of the study. The results observed suggest that, although not significant when considered isolated in the first regression (p-values: 0.376), when interacted with the Financial Crisis, bank-affiliation of the PE firms exercises a negative influence over the solvency of a private company with an estimated coefficient of -0.1123, almost at a 95% confidence level. This negative influence of Bank Affiliation in the crisis scenario connects with the findings of Fang, Ivashina, & Lerner (2013) that reported higher probability of default for bank-affiliated companies suggesting that banks involvement in Private Equity industry might have aggravated the Financial Crisis in this end, regardless of the frequent argument that Collaterazed Loan Obligations themselves were not responsible for the crisis.

Bank affiliation of one or more of the main co-investors on the other hand, proves to have no relevant effects, even when analyzed in interaction with the Crisis, suggesting little potential of contamination by bank-affiliated PE firms through second degree relations.

In opposition to what was observed for ROA, for Solvency we observe a negative – but much less intense – correlation with productivity and a positive influence of the Size over the solvency of companies (confidence interval above 99%) due to the natural link between Total Assets and solvency in the balance sheet, but also possibly as result of the greater availability of collateral to be used in loans acquisition. For similar reasons, the highly significant and negative coefficients observed for Gearing were presumable and logical.

24

The relevance of the industry where the company is inserted for its solvency reflects not only natural differences of leverage and operating models across industries but may also be a reflex of the different intensities of the crisis effects over their businesses (an interaction variable could be useful for analyzing such effect, although out of the scope of our study). Finally, the individual crisis variable proves to be meaningful at a 99% confidence interval, in line with what theory would support.

The between and overall R-squared observed for the regression on Solvency when regressed on factors considered individually were the highest observed - both at a level above 40%. After the inclusion of the interacted variables, we have a loss of overall explanatory power, compensated with a much higher “within” R-squared, suggesting a strong explanatory power for both changes across the years analyzed, reinforcing the meaningfulness of the findings for this dependent variable.

Financial P&L

Our third and last set of estimations leads us to ultimately discard the possibility of influence of bank-ownership of the PE firm over the financial performance of private equity companies as we observe no significant coefficients also on what regards their Financial Profit and Losses. The P-values for Bank-Affiliation and Co-investor Bank Bank-Affiliation in for the full period were 0.466 and 0.410 respectively. The estimation of the coefficients for their respective interactions with the Crisis variable also results in statistically insignificant results, leading us to fail to reject null hypothesis for H5 and H6.

We observe a negative and meaningful relation between productivity and Financial P&L, which can be attributed to the greater necessities of productive companies for funding and working capital. On the other hand, Size exercises a very strong influence over the financial results of the companies, which we attribute to the improved position to negotiate loans at lower rates and the lower riskiness of larger companies, which reflect in their credit spreads and consequently in lower financial expenses.

25

Gearing presents no statistically influence over the Financial P&L of companies, suggesting the main driver of financial expenses in the crisis scenario to be the cost of debt rather than its amount. The dummy representing the external factor Financial Crisis efficiently captured the effects of the market change, reporting p-values significant at a 99% confidence level. Finally, the dummies used to control for industry and country show to have insignificant influence for the Financial P&L in the crisis period. The R² of our regressions of Financial P&L dependent variable present very high explanatory power for both changes across individuals (above 47.76%) and for overall changes (44.54%) when considered in interaction with the crisis variable.

Robustness Check

Given the economically counter-intuitiveness of some of the findings, we believe that appropriate modifications and robustness improvements in the approaches utilized could have potentially led to substantially different results.

Among the most relevant points for future revision, we can mention the time deltas utilized in the analysis (1 year), which might be inappropriate for a period of so much volatility. However, although we acknowledge that quarterly-reported financials would have increased our number of observations and certainly result in more realistically estimated relations, data availability limitations for the companies made this option not available. A further note regards the probable presence of autocorrelation for the Return on Asset variable in the case of the use of shorter time intervals, which would implicate in the necessity of the inclusion of time lagged variables in our model.

Next to that, despite of the attempt to control for firm-specific characteristics through our three explanatory variables (SIZ, PRD & GEA) the incorporation of the Fama-French factors in the model as a way to account for the individual risk involved in the companies could have been determinant, since

26

much of the effects of bank-affiliation might have been overlapped or off-set by those of risks not captured by our model, such as growth/value or momentum.

Another point of attention or improvement is an eventual selection bias or “window-dressing” of the companies in our sample, once considered that the intermittence of reporting observed for many of them might suggest that very negative results were not reported, this way eventually disguising negative effects that bank-affiliation might have had. In this sense, the simple fact that the companies provide their financial information to a private database itself might lead to biased conclusions, since bad performing companies tend not to make their financial information public.

Also, the analysis of the Financial P&L estimation requires assumptions relative to the credit profile of the companies that could be more solidly backed by the inclusion of a variable accounting for credit risk and spreads, thus eventually connecting our findings to the higher probability of downgrade observed by Fang (2013).

Finally, on what regards our core regressors and the way that they relate to previous literature, we understand that the use of more specific sub-classifications for both the involvement with banks would have greatly richened the model. As mentioned previously, Fang, Ivashina, & Lerner ( 2013) propose the classification of bank affiliation between “bank-affiliated”and “parent sponsored”, as way to define the different ways the deals are relatd to the sponsoring bank. Although such segregation was not possible based on the information available at neither Orbis, Thomson, Zephyr or Capital IQ (all PE databases that were inspected during the data capture phase), it is likely to have great influence over the mechanisms of bank affiliation certification, and therefore over our results.

27

Conclusion

After analysing the influence of bank affiliation of the PE firm and its main co-investors over the financial performance of portfolio companies, estimated through the dependent variables ROA, Solvency and Financial P&L we find significant influence of Bank Affiliation only for the Solvency of the portfolio companies, discarding positive or negative effecst for the other two dependent variables. Although this study presented a relatively different focus and scope, such results contradict to some extent the findings of previous works regarding the certification benefits resulting from the affiliation of the PE firm to a major shareholder in PE such as Osnabrugge & Orbinson (2001), Tykvová (2006) and Chen & Martin (2011), although in line with the indiference between bank-affiliated and non-bank-affiliated PE’s observed by the very recent and complete study presented by Fang, Ivashina, & Lerner (2013).

Still then, the findings on the influence of Bank Affiliation over the Solvency of company propose raises relevant questions as the bankruptcy as result of insolvency has serious impacts for the economy as a whole, spreading unemployment and contaminating whole industries by affecting suppliers, creditors and other stakeholders. It also reinforces the importance of efficient regulation, once considered the increasingly major role that Private Equity has played in developing markets - frequently fragile economies, subject to cyclical crisis.

On a methodological aspect, it ilustrates the extension of the effects of the financial crisis factor and its potential to substantially change estimators when incorporated in a model through different approaches.

On a secondary note, our results suggest that larger private companies had their results and solvency more largely impacted by the financial crisis than smaller companies, which we attribute to a presumed higher exposure and sensibility to the behaviour of global markets. It still confirms that, in the context

28

of the financial crisis, increases in gearing naturally had a relevant influence over the solvency of the companies analyzed while relatively little influence over the financial expenses consequent of this debt. We interpret, based on the findings, that the size of the firm has played a key role in this relation, by reducing the spreads paid by larger companies.

It is however of extreme importance not to discard the possibility that these findings might have somehow been influenced by methodological aspects and sample biasedness that could have led to different results under different conditions. Among these aspects, we attribute particular attention to the possibility that the failure to capture certain risks in the deals might have overestimated returns - or underestimated underperformance – of the deals analyzed, eventually being responsible for the homogenic performance observed across PE ownership types.

Second, but not less important, the possible presence of biases in the sample, as result of data availability constraints and the well-known selective reporting in the Private Equity industry might have had the effect of hidding unsuccesful deals, when these could have been strongly influenced by bank-affiliation.

Last, and as is frequently the case in Finance, it is fundamental also to question the direction of the causal relation taken as main assumption for this study: namelly, is it bank ownership of the PE firm that makes the financial performance of the companies better (or worse), or are the better (or worse) performing companies simply more likely to be selected by bank-affliated PE managers? As discussed, our findings confirm neither.

29

References

Bloom, N., Sadun, R., & Reenen, J. V. (2008). "Do Private Equity Owned Firms Have Better Managing Practices?". London School of Economics.

Brissette, R. K. (2011). "The Volcker Rule's Unintended Consequences". North Carolina School Banking

Institute Journal - Vol 15, Page 231.

Chahine, S., & Filatotchev, I. (2008). "The effects of venture capitalist affiliation to underwriters on short- and long-term performance in French IPOs". Global Finance Journal 18, 351–372.

Chen, T., & Martin, X. (2011). "Do Bank-Affiliated Analysts Benefit from Lending Relationships?". Journal

of Accounting Research Vol. 49 No. 3 , 633-675.

Cumming, D., Fleming, G., & Schwienbacher, A. (2006). "Legality and venture capital exits". Journal of

Corporate Finance 12, 214 – 245.

Davis, E. P., Grob, S., & Haan, L. d. (2007). "Pension fund finance and sponsoring companies: empirical evidence on theoretical hypotheses". DNB Working Paper No. 158.

Drucke, S., & Puri, M. (2005). "On the Benefits of Concurrent Lending and Underwriting". The Journal of

Finance Vol.60(6), pp.2763-2799.

Duarte-Silva, T. (2010). "The market for certification by external parties: Evidence from underwriting and banking relationships". Journal of Financial Economics, 568–582.

Dufour, D., Nasica, E., & Torre, D. (2013). "Syndication in private equity industry: comparing the strategies of independent and captive venture capitalists". Université Nice Sophia Antipolis.

Erhemjamts, O., & Raman, K. (2012). "The Role of Investment Bank Reputation and Relationships in Equity Private Placements". The Journal of Financial Research • Vol. XXXV, No. 2, Pages 183–210. Fang, L., Ivashina, V., & Lerner, J. (2013). "Combining Banking with Private Equity Investing". The Review

of Financial Studies / v 26 n 9.

Gummi, M. R. (2013). "Restricted Investment in Private Equity: The Volcker Rule's Incursion Into Banking?". Cornell LL.M. Graduate Research Papers. Paper 3.

Hellmann, T., Lindsey, L., & Puri, M. (2008). "Building Relationships Early: Banks in Venture Capital". The

Review of Financial Studies/ v 21 n 2, pp.513-541.

Huw Jones. (2014, Jan 6). EU bank trading plan stops short of U.S. Volcker Rule. Retrieved Sep 30, 2014, from Reuters : http://www.reuters.com/article/2014/01/06/us-eu-banks-idUSBREA050SE20140106

30

Joos, P., & Joos, P. (1998). "The Prediction of ROE: Fundamental Signals, Accounting Recognition, and Industry Characteristics". INSEAD Working Paper Series.

Körnert, J. (2006). "Liquidity and solvency problems during the banking crises of the National Banking Era". Ernst-Moritz-Arndt-Universität Greifswald, Rechts- und Staatswissenschaftliche Fakultät, No.

05/2006.

Lopez-de-Silanes, F., Phalippou, L., & Gottschalg, O. (2011). "Giants at the Gate: On the Cross-section of Private Equity Investment Returns". Working paper.

McCahery, J. A., & Vermeulen, E. P. (2013). "Recasting Private Equity Funds After the Financial Crisis - The End of ‘Two and Twenty’ and the Emergence of Co-Investment and Separate Account

Arrangements". European Corporate Governance Institute, Law Working Paper N° 231.

Osnabrugge, M. v., & Orbinson, R. J. (2001). "The influence of a venture capitalist’s source of funds".

VENTURE CAPITAL, 2001, VOL. 3, NO. 1, 25-39.

Pierret, D. (2014). "Systemic risk and the solvency-liquidity nexus of banks". Stern NYU.

Porta, R. L., Lopez-de-Silanes, F., & Shleifer, A. (1998). "Law and Finance". Journal of Political Economy,

vol. 106, no. 6.

Tykvová, T. (2006). "How do investment patterns of independent and captive private equity funds differ? Evidence from Germany". Swiss Society for Financial Market Research.

31

Appendix

Bank Affiliation per Country

Table 1

1.1 1.2

Country 0 1 Total Bank (%) Country 0 1 Total Bank (%)

United Kingdom 209 37 246 15% United Kingdom 206 40 246 16%

France 131 29 160 18% France 128 32 160 20% Germany 27 6 33 18% Sweden 58 7 65 11% Spain 24 2 26 8% Finland 37 5 42 12% Finland 40 2 42 5% Germany 29 4 33 12% Greece 1 2 3 67% Greece 5 4 9 44% Denmark 8 1 9 11% Denmark 6 3 9 33% Italy 24 1 25 4% Italy 22 3 25 12% Luxemburg 1 1 2 50% Belgium 33 2 35 6% Netherlands 18 1 19 5% Luxemburg 0 2 2 100% Greece 8 1 9 11% Latvia 0 2 2 100% Romania 9 1 10 10% Spain 25 1 26 4% Austria 2 0 2 0% Ireland 4 1 5 20% Belgium 35 0 35 0% Netherlands 18 1 19 5% Bulgaria 3 0 3 0% Romania 9 1 10 10%

Czech Republic 7 0 7 0% Austria 2 0 2 0%

Estonia 2 0 2 0% Bulgaria 3 0 3 0%

Croatia 4 0 4 0% Czech Republic 7 0 7 0%

Hungary 1 0 1 0% Estonia 2 0 2 0% Ireland 5 0 5 0% Greece 3 0 3 0% Lituania 1 0 1 0% Croatia 4 0 4 0% Latvia 2 0 2 0% Hungary 1 0 1 0% Portugal 2 0 2 0% Lituania 1 0 1 0% Sweden 65 0 65 0% Portugal 2 0 2 0% Slovenia 1 0 1 0% Slovenia 1 0 1 0% Slovakia 1 0 1 0% Slovakia 1 0 1 0% Total 631 84 715 9% Total 607 108 715 16%

32

Bank Affiliation per Industry

Table 2 2.1

Industry 0 1 Total Bank (%)

Other services 211 29 240 12%

Machinery, Equipment, Furniture & Recycling 104 13 117 11%

Wholesale & Retail Trade 83 9 92 10%

Metals & Metal Products 28 7 35 20%

Chemicals, Rubber, Plastics, Non-Metallic Products 44 4 48 8%

Food, Beverages and Tobacco 13 4 17 24%

Hotels & Restaurants 14 4 18 22%

Publishing & Printing 30 4 34 12%

Banks 6 2 8 25%

Construction 21 2 23 9%

Education & Health 22 2 24 8%

Insurance Companies 1 1 2 50%

Post & Telecommunications 17 1 18 6%

Primary sector 6 1 7 14%

Textiles & Wearing Apparel, Leather 9 1 10 10%

Gas, Water, Electricity 5 0 5 0%

Transport 11 0 11 0%

Wood, Cork, Paper 6 0 6 0%

Total 631 84 715 13%

2.2

Industry 0 1 Total Bank (%)

Other services 196 44 240 18%

Machinery, Equipment, Furniture & Recycling 102 15 117 13%

Wholesale & Retail Trade 79 13 92 14%

Chemicals, Rubber, Plastics, Non-Metallic Products 41 7 48 15%

Education & Health 19 5 24 21%

Post & Telecommunications 13 5 18 28%

Food, Beverages and Tobacco 13 4 17 24%

Metals & Metal Products 32 3 35 9%

Wood, Cork, Paper 3 3 6 50%

Construction 21 2 23 9%

Hotels & Restaurants 16 2 18 11%

Publishing & Printing 32 2 34 6%

Banks 7 1 8 13%

Gas, Water, Electricity 4 1 5 20%

Primary sector 6 1 7 14%

Insurance Companies 2 0 2 0%

Textiles & Wearing Apparel, Leather 10 0 10 0%

Transport 11 0 11 0%

Total 607 108 715 15%

Bank Affiliate

33

Variables Description

Table 3

Variable Definition Source

RoA (Returns on Assets) Ln ((Net income / Total Assets) * 100) Orbis

Solvency Ln ((Shareholders funds / Total assets) * 100) Orbis

Financial P&L Ln ((Financial revenue-Financial expenses)) Orbis

Productivity Ln (Operating Revenues/Employees) Orbis

Size (total assets) Ln (Total Assetss) Orbis

Gearing Ln (((Non current liabilities + Loans) / Shareholders funds) * 100) Orbis

Bank-Affiliated - dummy 1- if field "Firm Type" in Thomson One = "Bank Affiliated" ; 0 -

otherwise Orbis x Thomson One

Bank-Affiliated Co-Investor - dummy 1- if 1 or more, of the firms in the field "Top Co-investors" in

Thomson One is classified as bank-affiliated; 0 - otherwise Thomson One

Industry BvD Major Sector Orbis

Country Country ISO Code Orbis

D ep en d en t In d ep en d en t

34

Summary Statistics

Table 4

Variable

Observations

Mean

Std. Dev.

Min

Max

Ln (Return on Asset)

4027

1.852016

1.34202

-5.80914 4.57223

Ln (Financial Profit & Losses)

1739

5.164692

2.354827 -3.63666 12.90506

Ln (Solvency)

3384

3.42258

1.018138 -3.91202 4.60517

Bank Coinvestor - dummy

6435

0.151049 0.3581244

0

1

Bank Affiliate - dummy

6435

0.1174825 0.3220194

0

1

Crisis - dummy

6435

0.4444444 0.4969426

0

1

Ln (Gearing)

4605

3.460748

2.012748 -6.90776 6.90429

Ln(Total Assets)

5761

9.835918

1.282929 -6.36432 15.35216

Ln(Productivity)

5513

5.173495

-1.20368 11.35109 10.36833

Industry

6435

10.27972

4.24858

1

18

Country

6435

11.68671

5.517323

1

26

35

Dependent Variables – Distribution

36

Panel Regressions – Individual Effects

37

Panel Regressions – With Interaction

Table 7

Coef. Std. Err. P>|z| Coef. Std. Err. P>|z| Coef. Std. Err. P>|z|

Productivity 0.1834377 (0.02976) 0.000 *** -0.0185111 (0.01672) 0.268 -0.10421 (0.05305) 0.049 **

Size (total assets) -0.3061939 (0.03101) 0.000 *** 0.1364772 (0.01884) 0.000 *** 1.126377 (0.05917) 0.000 ***

Gearing -0.1572424 (0.01239) 0.000 *** -0.2926242 (0.00866) 0.000 *** 0.031294 (0.02430) 0.198

Crisis - dummy -0.0123454 (0.04012) 0.758 0.0690517 (0.02067) 0.001 *** -0.31014 (0.09287) 0.001 *** Bank-Affiliated - dummy 0.1678073 (0.12451) 0.178 0.0787416 (0.08637) 0.362 -0.15314 (0.25134) 0.542 - Crisis x Bank-Affiliated - dummy -0.1167825 (0.10197) 0.252 -0.1123325 (0.05829) 0.054 * -0.00752 (0.24102) 0.975 Bank-Affiliated Co-Investor - dummy 0.1263648 (0.11805) 0.284 0.0323545 (0.07661) 0.673 0.043719 (0.22604) 0.847 - Crisis x Bank-Affiliated Co-Investor - dummy -0.0256574 (0.09738) 0.792 -0.0518162 (0.05330) 0.331 0.359254 (0.24923) 0.149

Industry -0.0254647 (0.00890) 0.004 *** -0.0128774 (0.00591) 0.029 ** 0.004951 (0.01805) 0.784 Country -0.0076565 (0.00703) 0.276 0.0037818 (0.00449) 0.400 0.005842 (0.01473) 0.692 Constant 4.671184 (0.30200) 0.000 3.670635 (0.18156) 0.000 -5.73147 (0.62698) 0.000 R² (%) -within 4.51 33.18 3.41 -between 19.49 19.48 47.76 -overall 14.86 21.30 44.54

Random Effects OLS. Standard Errors in Parentheses. * Significant parameter at 10% level, **Significant parameter at 5% level, *** Significant parameter at 1% level