Corporate debt maturity : how can gap-filling theory explain the maturity choices in debt issues by corporations

(1)

Corporate debt maturity: How can gap-filling theory explain the maturity choices in debt issues by corporations. ABS-UvA | 1

Corporate debt maturity:

How can gap-filling theory explain the maturity choices in debt issues by corporations.

Jaap Dansen

College card number: 10279547

Thesis Executive Master of Finance and Control 01/10/2015

Supervisor: Dhr. Dr. J.E. Ligterink

(2)

Management Summary

This paper studied the maturity choices made by corporations when they issued new debt. By testing the theoretical framework of gap-filling, the influence of government debt maturity on corporate maturity choices was analysed. Corporations were expected to change their maturity of debt towards the maturity buckets left empty by the

government in that period, in order to capture illiquidity premiums. A negative correlation between, government debt maturity of new issues and corporate debt maturity of corporations, is not found. This paper therefore rejects the idea that the filling of government debt maturity buckets as described by gap-filling theory is an important determinant of corporate maturity choices. This paper used a dataset consisting of 310.947 debt instruments to create a detailed proxy for the maturity of corporate and government debt, using data from 1993 until 2014, the new proxies replace a high level proxy used by Greenwood, et al. This improved proxy was one of the topics for further research addressed by the creators of gap-filling theory. The idea, at the start of the paper, was that by improving the measurement of the variables of gap-filling the theory would be enhanced and strengthened. Through this, our understanding of the variables influencing corporate debt maturity choices would be enriched. Based on the findings in this paper gap-filling theory however is unable to explain the factors influencing corporate debt maturity choices when the improved proxy is used. Results indicate that it still is relevant for finance professionals to monitor market factors as results show that they influence debt maturity choices of firms. Further research on the exact impact of; the level of government debt issued as percentage of total debt issued, term spread and credit spread on debt maturity decisions by corporations can create value through optimisation of debt maturity choices. Focussing on government debt maturity or turning away from the impact market phenomenon have on corporate debt maturity however seems suboptimal.

(3)

Corporate debt maturity: How can gap-filling theory explain the maturity choices in debt issues by corporations. ABS-UvA | 3 Contents 1) Introduction ... 4 2) Theoretical Framework ... 10 3) Methodology ... 20 4) Results ... 36

5) Discussion and conclusion ... 51

6) Bibliography ... 57

Appendix 1) Screening criteria Data set: ... 60

Appendix 2) -Screening criteria Data set: ... 61

Appendix 3) - Determining of corporate debt maturity Greenwood, et al. proxy (GrCDM) ... 63

Appendix 4) - Robustness regression 1a and 1b ... 63

Appendix 5) – Additional regression analysis data hypothesis 1a and 1b ... 72

Appendix 6) - Robustness regression 2a and 2b ... 74

Appendix 7) – Additional lags and information hypotheses 2a and 2b. ... 79

Appendix 8a) - Robustness of control variables ... 83

Appendix 8b) - Robustness and additional analysis of control variables table 10 ... 88

Appendix 8c) - Robustness and additional analysis of control variables table 11 ... 89

Appendix 9) - Robustness regression 3 ... 90

(4)

1) Introduction

The seemingly simple question “What is the best capital structure for a firm, and consequently how should it optimize its debt structure?” remains often unanswered. While multiple theoretic frameworks exist, there is no general framework that fully explains corporate decision-making on new debt issues. When trying to understand why a corporation chooses a specific moment, size and maturity structure for new the

number of theoretical frameworks that provide answers even gets sparser. This paper hopes to contribute to the existing research on debt maturity choices by looking at the effect that government debt maturity choices have on maturity choices of corporations. While effects of government policy on the economy and equity markets have been discussed in length, the effect of government choices on the structure of corporate bond issues has received less attention. (Joyce, et al., 2012)

One of the more recent explanations for corporate capital structure decisions is the market-timing hypothesis by Baker & Wurgler (2002). They claim that firms look at the market conditions and determine financing decisions based on the height of stock prices and interest rates. Debt financing decisions could therefore be explained by price setting in markets (Baker & Wurgler, 2002). Greenwood, Hanson and Stein, (2010) builds on the market timing theory to the explain debt maturity choices by corporations. The main idea behind their theory is that corporations change the maturity and moment of issuing new debt securities depending on the gaps in maturity buckets in the total bond market. In their paper, they showed that a negative correlation exists between the length and amount of government debt issued on the one hand and the characteristics of corporate debt issues in the same timeframe on the other hand.

The goal of this research paper is to determine the tradeoffs for the optimal timing and maturity buckets firms can pick when issuing new corporate bonds. The focus will be on improving maturity characteristics for bond issues in relation to

changes in government bond markets. This will help improve capital structure steering by firms and their finance professionals.

This paper will contribute to the existing research by analyzing actual debt issues in the period 1993-2014. This paper uses improved proxies for debt maturity of new

(5)

issues by using actual maturities of debt securities. The improvement of debt maturity measurement was one of the points which addressed by Greenwood, et al. that needed further research. This paper further looks at more recent data up until the year 2014, Greenwood, et al. analyzed data up until the year 2005. Lastly, this paper tests the findings and theoretical framework developed by Greenwood, et al. through the comparative study performed. By doing this we try to improve the theoretical

framework that explains debt maturity choices. Increasing our knowledge of the impact of government debt maturity choices on corporate debt issues can improve corporate debt decisions. Better understanding of debt maturity choices is fundamental for improving the managing, steering and optimization of a firm’s new debt issues, debt structure and in a greater perspective capital structure.

(6)

Research question and relevance

Corporate debt is an important category of corporate funding and has been under constant study in relation to the optimal capital structure of organizations. (Goyal, 2009) What however has received far less attention is the choices made by firms when issuing new corporate debt. Much is unknown on the dynamics that determine the maturities corporations choose for new debt issues. Both the dynamics in the total debt market (government + corporate) and firm balance sheet strength influence decision-making related to choices for issuing longer or shorter corporate debt following the theory of Greenwood, et al. (2010). More research and improved testing of this theory are needed to better understand the framework and detailed workings over a longer time span. (Greenwood, et al., 2010)

From a CFO or financial expert perspective monitoring debt issues and understanding the market dynamics within the total debt market is crucial for safeguarding the chosen capital structure. Maturity influences the funding mix and the target risk profile of a corporation, it must be closely monitored by management if they want this mix. (Rudani, 2013) Its impact on the risk and leverage of an organization necessitates that choices surrounding new bond issues are understood and reviewed by the financial

stakeholders within the organization. This in order to align capitalization choices with the firm strategy and optimize choices based on a risk and reward tradeoff. (Barton &

Gordon, 2006) A financial expert in a CFO, Treasurer, analyst or controller1_{role must be}

able to understand and monitor these financing decisions in order to successfully fulfill his role of decision-maker and translator of corporate strategy. (Strikwerda, 2014)

In addition to the relevance of corporate bonds from a funding perspective, controllers can also come across these instruments in the form of an investment category. For institutional investors (banks, insurers, and pension funds) corporate bonds are one of the main investment classes. In order to make a sufficient return on their liabilities this type of firm determines the best portfolio of investments based on several asset categories. When determining the optimal strategic allocation of assets, the characteristics and correlations of bonds are a key input for determining the optimal

1

CFO’s, management accountants, CFA’s, treasurers and controllers are financial experts and throughout this paper I also refer to financials in this context. Controllers in this context are Dutch finance professionals, preferably which an RC title.

(7)

investment portfolio. For these firms performance is strongly related to the risk taken by the allocation in the portfolio, monitoring the internal allocation and market

developments are therefore crucial for understanding the performance of this type of organization. (Hoevenaars, et al., 2008) For finance professionals within these

industries understanding the interrelation of new corporate debt maturities with new government bond maturity is relevant for making good investment decisions.

Analysing the maturity characteristics of corporate debt issues over time in the period 1993-2014 provides an opportunity to increase our knowledge of corporate choices. Controllers can use these insights to determine the best timing and maturity for new corporate bonds issues. This can improve capital structure decisions adding value for the organization and it improves steering towards the desired capital structure.

The research question answered in this paper is:

To what extent do corporations match new debt issues to market conditions (liquidity gaps in government bond markets) in the US market between 1993-2014?

The sub questions used for answering this central research question are:

a) a: What role does the controller have in funding and capital structure decision making within firms.

b) b: How does gap-filling theory explain the characteristics chosen by corporates when they issue new debt.

c) c: What is the expected behavior in the period 1993-2014 of the corporate bond markets based on the theoretic framework (based on B).

d) d: Can statistical evidence be found which proofs the expected relations between corporate bond and government bond issues for the period 1993-2014.

1) Hypothesis 1) Corporate bond issuers will fill in supply gaps in total bond supply resulting in a negative correlation between the maturities of corporate and

government bonds issued.

2) Hypothesis 2) The higher the percentage of government debt in relation to total debt the stronger the gap-filling effect.

(8)

3) Hypothesis 3) Firms with stronger balance sheets are more flexible and therefore more capable of filling in the gaps in the maturities of the bonds issues.

e) To what extend does gap-filling theory explain the maturities chosen of corporate bond issued.

f) How can controllers use the insights and ideas from gap-filling theory to optimize decision making related to new bond issues.

Relevance

All firms need capital in order to run their business, make investments and repay financial obligations. An important form of financing, mainly for larger organization, is through the issuing of corporate bonds. On the organizational level, multiple models exist which explain the optimal capitalization strategies individual firms should use. (Graham & Campbell, 2001) Common practices are; the matching of maturities, internal limits for working capital and monitoring of solvability and liquidity ratio’s. (Brealey, et al., 2004) These internal models however seem unable to explain the large changes in the volume of debt issued over the years, and their maturity characteristics.

(Greenwood, et al., 2010) In order to determine the best moment to issue debt through bonds and the proper characteristics for the issues more research is needed.

This paper creates additional insight of the dynamics of the corporate debt market over a multiyear period and its interconnectedness with the government debt market. In the years 1999-2001 Greenwood, Hanson and Stein found large fluctuations in long term U.S. corporate debt issues as percentage of total non-financial debt issues moved form 24,7% to 19,9% and back to 30,1% percent. Based on their findings they developed the theoretical framework on gap-filling. The main concept of this theory is that corporations change the maturity of new debt in order to fill maturity gaps left by government debt issued in the same period. This paper will test their gap-filling theory in order to see if their ideas hold in the period 1993-2014 in a pre and post crisis period, as to determine their practical use. The insights will be linked to decision making on the optimisation of funding in firms, and value for the controller of this process.

(9)

Report structure

After illustrating corporate funding choices from a strategic and financial expert perceptive in the introduction, the theoretical framework will be explained in chapter 2. The focus in this paper will be on the gap-filling theory of Greenwood, et al. with a sidestep to other theories which explain financing structures and funding decisions in firms. After the creation of the theoretical context, the method of research will be described. This paper focusses on testing the findings of Greenwood, et al. through an improved measurement of corporate and government debt. It is therefore important that this paper first analyses if the data set used is congruent with the findings of

Greenwood, et al. before improving the proxy of debt maturity. Where Greenwood, et al. use a proxy for debt maturity from flow of funds data this paper looks at all individual debt instruments issued between 1993-2014 available through Thomson Reuters Capital IQ. Based on the actual maturity this paper is able to determine a more detailed proxy for corporations and government debt maturity of new issues quarterly. This is used to test the assumptions of gap-filling through the use of OLS regression analysis. In the methodology chapter 3, this paper discusses the dataset and research method in more detail. Before moving to the results in chapter 4, this paper addresses robustness of analysis. The results will be structured in order to properly test the three assumptions underlying gap-filling theory as described by Greenwood, et al., The fifth and final

chapter contains the discussion and conclusion of this paper. In the appendixes, information is added on data selection criteria, robustness analysis and regression analysis.

(10)

2) Theoretical Framework

This chapter will discuss the existing theoretical concepts and ideas that try to explain the choices corporations make when issuing new debt. After a broader theoretic introduction, this paper will focus on market timing theory, which is at the basis of the theoretical framework of gap-filling as developed by Greenwood, et al. The chapter will continue by addressing debt maturity choices of corporations and the government in order to determine the underlying incentives for corporations to change the maturities of debt issues as described by gap-filling theory. In addition, the impact of the corporate to total debt ratio on these incentives and the type of firms most likely to behave in a gap-filling way will be discussed. These theoretical frames will lead to the hypothesis, which are the end of this chapter the it serves as a basis for the research carry out in the following chapters. This chapter will end with an explanation of testing method of the individual hypotheses.

There currently exists no theory, which is able to explain debt maturity choices in all circumstances and time periods. Three main theoretical frameworks that could provide explanations for this specific part of a corporation's capital structure are; 1) trade off theory, 2) Pecking order theory, and 3) Market timing theory. The underlying themes in all three are asymmetric information, agency problems, taxes and bankruptcy costs. (Miglo, 2010)

Trade of theory sees capital structure as a tradeoff between the benefits from lower taxes and the rising bankruptcy costs, which are associated with higher levels of debt. An organization will try to maximize value by increase debt levels up to the level that the tax benefit of an additional unit of debt equals the increase in the cost related to a higher bankruptcy risk. Bankruptcy should be seen as both the direct costs in the form of legal fees, but also indirect costs as decreased credibility towards customers. The level of debt a firm strives to set the result of maximizing the value of tax benefit by setting the optimal level of promised debt payments. (Miglo, 2010)

Changes in the level of debt and the type of debt issued should be attributed to the internal trade-off between tax benefits and the cost related to bankruptcy. However, several scholars find that changes in debt level, which are theoretically expected when the current level is not equal to the optimum of formula 1, are not in line with research

(11)

outcomes. (Fama & French, 2002) More resent research adds the transaction costs associated with issuing or changing debt structures in the equation. By subtracting the cost of transactions of the value of the tax benefit when increasing debt, market

circumstances, which influence transaction costs, become relevant. Bednarek, Issler & Patel (2014) relate the choice of the debt maturity of new issues by firms to the

transaction costs in trade-off theory. Their model predicts that the term structure of debt is a dominant factor for maturity decisions. The term spread is the difference between short-term and long-term interest rates. Bednarek et al (2014) see the width as a determinant of debt maturity choices. Corporations when faced with a narrow term spread face higher transaction costs, this higher level of costs will tilt the tradeoff so that the costs predominantly outweigh the benefits due to higher bankruptcy costs. Firms will therefore issue shorter maturities when the term spread for credit is high, because it means that expected bankruptcy probability for corporations is low. While this paper expects that this is not the main determinant for corporate debt maturity decisions this paper will look at term and credit spread as a control variable.

The pecking order theory focusses on the asymmetry of information between the decision makers in the firm and the share- and stakeholders. As management has an information advantage which potentially be used against shareholders even high quality firms have difficulty to broadcast the true value and quality of their organization. The reasoning behind this is that the medium of low quality firms will also try to convince shareholders that they are a high quality firm. In order to make the distinction between good and bad performing forms investors and stakeholders therefore look for signals given by organizations that can show to which category they belong. The choice to issue debt or the use of other types of funding can therefore be seen in this light. It provides a signal to stakeholders about the quality of the organization.

Different types of funding incorporate different levels of mispricing. Ranging from internal funding without mispricing to equity, which has the biggest negative impact on the value of existing shareholders and is thus associated with high mispricing costs. Theoretically, all firms try to appear as high quality firms, but only the strongest are able to constantly provide the associated signals. The choice to issue new debt and the

selected maturity is one of these signals and can therefore be seen as a characteristic of the quality of an organization. High quality firms have sufficient internal funding and

(12)

would first use retained earnings, secondly as an alternative they finance with debt, and thirdly issue new equity for financing purposes. Corporations are thus assumed to have a standard preference and choice alternative is their first choice is no longer available. (Miglo, 2010) (Myers, 1984) The maturity of debt issued can also be interpreted in this light, as short-term debt is less costly than long-term debt. Corporation of high quality will theoretically prefer to issue debt with shorter maturities. (Frank & Goyal, 2003)

Market timing theory is based on the idea that the decision to issue debt or equity depends on the market performance. Managers wait for the right market conditions and change the type of debt they issue based on the debt coupons and stock return when additional capital is required. (Baker & Wurgler, 2002) (Miglo, 2010) An important difference between market timing and trade-off theory is that market timing sees the market conditions as a determinant for the timing and type of debt issued. Corporate capital choices change in accordance with fluctuations in the market place. Capital structure decisions of firms are therefore the accumulated outcome of all individual cost benefit choices made through time at different market circumstances. Each new choice looks at optimizing the capital structure by looking the external market opportunities and the internal costs. Researchers found that firms are more likely to issue equity in time periods when markets to book values are relatively high and repurchase equity when these values are low. (Baker & Wurgler, 2002) The choice between short and long-term debt, thus the maturity of debt, is also found to be correlated with specific market circumstances. Studies by Barclay and Smith (1995) and by Jose Guedes and Tim Opler (1996) consistently found that the maturity of debt issues are higher when term spread, difference one year and ten year treasuries, is higher. Their findings are in line with CFO questionnaire research done by other authors; Graham, Cambell and Harvey (2001) who found that profiting from chaep debt or waiting for the market cost of debt to decrease was the third most imporant factor affecting their debt maturity choice. The most important determinant was matching of the companies debt with their assetsm and the second he managing the firms liquidity risk. (Graham & Campbell, 2001) While Graham and Campbell agree to the economic logic of matching debt maturity to assets and the limiting of liquidity risk they see no economic rational in the choice of CFO’s to change maturities when the cost of debt in the marketplace changes. In a properly funtioning market changes in the yield on debt paper would reflex the underlying risk and a CFO

(13)

would have no special forecasting power for the cost of debt. In their reasoning the timing of the market would therefore be caused by other variable than a special forecasting ability of managers.

The theory of Greenwood, et al. (2010) is grounded in market timing theory and tries to explain why managers try to change debt maturity decisions following

developments in the debt market. The authors have developed and tested a framework which explains maturity choices of corporations. In addition, they focused on explaining the strong changes in the maturities of corporate bond issues in a multi-year time frame as previous research had not yet found satisfying answers to significant changes in corporate maturity structures of time. The main idea of gap-filing theory is that corporations act as macro liquidity providers in the sense that they fill liquidity gaps that are the result of changes in the maturity of government debt issues. In important underlying assumption of Greenwood, et al. based on literary research is that there exists a stable demand for long-term debt by pension funds and insurance companies. These financials have long-term liabilities towards pension and insurance holders and try to match these liabilities with similar long-term return generating investments. From a demand and supply perspective the market is characterized by a strong and stable demand for debt with a longer maturity. Government debt, which is assumed to represent exogenous supply shifts, is the main source of supply for investors as government debt due to its relatively longer maturity and large supply. This holds especially when comparing to corporate debt (U.S.) to government debt (U.S.) as done in the study of Greenwood, et al. This is also the prerequisite of this paper. (Greenwood, et al., 2015) The debt structure of government (U.S) is governed by the U.S. Treasury department who strives towards regular and predictable pattern of debt payments based on the underlying deficit stream. This has been found to be the dominant reason for new government debt maturity characteristics although academics more recently voice alternative decision metrics. Such a theoretic alternative is that the treasury

department tries to optimize the maturity of newly issued debt, as a response to changes in interest rates and primary deficits seems not to be the case. (Berman, 2013) This leads to the assumption that government debt can be seen as an exogenous effect by Greenwood, et al. and this is the assumed in the rest of this paper. As it is not possible to distinguish a government debt instrument, which is uncorrelated with demand side

(14)

factors, this paper follows Greenwood and Vayanos (2008) and controls for the extent to which the debt to GDP, as a total debt demand proxy, explains gap-filling behavior. When moving to the actual issuance of government debt by the treasury department (U.S.) the maturity of issues can be categorized in specific maturity buckets (for example 5 yr or 10 yr). Following the targeting of long-term target of the treasury department for regular and predictable debt payments, debt maturities are altered over time to match the desired payment level of total debt. The government supply therefore fluctuates over time, and as a consequence, the size of government to total long-term debt, overall supply in long-term debt changes over time. (Greenwood, et al., 2010) The result is an unbalance in demand/supply due to the constant long-term demand by insurance and the fluctuating supply in long-term debt by the government. Greenwood, et al. find that the arbitrage by broker-dealers and hedge funds are unable to fill the created gap due to their limited capital and the undiversifiable nature of the trade. They calculate that in order for arbitrage by these parties to fill in one standard deviation increase in the maturity of government debt (9% with a mean of 59%) more that 67% of total assets of this industry should be targeted decreasing the gap. (Greenwood, et al., 2010)

Greenwood, et al., theorize and proof in their study that this unbalance in supply and demand creates opportunities for corporations who are able to change the maturity of new debt issues. By adjusting the maturity of their new debt issues corporations can lock in a liquidity premium when they choose to issue maturity buckets that are not filled by the government in a given time period. While corporations are assumed to have an ideal target maturity for matching their liabilities, they are remain flexible and can make a “deviation form target’ cost versus liquidity premium tradeoff. The underlying idea is that de strong stable demand for maturity buckets by institutional investors leads to a better price for market participant who can fill in the maturity bucket that the

government leaves empty that time period (say first half of 2015). The actual reaction by corporations however lags a bit as corporations need time in order to issue new debt paper. Greenwood, et al. assume that firms need 3 to 9 months for new debt issues, this results in the expectation that the minimum lag for gap-filling behavior surpasses this timeframe. A firm, which would normally solely look at internal asset-liability matching, should now weight the cost of deviating from the internal optimum with the additional premium received by gap-filling behavior. The tradeoff between the costs of an

(15)

suboptimal maturity match, higher of lower illiquidity risk than targeted and decreased steering of projects due to a sub optimal internal debt maturity could be outweighed by a high illiquidity premium of gap-filling. (Greenwood, et al., 2010) In the underlying table a simplification of the idea underlying gap-filling is given. In the example, the total amount of debt issued by the government is kept stable at $ 100 trillion dollar. The maturity however fluctuates over time period 0 and 1. Assuming that on average t = 0 represents a situation without liquidity premiums due to gap-filling corporations focus on their internal debt maturity targets. The influence of market circumstances and gap-filling opportunities are apparent in t = 1 as government based on total target payments increases the amount of debt issued in the 1 year maturity bucket and decrease the 5 years bucket. The result is that corporations receive a better price when issuing new debt with a maturity that mimics the vacant maturity bucket and are therefore incentivized to do so if it offsets the internal cost from deviating from the target distribution in t = 0.

(16)

Table 1: Illustrative example of gap-filling

A numeric example of gap-filling shows that when governments change their split between debt maturity buckets, corporations react and fill in the black spots in order to lock in the illiquidity premium associated with the unbalance of demand and supply in that specific bucket.

US government issues $ 100 dollar ( x trillion $) t = 0 t = 1

1 year $ 20 $ 40

5 year $ 20 $ 0

10 year $ 20 $ 20

20 year $ 20 $ 20

30 year $ 20 $ 20

US corporate issues $ 25 dollar ( x trillion $) t = 0 t = 1

1 year $ 5 $ 0

5 year $ 5 $ 21,25

10 year $ 5 $ 1,25

20 year $ 5 $ 1,25

30 year $ 5 $ 1,25

In line with the above illustration, gap-filling theory assumes a simple cost/benefit trade off made by corporations who are able to capture the illiquidity premium existing due to gap-filling. While grounded in market timing theory gap-filing theory does not attribute special forecasting power to corporations, which is assumed by other academics, gap-filling instead sees a market mechanism, which rewards flexibility in debt maturity. (Greenwood, et al., 2010) Through adapting and valuing supply and demand conditions within the financial marketplace, gap-filling shows how corporations are incentivized to change their capital structure. The capability to

optimize the internal debt target while incorporating external market opportunities into this tradeoff determines the value the firm can add using liquidity premium

opportunities. (Vayanos & Vila, 2009)

Greenwood, Hanson, & Stein (2010) tested their theory using a dataset, which looks at the period 1963-2005 and found a strong negative correlation between

corporate debt maturities and that of government debt. They used several data sources (Capital IQ, Orbis, flow of funds, individual treasury reports) to cross check this relation. Based on their findings using a proxy for de maturity of newly issued government and corporate debt they document three important relations of gap-filling. The first and most important is that when government decreases the maturity of new debt corporations do the opposite and increase the maturity of new debt. This inverse

(17)

relationship is found throughout the long time frame of 1963-2005 and holds when controlled for market conditions as credit spread, debt to gross domestic product. Adjustments for mortgages, financial firms, swaps and investments by government or corporates do not seem to have a significant altering effect on the baseline outcome of this relationship. Greenwood, et al. thereby underline the use of the direct link between government and corporate debt maturity and proceed to look at additional relations in gap-filling. The second relation they find concerns the strength of the gap-filling by corporations. They find, in line with their theoretical model, that a larger supply of government debt (U.S.) as a percentage of total debt in a specific time period results in stronger gap-filling behavior. The total level of government debt issued and corporate debt issued is thus not assumed to be constant, and of influence on the first relation of gap-filling. This theoretically seems logical as new government debt issued, as a

percentage of total debt issued would indicate a larger difference in supply and demand. This supply shock would create a larger disequilibrium, which would lead to a higher illiquidity premium for corporations who can fill in “larger” gaps left by the government. The higher liquidity premium would at firm level create more individual cases in which the risk reward tradeoff would favor the alteration of the corporates debt maturity in new issues in order to seize the premium resulting from gap-filling. This association is also found within the data set of Greenwood, et al. while using government to total debt and debt-to-GDP as a proxy for the total debt supply by government. The third

association found by Greenwood, et al. concerns the time of corporations who are most likely to pursue gap-filling. The assumption underlying their theoretic model is that corporations who have more flexibility in their capital structure are face lass costs when shifting debt maturities towards maturity buckets left empty by the government. The findings of Greenwood, et al. are not fully congruent and therefore need to be looked at with some reservations. While the solvability measures used by Greenwood, et al. are in line with their model, the proxy of leverage is not. Still they conclude that based on a solvency basis they see sufficient evidence that firms with stronger balance sheets are the more like to act as gap fillers. The underlying concept is that the costs for stronger, from a debt maturity perspective more flexible firms, is lower than that of firms with weaker balance sheets. The tradeoff between costs form divergence from the internal optimum maturity is therefore more often offset by the illiquidity premium of gap-filling.

(18)

Their strong position lowers refinancing risk, interest rate risk and bankruptcy costs and alters the tradeoff between these costs and the liquidity premium of gap-filling more often in favor of changing the maturity of new debt issues.

Gap-filling theory provides additional insight in a specific capital structure question, which is, how do corporations determine their debt maturity. This helps to fill the theoretically underdeveloped part of market timing theory and the gap between theory and practice. (Graham & Campbell, 2001) The theoretic framework of gap-filling is confirmed based on Greenwood, et al.’s empiric findings. The long time frame used within the study is one of the strengths of as is its generalizability due to the use of the full range of corporate and government debt levels. There however also all are several limitations to the method and findings of Greenwood, et al. Firstly Greenwood, et al. use proxies both for the maturity of newly issued government- and corporate public debt. As a proxy for the maturity they split corporate debt in longer (long-term) - and shorter (short-term) than one year debt. The amount of new issued is also a proxy determined by the change in the total level of long (short) term funds plus one-tenth of the level in the previous year are reported by the U.S. Treasury department (flow of funds).

Greenwood, et al. recon that more detailed study towards actual debt maturities would further improve the quality of their study and see this as one of the recommendations for future studies. The second main point of critique towards the study of Greenwood, et al. is that in order to match the data available on corporate debt (yearly) with the

monthly data on government debt they lose part of the underlying detail. When the maturity of newly issued government debt is measured on a yearly basis the reaction of corporations, which theoretically should act 6 to 12 month post government, gap

creation becomes less straightforward. In order to properly test the connection corporate and government debt maturity on new issues should be measured semi-annually or even quarterly as it is more in line with the theoretical framework.

In order to better, understand the debt maturity choices by corporations this paper will add to existing evidence on gap-filling theory. The aim of this paper is to validate the gap-filling theory hypothesis developed by Greenwood, et al. using actual maturities of debt paper in a more recent timeframe; 1993-2014. Doing this will not solely improve insight in gap-filling in a more recent time frame it also counters same of

(19)

the limitations of the Greenwood, et al. research. Instead of using the delta between long-term debt liabilities in a given year as proxy for debt maturity individual fixed income debt issued will be analyzed in coming chapters. As debt issued by government and corporations is the variable at the center of gap-filling theory. This paper will also look at the maturity of corporations and government on a quarterly basis instead of a yearly. As Greenwood, et al. already wrote in their paper they were unable to obtain this desired detail of debt instruments but see it as relevant for further research. This paper tries to fill that gap without distancing itself from the ideas and overall research method of Greenwood, et al. (Greenwood, et al., 2010) In order to do this the hypotheses of underlying gap-filling theory will be assumed to hold and tested using a new, more detailed and recent dataset. The three hypotheses based on the framework of Greenwood, et al. are:

Hypothesis 1) Corporate bond issuers will fill in supply gaps in total bond supply

resulting in a negative correlation between the maturities of corporate and government bonds issued.

Hypothesis 2) The higher the percentage of government debt issued in relation to total debt issued the stronger the gap-filling effect.

Hypothesis 3) Firms with stronger balance sheets are more flexible and therefore more capable of filling in the gaps in the maturities of the bonds issues.

These hypotheses will be tested using both univariate and multivariable regression analysis for the dataset time frame 1993-2014. Using capital IQ data of individual debt issues will aggregated to create an average maturity for each variable on a quarterly basis. The next chapter will go into further detail regarding data and

(20)

3) Methodology

The third chapter will start by explaining the dataset used for testing gap-filling theory in the period 1993-2014. After discussing the various variables used and their origin, this chapter will continue by comparing the created dataset with the data of Greenwood, et al. The relevance of this study lies in the use of improved proxies for the maturities of new issues, but as it is grounded on the works of Greenwood, et al, This paper will therefore first determine the overall comparability with their study. This chapter ends with an explanation of the tests that will be run for each hypothesis and the expected outcomes based on the theoretical framework of gap-filling.

Data

This paper uses two samples consisting of individual corporate and government US debt instruments issued in the period 1993 – 2014 available through the Capital IQ database. For the first and second hypothesis the data available in the Capital IQ Fixed Income Screening Report which has sufficient detail to determine the extent to which a relation exists between the maturity of new government debt issues and new corporate debt issues. Individual debt issues from this source are added on a quarterly basis. The maturity date and volume of the new issues is then used to determine the average debt maturity and weighted average debt maturity on a quarterly basis for new government and corporate debt. The offering date is set at the first of each month for aggregation purposes. The maturity of individual issues is calculated in years based on the offering date (set at the first of the month) minus the maturity date after which the total is divided by 365 days. By looking at the relation between the debt maturity of

government bonds and corporate bonds at each data point during the period 1993-2014 the correlation between government and corporate debt can be found. Data points each contain the debt issues of the previous quarter starting in Q2 1993 and ending in Q4 2014. The total number is 87 point for which an average and weighted average of all government debt issues and corporate debt issues is calculated. The expectation is that the weighted average of debt is the best variable for our analysis. You could however reason that it is the average price that determines the benchmark for new issues

because it is predominately based on multiple transactions and less on the impact of one or a few issues of large parties who might have more influence in the marketplace. This

(21)

paper therefore tests the hypothesis using average maturity of corporate and

government debt in a regression analysis and separately tests the weighted average maturity of debt issued. For government bonds, this paper uses the following categories of debt from Capital IQ; government bonds, government bills, government notes,

government strips – interest, government strips – principal and government trust certificates. In the timeframe of 1993 – 2014, this adds up to 3.388 US government debt instruments issued and incorporated in the dataset. For corporate debt, this paper adds two filters so both government and corporate debt issuers are located in the United States. This to maintain comparability with the study of Greenwood, et al. This is done by filtering on parent head quarter within the dataset. The second filter is created using the available industry classification in which banks are excludes from the dataset. In order to align this paper with the findings of Greenwood, et al. banks who are seen to have different funding requirements than other corporations. The following Capital IQ categories are includes in the dataset; corporate inflation indexed, corporate

debentures, corporate MTN, corporate MTN zero, corporate pass thru Trt, corporate PIK bond, corporate strip, corporate zero, inflation indexed security, corporate insured debenture. These corporate debt issues add up to a total of 307.559 instruments

incorporated in the dataset. The total amount of debt issued, which is used in hypothesis 2, is based on the combined amount of debt from corporate and government debt in each individual time period.

For the thirds hypothesis additional firm specific information is needed to

determine the balance sheet strength of an individual corporate issuer of new debt. For this purpose, additional dataset is created based on the combined data of the Capital IQ dataset, of hypothesis 1 and 2, and additional information from the Orbis database. Balance sheet strength is measured using the solvability ratio (asset based) available in Orbis which provides data up to the year 2005. In order to combine the Orbis data with the existing dataset this paper checked the availability of the solvency ratio for all companies in the dataset of hypothesis 1 and 2. This resulted in a subset of 49

companies who issued debt in the years 2005-2014 and were includes in the Capital IQ dataset and of who the solvency ratio was available for all years through Orbis. In total 148 debt issues by corporations are incorporated in this sub dataset. Per time period, quarterly, this paper then derived the average, weighted average of new debt issued by

(22)

corporations in combination with the average solvability ratio of those who issued during each time period. The average and weighted average government debt maturity of new issues is equal to those calculated in the dataset for hypothesis 1 and 2. In line with the analysis of Greenwood, et al., additional control variables are added to

determine the extent to which the corporate debt maturity is determined by other variables than government debt maturity. Additional data is therefore added to the dataset of hypothesis 1, 2, and the subset of hypothesis 3. In order to test the impact of overall economic growth this paper uses the inflation on a quarterly basis derived by taking the average of the inflation each quarter based on the public information made available by the bureau of labor statistics (U.S.) The inflation used through this source is grounded in the the development of the consumer price index. A second control variable is the spread on corporate bonds as to control for the issuing of shorter maturities by corporates when the term spread for credit is high in absolute terms. The proxy used is the moody’s Baa yield on seasoned corporate bonds – all industries minus the ten-year U.S. treasury constant maturity rate based on GS10 published by the board of governors of the federal reserve system. The third control variable is the term spread, which controls for the height of bankruptcy costs over time, as described in the previous chapter higher bankruptcy costs could be another reason for firms to decrease the maturity of the debt issued. For the term spread the difference between the one and twenty-year, U.S. treasury constant maturity rates are used. This variable is based on the GS1 and GS20 series of the board of governors of the Federal Reserve System. The data is not seasonally adjusted and the unit type is the natural log of percent in line with the variable used by Greenwood, et al. The fourth control variable is the total public debt as a percentage of the gross domestic product, which controls for overall demand of long-term debt, this variable is, based on the Federal Reserve Economic data set using code GFDEGDQ188S. This is the total public debt as a percentage of gross domestic product on a quarterly basis and seasonally adjusted.

Data comparison with study of Greenwood, et al.

Before examining the hypothesis, this paper will look at the comparability of our dataset and independent variables of Greenwood, et al. While the measurement of our government debt maturity of newly issued debt instruments ought to be more precise

(23)

the outcomes of analysis are expected to be roughly in line with the results of Greenwood, et al. In the coming analysis this paper will therefore compare the independent variable used in the original study of Greenwood, et al. with the

independent variable used in this study. This paper will add a proxy of corporate debt maturity based on the change in the flow of funds as was done by Greenwood, et al. The goal is not only to compare the underlying data and make sure the independent

variables are in sync, but also to determine the extent to which the negative association between government debt maturity and corporate debt maturity exists.

This first analysis will looks at four variables, two are used by Greenwood, et al (period 1963-2005) and two based on our dataset. The first one from Greenwood, et al.

is corporate debt maturity with a maturity bigger than 1 (variable name: DCL/ DC_{), the}

second is government debt maturity (variable name: GrM). The third variable is derived from our own dataset is, 3) the weighted average government debt maturity of new issued (variable name: AGDM) and so is the fourth which mimics Greenwoods, et al’s, proxy for 4) corporate debt maturity (variable name: GrCDM) For this last proxy this paper separated government debt issued with a maturity longer than one year (long-term debt) and debt issued with a maturity up to one year (short-(long-term debt). The actual maturity of government debt was already the main variable in our dataset. For the measurement of corporate long-term debt (GrCDM) a variable was added based on the quarterly Financial Accounts of the United States.-Z.1 (formerly called the Flow of Funds Accounts of the U.S.) which also provides corporate flow of funds data reported by the

US treasury department.2_{In table 2 the four proxies are compared which while}

representing slightly different time periods are very comparable indicating the suitability of our variables and dataset as a comparable reproduction of the data reported by Greenwood, et al.

The expected outcome of this initial analysis is that our detailed calculation of the independent variable, government maturity of debt securities issued (AGDM), provides similar outcomes in terms of Beta’s when tested for the association with corporate

2_{Flow of funds data is the reported data from the US Department of Treasury which are the quarterly Financial}

Accounts of the United States.-Z.1 (formerly called the Flow of Funds Accounts of the U.S.). source: treasurydirect.gov

(24)

term debt (GrCDM) as Greenwood, et al. found. It is expected that an increase in the maturity of government debt issues would be associated with a lower maturity in debt issues by corporations. The findings of Greenwood, et al. indicate that this association, which tests the effect of gap-filling, can be found in their dataset, this paper therefore expects the same.

Looking at the variable characteristics in table 2, results show that our mean and standard deviation of the corporate long-term issue share (GrCDM) similar is to the original proxy of Greenwood, et al. In addition, the average government maturity of debt securities issued (AGDM) when comparable to the of Greenwood, et al. shows

comparable output. Only the maximum outliers seems higher in our sample, which is due to less averaging of data and due to the inclusion of the period after 2005.

Table 2 Summary statistics

Characteristics of depended and independent variables form the study done by Greenwood, et al. and for comparable variables used in this paper.

variable Mean SD Median Min Max

Greenwood, et al. 1963-2005

Long term/total corp. debt level DCL/ DC _61.51 _4.97 _60.93 _53.46 _73.12

Gov. debt maturity in years GrM 4.51 0.9 4.57 2.82 5.75

This paper 1993-2014:

Proxy corporate debt Greenwood GrCDM 66.47 5.5 67.2 58.9 74.9

Weighted average gov. debt maturity

AGDM 4.27 1.1 4.13 2.4 8.1

After a first view at the variables used in this study and their comparability to those of Greenwood, et al. this paper will look if the association found by Greenwood at al also holds for our dataset. This paper wants to test if the independent variable, the weighted average government debt maturity (AGDM) as a similar association in our sample to the corporate debt maturity (GrCDM) proxy of Greenwood, et al., using univariate linear regression. In the following graph we will first test for normality in order to determine are possible linear regressions.

(25)

Graph 1)

Normality test of proxy for the variable of corporate debt maturity (GrCDM) based on flow of funds data and for the variable of government debt maturity (AGDM)

Based on graph 1 we must conclude that the proxy (GrCD) for long-term corporate debt maturity is not normally distributed for the period 1993-2014 due to two large outliers on the 59 and the 73 bucket. The data shows three peaks therefore linear regression analysis a less suitable statistical method for analysis. When looking at the average long term government debt maturity and the total average government debt maturity both are normally distributed so while weak linear regression statistically is can be used. Nonetheless this paper will, in addition to linear regression analysis add a logistic regression analysis in order to determine the association between the

government and corporate debt maturity variables. By dividing the corporate long-term maturity (GrCDM) into three quintiles; 0-61.99, 62.00-70.00 and 70.01-100 we are able to adapt our statistical analyses too correctly incorporate the two outliers shown in graph 1. By comparing the quintiles with one another, using two dummy variables, the association between corporate long-term debt maturity (GrCDM) and government debt maturity (AGDM) can be tested.

In table 3, the results of the linear regression of the independent variable for government debt maturity (AGDM) and the dependent variable for corporate debt maturity (GrCDM) are show at baseline (Lag 0) and for a lag of two period. Lags are used to mimic the time needed by corporations to set up a debt issuing programs through which that change the maturity of debt they issue. As described in the chapter on the theoretical framework corporations are expected to need 3 to 9 months for a new debt

(26)

program. In our regression analysis this paper will therefore control for a small delay in the reaction of corporations.

Table 3) The maturity of corporate and government debt univariate regression. This table shows the results of the linear regression analysis with the weighted average government debt maturity (AGDM) as independent variable and the Greenwood, et al. proxy for corporate debt maturity (GrCDM) as dependent variable. Lags are added to control for the reaction time needed by corporations. In additional analysis lags 1-14 showed significant outcomes indicating that outcomes are statistically to similar for this time of analysis.

95% CI, 95% confidence interval; LB, lower bound; UB, upper bound; β, beta; se, standard error; t, t-statistics; GrCDM, Greenwood, et al. corporate debt maturity proxy; AGDM, weighted average government debt maturity

Dependent variable: GrCDM at: β 95% CI LB 95% CI UB p r2 _se _t

Lag 0 0.55 1.91 3.90 1.91 0.23 0.50 5.80 Lag 1 0.56 0.02 0.04 <0.001* 0.23 0.05 5.96

The outcome seems not in line with the expected betas as the association is positive instead of negative as found by Greenwood, et al. In addition a significant association is found for all lags between 0 till 14 this seems highly unlikely and

combined with the fact that the GrCDM was not normally distributed we will not build on these outcomes of the regression analysis.

In order to deal with the non-existence of a normal distribution can use the logistic regression analysis. The corporate long-term debt maturity (GrCDM) which we divided into three quintiles ranging from 1) lowest-61.99 2) 62.00-70.00 and 3) 70.01 – highest with an equal number of data points in each tertile. These quintiles are all normally distributed and can therefore be used to create two dummy variables. (see appendix 3) The first dummy variable is GrCDMdummy1, data is restructured to the extent that all observation that fall within the first tertile receive a score of 1 and observations in the second and third tertile receive a score of zero. The second dummy variable is GrCDMdummy2, data for this variable is restructured so that all observation that falls within the last tertile receive a score of 1 and all observations in the first and

(27)

second tertile receive a score of zero. With a logistic regression analysis, we can see if a higher level of government debt maturity is associated with corporate bond issues, which have a lower maturity and thus fall in the lower tertile. In the two logistic regression analysis, one for each dummy the expected relation between the weighted average government debt maturity (AGDM) and the proxy for corporate debt maturity of Greenwood, et al. (GrCDM) is that a higher AGDM would be linked to GrCDM in the lower tertile and around. The second dummy (GrCDMdummy2) is therefore expected to have a higher odds ratio indicating the relation is stronger because the maturity of corporate debt measured as percentage of total corporate debt within this tertile has more extreme outcomes.

(28)

Table 4) The maturity of corporate and government debt, logistic regression.

Results of logistic regression analysis with weighted average government debt maturity ( AGDM) as independent variable and two dummies in order to test the dependent variable of corporate debt maturity calculated based on Greenwood, et al.’s method. GrCDMdummy1 and GrCDMdummy2 are thus both dependent variables. The odds ratio is the outcome variable for this analysis it indicates (GrCDMdummy 2) the odd that an increase of 1 year in the maturity of government is combined with a corporate debt maturity which lies in the third tertile and (>higher than 70%) for GrCDM dummy 2 this is 2,38.

95% CI, 95% confidence interval; LB, lower bound; UB, upper bound; OR, Odds ratio se, standard error; t, t-statistics; Wald, Wald chi-square test ; GrCDMdummy1, dummyvariable 1 Greenwood, et al. corporate debt maturity proxy; GrCDMdummy2, dummyvariable 2 Greenwood, et al. corporate debt maturity proxy AGDM, weighted average government debt maturity

Dependent variable GrCDM variable Odds Ratio 95% CI LB 95% CI UB p r 2 _se _Wald GrCDMdummy1 Dependent 0.48 0.29 0.78 0.004* 0.15 0.26 8.34 GrCDMdummy2 Dependent 2.38 1.42 3.99 0.001* 0.21 0.24 10.75

(29)

The results of table 3 show an association between government maturity of new issues and the corporate debt maturity, using the dummy variables GrCDMdummy 1 and GrCDMdummy 2. The odds ratio for GrCDMdummy1 has an odds ratio of 0.475

indicating that when the maturity of government debt issued increases with one (= 1 year) the odd that the long term corporate debt GrCDM has a maturity which lies the tertile lower that <62% of all values is 0.475 odds. (In percentages the impact is

approximately -32.2% as 0.475 odds is 67.8%) This decrease in corporate debt maturity when government debt maturity rises is in line with the findings of Greenwood, et al. and the expected outcomes. An increase in the weighted average government debt maturity (AGDM) would lead to increased gap-filling by corporations in lower maturity buckets due to the liquidity premium in these buckets. This increases the likelihood that corporations change the maturity of new debt issues, this result in the outcomes of table 4 as the odds that corporations have issued debt maturity, which fall in lower buckets increases with rising government maturities on new issues. The results of the second dummy variable of corporate debt maturity GrCDMdummy2 is higher than that of the first dummy for corporate debt maturity GrDMdummy1. This is in line with our theoretical framework and expected outcome as this variable looks at the odd that corporations issue debt maturities, which are in the highest tertile, which is the top 70% of the range. Results indicate that the odd that corporations issue debt in maturities in the highest tertile (>70%) accompanying an increase of government debt maturity of 1 year had an odds ratio of 2.78. This indicates that when the maturity of government debt issued increases with one year that the odd for long-term corporate debt to be higher, and lie in the higher tertile is 02.78 odds. (approximately -76.7% as 2.78 odds is 33.3%) The higher the government debt maturity (AGDM) the lower the chance that

corporations will issue debt maturities in the highest tertile indicating that the debt has the highest maturities. The logistic regression analysis confirms the relation between weighted average government debt maturity (AGDM) and Greenwood, et al. proxy for corporate debt maturity (GrCDM). In line with the regression analysis, an association is found between government debt maturity (AGDM) and the proxy for corporate debt maturity issued (long term debt as percentage of total debt (GrCDM)).

In the above section this paper has done a comparative analysis between the concept and proxies used by Greenwood, et al. in their theoretical model and dataset

(30)

1963-2005 and the dataset used within this study 1993-2014.

The conclusion from several regression and logistic analysis is that the association underlying gap-filling theory holds within the timeframe 1996-2014 when the proxy for corporate debt maturity of Greenwood, et al. is used in combination with the

government debt maturity based on our dataset. In the next part, we will therefore improve analysis used by Greenwood, et al. by adding a more detailed variable for corporate debt maturity. The actual maturity of newly issued corporate and government debt instruments will used to test the hypotheses set in the previous chapter.

In the coming chapter, the maturity of corporate debt issues is the dependent variable in each regression. Corporate debt maturity of new issues is assumed to be influenced by the maturity of government debt issues in a given time period as

described in the theoretical part of this paper. The maturity of newly issued government debt is the independent variable. In the second hypothesis the size of government debt in relation to the total debt issued in that time period is an interaction variable which is expected to increase the association between corporate and government debt

maturities. For the third hypothesis, the solvability ratio (asset based) is used as a proxy for the balance sheet strength of corporations. This variable is also an interaction

variable as we expect that stronger corporations are more likely to act as gap-fillers which would result in a higher association for firms with higher solvency ratio’s. In the next pages the hypothesis will be translated into regression, formula’s after which the expected outcomes are discussed.

Hypothesis 1) Corporate bond issuers will fill in supply gaps in total bond supply

resulting in a negative correlation between the maturities of corporate and government bonds issued.

When governments choose longer maturities when issuing debt, corporations will tilt their debt issuance away from longer-term maturity and visa versa. The underlying reasoning is that corporations can lock in a premium when they adabt the maturity of their debt contrary to the government. When this illiquidity premium ofsets the cost of changing the target debt maturity of the corporation it will benefit by

alterning the maturity of new debt issues. By testing the extent to which corporate debt marutity is negatively associated with governmetn debt maturity we can test if we can

(31)

find supportive evidence in line with the findings of Greenwood, et al. of the existence of the gap-filling. The expected outcomes of thes regression model is a negative beta

between government debt maturity and corporate debt maturity issued. A negative beta indicates that a higher maturity of government debt issued will result in a lower

maturity of corporate debt due to gap-filling. Corporate bond issuers will fill in supply gaps in total bond supply resulting in a negative correlation between the maturities of corporate and government bonds issued.

Based on the time corporations need to issue new debt Greenwood, et al. theorised a negative assotional between governmetn maturity of debt issues and corporate debt maturity of new issues is likely to ocure with a lag of 3 to 9 months. In their results Greenwood, et al. found that for lags up to two years the expected

associations appeared. In this paper we will control for the lags in the reaction of corporate debt maturity issues up to ten quarters, while focussing on the first three in line with the theoretical framework. If results provide additional insight they will be presented in the results.

The first hypothesis will be tested using two regression formulas so we can look at both the weighted average maturities of new issues and at the average maturities of issues. The average maturity of debt issued in a time period is calculated by taking the average of all debt issued in that period and taking the average of the maturity of these issued. For example if in period there would be 3 debt issues one with a maturity of 2 one with a maturity of 4 and one with a maturity of 6 the average debt maturity for corporation would be 4 in this period. For the weighted average we add the size of the debt issued in order to calculate a weighted average in the given period. By looking at both measurements of debt maturity we can control for one or a few large debt issues in time periods which might be issued on other conditions that the majority of debt issues. The regression formulas for hypothesis 1 are:

Regression formula 1a : CDMt = intercept + β * GDMt ± ε

CDM = average corporate debt maturity of new issued per time period GDM = average government debt maturity of new issued per time period t = time periods in quarters

(32)

Regression formula 1b : ACDMt = intercept + β * AGDMt ± ε

ACDM = weighted average corporate debt maturity of new issued per time period AGDM = weighted average government debt maturity of new issued per time period t = time periods in quarters

ε = standard error

Hypothesis 2) The higher the percentage of government debt issued in relation to total debt issued the stronger the gap-filling effect.

The size of government debt as a percentage of total debt influences the association between corporate debt maturity and government debt maturity. The

liquidity premium is expected to be higher when the government debt as a percentage of total debt issued in a period is higher. The tradeoff between costs for deviating from the internal target maturity of new debt and the illiquidity premium for gap-filling will therefore more often be in favor of the later. A higher percentage of government debt issued as a percentage of total debt issued will thus increase gap-filling and would lead to a stronger and more negative association and beta. With an interaction, term for the size of government to total debt issued per period we will test this relations for both average debt maturities and weighted average debt maturities.

Regression formula 2a : CDMt = intercept + β * GDMt + GDMt * (GDt / TDt) ± ε

CDM = average corporate debt maturity of new issued per time period GDM = average government debt maturity of new issued per time period GD = total government debt issued per time period

TD = total debt by corporate and governments issued per time period t = time periods in quarters

ε = standard error

Regression formula 2b : ACDMt = intercept + β * AGDMt + AGDMt * (GDt / TDt )± ε

ACDM = weighted average corporate debt maturity of new issued per time period AGDM = weighted average government debt maturity of new issued per time period GD = total government debt issued per time period