The Effect of Unfunded Pension Liabilities on State Borrowing Costs Part 1

The Effect of Unfunded Pension Liabilities on State Borrowing Costs 2012 Submitted to the Department of Economics in partial fulfillment of the requirements for the degree of Bachelor of Arts with honors The Effect of Unfunded Pension Liabilities on State Borrowing Costs Abstract My research examines the relationship between state liabilities, with a focus specifically on unfunded pension liabilities, and state borrowing costs. This study is also important for examining the market discipline hypothesis, which states that sovereign bonds’ default risk premium can be used to discipline irresponsible sovereign borrowers. My study provides some hope that financial markets are providing discipline to states that incur unfunded pension liabilities. Investors consistently price in these unfunded liabilities in risk spreads, and in my regressions, pricing in unfunded pension liabilities even appears to be more consistent than pricing in debt. The results in this study suggest that, at least during non-crisis times, investors are treating unfunded liabilities as they would any other state debt. However, my qualitative research of both unfunded pension liabilities and deficits suggest that there are reasons for caution as well. My analysis finds that only at a time when both these fiscal issues become dire do state governments feel the need to respond to these liability issues. The nonlinearity of the credit supply curve may not provide enough incentive to curb these borrowers in the pre-crisis phase and may provide a disciplining effect only during a crisis. If high unfunded pension liability or deficit states are forced to deal with their problems during a crisis, this may lead to massive pension benefit and expenditure cuts that could severely hurt the citizens in these states. iii Table of Contents Chapter 1. Introduction .............................................................................................1 Chapter 2. Unfunded Pension Liabilities: How Big are They and is This a Problem?.......................................................................................................................4 2.1 Calculating Public Pension Liabilities.........................................................5 2.2 The Future of Public Pension Liabilities.....................................................7 Chapter 3. The Credit Supply Curve: Theoretical Framework and Empirical Studies ..........................................................................................................................8 3.1 The Credit Supply Curve and Bond Yields: Theoretical Framework..........9 3.2 Credit Rationing and Nonlinearity.............................................................11 3.3 Effect of Fiscal Characteristics on State Bond Yields: Empirical Studies..............................................................................................................12 Chapter 4. Empirical Strategy..................................................................................14 4.1 The Basic Empirical Model.......................................................................14 4.2 Specification Issues....................................................................................15 4.3 Specifications in Previous Studies.............................................................17 Chapter 5. Data and Summary Statistics................................................................18 5.1 Data...........................................................................................................18 5.2 Summary Statistics.....................................................................................21 Chapter 6. Discussion................................................................................................25 6.1 A Simply Panel Data Estimation of the Credit Supply Curve...................25 6.2 Addressing the Simultaneity Issue in Time-Series Variation....................35 6.3 The Effect of the Financial Crisis on State Indicators...............................37 Chapter 7. Conclusion...............................................................................................42 iv Chapter 1. Introduction Whether the market can provide effective controls for profligate sovereign borrowing has been an area of much debate. Curbing excessive borrowing becomes especially important for sovereign borrowers because, unlike an individual bankruptcy that is usually caused by a negative net worth and has laid out legal procedures, a sovereign default and its procedures are not as clearly identifiable or straightforward. A sovereign borrower may, in fact, be unable to pay its debt or it may have the ability to repay but simply decide not to pay. Legal enforcement is difficult because most of a sovereign country’s assets are within the country itself and cannot be relied on in the event of default procedures. (Panizza et al., 2009, p. 653) The market discipline hypothesis is an oft-cited mechanism of preventing potential sovereign debt repudiation. The hypothesis states that sovereign borrowers face a nonlinear supply curve that curbs wasteful sovereign borrowing. As sovereign governments increase their levels of debt, the cost of borrowing increases at an increasing rate. As explained by Eaton and Gersovitz (1981), sovereign borrowers face this nonlinear supply curve due to credit rationing on the part of lenders.1 According to the market discipline hypothesis, the market itself can discipline sovereign borrowers and prevent sovereign defaults. However, just by looking at recent history, the market discipline hypothesis does not appear to hold true. One of the more striking examples of a lack of market discipline is the current sovereign debt crisis. As seen in Figure 1.1, after years of stability, markets 1 This is discussed in greater detail in Section 3.2. 1 Figure F 1.1: The T Sovereiign Debt Crrisis Source: Blloomberg New ws su uddenly becaame troubled d by the grow wing debt isssues in seveeral Europeaan countries, caausing a corrrespondingly y sudden jum mp in yields. There wass no gradual discipliningg of th hese irrespon nsible borrow wers, but a sudden s anxieety that serveed to intensify borrowers’ prroblems, esp pecially in Greece. G A lacck of early cchecks on Grreek borrow wing has led tto seevere austeriity measuress and GDP contractions ffor the past ffew years, bboth of whichh do not appear to be ending in n the near fu uture. It mayy be that durring the pre-ccrisis periodd, market m discip pline is too leenient and du uring a crisiss it is too haarsh. Althou ugh a crisis on the magn nitude of Greeece has not occurred in the United States, US staates have beeen having seerious debt isssues of theiir own. Desppite balanced bu udget laws, states, such as California and Illinoiis, have longg-term structuural imbalannces between revenues and exp penditures. In many staates, policy-m makers use thhe cyclical su urplus generrated during boom years to permanenntly reduce ttaxes or incrrease spendinng, faailing to see the temporaary surplusess as cyclical.. Downturnss, such as the period fo ollowing thee 2008 financcial crisis, ex xacerbate theese structuraal problems, and policy- 2 makers fail to undergo the necessary policy measures to combat these fiscal problems. California, the state with one of the largest total and structural deficits, addressed onethird of its almost $20 billion shortfall in fiscal 2011 with one-time or temporary fixes. (Murray, 2011) These growing fiscal concerns raise the question of whether markets can effectively discipline US state governments. Besides structural and cyclical deficits, unfunded pension liabilities are crucial for the analysis of states’ risk of default. Due to the 2008 crisis, these unfunded liabilities have dramatically increased. Furthermore, these are off-balance sheet liabilities and, as discussed in Chapter 2, they are difficult to calculate and most likely underestimated by states. Economists’ evaluations of the total magnitude of these liabilities range from $1.4 to $3.0 trillion. In many states, the unfunded pension liability exceeds the state’s bonded debt. The magnitude of states’ total liabilities raises concern for a potential crisis on the state level. Although many studies have analyzed the effect of state debt on borrowing costs, few have included unfunded pension liabilities in the analysis. The inclusion of this new important variable as well as the opportunity to analyze the effects of the financial crisis on borrowing costs allows my study to add to previous literature. The central issue examined is whether investors factor unfunded pension liabilities into state bonds’ default risk premiums, and, if they do, whether investors factor in these liabilities and bonded debt equally. I find that the ratio of unfunded pension liabilities to Gross State Product (GSP) does have a significant effect on state bond yield spreads of 11.2 basis points with an increase of one within standard deviation of the mean. I also find that an increase of one within standard deviation of debt to GSP is found to leads to an increase of only 4.2 3 basis points in yield spreads. This estimate, however, may be biased downwards as a result of simultaneity problems. In spite of these problems, the null hypothesis that these two estimates are equal cannot be rejected. A subsequent question in my study is whether the financial crisis caused a structural break in the effect of the parameters on yield spreads. I find suggestive evidence that although there does not appear to be a structural break across all parameters, the effect of unfunded pension liabilities on spreads appears to have decreased during the crisis. Lastly, my study analyzes whether markets can effectively discipline over-indebted state borrowers through a nonlinear supply curve. My study fails to find evidence of a nonlinear supply curve. Furthermore, through a qualitative analysis of both unfunded pension liabilities and state deficits before and during the financial crisis, my findings suggest that the market only disciplines profligate borrowers during periods of crisis. The paper is divided into seven chapters. Chapter 2 explains the debate concerning the size of the unfunded pension liabilities with a specific focus on the discount rate issue. Chapter 3 provides a theoretical framework for understanding interest rates and credit rationing and describes previous empirical studies that analyze the credit supply curve. Chapter 4 outlines the study’s empirical strategy and important specification issues. Chapter 5 summarizes the data, and Chapter 6 discusses the results. Chapter 7 then concludes and suggests policy implications. Chapter 2. Unfunded Pension Liabilities: How Big are They and is This a Problem? Within the past three years, a significant amount of research has been dedicated to calculating the magnitude of unfunded pension liabilities at the state level as well as analyzing the potential impact that these liabilities will have in the future. Although 4 these articles do not provide the theoretical or empirical analysis for how unfunded pension liabilities affect yield spreads, they are crucial for understanding the nature of these liabilities and the fiscal significance of analyzing the effect that these liabilities have on state operations. 2.1 Calculating Public Pension Liabilities State public pensions, unlike pay-as-you-go systems such as Social Security, work on a funded system, in which current assets should equal discounted future liabilities. In other words, the contributions that active workers make today are invested and then used to pay for the future benefits of these active workers. The unfunded liability, therefore, is the discounted future liability minus current assets.2 The controversial part of the calculation is calculating the discounted future liabilities, specifically regarding the correct discount rate. Like any future liability, pension liabilities must be discounted to their present value using a chosen discount rate. The Government Accounting Standards Board (GASP) ruling 25 mandates that states use their expected rate of return as the discount rate, which, depending on the state, is between 7.0% and 8.5%. According to the Pew Center for the States, the calculation as of 2009 for total unfunded pension liabilities across states using states’ various discount rates is $660 billion. Most economists believe that states do not use a low enough interest rate. Many argue that the liabilities of the public pension system need to be discounted using a discount rate that accounts for the risk associated with those liabilities. Brown and Wilcox (2009) explain that since the pension obligations are protected in most state 2 Many states use asset-smoothing techniques, in which they average out the pension fund assets over five or ten years. Using this asset calculation, the unfunded liability is the discounted future liabilities minus the smoothed assets. 5 constitutions, these obligations are “enforceable contractual relationships,” and therefore, they advocate the Treasury rate as the discount rate. In addition, they suggest the use of interest-rate swap contracts or a combination of swap contracts and Treasury rates, since interest-rate swap contracts have little credit risk. Novy-Marx and Rauh (2010) maintain that the risk of pension liabilities is at least as low as the risk of state General Obligation (GO) bonds. Brown and Wilcox (2009) agree. In fact, they argue that the risk may actually be zero, due to the constitutional, statutory, and common laws that guarantee these cash flows. Novy-Marx and Rauh (2010) re-calculate the pension liability numbers given in each state’s Comprehensive Annual Financial Report by first using the “taxable muni-rate,” the yields of zero-coupon state GO bonds of comparable maturity adjusted for the fact that pensions are not tax-exempt, and second using the riskless rate, specifically Treasury yields. Using the former method of calculation, unfunded pension liabilities total $1.4 trillion and using the latter method $2.6 trillion. Biggs (2010) calculates the total unfunded liabilities as well by pricing a put-option for how much taxpayers are willing to pay now to prevent a potentially larger future payment. Using this model, the total unfunded pension liabilities are $3.0 trillion. Although Brown and Wilcox (2009), Novy-Marx and Rauh (2010), and Biggs (2010) strongly argue for a lower discount rate, Lav and McNichol (2011), from a noneconomic public policy perspective, do not believe that these discount rates need to be changed. Arguing in favor of GASP standards, they claim that that the discount rate of around 8% is a correct assumption since 8% is the long-run trend for pension funds’ returns, and the volatility seen in the recent years does not affect the long-run consistency of these returns. However, Biggs (2010), to counter this type of argument, uses a Monte 6 Carlo simulation of asset returns based on state assumptions, and, due to the fact that an 8% return necessitates riskier investments, he finds that over the next 65 years there is a 40% probability that current assets will cover accrued benefits. 2.2 The Future of Public Pension Liabilities Although there is a range of opinions regarding the gravity of the state pension system’s unfunded liabilities, the economists in the debate generally acknowledge some level of unsustainability of the system. Rauh (2009) uses various investment return assumptions, specifically 6%, 8%, and 10%, and finds that the assets would be depleted by 2021, 2025, and 2030, respectively. Using the customary 8% expected return, he finds that in order to not exhaust the existing assets, states would need to contribute $100 million annually to cover the existing unfunded liabilities in addition to the necessary contributions for new liabilities. Novy-Marx and Rauh (2011a) look at various possible policy changes, such as increases in the retirement age, reduction or elimination of cost of living adjustments, and buyout rates of early retirement. They find that even the most extreme adjustments, many of which are politically unfeasible, do not dramatically decrease the unfunded liabilities. Munnell, Aubry, and Quinby (2010) have a more sanguine view of the situation than Rauh (2009) and Novy-Marx and Rauh (2011a), although they point to the dramatic difference between states. On average, states and municipalities spend 3.8% of their budget on public pension contributions, which, if liabilities amortized over a 30-year period, will need to increase to only 5.0% using an 8% discount rate and to 9.1% using a 5% discount rate. However, for three states that consistently do not pay their full Annual Required Contribution (ARC) and have low funded ratios, Munnell, Aubry, and Quinby 7 (2010) find that the budget percent would need to rise to 8% and 12.5%, using an 8% and 5% discount rate, respectively. Their analysis implies that the unfunded public pension problem may just be specific to certain states. On the policy side of the debate, Iris J. Lav and Elizabeth McNichol (2011), as mentioned above, argue that the discount rates used in these studies are too low, which exaggerates the funding issue. These authors provide a spectrum of arguments in the debate over whether we should be concerned about the unfunded state pension liabilities. A contentious point that runs throughout these articles is, once again, the appropriate discount rate, which is central to the determination as to whether these liabilities are a fiscal problem. To add to the existing literature, my study will utilize a different method of testing the significance of these unfunded pension liabilities. Instead of looking directly at the magnitude of the liabilities, I analyze how bond yields are affected by these liabilities to see whether these liabilities are increasing state borrowing costs. This study uses the unfunded pension liabilities numbers given by the states with the understanding that these numbers are most likely an underestimation of their true value. Chapter 3. The Credit Supply Curve: Theoretical Framework and Empirical Studies This chapter provides the theoretical framework underlying my empirical analysis of the state credit supply curve and summarizes previous studies in this area. This section first explains the general nature of the credit supply curve for states and how default risk is factored into bond yields. This theoretical framework is largely based on Goldstein and Woglom (1991). The second section discusses credit rationing and its implications for credit supply, with a specific focus on the sovereign credit supply curve. The third 8 section gives a brief literature review of previous studies that examine the credit supply curve with a focus on the studies that analyze unfunded pension liabilities. 3.1 The Credit Supply Curve and Bond Yields: Theoretical Framework The credit supply curve is the supply of funds that investors lend to state governments. State borrowing is generally assumed to be a small part of the entire credit market, and so states act as price takers, which means that they take the expected, riskadjusted yield to maturity as given. Nevertheless, there is great variation between states’ promised yields to maturity. States are not price-takers regarding the promised yield because any increase in debt levels increases the state’s probability of default, and so there is a spread of the promised yield over the expected, risk-adjusted yield. The yield to maturity of a bond is the borrowing cost for the bond issuer. For the case of a bond with default risk, the yield to maturity is called the promised yield because of the possibility that the return is not realized. Excluding other complication (to be discussed below), the promised yield of a bond with default risk for a given maturity can be decomposed into the expected yield to maturity and the compensation for the probability and consequences of default. In this simple form, the expected yield is equal to the expected yield of a riskless bond.3 The expected yield of the riskless bond, since there is no risk, equals the riskless yield to maturity. Therefore, a simply relationship between the yield of a riskless bond and the promised yield of a bond with default risk can be written as: 1 1 1 3 In reality, the expected yield of a bond with risk does not equal the expected yield of a riskless bond because investors need to be compensated for the fact that they are taking on risk. This would make the expected yield higher than the expected yield of a riskless bond. Preferential tax treatment also makes these two expected yields different. 9 where is the riskless yield to maturity, default (the default spread), and is the compensation for the probability of is the probability of default. This equation shows that the difference between the promised yield to maturity, which is , and the riskless yield to maturity, , depends entirely on the probability of default. Rewriting the above equation, the default spread is defined as: 1 1 In the above equations, it is assumed that the bondholder will either get full repayment or no repayment. This is usually not the case since the consequences of default can differ greatly, but the basic analysis remains the same. Another complication to the above equations is that investors need to be compensated for taking on more risk. While the default spread in the above equations only compensates for the expected loss of a bond, the default spread, in reality, also compensates for the increased riskiness of the investor’s overall portfolio due to the possibility of the bond’s default. This compensation is necessary due to the investor’s risk aversion. If given the choice between two assets that have the same expected return but different levels of default risk, investors prefer the less risky asset. In my study, all these components of the bond yield spread – compensations for the probability of default, for the consequences of default, and for increased risk – are combined together to form the default spread. I focus in this study on debt, and liabilities more generally, as important determinants of the default spread. The last complication to the above two equations is the preferential tax treatment of in-state bonds. In many states, there is no state tax for in-state bonds but there is an income tax for out-of-state bonds. This preferential treatment lowers the yield to 10 maturity for state bonds, and for most years, allows for AAA state bond yields to be below Treasury yields. Tax issues are not directly addressed in my study, but the effect of different taxes is relatively small. Furthermore, as explained in Section 4.2, my regression strategy in large part accounts for tax differences. 3.2 Credit Rationing and Nonlinearity Credit rationing is an important theory underlying the relationship between liabilities and interest rates. Credit rationing is when lenders limit the supply of credit even if the borrower is willing to pay a higher interest rate. Both Stiglitz and Weiss (1981) and Eaton and Gersovitz (1981) are seminal works that model the credit supply curve. Stiglitz and Weiss examine the effect of higher interest rates on a lender’s expected return. Due to the adverse selection effect (a higher interest rate attracts riskier borrowers) and the incentive effect (the incentive of borrowers to invest in riskier projects with higher expected returns), Stiglitz and Weiss conclude that the lender’s expected return does not increase as fast as the interest rate and that after a certain interest rate the lender’s expected return starts to decrease. Eaton and Gersovitz look specifically at sovereign countries that have the possibility of repudiating their debt, but are prevented by the likelihood of future exclusion from markets. Over a certain debt level, lender countries refuse to lend any more to a borrowing country because beyond that debt level, the benefit from repudiating the debt outweighs the cost of exclusion from future markets for the borrower countries. Both of these studies show that the credit supply curve can be backward bending when current borrowing affects the likelihood of default. At a certain promised interest rate, investors would refuse to lend anymore to the borrower, no matter how high the 11 interest rate, and instead, investors would require a lower debt level to offset the riskiness of such a high interest rate. Up to that point of credit rationing, as Bayoumi et. al (1995) explain, interest rates rise with increasing debt levels at an increasing rate, causing the credit supply curve to be nonlinear. The market discipline hypothesis is related to the theory of credit rationing. As mentioned above, this hypothesis states that due to the nonlinear supply curve, credit markets will naturally discipline profligate borrowers. Borrowers will be incentivized to not undergo wasteful borrowing and markets, at a certain point, will no longer lend to spendthrift borrowers. Both credit rationing and the market discipline hypothesis are analyzed in my study. However, as mentioned before as well, the market discipline hypothesis is called into question due to the market’s recent inability to discipline sovereign borrowers. A nonlinear supply curve that becomes increasingly steep with higher debt levels may be able to discipline borrowers, but if the market abruptly withdraws credit, the market may not be able to discipline state borrowers very effectively. 3.3 Effect of Fiscal Characteristics on State Bond Yields: Empirical Studies Drawing from the theoretical studies of Stiglitz and Weiss (1981) and Eaton and Gersovitz (1981), many empirical studies analyze the fiscal determinants of state bond yields. To assess the role of fiscal discipline in the newly formed European Union, a group of studies in the 1990’s and early 2000’s analyze fiscal discipline in US states as a comparative case. These studies (Goldstein and Woglom 1991; Eichengreen and Bayoumi 1994; Bayoumi et. al. 1995; Poterba and Reuben 1999; Poterba and Reuben 2001; Lowry and Alt 2001) examine the effect of deficits, debt levels, fiscal institutions, 12 and other state characteristics on the yield of state GO bonds. All these studies find that debt levels and fiscal institutions, such as debt restriction laws, deficit restriction laws, and other fiscal constraints, have the expected effect on bond yields, although there has not been support for a strong deficit effect on bond yields. A couple studies in the past few years have looked at the effect of unfunded pension liabilities on state bond yields, although this literature is relatively small. NovyMarx and Rauh (2011b) perform a case study that looks specifically at the last quarter of 2008 when the asset value of pension funds dropped dramatically, severely increasing the unfunded ratios. They find that for every 10% loss in pension fund assets there was a 7 basis point increase in bond yields for AA rated states and a 15 basis point increase for states rated AA- or below. In a panel study, Munnell et. al. (2011) analyze the effect of various state managerial, financial, economic, and debt-related characteristics on bond yields between 2005 and 2009. They include in this analysis is the percent of the ARC payment paid by the state, and find that the percent paid has a significant downward effect on bond yields. However, the effect is small, with a one standard deviation increase in ARC payment corresponding a 3 basis point decrease in bond yields. My study builds off this small body of research. Instead of an event study as in Novy-Marx and Rauh (2011b), my study uses panel data. My study also spans a longer time frame than Munnell et. al., extending back to 1997. In contrast to Munnell et. al. as well, this study tests the effect of the unfunded pension liabilities itself rather than ARC payment. Also, as described in Chapter 5, my study uses state bond yield indexes rather than the bond issuances themselves in order to have a more comparable yield data set. 13