Real interest rates were extraordinarily high in the 1980s by historical

Transcription

1 Journal of Economic Perspectives Volume 6, Number 1 Winter 1992 Pages Real Interest Rates and the Savings and Loan Crisis: The Moral Hazard Premium John B. Shoven, Scott B. Smart, and Joel Waldfogel Real interest rates were extraordinarily high in the 1980s by historical standards. From 1926 to 1981, the average real rate of interest on short-term Treasury bills was 0.1 percent, whereas the real rate averaged 4.7 percent for the final nine years of the decade (Ibbotson, 1989). The break in the time series appears equally sharp if one examines the monthly data from 1975 to Figure 1 shows the real rate of return on six-month Treasury bills over that period. 1 The graph shows that the highest real rates in the late 1970s were roughly comparable to the lowest rates during the 1980s. Very few time series illustrate such a sharp shift upwards. We have done some simple statistical tests on this time series; the results confirm what is apparent in the figure. Real interest rates shifted upwards by four or five percentage points in approximately The question is why. In this paper we review some of 1 These are realized real rates, calculated by converting both CD and Treasury rates to an annual yield basis and then subtracting the actual rate of inflation over the term of the investment. The dramatic jump displayed in Figure 1 also holds for expected real rates using lagged inflation rates or survey data to estimate expected inflation. For presentation purposes, the real interest rates shown in Figure 1 are smoothed via a 3-month moving average. 2 In regressions of the form, r t = c + αδ 80 + ε t where r t is the real treasury bill or CD rate in period t, c is a constant term, δ 80 is a dummy which is 1 in the 1980s and 0 before, ε t is an error term, and α measures the difference between real rates in the 1980s and the preceding period the point estimate of α is four to five hundred basis points and always significant at the 95 percent level. The difference in rates is significant whether one divides the sample at 1980, 1981, or 1982, although the difference in rates is largest when 1980 is the break point. John B. Shoven is Professor of Economics, Stanford University, Stanford, California. Scott B. Smart is Assistant Professor of Finance, Indiana University, Bloomington, Indiana. Joel Waldfogel is Assistant Professor of Economics, Yale University, New Haven, Connecticut.

2 156 Journal of Economic Perspectives Figure 1 Six Month Treasury Bill Rate the more popular explanations and point out that they are somewhat inconsistent with the facts. We then present a new explanation which may partially account for the dramatic increase. The foremost conventional explanation blames high real interest rates on the large federal government deficits of the 1980s. A related hypothesis attributes the jump in real interest rates to the combination of expansionary fiscal policy and the anti-inflationary shift of Federal Reserve policy in October A third explanation involves the slow adjustment of inflation expectations to the disinflation of the early 1980s; this argument holds that anticipated real returns were not as high as realized returns, because inflationary expectations still reflected investors' experience with double-digit price increases in the late 1970s. While each of these theories is a plausible cause of an increase in real rates, we question whether individually or collectively they can fully explain the timing, magnitude and persistence of the shift shown in Figure 1. Our new explanation for high real interest rates in the 1980s suggests that the upward shift in rates may be directly connected with the decade-long crisis in the savings and loan industry and the federal government's handling of that crisis. While regulators hesitated to recognize and address the problem of growing numbers of insolvent S&Ls, owners and managers of troubled thrifts responded to the incentives provided by underpriced deposit insurance by offering higher and higher rates in an attempt to attract new funds. 3 3 James Barth (1991) reports that the 205 S&Ls "resolved" by the FSLIC during 1988 had reported being insolvent for an average of three and a half years, and some had been insolvent for

3 John B. Shoven, Scott B. Smart and Joel Waldfogel 157 Depositors, anticipating that the government would protect their investments, actively sought out higher yields in local and national markets. The end result was that the rates offered by Treasury securities rose to compete with these quasi-risk-free substitutes sold by savings and loans. This added (and, indeed, continues to add) significantly to the federal government's borrowing costs. We calculate this increased cost under various assumptions about the effect of the S&L crisis on real interest rates. The Traditional Explanations Elementary macroeconomics teaches that large federal budget deficits cause high real interest rates. The argument underlying this view is straightforward: the increase in government borrowing to finance budget deficits is not fully offset by increased private saving or capital flows from abroad, so the interest rate must rise. As intuitively appealing as it might be, this proposition may be questioned on both empirical and theoretical grounds. It falls short as an empirical explanation in this case, in part because of the mismatched timing of deficits and high real interest rates. Real interest rates began to rise substantially in , but the dramatic shift in the U.S. budget deficit did not occur until From 1974 to 1981 the average annual deficit was $55 billion. From 1982 to 1989 the deficit averaged $176 billion per year (Economic Report of the President, 1990). Robert Barro (1974) questions the theoretical basis for the conventional view. He argues that consumers anticipate the higher future tax burdens associated with higher deficits and offset government borrowing with reductions in their own consumption. Altruistic parents increase their saving today to finance their children's consumption, neutralizing the effects of fiscal policy. Consistent with this hypothesis, Plosser (1982, 1987) and Evans (1987a) find no relationship between actual or expected budget deficits and interest rates in the U.S. or five other industrialized countries (Evans, 1987b). Using monthly data from (and many subperiods), Evans regresses real interest rates on measures of the deficit, government spending, and the money supply and finds no evidence of a positive relationship between interest rates and deficits. These results persist when Evans aggregates the data over time, and when he uses instrumental variables to estimate the equations in his model. 4 A second view of the increase in real interest rates advanced by Blanchard and Summers (1984) and Huizinga and Mishkin (1986) argues a change in up to ten years. Moreover, in the year prior to resolution, the largest of these failed thrifts were offering an average of 100 basis points more on their deposits than their competitors. 4 In his exhaustive review of the literature, Bernheim (1987) criticizes the theoretical foundations and empirical evidence for the proposition that deficits do not affect interest rates or other real variables. For other criticism, see Bernheim and Bagwell (1988), Bernheim, Shleifer, and Summers (1985), and Poterba and Summers (1987).

4 158 Journal of Economic Perspectives Federal Reserve policy in October 1979 was the primary force behind the increase in real rates. They present evidence that the change to a "tight" money regime shifted the stochastic process governing real rates. To buttress this argument, Mishkin (1988) points to a similar episode in the 1920s in which an unusually large increase in real rates followed a move to tighter money. The difficulty in timing here is that although monetary policy eased in the later 1980s, real interest rates remained at very high levels. Expectations provide the basis for a third explanation for high real rates in the 1980s. According to this theory, lenders who had earned very low or negative real rates of return on their investments in the inflationary '70s were slow to adjust their expectations of inflation. Because of the relatively rapid pace of disinflation in the early 1980s, the high real rates of return earned by investors in Treasury securities in the 1980s were just as much a "surprise" as the negative returns during the previous decade. While it is plausible that investors did not anticipate the rapid pace of disinflation during , the rate of inflation was relatively stable throughout the mid and late '80s: the December-to-December changes in the Consumer Price Index were 3.9 percent in 1982, 3.8 percent in 1983, 4.0 percent in 1984, and 3.8 percent in Real interest rates on Treasury bills, however, remained significantly higher than in previous decades. Explaining this as the result of a consistent inflation "surprise" strains credulity. A comparison of the actual year-to-year change in the Consumer Price Index with the change predicted by Eggert's Blue Chip Consensus (the average of about 40 forecasting services' predictions) illustrates that these forecasting services consistently overestimated inflation through most of the 1980s. However, the forecasting errors fell in the middle of the decade as inflation rates stabilized. Furthermore, since the size of the prediction errors was at most 2.5 percentage points in any given year, most of the basis point jump in rates must be attributable to other factors. We do not wish to dismiss any of the above explanations. All have some merit, and one could combine them to argue, for example, that the monetary policy shift in 1979 initially caused real rates to increase and that deficits kept rates high throughout the decade. Yet given the empirical difficulties in showing that deficits have affected real interest rates in any historical period, we believe that the persistence of high real rates requires further explanation. In the next section, we present evidence consistent with our theory that the savings and loan crisis contributed to high real interest rates in the past decade. The Moral Hazard Premium While the attention devoted to the S&L crisis by the popular press and by legislators probably peaked in the late 1980s, several economists documented the existence of the industry-wide problem much earlier (Kane, 1981; Carron, 1982). Savings and loan institutions experienced a dramatic decline in the

5 Real Interest Rates and the Savings and Loan Crisis 159 market value of their assets (primarily fixed-rate mortgages) in the 1970s. By 1980, the net worth of the entire industry measured at market value was actually negative (Carron, 1982, Table 1 4). Also in 1980, policy-makers began the gradual process of removing rate ceilings on the liabilities (deposits, CDs, and so on) of banks and S&Ls with the passage in March of the Depositary Institutions Deregulation and Monetary Control Act of 1980 (DIDMCA). Perhaps more important than easing restrictions on the rates financial institutions could offer their depositors, DIDMCA also increased the scope of federal deposit insurance from $40,000 to $100,000 per account. This legislation, in combination with the deteriorating financial condition of firms in the industry, gave S&Ls the means and the incentive to compete aggressively for funds. Considering the assurances provided to investors by FSLIC deposit insurance and the incentives of managers of weakened thrifts, it is hardly surprising that competition for deposits intensified. There is a moral hazard problem in a troubled thrift (Kane, 1989, 1990; Mishkin, this issue). Thrift institutions that are down, but not out, have every reason to take extra risks. The bigger the risk and the higher the stakes, the greater the chance that the firm might be saved. Of course, there also is a great chance that things will just go from bad to worse, but neither the owners (who may no longer have any of their money at stake at this point) nor the depositors (who are insured) have anything to lose. In their quest for money to play out the high-stakes, high-risk strategy, the troubled savings and loans bid up deposit interest rates. But due to the availability of federal deposit insurance, these high-yield institutions offer assets which are nearly perfect substitutes for deposits in safer institutions. Naturally, the safer institutions have to follow the movement up in yields if they want to remain competitive in the market for funds. The yield on other safe assets, such as Treasury bills, must also increase to remain as attractive as certificates of deposit (CDs) at high-risk institutions. We refer to this increase in yields on competing safe assets as the "moral hazard premium." The argument that Treasury security yields have carried a moral hazard premium in the 1980s does not require that Treasury bills and CDs be perfect substitutes. However, it does require that they be close substitutes in the sense that the demand for one of them decreases significantly when the terms offered on the other improve. This section presents three types of evidence supporting the view that CD rates have put upward pressure on Treasury bill rates: CDs and Treasury bills have similar intrinsic characteristics; CD and Treasury bill rates have moved together; and deposit flows have responded to the high rates offered by S&Ls. Treasury Bills and CDs Have Similar Characteristics Treasury bills are generally considered to be risk-free securities. While unpredictable inflation causes the real return to be somewhat uncertain and

6 160 Journal of Economic Perspectives interest rate variability may cause uncertainty regarding the return when the bill is not held until maturity, the full three-month or six-month nominal return is safe and predictable. Insured certificates of deposit have turned out to be extraordinarily safe, too. Such deposits have been honored at all of the troubled thrifts in the past decade. Even the deposits over the $100,000 insurance limit have often been protected, such as in the highly visible cases of Continental Illinois in 1984 or America Savings and Loan in Not only has depositors' money been safe, but it has been available in a timely manner. Few of the failed institutions have been closed for even an hour. Most often CD contracts of failed thrifts are completed by an acquiring institution. In a few cases where an acquisition has not been feasible, the principal and interest on CD accounts have been paid out prior to maturity. All in all, insured CDs have proven to be extremely safe investments, comparable in that regard with Treasury bills. Though CDs and Treasury securities have very similar risk characteristics, CD rates are typically higher than yields of Treasuries of the same maturity. Differences in the tax treatment of the two securities can explain the existence of this spread. Treasury bill income is exempt from state-level tax, while CD income is taxable at both the state and federal levels. Thus, even if Treasury bills and CDs had identical risk and maturity characteristics, thrifts would need to offer higher rates than are available on Treasury bills to attract CD investors in states with income taxes. For example, suppose that an investor who deducts state taxes on a federal tax return faces a state tax rate of 6 percent. When the nominal rate on CDs is 8 percent, that investor will be indifferent between investing in T-bills and CDs only if the spread between the two instruments is 48 basis points. 6 The costs of obtaining CDs and Treasury bills also differ. One can invest in CDs at very little cost through a local bank or thrift. Investing in high-yield CDs offered by nonlocal institutions became easier during the 1980s as major financial publications began printing lists of the institutions currently paying the nation's highest yields. Both the Wall Street Journal (November 2, 1988) and Money Magazine (November 1986) published articles in which the author urged investors to seek out these "no-risk, high-yield" investments. Today, there is even a number that plays a recording listing the current highest yields in certain major U.S. markets. 7 To invest in Treasury bills, one can avoid 5 The Southern Finance Project (1990) reported that accounts over $80,000 represented about one-third of the total value of protected deposits in a sample of 54 large failing thrifts. 6 It is interesting to note that the average nominal yield from on CDs was 9.6 percent, while the average spread was approximately 56 basis points. Given these values, the marginal tax rate "implied" by maximization of after-tax returns is 5.83 percent. In 1984, marginal state tax rates ranged from 0 percent to 14.1 percent for an investor with a $40,000 income. Across states, the average marginal rate was 5.3 percent, and the median marginal rate was 5.4 percent. 7 Brokers have also become much more active in channeling funds from investors to S&Ls. The Federal Home Loan Bank Board attempted to stem the flow of this so-called "hot money" by limiting its insurance coverage, but their proposal was struck down by the U.S. Court of Appeals.

7 John B. Shoven, Scott B. Smart and Joel Waldfogel 161 commissions by dealing either directly with the Treasury, or with a Federal Reserve bank or branch. Otherwise, a Treasury bill investor pays a brokerage commission. It is not clear which investment has higher transaction costs, but the costs are different. However, Treasury bills and CDs are available in similar denominations and maturities. Advertisers, as well as the business press itself, have described CDs as a risk-free alternative to Treasury bills. In the investment world, there is widespread and well-founded belief that CDs have risk characteristics similar to those of Treasury bills. CD and Treasury Bill Rates Move Together If Treasury bills and CDs are perfect substitutes and if financial markets equilibrate instantaneously, then CD and Treasury bill rates should move together, with a tax-induced spread between them. We would thus expect a high correlation between CD and Treasury bill returns. If Treasury bills and CDs are not close substitutes, then investors' required rates of return on the two instruments may diverge. Substitutability disciplines the relationship between movements in the two series. The correlations of nominal CD and Treasury bill rates from are 0.99 and 0.98 for three- and six-month instruments, respectively. First differences in these rates have correlations of 0.93 and 0.95, respectively. 8 In other words, the series move together. However, the spread between CD rates and Treasury bill rates has not been constant. Between 1975 and 1986, the spread averaged about 50 basis points for six-month CDs; in late 1987 and 1988 the spread persisted at a high level of 82 basis points; since 1989, the spread has averaged roughly 70 basis points. The large spread of may have reflected an additional risk premium that investors required for holding CDs. The fact that the spread fell in 1989 could reflect investors' perceptions that the 1989 S&L legislation (FIRREA) reduced the risk that the government would not honor its promise of deposit insurance. A risk-related spread that changes over time is not consistent with the notion that CDs and Treasury bills were perfect substitutes over this period. 9 However, all that we require to make our argument that high CD rates drive up the government's cost of borrowing is that when CD rates rise enough, insured deposit institutions attract funds that might otherwise have bought Treasury bills. The next subsection examines another part of this question, whether marginal thrifts are able to attract funds with high CD yields. 8 We obtained data on Treasury bill and certificate of deposit interest rates from the Department of the Treasury and the Federal Reserve Board of Governors respectively. 9 Alternatively, part of the movement in spreads during this period might be explained by the changes in state tax rates.

8 162 Journal of Economic Perspectives Interpreting Deposit Flows We have argued that deposit insurance made CDs very close substitutes for T-bills, and that moral hazard problems associated with the S&L crisis increased CD rates, making them very attractive investments vis-a-vis T-bills. One conceivable mechanism by which developments in the CD market could have affected T-bill rates is a movement by some investors out of T-bills and into CDs. Thus, one might conclude that if our theory is valid, one should observe funds flowing from T-bills to CDs, perhaps in massive quantities. Validation of our theory does not require such an empirical observation. Indeed, there are several reasons to expect only small changes in the relative quantities of T-bills and CDs. Consider the nature of Treasury auctions. The government simply announces a quantity of securities that it wishes to sell at a particular auction, and investors submit either competitive or noncompetitive bids for those liabilities. The government accepts the number of competitive bids that, in combination with noncompetitive bids, just exhausts its supply of securities for that auction. After the orders are filled, the quantity data reveal only that the Treasury sold its block of securities. The quantity of bills offered is determined primarily by the current shortfall in revenues and the stock of maturing liabilities. Pressures on T-bills from competing assets are likely to be observable only in price (rate) changes. Even if the demand for funds by the Treasury was (somewhat) elastic, it is not obvious that one would observe large quantity flows from T-bills to CDs. Consider this analogy: two gas stations on opposite street corners both sell a homogeneous product, both have self service and full service pumps, and so on. Now suppose that one station (station A) reduces its price. The new price is easily observable by potential customers and by the competing station (station B). If the price difference were sufficiently large, and if it persisted, then the likely outcome would be that station B would lose customers to station A (that is, quantity data would reveal a flow from the high-price to the low-price station). Naturally, the proprietor of station B is aware of the potential to lose customers, so station B matches its competitor's price cut. The two firms find themselves in the "bad" Nash equilibrium of a classic prisoner's dilemma. We emphasize that in the midst of such a price war, each station may retain its original clientele, and there may be little or no observable quantity flows between the two stations. The absence of an effect on quantities does not lead to the conclusion that there were no significant effects of the actions taken by the two stations. Similarly, it is plausible to expect that institutions offering the highest rates should be attracting new funds, at least in the short run. Below we present evidence that this is the case. However, it is inappropriate to conclude that only extremely large capital movements into S&Ls could have affected Treasury rates significantly. In equilibrium, as Treasury rates respond to pressure from

9 Real Interest Rates and the Savings and Loan Crisis 163 competing assets, financial flows between the assets may not appear unusually large by historical standards. The Flow of Deposits to High-Yield S&Ls Marginal thrifts might offer high rates on CDs for two reasons. First, they might offer high rates to offset the apparent riskiness of their securities. In this case, high rates would simply compensate depositors in high-risk thrifts for bearing additional risk. The high rates available at some institutions would not spread to government securities, nor even necessarily to other deposit institutions. 10 Second, marginal deposit institutions might offer high rates on CDs to attract above-normal deposit flows. For high rates to be contagious, marginal deposit institutions need to offer rates high enough not only to compensate investors for any additional risk, but also to entice them out of their current investments (at least if those current investments do not change their terms). Indeed, this is the central mechanism of our story: deposit institutions offer rates high enough to reduce the demand for government securities, thus raising the cost of all borrowing. There is some anecdotal and some systematic evidence that "bad," high-risk deposit institutions raise the cost of funds at healthy S&Ls. 11 On the anecdotal front, The New York Times (February 16, 1989) reports a cross-town CD rate war between Houston-area S&Ls. Two unhealthy S&Ls, Bancplus and Commonwealth Savings, repeatedly matched each other's rate increases, quoting new rates several times in only a few days. Executives at healthy Houston S&Ls complained that they were forced to keep pace with their financially-troubled rivals to maintain their deposit base. The funds that an S&L attracts when it raises its CD rates need not come only from its local market. During the 1980s, increasing amounts of CD deposits were placed with brokers who looked for the highest yields nationwide. Merrill Lynch, for example, reportedly sold $800 million in CDs for American Savings and Loan, a troubled California thrift. Drexel Burnham Lambert also specialized in selling CDs of Texas S&Ls. 12 Expressing his view 10 Lawrence White (1991) argues that the so-called "Texas premium" paid by thrifts in the state did not come about because those institutions were aggressively seeking new funds. Texas S&Ls paid higher rates, White argues, to compensate depositors for increased risk. At the same time, however, White recognizes that even healthy Texas thrifts were forced to pay a (smaller) premium. If solvent thrifts raise their rates to compete with insolvent ones, then it is plausible that this "contagion" effect could spread to thrifts in other regions and to other types of assets. 11 Stiglitz has suggested that this is a kind of Gresham's Law for S&Ls. 12 This story is reported in the Wall Street Journal, November 2, An official at one of the largest retail brokerage establishments informed us that his firm advised clients interested in low-risk investments to buy CDs rather than Treasury issues. In many cases these funds came from maturing Treasury securities.

10 164 Journal of Economic Perspectives that insolvent thrifts were able to attract huge inflows by raising rates, William Seidman, chairman of the Federal Deposit Insurance Corporation (FDIC), recently said, "Thrifts were essentially printing money through deposit insurance." Richard Breeden, chairman of the Securities and Exchange Commission (SEC) agreed: "The institutions that grew 2,000 percent or 3,000 percent weren't getting those funds because investors thought they were well-run businesses. Deposit insurance made those funds available" (Wall Street Journal, July 3, 1990). At a more systematic level, one can examine the movement of funds to states with deposit institutions offering high CD yields. Not coincidentally, states with large numbers of these high-paying thrifts also have many S&L failures. (Of course, an even better test would be to examine the reaction to increases in CD rates at the institution level, but we do not have firm-level data.) That funds chase high CD yields is an indication that high CD rates reflect more than a simple risk premium. It is possible to identify the states with high-rate institutions, and, furthermore, to see whether these states have experienced above-average deposit growth. Movement of deposits to institutions in high-rate states provides a mechanism for thrifts to raise the cost of federal borrowing. For each quarter between 1986:1 and 1990:2, we have tabulated the number of times each state's depository institutions are listed among the five highest yielding deposit institutions nationwide. 13 Each month, Money Magazine reports the four or five banks or thrifts offering the highest CD yields in the U.S., giving 12 to 15 institution-mentions per quarter. Three areas of the country stand out in the data: Texas, California, and the Northeast (Massachusetts, Maine, and Connecticut). Between late 1986 and early 1988, two-thirds of the institutions listed among the highest-yielding are in Texas, the state which would ultimately have the largest numbers of bankrupt S&Ls. Massachusetts appears on the list in 1987 as Texas's high-yield domination is starting to wane. By 1989, Massachusetts is the most prominent state on the list, contributing as many as half of the entries on the high-yield list per quarter. In 1990, as the New England economy continued to falter, threatening the viability of its financial institutions, other states from that region appear on the list. California, also a state with large numbers of S&L failures, steadily contributes a fifth of the high-yield institutions throughout the period. If the high rates offered by these deposit institutions were greater than necessary to compensate for risk, then the regions with high-yield institutions should have attracted disproportionate deposit flows. We examine the net deposit inflow into Federal Home Loan Bank Board (FHLBB) regions, scaled 13 We chose the period simply because the data to identify the institutions that were paying the highest rates were not available to us prior to Of course, each institution offers a variety of rates on different instruments. In this section, we use rates offered on six-month certificates of deposit.

11 John B. Shoven, Scott B. Smart and Joel Waldfogel 165 by each region's fraction of national thrift assets. If a region attracts 10 percent of national net thrift deposit inflow and the region's thrifts hold 10 percent of national thrift assets, then the region's ratio is 1. According to this asset yardstick, the region is in balance. When the ratio is above one, the region is attracting "more than its share" of deposits. 14 From 1975 to 1986, the Dallas region's (formerly the Little Rock region) ratio was persistently above 1. From 1980 to 1986, the ratio averaged over 1.5. In 1987, it fell to 1.4, and in 1988 the net inflow to Dallas region thrifts was actually negative. Thus, in 1988, even though Texas banks and thrifts offered some of the highest rates in the country, funds fled the region. This probably reflects regulatory activity. The federal regulators' "Southwest Plan," which took effect in 1988, sought to shrink and consolidate the Texas thrift industry in part through thrift liquidations (White, 1991, Ch. 8). In any case, the high CD rates were not enough to continue to attract funds to the region. Consistent with the steady appearance of California thrifts on our high-yield list, the San Francisco region had persistently high net inflows during the 1980s. Northeast institutions began to offer high yields in 1987, and net deposit inflows responded to these high rates. The Boston region's relative inflow ratio jumped from 0.46 in 1984 to over 1 in 1985 and In 1987 this ratio reached a height of 2.31, and in 1988 it remained high at 1.70, reflecting disproportionate deposit inflows to the Boston region. While we do not have the data to link deposit flows with interest rates paid by specific institutions, the evidence above indicates that states with large numbers of thrift failures and/or sagging economies (Texas, Massachusetts, California) tend to be the states where high-rate thrifts are located. Furthermore, we observe abnormally high deposit inflows at thrifts in these states. This channel offers a plausible way in which the combination of the S&L crisis and federal deposit insurance might have raised the costs of government borrowing. In the final section of the paper, we attempt a rough estimate of the magnitude of these additional costs. What is the Additional Cost to Government Borrowing? If high CD rates have driven up federal borrowing costs, as argued above, then it is interesting to ask what is the additional cost to the U.S. Treasury. Suppose that deposit institutions raise the rate on Treasury securities by some amount ε for ten years. Over the ten-year period, old debt taken out at low interest rates is gradually replaced with new high-rate debt, where the high rate is ε higher (in basis points) than the low rate. In the first year of the high-rate period, all existing government debt with maturity of less than a year 14 Data on regional deposit flows appears in the Savings and Home Financing Sourcebook from the Office of Thrift Supervision.

12 166 Journal of Economic Perspectives is replaced with high-rate debt; in addition, a portion of the longer-maturity low-rate debt is replaced with high-rate debt. After ten years, all of the old low-rate debt with maturity less than ten years has been replaced with high-rate debt. In this scenario, after ten years the rate falls to its original low level, and the fraction of outstanding debt at the high rate declines. If borrowing costs have risen by between 50 and 250 basis points and one discounts at 5 percent, then the present value of the additional cost is between $73.2 billion and $366.0 billion (in 1990 dollars). At a 10 percent discount rate, the bounds are $53.2 billion and $266.0 billion. These figures are significant even in comparison with the direct government costs of the savings and loan crisis. The impact of the high real interest rates of the 1980s on the economy has been linked with lower investment, slower economic growth, rising merger activity and corporate restructuring, and an increase in foreign holdings of U.S. assets. In addition, the deficit problem was made far more severe, possibly adding as much as $100 billion per year to federal government interest costs by the end of the decade. If the S&L crisis explains even a small fraction of the higher real interest rates, then the cost of the S&L bailout may be significantly increased. As long as insolvent thrifts are allowed to remain open, and, more importantly, as long as the price of deposit insurance does not reflect the risks taken by insured institutions, real interest rates will contain a moral hazard premium. We are grateful to Timothy Taylor, Carl Shapiro and Joe Stiglitz for helpful editorial comments. We also acknowledge the contributions of Mark Vaughan and seminar participants at Indiana University and the Tax Program Meeting of the National Bureau of Economic Research.

13 Real Interest Rates and the Savings and Loan Crisis 167 References Barro, Robert, "Are Government Bonds Net Wealth?," Journal of Political Economy, November-December 1974, 82, Barth, James R., "The Great Savings and Loan Debacle." Washington, DC: The AEI Press, 1991, Bernheim, B. D., "Ricardian Equivalence: An Evaluation of Theory and Evidence," mimeo, Stanford University, Bernheim, B. D. and K. Bagwell, "Is Everything Neutral?," Journal of Political Economy, April 1988, 96:2, Bernheim, B. D., A. Shleifer, and L. H. Summers, "The Strategic Bequest Motive," Journal of Political Economy, December 1985, 93:6, Blanchard, O. J. and L. H. Summers, " Perspectives on High World Real Interest Rates," Brookings Papers on Economic Activity, 1984, 2, Carron, A. S., "The Plight of the Thrift Institutions." Washington, D.C.: The Brookings Institution, Evans, P., "Interest Rates and Expected Future Budget Deficits in the United States," Journal of Political Economy, February 1987a, 95:11, Evans, P., "Do Budget Deficits Raise Nominal Interest Rates? Evidence from Six Countries," Journal of Monetary Economics, September 1987b, 20:2, Huizinga, John, and Frederic S. Mishkin, "Monetary Policy Regime Shifts and the Unusual Behavior of Real Interest Rates," Carnegie-Rochester Conference Series on Public Policy, Spring 1986, 24, Ihbotson Associates, "Stocks, Bonds, Bills and Inflation," 1989 Yearbook. Chicago: Ibbotson Associates. Kane, E. J., "Reregulation, Savings and Loan Diversification, and the Flow of Housing Finance." In Savings and Loan Asset Management Under Deregulation, San Francisco: Federal Home Loan Bank, 1981, Kane, E. J., "The High Cost of Incompletely Funding the FSLIC Shortage of Explicit Capital," Journal of Economic Perspectives, Fall 1989, 3:4, Kane, E. J., "Principal-Agent Problems in S&L Salvage," Journal of Finance, July 1990, XLV:3, Mishkin, Frederic S., "Understanding Real Interest Rates," American Journal of Agricultural Economics, December 1988, 70: Plosser, C. I., "Government Financing Decisions and Asset Returns," Journal of Monetary Economics, May 1982, 9, Plosser, C. I., "Fiscal Policy and the Term Structure," Journal of Monetary Economics, September 1987, 20, Poterba, J. M. and L. H. Summers, "Finite Lifetimes and the Effects of Budget Deficits on National Saving," Journal of Monetary Economics, September 1987, 20, Southern Finance Project, "Bailed out Thrifts: A Profile of America's Biggest S&L Failures," September U.S. Department of the Treasury, Office of Thrift Supervision, "Savings & Home Financing Sourcebook 1989," Washington, DC, U.S. Government Printing Office, Economic Report of the President, Transmitted to the Congress, Washington D.C., February 1990, 383. White, L. J., "The S&L Debacle: Public Policy Lessons for Bank and Thrift Regulation." Oxford: Oxford University Press, 1991.

14