The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
;< ■ ’ i Business Review Federal Reserve Bank of Philadelphia January • February 1 9 9 6 ISSN 0 0 0 7 - 7 0 1 1 Business Review The BUSINESS REVIEW is published by the Department of Research six times a year. It is edited by Sarah Burke. Artwork is designed and produced by Dianne Hallowell under the direction of Ronald B. Williams. The views expressed here are not necessarily those of this Reserve Bank or of the Federal Reserve System. SUBSCRIPTIONS. Single-copy subscriptions for individuals are available without charge. Insti tutional subscribers may order up to 5 copies. BACK ISSUES. Back issues are available free of charge, but quantities are limited: educators may order up to 50 copies by submitting requests on institutional letterhead; other orders are limited to 1 copy per request. Microform copies are available for purchase from University Microfilms, 300 N. Zeeb Road, Ann Arbor, MI 48106. REPRODUCTION. Perm ission must be obtained to reprint portions o f articles or whole articles. Permission to photocopy is unrestricted. Please send subscription orders, back orders, changes o f address, and requests to reprint to Publications, Federal Reserve Bank o f Philadelphia, Department o f Research and Statistics, Ten Independence Mall, Philadelphia, PA 19106-1574, or telephone (215) 574-6428. Please direct editorial communications to the same address, or telephone (215) 574-3805. JANUARY/FEBRUARY 1996 WHEN THE BUBBLE BURSTS: PSYCHOLOGY OR FUNDAMENTALS? Lee E. Ohanian The prices of stocks, bonds, and other assets frequently fluctuate, and some times these fluctuations are quite large. Such price shifts have important eco nomic implications, including the possi bility that asset prices have predictive power for the business cycle. In this ar ticle, Lee Ohanian analyzes the volatility of security prices and discusses whether movements in asset prices reflect changes in the fundamental value of the asset or whether extreme price changes may be associated with changes in market psy chology. THE CYCLICAL VOLATILITY OF INTEREST RATES Keith Sill Interest rates change in response to a variety of econom ic events, such as changes in Fed policy, crises in financial markets, and changes in prospects for long-term economic growth and infla tion. But such events are sporadic, and interest rates show a more regular pat tern of volatility that corresponds to the business cycle. In this article, Keith Sill examines some facts and theory about the cyclical volatility of short-term and long-term interest rates. When the Bubble Bursts: Psychology or Fundamentals? Lee E. Ohanian* p J L rices for stocks, bonds, foreign exchange, and other assets frequently exhibit large fluc tuations on a daily and long-term basis. Per haps the best known example of asset-price volatility was the 500-point decline in the Dow Jones Industrial Average on October 19,1987. The 23 percent drop coincided with similar *Lee Ohanian is an assistant professor of economics at the University of Minnesota. He wrote this article while he was on the faculty at the University of Pennsylvania and a visiting scholar in the Research Department of the Phila delphia Fed. Lee thanks Rick Lang, Steve Meyer, Dean Croushore, Keith Sill, Len Nakamura, and Sally Burke for helpful comments. declines in the Tokyo, London, and Hong Kong stock exchanges and was nearly twice the mag nitude of the October 1929 crash that ushered in the Great Depression. October 19,1987, was not the only turbulent day on the New York Stock Exchange in recent history. Since 1987, there have been 16 trading sessions in which the Dow moved at least 90 points. Extreme price volatility is not confined to the stock market, nor is it strictly a short term feature of the market. High variability has characterized foreign exchange rates since currencies were allowed to float in the early 1970s. The U.S. dollar, which rose 20 percent between February 1984 and February 1985, fell 3 BUSINESS REVIEW 25 percent over the following year. Price vola tility has also characterized the markets for corporate and U.S. government debt in recent years. Once the haven of conservative inves tors, the bond market now frequently displays fluctuations equal to those in the stock and foreign exchange markets. For example, the price of the 30-year U.S. Treasury bond rose more than 40 percent between October 1985 and July 1986 and fell nearly 20 percent during the first half of 1987. These price fluctuations have important economic implications. Recent empirical stud ies suggest that asset prices have predictive power for the business cycle. In particular, low bond prices (high interest rates) tend to pre cede recessions, and high bond prices (low interest rates) tend to precede expansions. There are also potentially important eco nomic costs associated with asset-price vola tility. In particular, substantial price volatility will tend to increase the volatility of returns on assets. Since investors typically dislike risk, high volatility will tend to increase the average rate of return on capital demanded by inves tors; that may lead to lower investment, a smaller capital stock, and a lower standard of living. This article presents an analysis of the vola tility of security prices. The objective is to discuss issues associated with whether move ments in asset prices reflect changes in the fundamental value of the asset or whether these extreme price changes might be associ ated with changes in market psychology that may not be related to business conditions. MARKET FUNDAMENTALS There is an old debate associated with whether asset prices correspond closely to their fundamental values or whether market psychology and extraneous factors can cause prices to deviate substantially from an asset's fundamental value. This debate has focused on the interpretation of changes in security http://fraser.stlouisfed.org/ 4 Federal Reserve Bank of St. Louis JANUARY/FEBRUARY 1996 prices and their volatility. Many academic economists have argued that security prices efficiently reflect current and past information and that market prices are a good approxima tion of a security's fundamental value. Funda mental values are often referred to as market fundamentals. The fundamental value of an asset is de fined as the present value of the expected payoff from that asset. For example, consider a hypothetical asset that yields $1 per year for five years. The fundamental value of this asset would be the sum of the five yearly payoffs, discounted by the relevant interest rate. (Dis counting a future cash flow by an interest rate is required because a $1 payoff in the future is not equivalent to a $1 payoff today.) One can use the same logic to determine the fundamen tal value of a stock. Since the payoff from a stock is the dividend, one measure of the fun damental value of a stock is the sum of all (expected) discounted future dividend pay ments. Market fundamentals, combined with the efficient markets theory, provide a simple tool for interpreting fluctuations in security prices. According to the efficient markets theory, se curity prices fluctuate only as investors re spond to new information concerning changes in market fundamentals (the discounted sum of future cash flows).1 For example, suppose a pharmaceutical manufacturer announces that it has developed and tested a new product that successfully combats cancer. The efficient mar kets theory predicts that the price of the company's stock would jump immediately as investors re-evaluate the security in light of the new information. The extent of the price in crease reflects how the new information alters market fundamentals. An increase of 15 per cent in the stock price indicates that the dis 1For a readable discussion of security prices and the efficient markets theory, see Burton Malkiel's book. FEDERAL RESERVE BANK OF PHILADELPHIA When the Bubble Bursts: Psychology or Fundamentals? counted sum of expected future dividends is 15 percent higher, according to the theory. A popular version of the efficient markets theory states that security prices will follow a "martingale."2The basic idea behind the mar tingale model for security prices is that the difference between a stock's price today and a stock's discounted price tomorrow is com pletely unpredictable.3 Thus, the main impli cation of this model is that the best forecast for tomorrow's stock price will simply be today's price. Moreover, the efficient markets theory implies that whatever change occurs in the stock price tomorrow will be completely ac counted for by new information on market fundamentals. This theory makes a number of predictions for the behavior of asset prices. One important implication of the martingale model is that trading strategies designed to "beat the mar ket" cannot be systematically successful. This follow s from the fact th at for the m arting ale model, the probability that the price of a stock will rise in value tomorrow is the same as the probability that the price will fall. Moreover, this theory predicts that stocks cannot be iden tified as under- or overvalued, nor are there particularly good or bad times to purchase stocks. Another strong implication of this theory is that the dominant investment strat egy is a very simple one: buy and hold a diversified portfolio of assets. This theory has been widely applied to un derstanding movements in asset prices. Its popularity likely reflects the fact that it pro vides a simple way of using basic economic theory to evaluate security prices. Also, an 2The martingale model of security prices, which has also been called the random-walk model, comes from an assumption that investors care only about the expected rate of return on an asset, not the variability of the return. te ch n ica lly , this implication is for the change in price plus any dividend amount. Lee E. Ohanian important im plication of the theory— that changes in asset prices are unpredictable— seems to be fairly well supported by a large body of data. However, some of the strong assumptions embodied in the theory, such as the risk neutrality of investors, and the fact that some other features of the data are diffi cult to reconcile with the theory have led to criticisms of this model. Some critics of the efficient markets theory point out that the volatility of security prices seems much too high to be justified by changes in market fundamentals. Market traders and many financial analysts claim that new infor mation about market fundamentals provides only a partial explanation of observed price fluctuations. While they acknowledge that long term movements in securities prices corre spond to changes in fundamentals, they argue that short-term fluctuations are caused by shifts in market psychology or perhaps even by events that have no direct b ea rin g on bu sin ess p ro s pects or economic conditions. BUBBLES A bubble is defined as any deviation of an asset's price from its fundamental value. We can think of an asset's price as consisting of two components: one associated with market fun damentals and the other representing the bubble. The bubble theory suggests that secu rities may go through periods of under- and overvaluation relative to fair-market values. One reason for this may be investor overreac tion. In the pharmaceutical example described above, investors may be overly optimistic in evaluating the increase in the firm's profits. Of course, investors have strong incentives to correctly evaluate how product developments affect firm profitability. This reasoning sug gests that it's unlikely that investors will con sistently overreact to news about firms' profit ability. Bubbles may also reflect investors' reac tions to factors unrelated to fundamental eco 5 BUSINESS REVIEW nomic and business conditions. Hypotheti cally, individual investors may rush into the stock market because they believe everyone else is making money in the market. In this case, they prefer to buy stocks immediately rather than miss an excellent buying opportu nity. As a result, the anticipation of rising prices becomes a self-fulfilling prophecy, and market participants enjoy profits that may not necessarily reflect favorable business prospects. For example, investors know that the out come of the Super Bowl played each January has had a good track record in predicting the course of that year's stock-market performance. When a National Football Conference team has won, the stock market has frequently in creased considerably over the year, while a win for an American Football Conference team often presages a lower stock market. Even though the outcome of a football game has little, if any, effect on overall business condi tions, the business press and investor publica tions often cite this correspondence. As long as some investors are perceived to act on this statistic, others also may buy in anticipation of this higher demand and rising prices. If enough investors behave this way, prices rise and ex pectations become self-fulfilling. Certain types of bubbles can be difficult to explain in a sensible way. They are similar to Ponzi schemes and chain letters in that partici pants will benefit from the game as long as others can be found who are eager to play the game. Of course, Ponzi schemes crash as soon as individuals believe it will be difficult to find others willing to participate. Similarly, some types of bubbles imply that dramatic declines in security prices are the result of investors finally realizing that rising prices may never be justified on economic grounds. At that point, investors try to sell their assets and prices drop: the bubble bursts. While certain types of bubbles seem to be inconsistent with rational behavior, there is a class of bubbles called rational bubbles.4 A ra 6 JANUARY/FEBRUARY 1996 tional bubble reflects a self-fulfilling belief among rational investors that an asset's price depends on variables unrelated to market fun damentals. In this context, a rational investor is an individual who efficiently uses relevant information for assessing the value of a secu rity. Within the bubbles framework, the fact that investors are rational means that while bubbles can exist, obvious profit opportunities cannot. This simply means that if an easy profit opportunity were available, a rational investor would exploit it and quickly eliminate the opportunity. In other words, for simple types of bubbles, the expected rate of return on a security must be the same whether or not the price includes a bubble. This means that one key feature of a rational bubble is that the evolution of the bubble over time is restricted to rule out easy profit oppor tunities. For example, a situation in which all investors expect a security to double in price between today and tomorrow, but fall back to its original value the following day would not constitute a rational bubble. In this case, every one would rationally want to sell the security tomorrow, so that the price would fall before the following day. Alternatively, an asset could be overpriced 20 percent relative to its funda mental value and, thus, could exhibit a rational bubble, as long as both the fundamental value and the bubble component are expected to grow at the same rate. For example, suppose that market fundamentals for a security were expected to grow at 5 percent per year forever. The price of this security would have a rational bubble if the bubble component also grew at 5 percent per year. In this case, the rate of return on the security with the bubble component 4A large literature has analyzed rational bubbles. This review provides an analysis of some very simple examples. For an extensive review of this literature, see the Journal o f Economic Perspectives, Spring 1990, Symposium on Bubbles, pp. 13-102. FEDERAL RESERVE BANK OF PHILADELPHIA When the Bubble Bursts: Psychology or Fundamentals? Lee E. Ohanian erns, where trades were made without margin limits and, presumably, the flow of spirits facilitated transactions. Interestingly, specula tion apparently spread to common bulbs unaf fected by the mosaic virus. In the first week of February 1637, prices peaked, and common bulb prices rose 20-fold in one month. Then, prices fell dramatically. While historical data from this period are sketchy at best, Peter Garber of Brown University has estimated that common bulb prices lost about 95 percent of their peak values just three months after the crash. A century later, the bulbs were virtually worthless. The strikingly colored Semper HISTORICAL EPISODES Augustus bulb, which traded for about $60,000 OF DRAMATIC PRICE MOVEMENTS (in current dollars) in February 1637, com A number of historical episodes of extreme manded just 50 cents in 1739. Tulipmania was a costly lesson for the Dutch. price movements have been interpreted as bubbles. While these episodes and the circum Unfortunately, the British did not learn from stances surrounding them bear little resem this episode. In 1711, some holders of short blance to modern financial markets, they are term British government war debt agreed to interesting to analyze, since they may be help exchange that debt for equity shares in a new government-chartered, joint-stock company ful in understanding current experience. Perhaps the most famous episode occurred called the South Sea Company. In return, the in 17th century Holland with an unlikely asset: company received a perpetual annuity paying diseased tulip bulbs. Tulipmania, as it is often 6 percent annually on the same face value of called, began quietly when a nonfatal virus, debt that had been exchanged. The South Sea known as a mosaic, attacked tulip bulbs. The Company was also given a monopoly on all effect of the virus was to produce a variegated trade to the South Seas. Although initial trad flower of brilliant stripes and colors. The virus ing was fraught with mistakes and a war with affected only a relatively small number of Spain shut off most trading opportunities, the bulbs, and these bulbs became highly prized price of the stock rose modestly. By 1719, it appeared that peace with Spain was at hand, by collectors. As the prices of the mosaic bulbs began to and as a result, prospects for the South Sea rise rapidly, investors as well as horticulturists Company looked better than ever. In 1720, many additional holders of govern began acquiring them. The increased demand for the bulbs resulted in even higher bulb ment debt traded the debt to the South Sea prices and large profits for existing owners. Company in exchange for new stock. The com Charles Mackay, who described this episode pany was expected to consolidate the debt and in his book, noted that "nobles, citizens, farm receive a steady stream of interest payments ers, mechanics, seamen, footmen, maid-ser on the government obligation. At this point, v an ts, even chim ney sw eeps and old the stock's price rose from 130 pounds to 300 pounds per share. After Parliament approved clotheswomen dabbled in bulbs." By 1635, tulipmania had engulfed the coun this plan, a new stock offering at 300 pounds try. Futures markets sprang up in local tav quickly shot up to 340. Fights among investors would be identical to the rate of return on the security without a bubble. Bubble interpretations have been popular with professional investors and the financial press for many years. In his introduction to Charles Mackay's Memoirs o f Extraordinary Popular Delusions and the Madness o f Crowds, the noted investor Bernard Baruch wrote, "All economic movements, by their very nature, are motivated by crowd psychology...Men think in herds; it will be seen that they go mad in herds, while they only recover their senses slowly, and one by one." 7 BUSINESS REVIEW JANUARY/FEBRUARY 1996 eager to buy the offering were common. The Critics of the efficient markets theory point next offering came out at 400, and the next at out that the theory cannot account for ob 500, with an option to buy at just 10 percent served volatility in security prices. Neverthe margin. When the stock hit 800, half of the less, the implication of the theory that changes members of the House of Lords and the House in asset prices are unpredictable has received a fair amount of of Com m ons em pirical sup plunged in. Soon PLEASE DON'T EAT THE TULIPS! port. For ex the price hit 1000 ample, a num pounds per ber of experi share. At this In his book, Charles Mackay relates an anecdote that m ents have point, the direc shows just how seriously the Dutch took their tulips. been conducted tors of the com Mackay describes an incident in which a young sailor in which stock pany began sell notified a merchant of the arrival of a shipment of new ing, w hich re p ortfolios goods. For bringing the news, the sailor was summarily picked by Wall sulted in rapid rewarded with a breakfast of herring. It so happened that the sailor noticed the ideal condiment for his herring, an liq u id atio n of Street's leading onion, perched on the merchant's counter and helped South Sea shares. money manag himself to it. To the merchant's— and ultimately the ers were com Parliament ul sailor's— distress, the "onion" was actually a prized pared over time timately passed Semper Augustus bulb. The merchant pressed charges, against a port the Bubble Act, and the unwitting felon spent several months in prison. folio chosen by w hich p ro h ib Of course, who knows how much— if any— of this throwing darts ited the issuing story is true. Peter Garber, for one, points out that an at a stock page of stock certifi astute merchant would hardly leave such a valuable from the Wall cates by compa object lying around, especially within easy reach of a Street Journ al. nies. So strong random guest. Nonetheless, it underlines the frenzy cre The martingale was the British ated by the speculation in tulip bulbs in 17th century Holland. model predicts aversion to a re that portfolios peat bubble that chosen at ran this law was in force for the next century: British companies dom should perform, on average, about the were not allowed to issue stock until 1825.5* same as those chosen by portfolio managers. In many of these experiments, random picks do BUBBLES VS. MARKET FUNDAMENTALS: just as well as many of Wall Street's leading EVIDENCE FROM MODERN TIMES traders. Moreover, critics of the bubble theory point Although not accepted universally, many economists agree that prices during these his out that technical analysis, which is the prac torical periods reflect some bubble compo tice of trying to identify systematic patterns in nent. Are bubble explanations of extreme price security price movements, should be useful in movements confined to just a few historical choosing securities if bubbles are present. The episodes, or might bubbles be relevant for basic idea is to plot security prices over time and use past price behavior to predict future today's financial markets? prices. Patterns often considered important for predicting future price movements include the "inverted head and shoulders," "triple top 5See Charles Kindleberger's book for a more in-depth treatment of the South Sea bubble. double bottoms," and "piercing necklines." In http://fraser.stlouisfed.org/ 8 Federal Reserve Bank of St. Louis FEDERAL RESERVE BANK OF PHILADELPHIA When the Bubble Bursts: Psychology or Fundamentals? general, these approaches have not signifi cantly outperformed randomly chosen strate gies or buy-and-hold strategies. Nevertheless, several observations from the stock market do challenge efficient markets explanations. One of the best known patterns is the January effect, which refers to the first two weeks of January when stock returns tend to be unusually high. This is also a period when stocks of smaller companies, such as those that tend to trade on the over-the-counter market, outperform larger, well-known issues. While selling stocks because of end-of-the-year tax considerations may play a role in explaining the January effect, it cannot completely ac count for the anomaly. The January effect was present in the United States even before in come taxes. Some economists have made another obser vation that challenges the market fundamen tals theory: the underpricing of initial public offerings (IPOs). An IPO is the initial sale of equity shares in a company that was privately held. Brokers allocate the initial offerings of shares to customers, and after the initial offer ing, these shares are traded on public ex changes. For many IPOs, the initial rate of return is enormous.6 In a 1988 paper, Roger Ibbotson, Jody Sindelar, and Jay Ritter re ported that between 1977 and 1987, the aver age initial return, which is defined as the per centage increase from the offering price to the end-of-first-day bid price, is over 20 percent. On an annualized basis, this rate of return would be in the neighborhood of over 1000 percent. These enormous returns suggest to some observ ers th at the shares are in itially underpriced. There does not appear to be a 6For example, in August 1995, Netscape, a company that produces software for the Internet, had an IPO with an offering price of $28 on Tuesday and closed at $58.25 on Wednesday. Lee E. Ohanian generally accepted theory of this observation, and it is somewhat puzzling as to why issuing firms would agree to deal with underwriters who underprice the security.7 TESTING FOR BUBBLES AND EXCESS VOLATILITY IN ASSET MARKETS The tulipmania and the South Sea bubbles are striking examples of how prices may di verge from fundamental values. Many econo mists think it unlikely that similar episodes could occur today. If there are bubble or nonfundamental components in asset prices, chances are they will be much less dramatic and harder to distinguish from market funda mentals. Until recently, claims that prices were out of line with market fundamentals were conjec tures, substantiated by little more than anec dotal evidence. However, recently developed statistical tests may help shed some light on the debate. A number of tests have been devel oped, and two widely used tests will be dis cussed here. Robert Shiller of Yale University developed and implemented one popular test that has been used to evaluate whether prices are con sistent with market fundamentals. Shiller con structed an economic model of the fundamen tal price of an asset. The test compares the volatility of the observed security price with the volatility of the fundamental price. These tests are typically called variance bounds tests, since the basic idea is to determine whether the observed variability of market price is consis tent with the observed variability of market fundamentals. For stocks, the model assumes that the price an investor would be willing to pay today depends on the total return (the dividend and price appreciation) he expects to receive from 7For additional discussion of asset market anomalies, see Richard Thaler's 1992 book. 9 BUSINESS REVIEW the stock tomorrow. In turn, the price in the following period depends on the dividend and price appreciation he expects to receive two periods from now, and so forth. This logic implies that the fundamental price of a stock today will depend on all expected future divi dends adjusted by an appropriate discount rate (interest rate). This analysis suggests that today's share price is a predictor of future returns. If the market price is consistent with market fundamentals, the share price should equal market fundamentals. In this case, the volatility of predicted cash flows (the market fundamentals price) cannot exceed the volatil ity of actual cash flows (the returns). Using data on dividends and prices, we can compare the historical volatility of the predicted cash flows to the actual cash flows. A constructed series represents the sum of discounted dividends from stocks listed in the Standard & Poor's 500 graphed against the price of the S&P 500 since 1871 (Figure 1). Clearly, stock prices are many times more volatile than the present value of discounted dividends. Given the relatively stable history of dividends over the last century, market fundamentals, constructed this way, clearly cannot account for the extreme volatility of asset prices. One interpretation is that stock prices are too volatile relative to observed changes in cash flows and that some factor unrelated to business conditions is responsible for the bulk of asset price fluctuations. However, there are some important caveats associated with interpreting these tests. First, there is no unique way to determine how investors discount future cash flows. The typi cal procedure carried out in these tests (and in Figure 1) is to assume that the discount factor (interest rate) is constant, which may not be true. Second, we cannot observe people's ex pectations of future dividends directly, so we must infer them. It is common to simply as sume that today's stock price is exactly equal to the future discounted sum of dividends. But http://fraser.stlouisfed.org/ 10 Federal Reserve Bank of St. Louis JANUARY/FEBRUARY 1996 this practice leads to difficulties in evaluating whether market fundamentals are consistent with price data. Instead, Robert Flood, Robert Hodrick, and Paul Kaplan, in a 1986 paper, suggested that apparent violations of variance bounds tests reflect errors in the model. That is, the test depends on the underlying eco nomic model being correct. Of course, this is a very strong assumption, and test results may simply reflect misspecification of the economic model. While there may be bubble compo nents to asset prices, this type of test will not likely resolve the debate. An alternative approach for testing whether variations in security prices are consistent with variations in market fundamentals is to deter mine whether the trend rate of growth in the asset price is similar to that in market funda mentals. Specifically, if market fundamentals are growing at a slower rate than the price of the corresponding asset, we may reasonably conclude that prices include a particular type of bubble component. This procedure can be used to detect the presence of bubbles that grow continuously over time. In 1985, Jam es H am ilton and Charles Whiteman, and in 1988, Behzad Diba and Herschel Grossman conducted tests along these lines. To determine whether market prices grow at a faster rate than market fundamen tals, we must evaluate the trends in the data. First, we test the data on annual stock prices and annual dividends to see if there are trends. If both series have trends, the series are "differenced." For example, to calculate the differenced data for market prices, subtract the price of the asset last year from its price this year. The differenced data for market prices and dividends are then tested for trends. If both of these differenced series have trends, the series are differenced again, and the trend tests are repeated . This process of su ccessiv ely differencing the data continues until the trans formed data do not have trends. If market FEDERAL RESERVE BANK OF PHILADELPHIA Lee E. Ohanian When the Bubble Bursts: Psychology or Fundamentals? FIGURE 1 Detrended Stock Prices and the Present Value of Detrended Dividends 1871 -1994 Index P is the real Standard & Poor's Composite Stock Price Index, detrended by a long-run exponential growth factor. P* is the discounted present value of real dividends, detrended by the long-run exponential growth factor. Real values are calculated by dividing nominal values by the wholesale price index. Source: Shiller, Market Volatility, Figure 5.1, updated by author. prices must be differenced more times than market fundamentals, we may reasonably con clude that a bubble is present in market prices. This analysis for dividends and stock price data, which appears in Figure 2, offers evi dence that both prices and dividends have trends, but when differenced once, both do not. This implies that prices over this period have not grown consistently faster than divi dends and provides evidence against the no tion that stock prices have included a growing bubble component. Although the analysis presented here was conducted with data only from the stock mar ket, these same tests can be used to evaluate data from the bond and foreign exchange mar kets. Briefly, the nature of these data are quite similar to data from the stock market. Like stocks, the variability of bond prices and ex change rates seems to be high relative to marll JANUARY/FEBRUARY 1996 BUSINESS REVIEW FIGURE 2 Stock Prices and Dividends 1871 -1994 1871 1881 1891 1901 1911 1921 1931 1941 1951 1961 1971 1981 1991 "Stock prices" is a logarithmic index of the real Standard & Poor's Composite Stock Price Index. "Dividends" is a logarithmic index of the real dividends on the real Standard & Poor's Composite Stock Price Index. Real values are calculated by dividing nominal values by the wholesale price index. Source: Author's calculations from data in Standard & Poor's Security Price Index Record. ket fundamentals. Moreover, there don't ap pear to be any differences in the trend behavior of market fundamentals and prices for either bonds or foreign exchange. CONCLUSION The extreme volatility of security prices has been a source of considerable interest since financial assets have traded in organized mar kets. It is important to distinguish between http://fraser.stlouisfed.org/ 12 Federal Reserve Bank of St. Louis market fundamentals and bubbles when ana lyzing the volatility of any security. If there are dramatic changes in fundamental economic factors, we would expect to see highly volatile security prices. If the volatility of security prices is considerably greater than the volatility of underlying business conditions, or if asset prices tend to grow much faster than the asset's associated cash flows, price movements may reflect a bubble component. FEDERAL RESERVE BANK OF PHILADELPHIA When the Bubble Bursts: Psychology or Fundamentals? The episodes of Dutch tulipmania and the British South Sea bubble provide dramatic examples of how prices may have deviated from fundamental values. Anecdotal evidence from recent periods provides no clear answer to the question of whether price movements may be due to bubbles. A number of statistical procedures have been developed to investi gate these questions directly, and these tests have been applied to stock market data. Unfor tunately, these tests often rely on assumptions that make interpretation of results very diffi cult. Test results that show differences be tween security prices and market fundamen tals may be due to bubble components, but Lee E. Ohanian they may also reflect errors in the model for market fundamentals. That is, a researcher may find evidence in favor of bubbles, but this may simply be due to the fact that the model for market fundamentals is wrong. Since market fundamentals are generically unobservable, it will always be difficult, if not impossible, to analyze data on asset prices and determine whether price movements can be entirely reconciled with movements in market fundamentals. We are left with the interesting observation that there are historical variations in asset prices that, at least, do not appear to be consistent with variations in underlying busi ness conditions. References Diba, Behzad T., and Herschel I. Grossman. "Explosive Rational Bubbles in Stock Prices?" American Economic Review, June 1988, pp. 520-30. Flood, Robert P., Robert J. Hodrick, and Paul Kaplan. "An Evaluation of Recent Evidence on Stock Market Bubbles," National Bureau of Economic Research Working Paper 1971, Cambridge, MA, 1986. Garber, Peter M. "Tulipmania," Journal o f Political Economy, 97, June 1989, pp. 535-60. Hamilton, James D., and Charles H. Whiteman. "The Observable Implications of Self-Fulfilling Expec tations," Journal o f Monetary Economics, 16, November 1985, pp. 353-73. Ibbotson, Roger G., Jody L. Sindelar, and Jay R. Ritter. Corporate Finance, Summer 1988, pp. 37-45. "Initial Public Offerings," Journal o f Applied Kindleberger, Charles P. Manias, Panics, and Crashes: A History o f Financial Crises. New York: Basic Books, Inc., 1978. Mackay, Charles. Memoirs o f Extraordinary Popular Delusions and the Madness o f Crowds. London: Bentley, 1841. Malkiel, Burton G. A Random Walk Down Wall Street: Updated fo r the 1990s Investor, 5th ed. New York: Norton, 1991. Shiller, Robert J. Market Volatility. Cambridge, MA: MIT Press, 1989. Thaler, Richard H. The Winner's Curse: Paradoxes and Anomalies o f Economic Life. New York: The Free Press, 1992. 13 The Federal Reserve Bank of Philadelphia Conference on Expectations in Economics: In Honor of the 50th Anniversary of the Livingston Survey October 3 & 4 ,1 9 9 6 Philadelphia, Pennsylvania The Research Division of the Federal Reserve Bank of Philadelphia is sponsoring a conference on Expectations in Economics: In Honor of the 50th Anniversary of the Livingston Survey to be held at the Philadelphia Fed on October 3 and 4,1996. The purpose of the conference is to bring together practitioners and researchers who are engaged in both theoretical and empirical work on expecta tions, including research using data from surveys of expectations. CALL FOR PAPERS If you are interested in presenting your research at the conference, please send a completed paper or detailed abstract by April 1 to the address below. Please note that we will not publish the papers presented at the conference; we will, however, publish a summary of the conference in our Business Review. We will pay modest honoraria and provide travel expenses for paper presenters. 14 Send papers or abstracts by April 1 to: Dean Croushore Research Department Federal Reserve Bank of Philadelphia Ten Independence Mali Philadelphia, PA 19106-1574 e-mail address: croushor@frbphil.org FEDERAL RESERVE BANK OF PHILADELPHIA The Cyclical Volatility of Interest Rates Keith Sill* T JL he variability of short-term and long term interest rates is a prominent feature of the economy. Interest rates change in response to a variety of economic events, such as changes in Fed policy, crises in domestic and interna tional financial markets, and changes in the prospects for long-term economic growth and inflation. However, economic events such as these tend to be irregular. There is a more regular variability of interest rates associated *Keith Sill is an economist in the Research Department of the Philadelphia Fed. with the business cycle, the expansions and contractions that the economy experiences over time. For example, short-term interest rates rise in expansions and fall in recessions. Long term interest rates do not appear to co-vary much with the level of economic output. The term cyclical volatility o f interest rates refers to the variability of interest rates over periods that correspond to the length of the typical business cycle. In this article, we will examine some facts and theory about the cycli cal volatility of short-term and long-term in terest rates. Why should we care about interest rate volatility? How do short-term and long 15 BUSINESS REVIEW term interest rates behave over the business cycle? What determines the cyclical volatility of interest rates associated with different maturities of government bonds? These ques tions are important to ask and answer as we seek a fuller understanding of the dynamics of the business cycle in market economies. WHY DOES INTEREST RATE VOLATILITY MATTER? The variability of interest rates affects deci sions about how to save and invest. Investors differ in their willingness to hold risky assets such as stocks and bonds. When the returns to holding stocks and bonds are highly volatile, investors who rely on these assets to provide for their consumption face a relatively large chance of having low consumption at any given time. For example, before retirement, people receive a steady stream of income that helps to buffer the changes in wealth associated with changes in the returns on their investment portfolios. This steady return from working helps them maintain a relatively steady level of consumption. After retirement, people no longer have the steady stream of income from working (though it will, in part, be replaced by pension income and Social Security), so a less volatile investment portfolio is called for. The lower volatility of investment returns allows retirees to maintain a relatively even level of consumption over time. Young investors, who are saving for retirement, are better able to absorb the risks of holding assets with highly volatile prices and returns. They can weight their portfolio more heavily toward risky stocks and bonds because they are receiving a steady return from working. For holding these riskier assets, the young investor will be rewarded with a higher average return on the invest ment. Just as individuals care about managing risk in their investment portfolios, so do firms. To manage risk, firms must pay attention to interest rate volatility and the composition of 16 JANUARY/FEBRUARY 1996 their portfolios. Many business firms hold portfolios containing large numbers of assets and, thus, are interested in quantifying the risk of losing large sums of money. As risks in the economy change, the expected gains and losses from the investment portfolio change. Mea suring this risk involves knowing how volatile prices of and returns on assets are, as well as how the returns on different assets change together over time. The volatility of interest rates is likely to be an important component in quantifying risk and guiding the investment decisions of these institutions. Interest rate volatility also has implications for how the prices of certain types of assets are determined. Options are assets that give inves tors the right, but not the obligation, to buy (call options) or sell (put options) other assets (such as stocks or bonds) at a prespecified price at or before some prespecified time in the future. For options purchased on interest-bear ing securities, modern finance theory demon strates that the option price depends on the volatility of returns on the underlying asset. The volatility of interest rates is related to the volatility of returns on these assets. Thus, interest rates and their volatility have important implications for how both individu als and firms make investment decisions. These investment decisions are part of the process w hereby resou rces are allocated in the economy. To begin, we'll briefly discuss how bond prices, interest rates, and maturities of bonds are related and how interest rates can be determined from bond prices. INTEREST RATES, BOND PRICES, AND THE TERM STRUCTURE There is a very close connection between bond prices and interest rates. We will focus on interest rates calculated from prices of traded U.S. government securities and show how the interest rate on a particularly simple type of security can be derived solely from its price. We focus on yields derived from U.S. governFEDERAL RESERVE BANK OF PHILADELPHIA Keith Sill The Cyclical Volatility of Interest Rates merit securities because these assets are backed by the full faith and credit of the government and, therefore, have virtually no default risk. The U.S. government issues securities of many different maturities: the maturity is the length of time until the final payment on the security is made by the issuer. Treasury bonds are fixed -cou p on secu rities w ith in itial maturities of more than 10 years. Treasury notes are fixed-coupon securities with initial maturities of from two to 10 years. Treasury bills are securities that are sold at a discount from face value and have initial maturities of a year or less. If we know a bond's current price and the payments that the bondholder will receive over the course of the bond's life, we can calculate the implied interest rate on the bond. This interest rate, called yield-to-maturity, equates the current price of the bond to the present value of the bond's payment stream.1* The relationship between the maturity of bonds and the interest rates implied by bond prices is called the term structure of interest rates. A plot of the relationship between interest rates and bond maturity, called the yield curve, can take a variety of shapes (Figure 1). Typically, interest rates on short-term bonds are lower than interest rates on long-term bonds, in which case the yield curve is upward sloping, as shown in the figure for the fourth quarter 1987. But sometimes the yield curve inverts, in which case interest rates on short-term bonds are 1For more detail on how yield-to-maturity is calcu lated, see my article in the July/A ugust 1994 Business Review. FIGURE 1 Term Structure of Interest Rates Interest Rate 17 BUSINESS REVIEW higher than interest rates on long-term bonds, as shown in the figure for the third quarter 1989. The shape of the typical yield curve shows that interest rates often vary with maturity. We might also suspect that the volatility of interest rates varies with maturity. But before we turn to how volatility is measured and how v olatility is related to m atu rity , le t's clarify the relationship between interest rates and the price of a particularly simple type of bond. Interest Rates and Bond Prices. Interest rates on certain types of bonds can be derived solely from the bonds' price and maturity. Let's look at a particular type of bond called a discount, or zero-coupon, bond. A discount bond sells at a discount from its face value and makes no interest payments over its lifetime. When the bond matures, the bondholder re ceives the bond's face value. For example, a one-year Treasury bill with a face value of $10,000 is a discount bond that promises to pay the holder $10,000 in one year's time. Such a bond may sell for a current price of $9434, in which case the implied interest rate on the bond is 6 percent (($10,000-$9434)/$9434 = .06). Clearly, as the current price of the bond changes, the implied interest rate will change. For example, suppose the current price of the bond falls to $9009. Then the implied interest rate on the bond is 11 percent (($10,000-$9009) / $9009 = .11). So, as the price of the bond falls, the interest rate rises; as the price rises, the interest rate falls.2 The U.S. Treasury does not issue discount bonds with maturities greater than one year. 2There is a simple relationship between the interest rate on a discount bond and the price of the bond. Suppose the price today of a bond that pays off $100 in five years is $75. The five-year interest rate on the bond is 33.3 percent [(100-75)/75 = .333]. The average annual interest rate on the bond is 5.9 percent since [$75 x (1.059)5 = $100], Gen eralizing this idea, if the interest rate is r on a bond paying $1 in j years, the current price of the bond is $1 / (l+r)>. 18 JANUARY/FEBRUARY 1996 However, financial market participants create pure discount bonds from long-term, coupon paying Treasury bonds by "stripping" the cou pon (semiannual interest) payments from the principal payment and selling the components as separate discount securities. In February 1985, the Treasury announced the STRIPS (sepa rate trading of registered interest and princi pal o f securities) p ro g ram , which facilitated the "stripping" of long-term Treasury bonds. Under the STRIPS program, all newly issued Treasury bonds and notes with maturities of 10 years or longer are eligible for stripping. The prices of these pure discount bonds can be found in publications such as the Wall Street Journal. Since there is a clearly defined relationship between interest rates and prices for discount bonds, we need to refer to only one of these elements, not both. When we consider dis count bond prices, we can easily derive the implied interest rates. Similarly, when we talk about the volatility of discount bond prices, we will easily be able to make inferences about the volatility of interest rates. Trends and Cycles in Interest Rates. We can plot the interest rate on discount bonds with a 10-year maturity from 1959 to 1990 (Figure 2).3 Notice that, overall, the interest rate tended to rise from 1959 to the early 1980s, after which it generally declined. From Figure 2 we can discern two types of variability in interest rates and hence in dis count bond prices: long term and short term. Long-term variability refers to broad trends in interest rates, such as the upward trend until the early 1980s and the downward trend since then. Short-term variability refers to how in terest rates vary around these long-term swings. Since our focus is on the business-cycle volatil ity of interest rates, we would like to remove 3The data plotted in Figure 2 are yields on discount bonds from the dataset compiled by McCulloch and Kwon. FEDERAL RESERVE BANK OF PHILADELPHIA Keith Sill The Cyclical Volatility of Interest Rates FIGURE 2 Ten-Year Interest Rate and Long-Term Trend 59 61 63 65 67 69 71 73 that part of interest rate volatility associated with swings of longer duration than the typical business cycle. The National Bureau of Economic Research defines minor cycles as recurrent fluctuations lasting from two to four years and major cycles as recurrent fluctuations lasting about eight years. Figure 2 clearly shows long-term trends in interest rates that are of greater duration than typical business-cycle lengths. In Figure 2 we've plotted a long-run trend that is fitted to the interest rate data. This long-term trend is chosen in such a way that it removes the swings in interest rates associated with peri ods longer than about eight years.4* The re maining short-run variability then corresponds more closely to variability that is part of the business-cycle movement in interest rates. We will define the difference between the actual interest rate and the long-run trend as the 75 77 79 81 83 85 87 89 cyclical component of the interest rate (Figure 3). The long swings in interest rates have been taken out, and all the variability in interest rates is around zero because this figure plots deviations from the long-term trend. When the interest rate is zero in Figure 3, we are on the long-term trend line in Figure 2. Henceforth, when we refer to interest rate volatility, we will be referring to the variability of this shortrun component. MEASURING THE VOLATILITY OF INTEREST RATES We will measure interest rate volatility us ing a statistic called the standard deviation. 4The fitted trend is calculated using the HodrickPrescott filter. More details on how this filter works are provided in the paper by Robert King and Sergio Rebelo. 19 JANUARY/FEBRUARY 1996 BUSINESS REVIEW FIGURE 3 Cyclical Component of Ten-Year Interest Rate Percent 3.0 2.0 1.0 0.0 - 1.0 - 2.0 59 61 63 65 67 69 71 The standard deviation measures how dis persed a variable is around its average value. If the standard deviation is high, observations on a variable tend to be far away from the variable's average value. If the standard de viation is low, observations on the variable tend to be clustered around the average value. Therefore, as the standard deviation increases, there is a greater chance that we will see large changes in the value of the variable.5 The volatility of interest rates can be calcu5The standard deviation is calculated as the square root of the variance of a variable. Suppose we have n observa tions on a variable {X1,X2,...Xn}. Denote the average value of X by X. The variance of X is then the average of the squared deviations of X from its mean: V(x) = [ (X,-X)2 + (X2-X)2 + ... (Xn-X)2]/n . The standard deviation of X is the positive square root of V(x). 20 73 75 77 79 81 83 85 87 89 lated over the entire term structure of interest rates: we simply use historical data to calculate the standard deviation of interest rates for each maturity. Table 1 presents the relation ship between interest rate volatility and the maturity of bonds as well as the standard deviation of the associated prices for discount bonds.6 The table shows that short-term inter est rates are more volatile than long-term in terest rates and that long-term discount bond 6More specifically, we calculate the standard deviation of detrended yields and logarithms of bond prices using quarterly data over the period 1959:Q1 to 1990:Q1. We use the logarithm of the discount bond price because it is proportional to the yield-to-maturity of the bond, with the factor of proportionality equal to the maturity of the bond. The interest rates and bond prices are detrended using the Hodrick-Prescott filter. FEDERAL RESERVE BANK OF PHILADELPHIA Keith Sill The Cyclical Volatility of Interest Rates prices are more volatile than short-term dis count bond prices.7 In describing the cyclical volatility of inter est rates we would like to know not just how much interest rates vary but also how they vary with the state of the economy. During recessions, real output is declining; during expansions, it's rising. We can get an idea of the behavior of interest rates over the business cycle by evaluating how interest rates and the level of real output co-vary over the business cycle. The correlation coefficient is a measure of the strength of the co-variation between two variables, and it can take on values between 7There is a direct relationship between bond-price vola tility and the volatility of the interest rate on the bond. Using the relationship between bond prices and interest rates in footnote 2, it can be shown that the standard deviation of the interest rate on a j-period bond is approxi mately equal to the standard deviation of the logarithm of the j-period bond price divided by j. minus one and one. When the correlation coef ficient between two variables is positive and close to one, the two variables track each other closely and move in the same direction: when one variable is high, the other variable is very likely to be high. If the correlation coefficient is negative and close to one, the two variables track each other closely but move in opposite directions: when one variable is high, the other is likely to be low. When the correlation coef ficient is zero, the two variables do not track each other closely in either direction. The cyclical component of short-term inter est rates has a positive contemporaneous cor relation with the cyclical component of real output. So when current output falls, short term interest rates tend to fall, and when cur rent real output rises, short-term interest rates tend to rise (Table 2). The strength of this correlation between output and interest rates tends to decline as the maturity of the bonds increases. By the time we get to bonds with 10- TABLE 1 Interest Rate and Bond-Price Volatility 1959:Q1 - 1990:Q1 Time to Maturity Standard Deviation of Detrended Interest Rate Standard Deviation of Detrended Discount Bond Price 1 quarter .0032 .0032 2 quarters .0031 3 quarters .0030 .0063 .0092 1 year .0029 .0117 2 years .0025 .0205 5 years .0021 .0426 10 years .0018 .0735 Standard deviations are calculated from the term structure data in McCulloch and Kwon (1993). Standard deviation is of the logarithm of discount bond prices (see footnote 5). 21 BUSINESS REVIEW JANUARY/FEBRUARY 1996 TABLE 2 Correlations of Interest Rates and Real GDP 1959:Q1 - 1990:Q1 Correlation of Detrended Interest Rate in Period t with Detrended Output in Period: Time to Maturity t-1 t t+1 1 quarter .42 .35 .11 2 quarters .41 .34 .09 3 quarters .40 .32 .08 1 year .37 .29 .05 2 years .27 .20 -.02 5 years .11 .04 -.15 10 years .02 -.05 -.21 Output is measured as the logarithm of real GDP. Interest rate data are from the term structure data in McCulloch and Kwon (1993). year maturity, the contemporaneous correla tion is negative, though quite small. This im plies that there is little co-variation between the cyclical movements in current real output and the cyclical movements in long-term inter est rates. These facts can be expressed by saying that short-term interest rates are procyclical and long-term interest rates are acyclical. The results in Table 2 suggest signifi cant business-cycle variability in short-term interest rates but relatively little business-cycle variability in long-term interest rates. The last column of Table 2 shows that the correlation between current interest rates and real output one quarter into the future is posi tive for short-term and negative for long-term interest rates. This fact suggests that upward movements in short-term interest rates are associated with upward movements in nearterm output, but that higher long-term interest rates forecast lower near-term output. The 22 first column of Table 2 shows the correlation between current interest rates and the level of real output one quarter in the past. These correlations suggest that increases in current output are associated with increases in future interest rates. We can also make some deductions about the shape of the yield curve over the business cycle using the data in Table 2. We have seen that short-term interest rates tend to move up when output moves up but that the correlation tends to decline as the maturity of the bond increases. Thus, when current output rises, the yield curve tends to flatten, since short-term interest rates tend to rise and long-term inter est rates move relatively little. Similarly, when current output declines, the yield curve tends to steepen, since short-term interest rates tend to fall with output and long-term interest rates tend to remain about the same. We have seen how the volatility of interest FEDERAL RESERVE BANK OF PHILADELPHIA Keith Sill The Cyclical Volatility of Interest Rates ity, and the deviations of the interest rate on 10-year bonds from the trend line have been large and persistent. In fact, the results in Table 4 show that interest rates at all maturities may have been more variable since that time. The table shows the standard deviation of interest rates using the same data, but the sample is divided into two subsamples: from first quarter 1959 to first quarter 1979 and from second quarter 1979 to first quarter 1990. We see that interest rates at all maturities have been more volatile since 1979. This result suggests the possibility that some structural change in the economy has affected the variability of interest rates and bond prices.8 rates changes with maturity and how interest rates move in relation to real output over the business cycle. But how are interest rates on bonds of different maturities related to each other? In general, interest rates on bonds of different maturities are highly correlated with each other, with the highest correlations occur ring between bonds of similar maturities (Table 3). Let's take the case of the interest rate on a security with one-quarter maturity. We see that the one-quarter interest rate is most highly correlated with the interest rate on a bond with two-quarter maturity, and that the correlation declines, though remains strong, as we com pare bonds w ith in creasin g ly d ifferen t maturities. These correlations suggest a ten dency for the entire yield curve to shift up and down, while allowing for the possibility that the shape of the yield curve can change. Finally, if we re-examine Figure 3, we might suspect that the measured volatility of interest rates depends on the period we're looking at. Since the late 1970s, long-term interest rates appear to have shown more short-run variabil 8For short-term interest rates in particular, higher vola tility after 1979 may reflect a change in the way that the Federal Reserve implements monetary policy. After late 1979, and especially between late 1979 and late 1982, monetary policy placed less emphasis on smoothing short term interest rates. Thus, after 1979, short-term interest rates were more likely to reflect changes in the state of the economy. TABLE 3 Cross Correlations of Detrended Interest Rates 1959:Q1 - 1990:Q1 Maturity 1 quarter 1 qtr 2 qtr 3 qtr 1 year 2 years 5 years 10 years 1.0 2 quarters .99 3 quarters .97 .99 1 year .96 .99 .99 2 years .90 .94 .96 .98 5 years .79 .84 .87 .90 .96 10 years .69 .75 .78 .81 .90 1.0 1.0 1.0 1.0 1.0 .98 1.0 Interest rate data are from McCulloch and Kwon (1993). 23 JANUARY/FEBRUARY 1996 BUSINESS REVIEW TABLE 4 Interest Rate and Bond-Price Volatility Time to Maturity Standard Deviation of Detrended Interest Rate Standard Deviation of Detrended Discount Bond Price 59:1-79:1 79:2-90:1 59:1-79:1 79:2-90:1 1 quarter .0026 .0040 .0026 .0040 2 quarters .0026 .0040 .0051 .0079 3 quarters .0025 .0039 .0075 .0116 1 year .0024 .0037 .0094 .0148 2 years .0019 .0034 .0151 .0275 5 years .0014 .0030 .0278 .0608 10 years .0011 .0027 .0442 .1080 Standard deviations are calculated from the term structure data in McCulloch and Kwon (1993). Logarithms of discount bond prices are taken before the standard deviation is calculated (see footnote 5). WHAT DETERMINES INTEREST RATE VOLATILITY? The postwar data imply that prices of long term discount bonds are more variable than those of short-term discount bonds and that long-term interest rates, measured by yield-tomaturity, are less volatile than short-term rates. In addition, we find that short-term interest rates are procyclical, while long-term interest rates vary little with current output. What economic factors influence interest rate vari ability? If we can isolate some economic deter minants of the levels of interest rates and bond prices, we will be well on our way to finding determinants of this variability. Determination of Short-Term Interest Rates. A standard economic model will help us think about how the interest rate on short term discount bonds is determined. Let's con sider the case of a discount bond that will pay off $100 with certainty in one year. Suppose a prospective bond buyer expects her real in http://fraser.stlouisfed.org/ 24 Federal Reserve Bank of St. Louis come over the coming year to be higher than usual (real income refers to income adjusted for any change in the general level of prices over time). In that case, she has less of an incentive to increase her savings by purchas ing a bond today. In fact, she may well decide to borrow against some of her expected in crease in income. If all prospective bond pur chasers expect higher real income over the coming year, demand for current one-year bonds will fall, and their prices will fall as well, which means that the one-year interest rate will rise. On the other hand, investors may decide to hedge against the risk of lower future income by purchasing bonds today that pro vide a guaranteed future payoff. If current real output (and thus aggregate real income) is low, investors may expect fu ture output to be low, because there is some persistence to output movements. Hence, a downward movement in current output is consistent with a downward movement in FEDERAL RESERVE BANK OF PHILADELPHIA The Cyclical Volatility of Interest Rates current short-term interest rates if people ex pect output and income in the near future to be low as well. This theory is consistent with procyclical movement in short-term interest rates. The yield curve tends to flatten when out put is high and tends to steepen when output is low. Suppose we are currently in a boom, but people expect a recession in one year. Inves tors may buy one-year bonds to hedge the risk of low future income, and they may pay for these bonds, in part, by cashing in their shorterterm assets. This portfolio reallocation tends to lower one-year interest rates and raise shorter-term interest rates, thus leading to a flatter yield curve. Empirical studies have found that the shape of the yield curve does help predict recessions and expansions.9 Expected inflation is also a determinant of interest rates. Consider again the case of a discount bond that pays $100 with certainty in one year. Suppose now that prospective bond purchasers expect inflation to rise over the coming year. When inflation rises, the current price of one-year bonds will fall because inves tors realize that their dollars buy less when prices rise. For example, if the price of a cup of coffee one year from now is $1, bondholders can buy 100 cups of coffee with the $100 that the bond pays off. But if the price of a cup of coffee is expected to rise to $1.05, bondholders will be able to buy only 95 cups of coffee. To be compensated for the loss in purchasing power, investors must get a higher dollar return on their investments. Thus, bond prices will fall and interest rates will rise when expected in flation rises. This model suggests that when expected income or expected inflation rises, bond prices will fall. This fall in bond prices translates into higher interest rates. So, when we think about how short-term interest rates are determined, 9See the article by Campbell R. Harvey. Keith Sill we want to think about people's forecasts for real income growth and inflation. Any current economic variables that help to predict real income growth and inflation will help to deter mine current short-term bond prices and inter est rates. D eterm ination of Long-Term Interest Rates. Long-term interest rates can be linked to short-term interest rates by the expectations theory of the term structure. This theory says that long-term interest rates are equal to an average of expected short-term interest rates plus a risk premium.10 The risk premium ac counts for the co-variation over time of vari ables like income growth and inflation that could influence the level of interest rates. The logic of the expectations theory of bond prices is most clearly seen in an example in which we ignore the risk premium. Take the case of an investor who has a two-year invest ment horizon. The investor can purchase a two-year bond, or he can purchase a one-year bond today and, when that bond matures, purchase another one-year bond. The expected return on these alternative investment strate gies should be equal. Since there is a direct relationship between interest rates on bonds and bond prices, the expectations theory also links long-term discount bond prices to ex pected short-term discount bond prices over the life of the long-term bond. In terms of expected future short-term bond prices, the same variables that affect short term bond prices basically determine long term bond prices and interest rates. Thus, expected future income growth and expected inflation are also determinants of long-term bond prices, but now the forecasts of income growth and inflation are for further in the future. It is still the case that if, over the life of 10For more detail on the expectations theory and risk premiums, see my article in the Ju ly / August 1994 Business Review. 25 BUSINESS REVIEW the bond, expected future income growth or expected future inflation rises, long-term in terest rates will rise. Including a risk premium does not alter these basic conclusions about the determinants of interest rates. However, the risk premium can be an additional source of variability for interest rates because it picks up some indirect effects of income growth and inflation on interest rates, as well as other risk factors. This model helps us think about why long term interest rates co-vary less with current output than do short-term interest rates. Cur rent movements in real output are much more closely correlated with output movements in the near future than they are with output movements in the far future. Since the pay ment stream on a long-term bond extends further out into the future than that on a short term bond, long-term interest rates are less likely to have a strong co-variation with cur rent output movements. Determinants of Interest Rate Volatility. The same basic economic factors that deter mine interest rates and the prices of bonds also determine the volatilities of interest rates and bond prices. This economic model suggests that expected real income growth and expected inflation determine bond prices and interest rates. It follows then that the volatility of ex pected real income growth and the volatility of expected inflation, as well as the correlation between the two, determine the volatility of interest rates and bond prices. The reasoning behind this conclusion is straightforward. Take the case of real income growth. We saw above that if real income growth is expected to be high, current bond prices will fall and interest rates will rise. The higher real income growth is expected to be, the higher interest rates will be. Thus, large changes in expected real income growth are associated with large changes in interest rates. When real income growth has high volatility, large changes in real income growth occur http://fraser.stlouisfed.org/ 26 Federal Reserve Bank of St. Louis JANUARY/FEBRUARY 1996 more frequently, and hence large changes in current bond prices and interest rates occur more frequently. When large changes in inter est rates occur more often, interest rates are more volatile. Similar reasoning holds for the case of inflation. When large changes in ex pected inflation occur, large changes in cur rent bond prices and interest rates occur also. So, more volatile inflation translates into more volatile bond prices and interest rates.11 W hat determ ines how volatile income growth and inflation will be? One factor is monetary policy. Take the case of monetary policy and inflation. Economists generally be lieve that a persistent inflation has its root causes in monetary policy, in particular, how fast the money supply grows relative to real income growth. If growth of the money supply is excessive, inflation is likely to be high. If we take growth of the money supply as the pri mary determinant of inflation, highly volatile growth in the money supply can lead to vola tile inflation. This does not mean that every change in the money supply necessarily leads to a change in inflation. Rather, if, on average, money supply growth becomes more volatile, inflation can become more volatile as well. As we have seen, the model then suggests that bond prices and interest rates will also be more volatile. Monetary policy could also have an effect on real income, although economists disagree on the mechanism by which this occurs. One theory is that workers write contracts with their employers that fix a nominal wage rate over the contract period. Workers and firms 11Higher volatility of income growth and inflation sug gests that price volatilities for both short-term and long term bonds will increase. Long-term volatility remains higher than short-term volatility because investors who buy long-term bonds have to make forecasts about future variables that are not relevant for determining the prices of short-term bonds. FEDERAL RESERVE BANK OF PHILADELPHIA The Cyclical Volatility of Interest Rates negotiate the contracted wage based, in part, on their expectations of what inflation will be over the contract period. Since monetary policy affects inflation, this requires workers and firms to forecast what monetary policy will be over this same period. If monetary policy and the price level turn out to be different from what workers and firms expected when they wrote the contract, employment and output could be affected because firms' demand for workers depends on the real wage rate that must be paid. If nominal wages are fixed by a contract and prices rise unexpectedly, real wages fall, and firms demand more workers and produce more output. If prices fall unex pectedly, real wages rise, firms lay off work ers, and output falls. Thus, variability of the money supply, through its impact on prices, could have an impact on the variability of real income. We can point to many other factors, besides monetary policy, as potentially influencing the volatility of output and inflation. For example, variability in weather can affect agricultural output as well as production in the economy. Changes in productivity due to the introduc tion of new technologies can influence the variability of output and inflation as well. A whole class of economic models, called realbusiness-cycle models, attempts to account for output volatility over the business cycle. These models assume that shocks to productivity are the main cause of business cycles.12 Shocks to current productivity affect peoples' forecasts of the future course of the economy and thereby affect their expectations about economic vari ables like real income growth and inflation. The more persistent productivity shocks are, the greater their effect on long-term interest rates will be, since output and inflation far into the future will be affected. 12See the article by Satyajit Chatterjee in the September/O ctober 1995 Business Review. Keith Sill Economic Models and Interest Rate Vola tility. This economic model for determining bond prices and interest rates suggests that investors' expectations of future real income growth and inflation are the primary determi nants of current bond prices and interest rates. There are, of course, other determinants of interest rates and interest rate volatility in the economy. But we can try to assess how well this view of interest rate determination ex plains the interest rate volatility that we ob serve in the actual economy. One approach to assessing how well a model performs is to use the model to simulate inter est rates and then compare the properties of the simulated interest rates to the properties of actual interest rates. For example, we can set up models and use them to simulate price data on discount bonds of various maturities. We can then calculate the standard deviation of these simulated data and compare it to the standard deviation of discount bond prices implied from the interest rates we observe in the economy. We can also examine how the simulated bond prices and interest rates co vary with simulated output and compare the correlations to the correlations we find in the actual data. In this way, we can assess the ability of the model to account for the cyclical volatility of interest rates. SIMULATION RESULTS In my 1994 working paper, I present an exercise similar to the one following. Briefly, in the model, expected real income growth and expected money growth determine current discount bond prices and yields. Expected money growth is assumed to be the primary determinant of inflation. The model also re quires some input on investor characteristics, such as how willing investors are to undertake risky investments. Table 5 shows the variabil ity of the bond prices and yields simulated by one particular version of the model and repro duces the variability of bond prices and yields 27 JANUARY/FEBRUARY 1996 BUSINESS REVIEW TABLE 5 Yield and Bond-Price Volatility From Model Simulations Time to Maturity Standard Deviation of Detrended Interest Rate Standard Deviation of Detrended Discount Bond Price Simulated Actual Simulated Actual 1 quarter .00362 .00322 .00362 .00322 2 quarters .00332 .00315 .00663 .00629 3 quarters .00303 .00308 .00909 .00923 1 year .00278 .00293 .01112 .01172 2 years .00202 .00256 .01619 .02051 5 years .00100 .00213 .02005 .04265 10 years .00051 .00184 .02045 .07348 Standard deviations are of the yields and logarithms of discount bond prices (see footnote 5). Actual discount bond prices are calculated from the term structure data in McCulloch and Kwon (1993). derived from actual interest rate data for com parison.13 The model generates data in which volatil ity of interest rates falls but bond price volatil ity rises with the maturity of the bond. Out to a maturity of about one year, the variability of the simulated bond yields and prices matches the variability of the data fairly closely. At a m aturity of three m onths, the m odel overpredicts the volatility of bond prices and yields about 12 percent. At a maturity of one year, the model underpredicts the volatility of bond prices and yields about 5 percent. These results suggest that the variability of income 13The model also replicates some of the correlations between discount bond prices and output as well as some features of the correlation patterns of bond prices. 28 growth and money growth account for a sub stantial portion of the variability of short-term discount bond prices and hence of short-term interest rates. For longer maturities, the variability of simu lated bond prices and yields underpredicts the volatility of actual yields and implied prices of discount bonds by a progressively larger amount. When we look at the historical data, the variability of implied prices for a discount bond with 10-year maturity is about 23 times larger than the variability of short-term dis count bond prices. But in the simulated data, the variability of 10-year discount bond prices is only about five times greater than the vari ability of short-term bond prices. Many reasons might explain why the growth of the money supply and the growth of real income do not account for much of the vari FEDERAL RESERVE BANK OF PHILADELPHIA Keith Sill The Cyclical Volatility of Interest Rates ability of long-term bond prices and yields. The basic model is designed to highlight the business-cycle variability of interest rates, and as we have seen, long-term interest rates do not appear to have a large business-cycle com ponent. In addition, the model is very simple, and so it is missing some important elements found in actual economies. For example, the model does not account for the fact that differ ent people have different beliefs about the future course of the economy or that people are continually learning about the economic envi ronment. Changes in fiscal and monetary poli cies may induce greater volatility in interest rates than the simple economic model accounts for. The expectations theory may be an inad equate model of the term structure of interest rates. Despite difficulties such as this, the model's implication that real income growth and money supply growth are factors that help to determine the volatilities of interest rates and discount bond prices does find some sup port, especially for shorter maturities, when we compare the model with actual data. CONCLUSION We have seen that the volatility of interest rates depends on the maturity of the underly ing bond: long-term interest rates are less vari able than short-term interest rates. Short-term interest rates are procyclical while long-term interest rates co-vary little with movements in output over the business cycle. Economic theory suggests that both the level and volatil ity of interest rates should be tied to economic variables such as income growth and inflation. Simulation results suggest that the volatility of both income growth and money growth ac counts for a large portion of the volatility of short-term discount bond prices. However, these same economic variables by themselves are able to account for only a small fraction of the volatility of long-term discount bond prices. References Chatterjee, Satyajit. "Productivity and the American Business Cycle," Federal Reserve Bank of Philadelphia Business Review (September/October 1995). Harvey, Campbell R. "Term Structure Forecasts Economic Growth," Financial Analysts Journal (M ay/June 1993). King, Robert G., and Sergio T. Rebelo. "Low Frequency Filtering and Real Business Cycles," Journal o f Economic Dynamics and Control, 17 (1993), pp. 207-31. McCulloch, J. Huston, and H.C. Kwon. "U.S. Term Structure Data, 1957-1991," Working Paper 93- 6, Ohio State University (1993). Sill, Keith. "Money, Output, and the Cyclical Volatility of the Term Structure," Working Paper 94- 14, Federal Reserve Bank of Philadelphia (July 1994). Sill, Keith. "Managing the Public Debt," Federal Reserve Bank of Philadelphia Business Review (July/August 1994). 29 Philadelphia / RESEARCH Working Papers The Philadelphia Fed's Research Department occasionally publishes working papers based on the current research of staff economists. These papers, dealing with virtually all areas within economics and finance, are intended for the professional researcher. The papers added to the Working Papers series thus far this year are listed below. To order copies, please send the number of the item desired, along with your address, to WORKING PAPERS, Department of Research, Federal Reserve Bank of Philadelphia, 10 Independence Mall, Philadelphia, PA 19106. For overseas airmail requests only, a $3.00 per copy prepayment is required; please make checks or money orders payable (in U.S. funds) to the Federal Reserve Bank of Philadelphia. A list of all available papers may be ordered from the same address. 95-1 Satyajit Chatterjee and Dean Corbae, "Valuation Equilibria with Transactions Costs" 95-2/R Sherrill Shaffer, "Structural Screens in Stochastic Markets" (supersedes Working Paper No. 92-23) 95-3 Franklin Allen and Douglas Gale, "A Welfare Comparison of Intermediaries and Financial Markets in Germany and the U.S." 95-4 Franklin Allen and Douglas Gale, "Financial M arkets, Interm ediaries, and Intertemporal Smoothing" 95-5 Gregory P. Plopper, "The Dynamics of the Exchange Rate Under a Crawling Peg Regime: A Game Theory Approach" 95-6 Franklin Allen and Douglas Gale, "Universal Banking, Intertemporal Risk Smooth ing, and European Financial Integration" 95-7 Paul Calem and Michael Stutzer, "The Simple Analytics of Observed Discrimination in Credit Markets" 95-8 Joseph Hughes, William Lang, Loretta Mester, and Choon-Geol Moon, "Recovering Technologies That Account for Generalized Managerial Preferences: An Application to Non-Risk-Neutral Banks" 95-9 Ana Castaeda, Javier Diaz-Gimenez, and Jose-Victor Rios-Rull, "Unemployment Spells and Income Distribution Dynamics" 95-10 Paul Calem and Loretta J. Mester, "Consumer Behavior and the Stickiness of Credit Card Interest Rates" (Supersedes No. 92-24/R) 95-11 Richard Voith, "Parking, Transit, and Employment in a CBD" 30 FEDERAL RESERVE BANK OF PHILADELPHIA 95-12 Gary Gorton and Richard Rosen, "Banks and Derivatives" 95-13 Sherrill Shaffer, "Translog Bias Under Declining Average Costs" 95-14 Alberto Trejos and Randall Wright, "Toward a Theory of International Currency: A Step Further" 95-15 Gerald Carlino and Robert DeFina, "The Differential Effects of Monetary Policy Shocks on Regional Economic Activity" 95-16 Paul Calem, "Mortgage Credit Availability in Low- and Moderate-Income Minority Neighborhoods: Are Information Externalities Critical?" 95-17 Leonard I. Nakamura, "New Directions in Information and Screening in Real Estate Finance" 95-18 William J. Stull, "Is High School Economically Relevant for Noncollege Youth?" 95-19 James McAndrews and George Wasilyew, "Simulations of Failure in a Payment System" 95-20 Keith Sill, "An Empirical Investigation of Money Demand: Evidence from a Cash-InAdvance Model" 95-21 Leonard Nakamura, "Is U.S. Economic Performance Really That Bad?" 95-22 Laurence Ball and Dean Croushore, "Expectations and the Effects of Monetary Policy" 95-23 Sherril Shaffer, "The Discount Window 95-24 Stephen Morris and Hyun Song Shin, "Informational Events That Trigger Currency Attacks" 95-25 Robert H. DeFina, Thomas C. Stark, and Herbert E. Taylor, "The Long-Run Variance of Output and Inflation Under Alternative Monetary Policy Rules" 95-26 Bernardino Adao and Theodosios Temzelides, "Beliefs, Competition, and Bank Runs" 95-27 Theodosios Temzelides, "Evolution, Coordination, and Banking Panics" and Credit Availability" 31 FEDERAL RESERVE BANK OF PHILADELPHIA Business Review Ten Independence Mall, Philadelphia, PA 19106-1574 Address Correction Requested