The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
R FEDERAL E RESERVE CLEVELAND V RANK OF E I W E C O N O M I C 1 9 8 6 2 Disinflation, Equity Valuation, and Investor Rationality. Some observers have suggested that the Great Bull Market of the 1980s is due solely to lower inflation. The effect of inflation on stock prices, however, is not quite so direct. Economists Jerome S. Fons and W illiam P. Osterberg examine theories of the relation between inflation and equity prices in light of recent experience. R E V I E W Q U A R T E R Economic Review n An exhaustible resource model, com bined with risks stemming from large government stocks, explains rising gold prices from 1968 to 1981. This paper posits that demand misestimation accounts for the extended decline in gold prices after 1981. is published quar terly by the Research Department of the Federal Reserve Bank of Cleve land. Copies of the issues listed here are available through our Public Information Department, 216/579-2047. Editor: William G. Murmann. Design: Michael Galka. Typesetting: Liz Hanna. Opinions stated in The Collapse in Gold Prices: A New Perspective. 4 Economic Review are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System . Material m ay be reprinted provided that the source is credited. Please send copies of reprinted materials to the editor. "1 ™7 “Don’t Panic”: A Primer J. / on Airline Deregulation. The effects of airline deregulation are current ly a much-debated topic. Recent empirical work on this issue is summarized and collated here for the layman. The conclusion is that deregulation has led to substantial aggregate benefits to society, but that the industry has yet to fully adapt to its new environment. IS S N 0013-0281 Disinflation, Equity Valuation, and Investor Rationality by Jerome S. Fons and William P. Osterberg Jerom e S. Fons and William P. Osterberg are economists at the Federal Reserve Bank of Cleveland. Tbe authors would like to thank Jam es Balazsy for research assist ance, and William T. Gavin, K .J . Kowalewski, Richard Kopcke and Richard Cohn for their helpful comments. Introduction Until the early 1960s, economists largely ignored the effect of inflation on the prices of corporate equities. Since revenues and costs were thought to be proportionately affected by changes in the price level, profits would expand so as to keep pace with inflation. As residual claims to the earnings of corporations, equities were seen as partial, if not complete, hedges against the effects of inflation. In the late 1960s and early 1970s, this notion was shattered. Despite a 95.2 percent rise in the consumer price index (CPI) from the end of 1966 to the end of 1977, the Standard & Poor’s Stock Index rose only 2.4 percent. The 52.5 percent decline in the real value of equities over this period led to the development of many the ories to explain the relationship between equity prices and inflation. Among the most widely received theories was one offered by Franco Modigliani and Richard Cohn (1979). They claimed that investors make valuation errors by ignoring the gains debt ors experience from inflation and therefore use the wrong measure of profits in pricing equities. Since inflation implies that the principal of the loan will be paid back in “cheaper” dollars, lend ers require an inflation premium in the coupon on the loan. This suggests that a part of the firm’s debt service is used to maintain the real value of the firm’s debt and should not be treated as an expense. Traditional accounting measures, how ever, treat the entire debt service as an expense. Modigliani and Cohn claimed that the measure of “true profits” consistent with rational valuation would equal accounting profits, plus the portion of the interest expense attributable to inflation. They reasoned that a more serious investor error involves the comparison of the dis count rate for a pure equity stream with nominal, rather than real, interest rates. In figure 1 we pre sent a time-series plot of the nominal interest rate on Aaa-rated corporate bonds and the earnings/ price ratio of stocks in the Standard & Poor’s Stock Index. At least since I960, these two series track one another well. Because long-term nom i nal interest rates are thought to be largely deter mined by inflation expectations, this comparison by investors further erodes the level of stock prices in an inflationary environment. Modigliani and Cohn showed that, in the absence of market imperfections, the real value of the firm should remain unaffected by inflation. Using a statistical model, they found that investors had indeed committed one or both forms of valuation error. In this paper, we review the model introduced by Modigliani and Cohn and the alternative analyses of other investigators. We then evaluate those analyses by examining the behavior of the rate of return required on equi ties from 1953 to 1985. Surprisingly, we find little evidence of valuation errors. In particular, we note that when reported earnings are adjusted in the manner prescribed by Modigliani and Cohn, capitalization rates for equities appear to follow real interest rates, though they may also respond to factors related to aggregate risk. anticipated I. A Fundamental Valuation Model Fundamental equity valuation models assume that the goal of the firm’s management is to max imize stockholders’ wealth. Projects are accepted only if they increase the market value of the equity, that is, if the present discounted values of the expected net cash flows from new projects are positive. The market value of the firm’s equity is found by discounting the cash flows distributed to stockholders at the rate stockholders could earn on alternative investment flows of equivalent risk.1 The distribution to stockholders, or divi dend, equals profits (revenue, less operating expenses and investment expenditures) minus interest payments on the firm’s debt. Following Modigliani and Miller (1958), we make assumptions sufficient to derive an expression for the value of the firm’s equity: a) capital markets are frictionless, that is, partici pants can borrow or lend at the riskless rate of interest and there are no taxes; b) the social costs of bankruptcy are zero; c) all firms are in the same risk class; and d) equity and default-free debt are the only types of claims on firms. EARNINGS/PRICE RATIO FOR THE S&P STOCK INDEX ANO MOODY’S Aaa-CORPORATE RATE V “(t) (1) = X/p Note that if the adjusted profits of the firm are expected to grow continually at a rate g, the firm’s value can be represented as: V “(t) = X/(p-g) (2) Given the above assumptions, Modigliani and Miller go on to show that a firm’s value is inde pendent of its capital structure. That is, rational investors will ignore the effects of the firm’s bor rowing and base their valuation on the firm’s cash flow from operations. The levered firm’s total market value, is defined as the sum of the market values of equity, and debt, Vl(t), Vl( t) = S(t) (3) S(t), + D(t) : D (t) The adjusted profits available for distribution to the stockholders of a levered firm differ from an unlevered firm’s adjusted profits at date by the firm’s interest expense, The expected rate of return to the levered firm’s stockholders, is simply: t, X(t), rD(t). i, i = [X(t) - rD (t)\/S(t) (4) Combining equations (1) and (3), Modigliani and Miller’s Proposition 1 states that: Xft) (5) = p[Vl(t)J = p[S(t) + D(t)\ Substituting (5) into (4) and allowing for earn ings growth at rate gives: g (6) / = d p +( p-r)d - g, D(t)/S(t), where = is the firm’s debt-equity ratio. The value of the equity of a levered firm can then be found by discounting the income stream available to stockholders at the appro priate rate (given by equation [6]). That is: FI GURE 1 The value of an unlevered (all equity) firm at date with expected adjusted profits is found by discounting the firm’s expected available net cash flow at the rate that is appropriate for the firm’s risk class (p ).2 Viewing the firm as an ongoing concern with a perpetual income stream its value is given by: X S(t) = [X(t) - rD(t)]/[p + (p-r)d - g] (7) S O U R C E : Moody's Investors Service; and Standard & Poor’s Corp. t, Vu(t), X, Following Modigliani and Cohn, suppose that at time 0 there is no inflation and that imme diately thereafter fully anticipated inflation begins at the rate p and continues forever. Adjusted prof- t- 2 ‘Adjusted profits’ refer to after-tax reported profits adjusted for the effects of inflation on inventory valuation and the value of actual depreciation deductions. In the N IP A these adjustments are referred to as ‘IV A ’ and ‘C Cadj,’ respectively. They are based on corporate tax records and assumptions about asset lives and replacement costs. For a discussion of the N IP A adjustment, see Grimm (1982). A problem with I This should be distinguished from the so-called book value of applying this adjustment to the S & P reported earnings index is that the equity, found by subtracting the book value of liabilities from the N IP A profits measure is based on “book" profits which vary somewhat book value of assets. from reported earnings, especially after 1981. its will rise continuously at the rate of inflation so t, X (0 )epl. X(t) that at any date the unlevered firm’s profits, w ill equal From equation (1), the value of the unlevered firm at date equals In other words, the value of the unlevered firm will not be affected by fully antici pated inflation. Rationally priced equity claims on such a firm are complete inflation hedges. t, Vu(t), real Vu(0)ept. Conventional accounting measures of a levered firm’s profits are distorted by infla tion. Accounting profits equal operating income, minus nominal debt expense. Assume that the nominal interest rate ( ) is approximately equal to the sum of the real interest rate ( r) and the expected inflation rate and that the firm’s debt remains fixed in real terms equals D[0] Also assume that the firm’s debt is structured so that it always pays the current rate of interest. The levered firm’s accounting profits, II, can then be written as: R (p) (D[t] epf).} n (0 = (8) 4 = [ X (0 - X(t) - RD (t ) (r+p)D(t)} = [X(0) - rZ)(0)] ex-ante, ex-post ept - pD (0)ept The firm’s accounting profits have been expressed in this form to illustrate the following essential points. 1) The portion of reported inter est expense attributable to inflation, should be added back to accounting profits to yield “true” profits. 2) At high enough inflation rates, accounting profits may become negative.4 True profits, II*, w ill therefore in crease at the fully anticipated inflation rate. That is: pD (0)ep!, (9) Equation (13) shows that, although the real value of a firm’s equity should be unaffected by infla tion, accounting earnings must be adjusted for inflation’s effect. Modigliani and Cohn hypothesized that investors failed to incorporate inflation in their valuations of equities. They tested this hypothesis by regressing a measure of stock prices on variables that enter either the numerator or the denominator on the right-hand side of expression (13). Their estimate of the coefficient on inflation im plied systematic misvaluation. In our attempts to replicate the results of Modigliani and Cohn, however, we found that the results were sensitive to assumptions regarding lag dis tributions used to construct proxies for or expected, values of key variables. In addition, our attempts to replicate the results of Modigliani and Cohn yielded a coefficient on inflation that differed from their estimate (see Appendix). Rather than update their empirical work, we take a different approach to evaluating the perfor mance of Modigliani and Cohn’s model.5 n*(0 = n(0 +pD (o)ept We utilize observable, observations on each of the relevant variables to simulate the model, calculating im plied values for p, the required real rate of return of a pure equity stream. To the extent that our measures of reflect expectations, our estimate of p is an required rate of return on a pure equity stream. Consequently, p is analogous to a real interest rate, adjusted for the risk in equity and the fact that the security is a perpetuity. By focusing on the time-series values of p, im plied by the model rather than the predicted equity values, we avoid much of the con troversy surrounding equity valuation having to g ante ex- = [X(0) - rD(0)]e** Substituting the levered firm’s true profit stream n*, into equation (7), we have: (10) 5 (0 = n*(0/[p + 3 If, as finance theory suggests, investors are concerned with after tax real rates of return, then one could replace fhr+p with fl*=fl(1-r)=r+p, where r is the marginal tax rate on interest income. (p r)d - g] Clearly, fixing r implies that the change in R* due to a change in p is not 1 for 1. This relates to Hendershott's (19 8 1) argument discussed below. real Equation (10) therefore indicates that the value of a firm’s equity is unaffected by inflation. Now substituting for accounting profits and rear ranging, equation (10) becomes: (11) [11(f) + p D (t)\/S (t) = P +( p-r)d - g 4 A n additional factor that is thought to offset the inflation-induced gain from debt service, pD(0)ep[ is the possible increase in the firm’s pension obligations. This argument requires that inflation be unan ticipated and is relevant only for defined-benefit pension plans (currently comprising roughly 75 percent of all pension assets). A defined-benefit pension is one in which contributions are determined by the benefits they will eventually yield. The obligation of the firm to restore under funded pensions, however, rests in part on the nature of the firm’s con This expression reduces to: tract with labor. Feldstein and Morck (1983) find that the stock market (1 2 ) n (r)/S (r) = p + (p-r)d - pd - g, appears to react favorably to firms with overfunded pensions and nega tively to underfunded pensions. They note, however, that most large, well-managed firms have traditionally had overfunded pensions. or, (13) S (t ) = II (t)/[p +( p-r)d - pd - g] A n empirical update of Modigliani and Cohn is presented by Townsend (1986). do with the appropriate form of the discount rate.6 The advantage of this approach is that we are able to see how p varies over time and, in partic ular, if it is correlated with inflation or real inter est rates. This does not contradict the assumption that at any point in time, all variables in the denominator of (13) are expected to remain con stant forever. W hile Modigliani and Cohn assumed that p is not affected by inflation, the theories discussed below allow p to be related to many factors, including the rate of inflation. In order to isolate p we can re write equation (13) as: (14) P = n (t) s(t) r+p)d + g 1 +d +( Using the definition of the nominal interest rate, K, we have: n(Q (15) p = S(t) Rd + g 1 +d + Equation (15) shows the relation between the required real rate of return on a pure equity stream in a given risk class and mostly observable variables. The only unobservable variable is the expected growth rate of reported profits. The var iable p may be viewed as a modified earnings/ price ratio, adjusted for inflation, leverage, and earnings growth. II. The Determinants of p Below we discuss three theories of the determi nation of the cost of a pure equity stream. Two of the explanations given for the behavior of p focus on a risk premium, while the third considers the relation between p and the real rate of return on bonds. In trying to explain the behavior of the stock market in the mid-1970s, Burton Malkiel (1979) adjusted corporate profits for inflation’s effect on corporate debt and found them to be steady in low- and high-inflation periods. He argued that the decline in real stock prices was caused by an increase in the risk premium embodied in the rate of return required by stockholders. The increased risk premium was due to economic developments of the early 1970s that led to a departure from the relative stability of the 1960s. He reasoned that investors thought policymakers could no longer “fine tune away” economic fluctuations and that long-run planning involved greater uncertainty. Although profits rose with the price level, their dispersion across industries also rose, in turn raising busi ness risk. The rising use of debt financing was another source of increased risk for the financial system. Finally, rising government regulation may have been perceived as reducing profitability. As evidence supporting the per ception of increased risk, Malkiel cites the rise in the “risk spread” between anticipated returns on equities and long-term government bonds, as well as between the yields on Baa-rated corporate bonds and government bonds. These widening spreads throughout the 1970s may suggest that investors believed the credit quality of firms was falling. According to Malkiel’s findings, we would expect to see a path for p that starts out low in the ‘50s and ‘60s and then turns higher in the mid-to-late 1970s. A related theory of the behavior of p involves the possibility of a disinflationary dis tress premium: real required rates of return on pure equity streams rise in a climate of disinfla tion. Firms may be under greater strain in a disin flationary environment as they are often unable to match declines in revenue with declines in expenses.7 This is particularly evident following a period of prolonged high inflation. Extreme examples of the upheaval associated with disin flation can be found in the oil and steel indus tries. Further, corporate defaults have generally been higher in disinflationary periods than in inflationary periods.8 This hypothesis implies that stockholders will require a premium whenever there are large reductions in inflation in order to compensate them for the increased credit risk. By this hypothesis, p should fall with increases in inflation and rise with disinflation. Hendershott (1981) attributes the valuation error noted by Modigliani and Cohn solely to investors’ comparisons of the expected real yield on equities, p, with the nominal yield on bonds. He claims that Modigliani and Cohn’s 7 This m ay be due to the existence of fixed labor and supply con tracts. A simple model introduced by Wadhwani (1986), on the other hand, suggests that the inflation premium in a levered firm’s debt service causes nominal debt expense to increase proportionately more than nominal revenue during inflation, forcing the firm to report lower accounting profits. Conversely, this expense will decrease more than proportionately during disinflation, resulting in higher reported, or account ing, profits. 6 The emphasis on p is also justified by the implications of work done by Shiller (1981) and others on the volatility of dividends and stock prices. The literature on stock volatility suggests that profits have much lower variances than stock prices. Thus, variation in p and 8 Fons (1986) investigates the correlation between "unanticipated” changes in the consumer price index and a measure of expected corporate default rates embodied in yield spreads. Though not statisti other factors influencing the rate at which profits are discounted could cally significant, the relationship between inflation surprises and an be expected to account for much of the variation in stock prices. implied default premium on low-rated corporate debt is negative. model implies that the after tax real bond yield falls as a result of inflation, while nominal yields remain constant. Since bonds and equities are substitute assets, the fall in the after-tax real yield on debt would lower the rate of return required by stockholders. The decline in the required yield on equity offsets the overpayment of taxes resulting from the inflation-induced understate ment of depreciation and inventory costs (see discussion ofFeldstein [1980] below), or increased risk premia noted by Malkiel, leaving the nominal value of stocks essentially unchanged. Hendershott claimed that there were other factors responsible for the decline in the real value of equities. First, there was a decline in savings due to lower real after-tax yields. Second, there was a decrease in the productivity of new capital due to higher regulatory costs and higher energy prices. In addition, Hendershott felt that an increase in the realized rates of return on non corporate assets, such as residential housing, may have induced investors to reduce their holdings of debt and equity. By Hendershott’s reasoning, p should decline in inflationary periods and rise with disinflation. Declines in productivity, how ever, would be reflected in a lower expected growth rate of earnings (g). 6 & Poor’s Index. The price index, based on as many as 500 different equities mostly traded on the New York Stock Exchange, is constructed in such a way that, when divided into the associated earnings index, the unwanted weighting factor can cels. The eamings-per-share index is constructed from the reported earnings over the past four quar ters of the firms in the corresponding stock index. We adjust for inflation-caused inventory valuation and depreciation errors by multiplying the earn ings index by the ratio of adjusted-to-reported after tax profits found in the National Income and Product Accounts (NIPA) (see footnote 2). The interest rate on corporate bor rowings is measured as Moody’s cross-sectional average yield on single A-rated bonds. This rating corresponded to the average quality rating (in terms of par value) of all publicly traded corpo rate debt as of December 1985. As was previously discussed, the nominal interest rate embodies inflation expectations. In using this measure, we avoid the problems encountered by Modigliani and Cohn in constructing an econometric proxy for expected inflation. The debt-equity ratio for nonfinancial corporations, was constructed from two sources. Data covering 1953 to 1961 was taken from Von Furstenberg (1977), in which the market value of debt is inferred from a present value relation. The 1961 to 1985 series for the market values of corporate debt and equity were constructed by the Board of Governors of the Federal Reserve System. In this case, the market value of debt is found by pricing all mortgages and long-term bonds at the average price of bonds traded on the New York Stock Exchange, ignoring such nontraded items as deferred taxes, leases, and pension obligations. An attempt was made in the estimation of the market value of equity (the listed values on all exchanges, times the number of corresponding shares outstanding) to avoid the double counting of firm ownership through stock holdings. The computation of p involves assumptions about the process generating the parameter One extreme is to let assume its realized value equal to the annualized growth rate of four-quarter reported earnings for each period. The volatile behavior of and p when gis measured this way can be seen in figure 2. We feel that such erratic movement in is unreason able since, in theory, is the expected perpetual growth rate of earnings. Presumably this pre cludes from being negative. An alternative way to measure is to utilize a time-series model to construct an insample one-period-ahead forecast of earnings growth. We modeled the quarterly growth of four-quarter earnings as following an ARMA(1,1) process. Using the forecast for at each date in d, g. 1950 1955 1960 1965 1970 1975 1980 1985 S O U R C E : Figures 2 , 3 . and 4 author's calculations. FI GURE 2 III. Data & Methodology. Quarterly observations for the period covering 1953 through 1985 on each of the following data series were used to construct estimates of p: ad justed earnings per share, stock prices, nominal corporate interest rates, aggregate debt-equity ratios, and earnings growth. The values for stock prices and earnings were taken from the Standard g g g g g g g the calculation of p yields the time-series plot of p presented in figure 3 This series is only slightly less volatile than the series constructed from actual growth rates. COMPUTED p SERIES WITH ARMA(1,1) EARNING GROWTH A time-series plot of p constructed with fixed at its average value is presented in figure 4. The required real rate of return ranged between 10 and 13 percent from 1952 through 1974, with moderate deviation. At the start of 1975, however, p began to rise slowly and then sharply in 1981. It peaked at the end of 1981 and again at the beginning of 1984. For comparison’s sake, setting equal to zero over the entire sam ple period produces values for p ranging between 4.5 and 7.5 percent from 1952 through 1976, topping out at 14.2 percent in mid-1984. g g FI GURE 3 A third alternative is to fix the growth rate of earnings at its average value over the entire sample period, 6.4176 percent.9 This procedure may be justified on the grounds that investors somehow possess perfect foresight of earnings growth and that they ignore short-run fluctuations. The infinite-horizon nature of the estimated model requires an unbiased estimate of perpetual earnings growth. It is possible, with the S&P data, to construct an estimate based on earnings growth as far back as 1926. The inclu sion of a persistent recession and a major war, however, would likely result in a less satisfactory estimate of expected earnings. COMPUTED p SERIES WITH FIXED EARNINGS GROWTH IV. Analysis of Computed p Series In this section, we analyze the behavior of p, computed with expected earnings growth fixed at its actual mean value. Our goal is to shed light on this component of equity valuation. By their nature, however, it is not possible to completely separate the implications of the various hypo theses discussed above. The computed value of p appears to support Malkiel’s hypothesis that p begins to rise in the mid-1970s due to the risk factors cited earlier. In addition, the rapid rise in 1981 could be explained by Bodie, Kane, and McDonald (1986), who concluded that there was a dramatic increase in the risk premium required in long term bonds in the early 1980s. They attribute this to the switch in operating procedures by the Federal Reserve in late 1979The disinflation hypothesis pre sented earlier suggests that p should vary inverse ly with the level of inflation. In figure 5, we pre sent plots of p and the rate of inflation. Note that the major upturns in p appear to coincide with the inflationary peaks occurring in 1974 and again in 1981. Smaller, previous inflation spikes do not, however, seem to be accompanied by any signifi cant movement in p. The same figure can be used to examine Hendershott’s claims. Conspicuously absent is the hypothesized decline in p as infla tion rises. The lack of noticeable downward movement in p during rising inflation eliminates much of the support for his arguments. His main conclusion, however, that p is tied to the real rate of return on debt, can now be addressed. 9 The average annual growth rate of adjusted earnings over the sample period w as 17.0 1 percent. The growth rate of this series since mid-1983 has been so great as to completely dominate this figure. It w as felt that the effects of this adjustment could not have been rea sonably foreseen over much of the sample period and, in fact, should ''wash1' over the long run. W e therefore chose to use the average annual growth rate of unadjusted earnings in the computation of p. COMPUTED p WITH FIXED EARNINGS GROWTH AND INFLATION 0.20 ___________________________________ 0 02 ---------- 1 ---------- '---------- 1--------- 1 ---------- 1--------- '---------- u 1950 1955 1960 1965 1970 1975 1980 1985 S O U R C E : Author's calculations; and U .S . Department of Commerce, Bureau of Labor Statistics. FI GURE 5 REAL Aaa-CORPORATE BOND YIELD S O U R C E : Moody's Investors Service. ex-post A plot of real long-term cor porate bond rates is shown in figure 6. This figure was constructed by simply subtracting 1 from the ratio of the gross yield on Aaa-rated corporate bonds to the previous year’s gross inflation rate at each date. Note that real required rates of return on fixed income securities reached unprece dented levels in 1981, the same year in which p significantly departs from its postwar behavior. Hendershott’s hypothesis, therefore, appears to explain the sharp rise in p that occurred in 1981. However, it does not shed light on the moderate increase beginning in 1975, but it does help explain the slight decline in p that occurs between the end of 1971 and the end of 1974. Though separate from the riskrelated hypotheses, Benjamin Friedman (1986) claimed that an increase in the government deficit, such as that beginning in early 1981, would drive down the realized rate of return on equity relative to either short- or long-term debt, thereby increasing the required rate of return on a pure equity stream. This theory then suggests that the rise in p is a function of deficits, thus explaining the sharp rise in 1981. Had we found no rational explana tion for the behavior of p, we would have searched for evidence of measurement errors related to corporate earnings. For instance, Feldstein (1980) claimed that biases in the tax system, rather than inflation-induced valuation errors, could explain the poor performance of the stock market. In particular, Feldstein emphasized that corporate capital depreciation deductions are based on historical, rather than current, costs. In inflationary periods, with a rising price of invest ment goods, this implies that the real value of depreciation deductions declines. This, in turn, implies that taxable profits (net of depreciation deductions) rise, causing real after-tax profits to fall. Feldstein also pointed out that nominal rather than real capital gains are subject to capital gains taxes. This implies that even if the nominal value of equities increased at the inflation rate, the real after tax yield on equities would decline. In contrast to Modigliani and Cohn, Feldstein viewed the stock market decline as a rational response to inflation. Modigliani and Cohn, in response to the criticism of Feldstein, discussed the possi bility of tax biases due to inflation. They noted that other analyses of the interaction of inflation and taxes have ignored the fact that firms are not taxed on the portion of returns used to depre ciate debt. They argue that this offsets the decline in real after tax profits that results from the decline in real depreciation deductions. They support this by noting that the share of corporate income paid as taxes has remained relatively con stant in inflationary periods. In their empirical work, as well as in our construction of p, an adjustment factor constructed from the National Income and Product Accounts was used that attempts to correct reported earnings for depreci ation and inventory distortions caused by infla tion. The NIPA adjustment, however, may mis state the lagged response of tax shelters to inflation. In addition, the analysis is complicated further by the fact that much corporate debt is fixed-rate and thus debt yields do not adjust instantly to inflation expectations. In figure 7 we present both unad justed and adjusted reported four-quarter earn ings per share using the NIPA data. For the early part of the sample period the two series are virtu ally identical. They begin to diverge at the end of ADJUSTED AND UNADJUSTED EARNINGS PER SHARE S O U R C E : Standard & Poor’ s Corp.; and U .S . Department of Commerce. 1972, with adjusted earnings falling somewhat below unadjusted earnings. The situation reverses dramatically, however, in 1983- At this point, ad justed earnings clim b far above unadjusted earn ings. Further study may shed light on the sensitiv ity of our results to the adjustment of earnings, especially for the period following 1981. V. Conclusion We conclude that equity prices respond rationally to such factors as real interest rates and risk. When we use the model of Modigliani and Cohn to compute the discount factor applied to a pure equity stream of a levered firm, we find no evi dence of inflation-induced valuation errors. The evidence presented, however, is consistent with the hypothesis that disinflation influences the risk premium applied to pure equity streams. FI GURE 7 REFERENCES Bodie, Zvi, Alex Kane and Robert McDonald, “Risk and Required Returns on Debt and Equity,” in Benjamin M. Friedman, ed., Chi cago: University of Chicago Press, 1986. Financing Corporate Capital Formation, Feldstein, Martin, and Randall Morck. “Pension Funding Decisions, Interest Rate Assumptions, and Share Prices,” in Zvi Bodie and John Shoven, eds., Chicago: University of Chicago Press, 1983. Financial Aspects of the United States Pension System. Feldstein, Martin. “Inflation and the Stock Market,” vol. 70, no. 5 (December 1980), pp. 839-47. American Economic Review, Fons, Jerome S. “The Default Premium and Cor porate Bond Experience,” Federal Reserve Bank of Cleveland, June 1986. Working Paper #8604, Friedman, Benjamin M. “Implications of Government Deficits for Interest Rates, Equity Returns, and Corporate Financing,” in Ben jamin M. Friedman, ed., Chicago: University of Chi cago Press, 1986. Capital Formation. Financing Corporate Grimm, Bruce T. “Domestic Nonfinancial Corpo rate Profits,” U.S. Department of Commerce, Bureau of Eco nomic Analysis, vol. 62, no. 1 (January 1982) pp. 30-42. Survey>of Current Business, Hendershott, Patric H. “The Decline in Aggregate Share Values: Taxation, Valuation Errors, Risk, and Profitability,” vol. 71, no. 5 (December 1981), pp. 909-22. American Economic Review, Malkiel, Burton G., “The Capital Formation Prob lem in the United vol. 34, no. 2 (May 1979), pp. 291-306. States," Journal of Finance, Modigliani, Franco, and Richard Cohn, “Inflation, Rational Valuation and the Market,” vol. 3, no. 2 (March-April 1979), pp. 24-44. Analysts Journal, Financial _________ and Merton Miller. “The Cost of Capi tal, Corporation Finance, and the Theory of Investment,” vol. 1958, no. 3 (June 1958), pp. 261-97. American Economic Review, Shiller, Robert J. “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?,” vol. 71, no. 3 ( J u n e 1981), pp. 421-36. American Economic Review, Townsend, Henry. “Another Look at the Modigliani-Cohn Equation,” vol. 42, no. 5 (SeptemberOctober 1986), pp. 63-66. Financial Ana lystsJournal, Von Furstenberg, George M. “Corporate Invest ment: Does Market Valuation Matter in the Aggregate?,” No. 2: 1977, pp. 347-408. Activity, Brookings Papers on Economic Wadhwani, Sushil B. “Inflation, Bankruptcy, Default Premia and the Stock Market,” vol. 96, no. 381 (March 1986), pp. 120-38. nomic Journal, Eco 9 APPENDIX Reestimation of Modigliani and Cohn’s Model In this section, we describe our attempts to repli cate the results of Modigliani and Cohn and then to reestimate their model, extending the sample period through 1984. Modigliani and Cohn estimated the following regression, which is im plied by expression (13), after taking the log of both sides: = a 0 = w l(L )U (t) + w 2(L)D IV(t ) a3 w3(L)[LF/E](t ) + a4D V F(t) - /1w 4(L)R (t ) + yw 5 (L )P (t) + u (t) s (t ) L + The variable is the lag operator and the parameters through represent coefficients on the lagged terms of the five forecasted vari ables. The distributed lag, embodies the assumption that expected, or profits equal a one-sided distributed lag of past profits. Profits were measured as described in the text and in Modigliani and Cohn. Although it is not unusual to view expected dividends as influenc ing stock prices, Modigliani and Cohn include a distributed lag of dividends, on the grounds that dividends provide information about future profits. They then restrict the coeffi cients of the distributed lag on dividends, so that a change in dividends has no permanent effect on firm value, given the history of profits. Divi dends were measured as dividends per share for the issues in the S&P 500, adjusted as described by Modigliani and Cohn. a distributed lag of the ratio of the labor force to employment, is included to provide a cyclical adjustment to the ability of past profits to predict future profits. The term is included as a measure of the risk premium entering the formulation of p. M odigli ani and Cohn measured as the 15-year moving-average deviation of the unemployment rate from 4 percent. We chose instead to use a 12-quaner moving-average. The distributed lag on the nominal interest rate, and the dis tributed lag on inflation, are included to measure the real rate, also a component w\ w5 wl(L)U(t), ex-ante, w2(L)DIV{ t), u5(L)LF/E, DVF(t) DVF u4(L)Rit), u^(L)P{t), tit), R( t) of p. is measured as the new issue yield on AA corporate bonds. is measured as the annual percent change in the CPIU. We used the current value and seven lagged values in each distributed lag. This choice of lag length differed from that of Modigliani and Cohn, but seemed only equally arbitrary. We maintained the following restrictions regarding the form of the distributed lags: a) the coefficients on profits sum to one, b) the coefficients on dividends sum to is quadratic, zero, c) the distributed lag on d ) the distributed lag on dividends is linear, e) the distributed lag on the nominal rate is quadratic, and f) the distributed lag on inflation is quadratic with the endpoints constrained to equal zero. The parameters to be estimated are «0, and the parameters in the distributed lags. The theoretical model of Modigliani and Cohn implies that the coefficient on the distributed lag of inflation, should equal where the debt-equity ratio and is the capitalization rate. Their estimate of -0.08, differs from a com puted value of 0.05. Thus, an increase in expected inflation reduced market values, although this should not have been the case if investors had been rational. In feet, Modigliani and Cohn calculated that a one percent increase in inflation would reduce the market value of equities by 13 percent. Thus, the market had been drastically undervalued due to inflationinduced valuation errors. When we attempted to replicate the results of Modigliani and Cohn, over the same sample period, we estimated to be .015. When the sample period was extended through 1984, how ever, the estimate of was -0.025. If the misvaluation of equities was being eliminated, the esti mate of over the longer period would have been closer to the theoretically predicted value than for the shorter period. Since our results not only differed from those of Modigliani and Cohn, but indicated worsening misvaluation, we chose to consider a different approach. P( t) LF/E a3, aA, Pi y, y, K y, d/K, y y y (d/K) d/K, d is The Collapse in Gold Prices: A New Perspective by Eric Kades “If all men were rational, all politicians honest and we had a world central bank issuing a single currency that was universally acceptable, then gold would drop to $20 an ounce— and be overvalued at that.” — Andre Sharon, gold analyst, quoted in Newsweek, Dec. 1 6 ,1 9 7 4 ; as quoted by George Seldes in Quotable Quotations. Eric Kades is an analyst in the Brian Gendreau of Morgan Guaran Financial Strategies Group at Gold tee Trust Com pany, Stephen Salant man, Sachs, & Co. Fie is a former of the Rand Corporation, and Mark analyst at the Federal Reserve Bank Sniderman of the Federal Reserve of Cleveland. The author thanks Bank of Cleveland for helpful com ments and corrections. Introduction The daily summaries and analyses of the gold market that appear in most newspapers support Mr. Sharon’s assertion. The press invariably attrib utes gold price movements to political uncer tainty, gyrating monetary policies, inflation hedg ing, and international liquidity concerns. This view implies that the demand for gold is highly volatile, subject to coups, sudden shifts in central bank behavior, oil flow interruptions, and other jolts to the world economy. GOLD PRICE AND INTEREST RATE TRENDS Real value of assets, in 1970 dollars If Mr. Sharon and the press were right, then economists would have little to con tribute to an analysis of even long-term move ments of gold prices, or to forecasts of price trends. These activities would be better left to political experts, to central bank analysts, and to other savvy observers in areas that are likely to generate surprises affecting gold prices. There would be no point in statistically estimating a demand function for gold, since demand for gold would be always be fluctuating randomly, not moving systematically. This conventional explanation of gold price movements is essentially a superficial one. W hile unexpected political and economic events undoubtedly influence daily gold prices, such events cannot explain long-run trends in gold prices. Before the Bretton Woods interna tional monetary system began to crumble in 1968, the price of gold was fixed at about $35 an ounce. The real price of gold, that is the nominal, or observed price divided by a price index, has followed two distinct trends since 1968 (see curve A in figure l) .1 From 1969 to 1981, the real price of gold rose rapidly, except for a few brief, but sharp, price dips, and for one extended slide. From 1981, until this year, the real price of gold fell — S O U R C E : London P .M . Gold Fixing; Consumer Price Index: U .S . Department of Labor. Bureau of Labor Statistics; and Board of Governors of the Federal Reserve System. FIGURE 1 1 The price index in this case is the C PIU. W e study the real price to correct for changes in the purchasing power of the dollar. 11 rapidly at first, then more gradually. This year has witnessed renewed strength in gold prices. As confusing as these long-run trends might seem, economists can offer an explanation that draws from previous studies and that is rooted in the unique characteristics of the gold market and of gold itself. In this study, we explain gold price movements in terms of a model for exhaustible resources, describing the qualities of gold that account for its spectacular price rise from 1968 to 1980. We show why m od els offered in previous studies must be modified to explain the in gold prices since 1981. The recent rise in gold prices can be interpreted as supporting the perspective we present. decline I. The Conventional Supply and Demand Analysis If viewed as any other good, gold price move ments would be analyzed in terms of conven tional supply and demand. The key factor in supply would be the marginal cost of extracting additional ore from the earth (or the cost of recy cling gold). These costs would follow predictable patterns and might be estimated effectively. More analytical difficulty could be expected to arise from unstable demand. W hile the demand for gold by industries and jewelers is stable, other buyers (inflation hedgers and liquidity seekers) have erratic gold-buying habits. This demand/supply view of the gold market turns on gold’s historical role as a defense against inflation and posits that a predic table relationship between inflation and real gold 12 price movements overwhelms any random influ ences from other types of gold buyers. Interna tional liquidity measures (money supplies and related stocks) and exchange rates also enter conventional gold-pricing models, but these may be thought of as leading and coincident indica tors of inflation among nations. But the contem poraneous correlation between inflation and gold prices is weak (see figure 2) and formal statistics show no strong relationship between gold prices and past inflation. There are more fundamental arbi trage arguments against a and reliable relationship between gold prices and inflation. If the real rate of return to gold were lower than riskless assets during deflationary times, no rational investor would include gold in his portfolio while prices fell. Since someone must hold existing stocks of gold, the rate of re turn (i.e., price changes over tim e) would be bid up to induce agents to hold existing stocks. Sim ilarly, if gold has an extraordinary rate of return during inflations (with little or no risk) then this return would be bid down by investors. Thus, gold cannot yield more/less than riskless government bonds during inflations/deflations. Such a regime would not be in equilibrium since people will want to increase/decrease gold hold ings and thereby shift the price of gold. Gold may well serve as an effective against inflation by purchasing power, but this factor cannot explain the long-term trends observed in the high real return to gold. systematic systematically systematically hedge preserving II. Gold: An Exhaustible Resource Gold is literally an resource. Thomas Wolfe, a former director of Treasury precious metals operations, noted, “ [g] old is among the few minerals that could reach a critical supply situation within this century, taking account of available reserves above and below ground.”2 World stocks have declined sharply, on net, since 1968 (see figure 3). Thus, gold’s very scarcity requires a unique method of determining its price. Economists do have a valid technique, which has existed since the 1930s, for analyzing prices of exhaustible resources.3 An ingenious modification exhaustible GOLD PRICES AND IN FLATIO N: A WEAK CORRELATION % Change, quarter-quarter, in gold prices % Change, quarter-quarter, in C P IU S O U R C E : London P .M . Gold Fixing; and U .S . Department of Labo r. Bureau of Labor Statistics. FIGURE 2 2 See Stephen Salant and Dale Henderson, “ Market Anticipations of Government Policies and the Price of Gold,” Economy, 3 Journal of Political vol. 86. no. 4 (August 1978), pp. 627-48. See Harold Hotelling, “The Economics of Exhaustible Resources," Journal of Political Economy, vol. 39. no. 2 (April 1931) pp. 137-75. SHRINKING WORLD GOLD STOCKS Gold mined and used, metric tons Net change, world gold stocks, metric tons rate. If the price rose faster, nobody would sell it, since it would be a superior investment over any time horizon. This price path could not constitute an equilibrium, however, since eventually the constantly rising price would cause all demand for manna to be choked off, and yet none would have been consumed. The good’s value would be too high for it to be liquidated. On the other hand, if the price of manna rose more slowly than agents would not want it in their portfolios; in the rush to sell it, the price would fall precipitously. Then the price would jump from a very low level in the initial period (as everyone rushed to sell) to infinity in the next period, because people would consume the manna, driving the supply toward zero. However, then an investor who bought at the low price in the first period and held on to the manna could charge a very high price in the second period, earning a rate of return above Only a path with prices increasing at the rate leaves people with no incentives to change their behavior and, thus, is the only equilibrium for pricing a simple exhaustible resource like manna. Curve B in figure 1 indicates such a price path. It shows how the value of a bond equal in price to an ounce of gold in 1965 would have appreciated at a constant real interest rate of 4 percent (arbitrarily chosen). It is easy to see that from 1965 to 1981 gold yielded a higher re turn. Curve C shows the real rate of return— the real value an investor would have achieved by purchasing U.S. Treasury bills and roll ing them over quarterly. Gold even more easily dominated this investment option over the period. rh, 68 70 72 74 76 78 80 82 rb. rb S O U R C E : Annual gold reports ( 1 9 7 8 - 8 5 ) , Consolidated Gold Fields, Ltd. FIGURE 3 of this technique, by Stephen W. Salant and Dale W. Henderson, explains the major gold price movements from 1968 to 1981, but does not fully explain the decline in gold prices from 19811986, which is the period that interests us here.4 We will use a simple allegory to explain how exhaustible resources should be priced. III. Pricing a Depletable Resource: A Simple Model A finite amount of “manna” is distributed among some lucky individuals. Once eaten, the manna is gone forever. What direction should manna prices take? Think of manna as an asset, just like stocks and bonds; we are concerned with only the in vestment characteristics of manna, not with the final consumers. Each asset in the economy has an annual rate of return. We expect the rate of return to be higher for more risky assets. Inves tors will demand a higher rate of return on risky “junk” bonds of an indebted corporation, for example, than they will require to hold the virtu ally riskless bonds and bills of the U.S. Treasury bearing a riskless interest rate With a fixed, known supply and stable demand, manna is a riskless asset. There fore it must yield the same rate of return as the riskless interest rate on government securities, The only way it can yield such a return (since it pays no dividends and earns no interest) is for its price to rise at a rate equal to the riskless interest rh. rb. 4 See “ Market Anticipations of Government Policies and the Price of Gold?" ex post IV. Gold: A Special Case No resource exactly duplicates the simple price path illustrated in curve B of figure 1. Unexpected events will change the supply of or demand for a good. These jolts, not readily addressed by eco nomic arguments, affect the price of a good ran domly, increasing the variance without changing the long-run path. Also, every commodity has its own peculiarities that alter its price trend systemati cally. Several such factors operate in the gold market and are often cited as important determi nants of price trends. The costs of m ining gold, for exam ple, rose between 1965 and 1981 and contributed to at least part of gold’s price increase. However, it can also be argued that mining costs could not be responsible for price increases in a market, like the gold market, where there are speculators. Rising extraction costs do not pre vent speculators from buying gold in one period and selling it in the next. Because speculators do 1 3 not pay extraction costs (and assuming storage costs are constant), competition among them will prevent the rate of gold price increases from ex ceeding the riskless rate of interest.5 The gold mar ket unquestionably includes many speculators. Another salient feature of the gold market is South Africa’s dominant, almost monopolistic role in gold production. Since the price of gold began to rise in 1968, the South African share of production has averaged near 75 percent, although it has fallen moderately in recent years. Such hegemony can raise prices above competitive levels, but, like rising produc tion costs, cannot account for observed rapid increases in gold prices. Any attempts by South Africa to raise prices faster than would create arbitrage opportunities that would force prices back down. Speculators would buy gold in one period and then, being w illing to accept a rate of return would undersell the South Africans in the next period. Salant and Henderson conclude that the only valid special factors in the gold market are the huge stocks governments hold and, particularly, the perceptions of speculators about what buying or selling actions governments will take. This, they argue, causes the price of gold to move systematically at variance with the simple exhaustible resource explanation. To see how this matters, think about the amount of gold available to satiate demand in a given period. Production levels w ill be relatively stable, because construction of large mines takes a long time. However, governments hold huge stocks of gold (now about 40 years’ worth of cur rent industrial, artistic, and jewelry demand; in 1970, governments held 25 percent more). If they decide to sell a significant amount of gold in a given period, the price will drop sharply. The “threat” of government sales means that since there is a chance that govern ment actions will have a severe impact on its price. Risky assets must give higher yields, on average, to compensate their owners. Comparing curve A (actual real gold price) with C (actual price trend for real return to three-month Trea sury bills), strikingly illustrates that gold did indeed command a return higher than the risk less interest rate from 1968 to 1981. There were a few exceptions, when the price dropped precipitously for short periods of time. These occasional price dips, however, fit precisely into the scenario that Salant and Henderson present. They are the announce- rh rb, 1 4 gold can no longer be considered a riskless asset, Gold that cost more to extract would not be mined until the price rose sufficiently to justify the expense. ment dates of government sales or news leaks of the likelihood of such sales. These events illus trate the riskiness of holding gold in the presence of government stocks that can depress prices temporarily. For example, arrow 1 in figure 1 marks the first announcement of possible Interna tional Monetary Fund (IMF) nation sales; arrow 2 shows the price decline caused by the first U.S. Treasury auction of gold since World War II. The price decline that lasted from 1975 to 1976 occurred while gold’s role in the international monetary system was being revised. These changes included provisions for large sales of IMF gold, permission for member nations to sell significant quantities of gold on the free market, and a major de-emphasis of gold’s monetary function. All these factors held down gold prices during most of 1975 and 1976. When direct depressing effects ended, prices rose again, and gold achieved superior rates of return— much higher than Salant and Henderson’s explana tion for the trends in gold prices is an elegant and convincing one for the period from 1968 (which marked the end of gold price-fixing) until 1981, but it breaks down after 1981. There has been a striking change in the behavior of gold prices since 1981. They fell, first sharply, then more gradually, with only short-lived reversals. In 1986 they again began to rise sharply. How, if at all, can these trends be reconciled with the rela tionship between the price of gold and with sales of government supplies of gold described above? The starting price of an exhaustible resource holds the key to our explanation. rb. V. Initial Price and Expectations of Demand The price of a depletable resource plays an important role in its price behavior. The price must increase at the rate (a risky asset like gold, of which governments hold large stocks, increases at a rate higher than In a “perfect world,” the initial price will be set so that the last ounce of gold is used via transactions completed up along a unique price path starting at the initial price and increasing at the set rate. A low initial price would result in greater demand at every date along the price path and the supply of gold would be depleted at a price that didn’t extinguish demand. Conversely, a high initial price would mean less demand for gold in each period; demand would drop toward zero, and gold stocks remain. Profits for owners in both cases would be lower than if the equilibrium price path were to emerge, so market forces tend to seek this unique initial price and price path. To calculate the correct initial price, it is essential to estimate the demand curve initial rb rb. for gold— that is, what demand will be for all prices. Incorrect estimation of the demand curve would lead to incorrect setting of the initial price and would necessitate later adjustment of the price to reflect the true demand. VI. An Unexpected Price Path Suppose that, in 1968, market participants esti mated a demand curve for gold, based on past demand and existing world stocks. In doing so, they implicitly calculated that if prices began to rise at a rate equal to (plus some risk pre m ium ), the world’s supply of gold would be exhausted just as the price rose to levels that would choke off demand. Market participants had many opportunities to observe demand in the price range from $35 to $100 an ounce (in real 1977 dollars) from at least the beginning of the twen tieth century until 1978. Although we do not have the data to plot these points precisely, we assume for this example that they fit a linear demand curve fairly well, as shown in figure 4. Based on these observations, gold speculators postulated that the same solid line that approximated rb POSSIBLE MISESTIMATION OF GOLD DEMAND Price estimated the demand curve for gold, given that demand at the very high prices that prevailed in 1980-81 had no precedent. But why has the decline in the real (as well as the nom inal) price of gold been so extended? Market participants are actively revising their estimates of demand at the prices where they first began to make serious errors, in the $200 to $400 range. First, the price fell precipitously as all speculators temporarily liquidated stocks in the knowledge that prices would fall. Most speculators were surprised when, after this initial price drop, demand was still too weak to support new price increases (consistent with the exhaustible resource model). Since 1983, when the price fall moderated, the market may be said to have been groping for a price path that would lead to a depletion of gold just as demand chokes off. Demand was weaker than expected in the inter mediate price range, and so the price continued to edge downward. The recent rise in gold prices may indicate that the bottom has been found, and that gold will yield superior returns. Perhaps a more fundamental ques tion is: why did people misestimate the demand for gold in the first place? Certainly there are many plausible explanations, and this paper does not attempt to establish one as being more cor rect than another. However, one possible expla nation is that their information was inadequate and inappropriate. People had virtually no basis for estimating the entire market demand curve, since the price had been more or less fixed for over 25 years. People did not even have estimates of the average expected demand at higher prices, let alone the variation to be expected about this average. Their estimate of market demand proved correct for prices that were not too far from observed values, but people systematically over estimated demand at higher prices. O il market analysts undoubtedly had similar difficulties fore casting demand after OPEC suddenly tripled prices in the early 1970s.6 Quantity demanded S O U R C E : Author. 6 Salant notes that rising real interest rates, along with incorrect de mand forecasting, can help explain w hy gold prices dropped after 1980. W e have implicitly assumed a constant real interest rate. Salant FIGURE 4 demand for prices from $35 to $100 an ounce would also be valid at higher prices. However, if true demand were represented by the broken curve, then it is obvious how this misestimation could produce an unexpected flagging in demand that would, in turn, cause the price decline in gold observed since 1981. Figure 4 illustrates just one of many ways that agents could have incorrectly points out that if, for whatever reason, the real interest rate rises, the price of gold would initially fall before increasing at a faster rate. W hy is this so? A higher real interest rate implies that gold prices must rise more rapidly. If no price decline occurred when real interest rates rose, the new higher gold price path would induce lower demand at every date than the original price path. But the original price path was set such that supply would be deplet ed just as a sufficiently high price choked off demand. If no price drop occurred when a higher interest rate prevailed, the stock of gold would not be exhausted; some owners would be left holding gold when high prices ex tinguished demand. This is not an equilibrium; such a prospect forces prices to jump down when the interest rate rises. Conclusion Two distinct regimes explain the unique behavior of gold prices since 1968. Between 1968 and 1981, prices increased according to the Salant and Henderson analysis; based on prices actually pre vailing during the 1968 to 1981 period (as high as $200 an ounce in real 1977 dollars) estimates of the demand curve for gold were roughly cor rect. However, incorrect forecasts of gold demand at higher prices meant that the price had to fall. The initial precipitous decline reflects the first reaction to this prediction. The continued m ild slide indicated that the market was edging down the demand curve in search of the price that fits the Salant and Henderson explanation of gold price determination. The turnaround in gold pri ces may well be telling participants that demand has been reestimated with enough confidence to justify a renewed upward trend in gold prices. “Don’t Panic”1 A Primer : on Airline Deregulation by Paul W. Bauer The old dictum says that if the Devil did not exist, the Church would have had to invent him. Similarly, if the regulator didn’t exist, the airline industry would have had to invent him—and did in 1938. A current question is what would happen to the industry were it totally deregulated. One thesis is that there would be a rush by existing and new entrants to those routes thought to be profitable. Other routes would be abandoned. Price competition would be destructive. With the essential link between economics and safety there would be an inevitable major air disaster, possibly involving a prominent Member of Congress. Public outcry and congressional responses would lead to the re-establishment of regulation. Since this was the sequence of events in the mid-30’s, why re-learn that lesson? This thesis has been challenged, but the lesson of history . . . cannot be totally ignored. Secor D. Browne, Chairman Civil Aeronautics Board (January 1972)2 Paul W . Bauer is an economist at the Federal Reserve Bank of Cleveland. The author would like to thank Randall W . Eberts, Jo e A . Stone, and others who provided useful com ments on an earlier draft of this paper. Introduction Former Civil Aeronautics Board (CAB) Chairman Browne’s statement 15 years ago can scarcely be interpreted as an unqualified endorsement of the government’s current policy of airline deregula tion. It does remind us, however, that the issue of airline regulation has been controversial for quite some time. The Civil Aeronautics Act (CAA) of 1938, enacted to counteract the alleged condi tions of competitive instability of an industry then in its infancy, began 40 years of pervasive government regulation by the now-defunct CAB. With passage of the Airline Deregulation Act (ADA) of 1978, the federal government com pleted an about-face in policy and reintroduced competitive forces into the market. For eight years now, the airline industry has been experiencing a great deal of turmoil, as evidenced by the large number of entries, mergers, and bankruptcies. Much of this turmoil, however, is not the result of deregula tion, but rather of the fuel price increase in 1979, 1 2 Sound general advice from The Hitchhiker's Guide to the Galaxy by Douglas Adams. Foreword to R .E.G . Davies' Airlines of the United States Since 1914, Putnam & Company Limited, London (1972). of the recession in the early 1980s, and of the air traffic controllers’ strike in August 1981. Even so, the regulation debate is heating up again as the events predicted by Mr. Browne seem to be unfolding—with such examples as the recent bankruptcy of Frontier Airlines, the financial prob lems of People Express and Eastern Airlines, and the crash of the Aeromexico airliner in southern California in August 1986. This paper analyzes the conditions that prevailed under CAB regulation and that led to the Airline Deregulation Act of 1978. These conditions are contrasted with the effects of deregulation observed so far. Finally, an attempt is made to predict the future evolution and per formance of the U.S. airline industry under deregulation. I. The U.S. Airline Industry Under CAB Regulation Between 1938 and 1978, the CAB maintained stria control over the two most important decisions airlines had to make: where to fly and how much to charge. This meant that airlines could only compete with one another by offering a higher quality of service (primarily more frequent flights 17 and other amenities). Studies have shown that CAB regulation led to more frequent flights and to lower load factors (the proportion of seats on a flight that are filled by paying passengers) than would be normal in a competitive airline industry.3 Since these actions resulted in high er costs for the airlines, and since the CAB was charged with maintaining the financial health of the industry (that is, preventing losses), it follows that fares were higher. In fact, the interstate carri ers subject to CAB regulation marked up fares 20 to 95 percent more than the intrastate carriers not subject to CAB regulation for similar routes.4 The General Accounting Office (GAO) estimated that passengers could save up to $2 billion dollars or more per year with competitive fares.5 18 II. The Theory Behind Deregulation Given fare markups of these magnitudes, why were the airlines’ earnings so mediocre? The an swer appears to be that regulated industries do not have sufficient incentives to control costs. Given the CAB’s mandate to maintain the health of the industry by raising fares whenever the airlines experienced hard times and the lack of a threat of competitive entry (the CAB had not al lowed the formation of a single new trunk airline from 1938 to 1978), a strong prima facie case exists for inadequate cost control. Using data from 1972 to 1978, Bauer (1985) found that, on average, airline costs during that period were 48 percent over the m inim um cost of providing the same service. Another example of the poor in centive structure can be found by analyzing labor costs. Providing a service product— transportation between two points—airlines could not stockpile their output in anticipation of a strike. Any output diverted by one carrier (either to other carriers, or to other transportation modes) as a result of the strike is a permanent loss to that carrier. Further, even when the strike is settled, the air line may lose some of its customers to other car riers. Regulated airlines could not offer large dis counts and free flights to lure their customers back, as United Airlines did after a strike in 1979. Under CAB regulation, strikes were very costly to the airlines, but higher labor costs could be 3 Douglas, George W . and Jam es C. Miller, (1974) Economic Regula tion of Domestic Air Transport: Theory and Policy, Brookings Insti tution, Washington, D.C . 4 5 T. E. Keeler, "Airlines Regulation and Market Performance," Journal of Economics 3 (Autumn absorbed by CAB fare increases or CAB approval to enter some profitable new route. Thus, there was little incentive for airlines to endure strikes. Given the evidence on fare mark ups and the suspicions about airline inefficiency, proponents of deregulation became convinced that elimination of CAB regulation, and a move towards more competition in the industry, would be beneficial to travelers and, ultimately, to the industry itself. Two basic tenets drive the model of the industry that proponents of deregulation had in mind: one, that the m inim um efficient scale size is reached at a relatively low level of output and, two, that new entry and the threat of new entry into the industry would ensure suffi cient competition to hold fares close to marginal cost and only allow firms to earn a normal profit.6 Numerous studies performed prior to deregulation, using various data sets from the late 1950s forward, found that larger airlines had no significant unit-cost advantage (measured in passenger miles) over smaller airlines. This research im plied that there was plenty of room in the U.S. airline industry for anywhere from 20 to 100 efficiently sized airlines (see White [1979]), and that there was little chance of concentration increasing in the industry if it were deregulated. The second tenet, that freedom of entry would severely lim it any market power that an airline may have, was being strongly sup ported by the new theory of contestable markets (see Baumol, Panzar, and W illig [1982]). Simply stated, this theory predicts that if market entry and exit involves no irrecoverable costs and can occur quite rapidly, the threat of entry is sufficient to ensure that firms in this market earn no more than a normal profit. The following illustrates how this result occurs. Suppose the firms in a contestable market decided to collude and to raise their prices. Although the strategy might work in the very short run, soon new firms not party to this agreement would recognize the opportunity for above-normal profits and w ould enter the indus try, driving prices back down. In a contestable market, even a monopolist w ould thus earn a normal profit, because if it tried to take full advantage of its monopoly power to earn more than a normal profit, another firm would enter and charge the lower price, capturing the entire market for itself. Clearly, not all industries in the economy can be considered contestable (the auto industry, for example, is definitely not). However, deregulation proponents considered Bell 1972), pp. 339-434. General Accounting Office, Report to Congress, Lower Airline Costs per Passenger Are Possible in the United States and Could Result in Lower Fares, February 1977, p. 11. 6 A normal profit is the minimum return required to keep the firm from shifting resources out of the industry. the airline industry a good candidate for contestability— once the artificial barriers to entry created by the CAB were eliminated. The following market characteris tics were considered to promote contestability: • Inputs used by the airline industry are all relatively mobile when compared to most other industries. Labor, energy, and mate rials can either be employed or let go on fairly short notice, as in most industries, but capital is much more mobile than in almost any other major industry. • Airlines can quickly shift planes from one route to another as the need arises. Further, since there is a ready secondary market for used aircraft— in fact, many carriers rent a sig nificant portion of their fleets— planes are fairly mobile from one carrier to another. • Ground facilities are usually rented, making them fairly disposable (acquisition is another matter, and will be discussed later). These properties are thought to make it relatively easy for incumbent airlines to begin service on new routes, so that if fares are too high on a given route, other airlines will enter those markets at lower passenger fares. These properties are also thought to facilitate the start-up of new airlines if existing lines are mak ing more than a normal profit. Thus, according to the contestable market view, there was little to fear on the part of consumers from airline deregulation. Even if the industry did evolve into a handful of firms, the contestable market theory predicted that they could only earn a normal profit and fares would be as low as possible. In summary, the proponents of deregulation predicted sharply lower coach fares, as fare markups would be bid down and airlines would strive to reduce their costs in the face of observed and potential competition. There would be some deterioration in service quality as flight frequencies would be reduced. However, this would in turn lower airline costs (by increasing load factors), thus further lowering fares, and pas sengers would receive the fare-service mix that they prefer. It was felt that there was no need to worry about increased concentration in the air line industry, because the m inim um efficient scale would be small enough to make room for many carriers. Besides, the threat of entry would be sufficient to hold fares down and service qual ity up, even on routes with few carriers. This divergence of prediction and reality can be traced to changes in the airlines’ operating strate gies that were induced by the increased freedom given to them by the elimination of CAB regula tion. These changes in strategy7occurred in the two areas mentioned earlier: where to fly and how much to charge. Market competition seems to have induced even more innovation than industry experts foresaw, leading to predomi nately beneficial changes in airline behavior. Fares As the CAB’s authority over fares was diminished, the airlines gradually developed a more complex fare structure to replace the relatively simple firstclass and coach-fare structure that existed under regulation. W hile an element of price discrimina tion certainly exists, most of the variation in fares is based on differences in the cost of serving the various classes of passengers.7 Fares are lower for travel outside the periods of peak demand. Examples include flying on weekends, flying in the middle of the day or late evening, and flying to locations that are out of season. A prime example of fare differences based primarily on cost is found between those who can book and pay for tickets in advance and those who cannot. It is costly for airlines to fly planes with empty seats, yet they intentionally have some slack in their systems so that they can accommodate lastminute travelers— for a higher price. These pricing strategies have enabled the airlines to increase both traffic and revenue far more than if a uniform pricing policy had been followed. The increase in the industry’s revenue passenger miles (RPM) and average load factor are plotted over time in figure 1. Both have increased since deregulation, although the effect of the recession in the early 1980s is clearly evi dent. Traffic increased 33 percent just from 1977 to 1979. As a result of this shift in pricing strategy, the average fare that passengers actually paid (adjusted for inflation) has fallen about 20 percent in the last 10 years, even though the standard coach fare has fallen very little. Though this is a far cry from the drop that had been expected given the fare markups and inefficiency that existed under regulation, it does represent a 7 For example, whether one stays over a Saturday night on a round trip has no effect on the airline’s cost of providing the service, yet it provides a very useful screening device enabling the airlines to charge higher fares to business travelers (who generally cannot meet this restric III. The Effects of Airline Deregulation The actual effects of airline deregulation, while being generally beneficial to date, have not mate rialized precisely as the proponents predicted. tion) and lower fares to pleasure travelers (who usually can). Thus the air lines can price discriminate between the two classes of consumers, taking advantage of the business travelers' higher price elasticity of demand (and the leisure travelers’ lower elasticity of demand) to increase their revenue and profits. quency on most routes (as a result of the increase in traffic) and by the lower fares (for those who could qualify for the discount fares); and the air line industry was able to increase its profits over what they would have been under regulation as the increase in load factors lowered costs. Routes 1945 1950 1955 1960 1965 1970 1975 1980 S O U R C E : Data from Bailey, Graham, & Kaplan (1 9 8 5 ). FIGURE 1 20 PASSENGER REVENUE PER RPM (in constant cents) OPERATING PROFIT MARGIN (% ) S O U R C E : Data from Bailey, 6raham, & Kaplan (1 9 8 5 ). FIGURE 2 considerable savings to travelers. A measure of the average fares paid by travelers, the average passenger revenue per RPM, is plotted along with the average operating profit in figure 2. All parties benefited to some extent by this new fare structure. The super-low fares enabled many leisure travelers to take trips they would not have considered before; business travelers gained by the increase in flight fre The other fundamental change in the airlines’ strat egies concerns the decision of where to fly. Few people inside or outside the industry foresaw the shift of the airlines to what is now known as a huband-spoke system. Since deregulation, instead of serving a hodgepodge of routes as dictated by the CAB, airlines organized their routes so that most of their flights now converge on one or two hubs. These hubs collect traffic from the “rim” cities, then the passengers change planes at the hub to go out on other flights to their final destinations. The potential benefits of this system were dem on strated to a small extent by Delta Airlines, which had a hub in Atlanta even under regulation.8 The hub-and-spoke system has enabled airlines to increase their load factors on flights both into and out of the hub, thus lower ing their costs and enabling them to lower their fares. An important side benefit is that flights can be scheduled more frequently because of the higher traffic density. Thus, instead of flight fre quencies decreasing under deregulation, as was generally predicted, they actually increased. Pas sengers are also more likely to be able to com plete their entire trip on one airline (which is advantageous to the airlines) and to avoid the inconvenience of changing planes at busy air ports (which the passengers like). Another benefit is that passengers can fly from almost any city to almost any other city without having to endure multi-stop flights. Usually a one-stop flight can be found, and routes with sufficient traffic density still receive nonstop service. How much are these innovations worth to consumers? Morrison and Winston (1986) estimated the total benefit of deregulation to con sumers to be $5.7 billion a year. For the average passenger, the benefits per trip were $11.08 and came from the following sources: a gain of $4.04 from lower fares, a loss of $0.96 from slightly increased travel time, and a gain of $8.00 from increased flight frequency. Morrison and Winston further estimate that airline profits would have been $2.5 billion higher than they were under regulation. Thus, airline earnings would have 8 The joke then was, “ It does not matter whether you are going to heaven or hell; you have to go through Atlanta first.” been even worse than they actually were (as reported in figure 1) had CAB regulation con tinued. These are substantial aggregate benefits. Passenger Concerns Even so, the gains of deregulation have not been shared equally by all travelers and, in fact, some may be worse off. Travelers who do not qualify for the discount fares and who must pay the full coach fare are probably worse off, unless the benefit from the increase in flight frequency is sufficient to offset this effect. Also, due to the oversupply of wide-body jets, which are ideally suited to carrying passengers coast to coast, fares for flights between 2,000 and 2,999 miles have fallen much more than other fares, so that travel ers on these routes have benefited proportion ately more than travelers on shorter routes. This is a temporary7benefit, however, and will last only until the airlines adjust their fleets. Finally, travel time for most flights involving large hubs has increased due to the increase in traffic. One of the early concerns of opponents and even of some supporters of dereg ulation centered on the availability of air service to small communities. Provision was made in the ADA for subsidies to help support air service to small communities for a period of up to 10 years, but many communities were not covered by these provisions. However, most small communities, far from losing service, have gained service. In gen eral, hedgehopping, multi-stop flights have been eliminated (lowering travel time), and flight fre quencies have been increased. Travel time for trips involving nonhubs has fallen from one to six percent on average.9 While service by trunk air lines has been replaced with service by commuter airlines in many cases (which is seen as less desir able), most of these commuter lines have their schedules coordinated with a major carrier at the connecting hub. When there is provision for on line ticketing, travelers can save approximately 25 percent over the interline fare. The few com m un ities that have lost all service have not had enough traffic to support scheduled carrier ser vice by any class of carrier. In these cases, sendee could be restored by government subsidies if the affected taxpayers deemed it desirable to do so. Beyond the basic issues of where to fly and how much to charge, there is the issue of whether the skies have become less safe under de regulation. Generally, the argument is that compe tition gives airlines an incentive to cut corners on 9 maintenance and to force pilots to fly more hours than is prudent. Under regulation, it was claimed that this was not a problem because the CAB en sured that the airlines were financially healthy so that they would not be as tempted to cut corners. So far, the safety record of the air lines is as good as ever, but there is the charge by some that the country has simply been lucky. There are two responses to this charge. First, it is bad for an airline’s business for its aircraft to be in volved in an accident that is shown to be a result of its own negligence. Not only is the public likely to avoid the airline, but the airline would also have lost a plane worth m illions of dollars and ex posed itself to even greater claims of liability.1 0 Second, and more important, one sure way of forc ing the airlines to perform proper maintenance is to step up inspections by the Federal Aviation Ad ministration (FAA). There may be a problem in do ing this, however. The number of airlines and air craft in service has risen dramatically since 1978, but the number of FAA inspectors has remained the same due to federal budget constraints. A related problem is that the number and the level of experience of the nation’s air traffic controllers has declined since deregulation as a result of the Professional Air Traffic Controllers’ Organization (PATCO) strike in the summer of 1981. Thus, if there is a poten tial safety problem, it is likely to arise from inadequate attention to inspection and flight con trol, not from deregulation. Industry Concerns As one might have surmised from the earlier dis cussion of strikes, labor leaders were also con cerned about the effects of deregulation. In fact, however, overall employment in the industry is up and compensation has kept pace with infla tion. According to data presented by Morrison and Winston (1986), from 1975 to 1984, pilots’ average real income fell a modest $500, dropping to $47,720 in 1977 dollars, while that of flight attendants increased $1800 to $14,428, and that of mechanics increased about $500 to $19,775. Industry employment has increased since the early 1970s. Employment declined from a 1980 peak until 1983 when it rebounded and continued the upward trend it followed from 1971 to 1978 (see Morrison and Winston [1986]). Though the average worker has not suffered A n airport is classified as a “nonhub" if its total enplaned revenue passenger miles represents less than 0.05 percent of the total U .S . market. 10 -1 It is assumed, of course, that the idea of preserving life also _L W enters into the issue. 21 22 under deregulation, many union workers have been forced to take wage- and work-rule conces sions, and some have had their careers inter rupted as they have been either laid off or let go by airlines performing poorly in the new compet itive environment. Two-tiered labor contracts have also been introduced. All this and the growth of the nonunion sector of the industry among the entering airlines have induced wide, and sometimes surprising, wage differentials between workers for different airlines, so that aggregate data on the welfare of workers is somewhat misleading.1 1 Finally, some firms may not have benefited from deregulation. There have been a number of bankruptcies in the airline industry since deregulation, most notably Braniff Airlines and Continental Airlines, which are both still fly ing after Chapter 11 reorganizations. Another air line (Frontier) is not flying, but is being acquired by Texas Air. In addition, there have been numer ous mergers, particularly in the last year. Cur rently pending are two large mergers involving Continental-Eastem-People Express-Frontier (by Texas Air) and Delta-Western, that would create the first- and fourth-largest airlines in the U.S., respectively. While business failures impose some costs, such as uncertainty and inconve nience on the part of consumers, the loss of jobs on the part of workers, and the financial loss to creditors and stockholders, failures are a neces sary force to ensure that firms operate efficiently in providing the services that consumers desire at a cost they are willing to pay. IV. Future Evolution of the Industry The current merger wave could be regarded as a natural process leading toward a competitive air line industry. Travelers prefer to have nonstop or one-stop flights with one carrier, rather than take a flight that would require them to endure two or more stops, or to change airlines at a busy airport. Providing such service requires a national route network with several regional hubs. In addition to the benefits for travelers, there also might be cost advantages to operating such a large hub network. Though the cost studies performed dur ing the regulatory period indicated that there were no scale economies in the airline industry, the cost inefficiencies present in the regulatory era may have distorted these estimates. Bauer (1985) used an econometric procedure that allowed for these inefficiencies and found evi dence of substantial returns to scale (contrary to n For example, unionized Western Airline workers earn less than Delta’s nonunion workers. Also, United’s unionized pilots earned 40 percent more than their ill-fated Frontier brethren. the cost studies that did not allow for ineffi ciency). This issue aside, there are definitely cost advantages to the extent that large hub-and-spoke systems lead to higher load factors. Currently, only United Airlines and American Airlines oper ate such networks. However, once the current wave of mergers subsides, there will be anywhere from six to eight such super-airlines, perhaps another four to six medium-sized carriers, and perhaps 10 to 30 regional carriers. Should the public be concerned about the potential anti competitive effects of these airline mergers? If the industry were per fectly contestable as discussed earlier, then the answer would be no. Many researchers have tested whether or not the implications of the the ory of contestable markets hold exactly; unfortu nately, no one has found that they have. Bailey, Graham, and Kaplan (1985), for example, found that on concentrated routes (routes served by only one or two carriers) airlines can raise fares five to 10 percent over what they could charge on nonconcentrated routes. There are two reasons why actual and potential competition have not lived up to their promise in the airline industry. First, capital—both physical and human capital— may not have fully adjusted to the new deregulated environment. The number of merger proposals re cently is evidence that the airline industry is not in a long-run equilibrium with respect to the number and size distribution of carriers. Given that it has been eight years since the formal dereg ulation process started, it appears that the transi tion from a regulated to a competitive market equilibrium will take longer than expected. A second reason for the apparent lack of competition on some routes is that entry into some concentrated markets is not as easy as was first expected. Many airports across the coun try7have severe problems with traffic congestion (for example, airports in Denver and Washington, D.C.); obtaining gates and takeoff and landing slots at these airports is difficult. Since gates and landing rights are “grandfathered” to the airline holding them as long as they are used, the air lines that have these scarce resources can earn monopoly returns from them. This creates a severe barrier to entry for airlines wishing to begin service on these routes. The importance of this problem was highlighted in the recent merger of Continental Airlines with Eastern Air lines. To get approval for the merger, slots at LaGuardia airport had to be sold to Pan-Am so that it could set up a competing shuttle service. Even at relatively uncongested airports, such as Cleveland Hopkins, airlines are reluctant to release unused gate space. Much of the impetus for the current merger wave is that airlines find it is easier to buy other airlines to expand (in an effort to reach the most efficient size) than it is to grow internally (and be forced to try to obtain takeoff and landing slots on their ow n).1 2 Given that the contestable market theory does not seem to apply on all routes, should consumers worry about the increasing concentration of the industry? Currently, the national four-firm concentration ratio (CR), the sum of the market shares of the largest four firms in an industry, has remained unchanged at 47 from 1975 to 1986. Depending on how the cur rent merger proposals are approved, it is likely that the resulting concentration ratio for the industry will be anywhere from 57 to 61. While this is high enough to cause concern, particularly in light of the fact that some individual city pairs now have even higher concentration ratios, there are reasons not to become alarmed just yet. First, even though the industry has a fairly small number of firms, and concentration is relatively high, fare and route competition has been intense since deregulation. There have been no accusations that the industry' as a whole is earning more than a normal profit. Further more, to the extent that only large airlines can provide the national route structure and the potential for nonstop and one-stop service that consumers prefer at the lowest cost, the level of concentration is only a reflection of the fact that there is only room for a limited number of effi ciently sized airlines in the market. If the ultimate effect of deregula tion is a national market with six to eight huge airlines, there still would be a great deal of com petition in the industry, even if many of the major cities are dominated by as few as two carriers. If one wants to fly from Cleveland to Los Angeles, for example, there may only be one or two air lines to choose from that provide nonstop service. However, one-stop service is a close substitute for nonstop service and, in that case, one would con ceivably have six to eight choices depending upon which hub city he or she prefened to change planes. On shorter routes, such as Cleveland to Chicago, the smaller regional carriers would pro vide additional competition to the major carriers and thereby put a check on fares.1 O n still short3 -1 O A further cause of the increased merger activity now is that _L Ld the Department of Transportation (D OT) has authority over air line mergers for the next two years, at which time the Department of Ju s tice (D O J) will have that responsibility. The D O T has been much more lenient than the D O J. If they cannot obtain space at the major airports on the route in question, they have the aircraft that can effectively utilize the smaller regional airports which, in some cases, may be more conve nient for passengers. er flights, Cleveland to Columbus for example, sur face transportation provides some additional com petition even if the market for air travel between those points is concentrated. Given the shortcom ings of the contestable market theory as applied to the airline industry, however, the disciplining effect of potential competition may not be enough to ensure competitive behavior. It may still be necessary for the Departments of Transportation and Justice to enforce current antitrust laws. In summary, at this point, the mar ket for air travel in the U.S. is not perfectly contes table and, on some concentrated routes, airlines are able to charge modest fare markups on the order of between 5 and 10 percent. This situation is likely to continue for the foreseeable future, until steps are taken to alleviate the congestion problems at certain airports. The next few years will probably witness an increase in the concen tration in the industry to the point where six to eight large airlines dominate the national market with a host of smaller regional and commuter lines filling a variety of special niches. There will be sufficient competition to ensure that travelers are better off than they were under regulation, but it remains to be seen how closely the industry will conform to the perfectly contestable ideal that was envisaged by proponents of deregulation. V. Conclusion Deregulation of the airline industry has been a painful experience for some travelers, workers, and firms. Large fuel price increases, the air traffic controllers’ strike, and recessions have made the process even more difficult. O n the whole, how ever, deregulation has been favorable. Far more individuals have benefited than have been hurt. Consumers are receiving better service for lower average fares; employment and compensation in the industry are up; and the airlines are generally earning higher profits than they would have under regulation. Yet, even eight years later, the industry is still adjusting to its new environment, and the final results of deregulation have yet to be determined. There are several steps that can be taken to ensure that the gains to date are not lost and that the costs of adjustment to deregulation are minimized. First, airport expansion is needed to help reduce one of the few barriers to entry that remain in the industry. Deregulation, by great ly increasing air travel through lower fares, made the congestion worse. The solution, however, is not to reduce air travel, but to expand the system. The federal government has a $3 5 billion fund that can be spent only on promoting air travel. This fund is financed by an 8 percent tax on air fares, but has become embroiled in the current federal budget problems. The money 23 could be spent to expand airport facilities, to modernize the air traffic control system, and to hire more FAA inspectors. These expenditures would enhance the competitiveness of the system by lessening the incentives for airlines to merge, as well as by improving their safety and reliability. Second, the U.S. Departments of Transportation and Justice should continue to enforce existing antitrust laws. W hile the compet itive discipline that free-entry into the industry offers should not be ignored, it is important that REFERENCES The Hitchhiker’s Guide to the Adams, Douglas. New York, NY: Pocket Books, 1979- Galaxy. Bailey, Elizabeth E., David R. Graham, and David P. Kaplan. Cam bridge, MA: The MIT Press, 1985. Deregulating the Airlines. 2 4 Bauer, Paul W. “An Analysis of Multiproduct Technology and Efficiency Using the Joint Cost Function and Panel Data: An Application to the U.S. Airline Industry,” Ph.D. Dissertation, Uni versity of North Carolina at Chapel Hill, 1985. Airlines of the United States since Davies, R.E.G. London: Putnam & Company Limited, 1972. 1914. Economic Regulation of Domestic Air Transport: Theory' and Policy. Washington, D.C.: Brookings Insti Douglas, George W., and James Miller. tution, 1974. Graham, David R., Daniel P. Kaplan, and David S. Sibley. “Efficiency and Competition in the Air line Industry,” vol. 14, no. 1 (spring 1983), pp. 118-138. BellJournal of Economics, these agencies not place too much faith in freeentiy to the exclusion of other factors, particularly in the short run. Finally, allowing foreign air carriers into the U.S. market (with reciprocal agreements for entry into their markets) should be consid ered as a way of further increasing industry com petition. These policies would help the U.S. to maintain its position as having the world’s fore most airline network. Keeler, Theodore E. “Airline Regulation and Market Performance,” vol. 3, no. 2 (autumn 1972), pp. 399-424. BellJournal of Econom ics and Management Science, Deregula tion and the New Airline Entrepreneurs. Cam Meyer, John R., and Clinton V. Oster, Jr. bridge, MA: The MIT Press, 1984. Airline Deregulation: The Early Expe rience. Boston, MA: Auburn House, 1981. Morrison, Steven, and Clifford Winston. The Eco nomic Effects of Airline Deregulation. ________ Washington, D.C.: Brookings Institution, 1986. U.S. General Accounting Office. “Lower Airline Costs per Passenger are Possible in the United States and Could Result in Lower Fares,” Report to Congress, February 1977. White, Lawrence J. “Economies of Scale and the Question of ‘Natural Monopoly’ in the Airline Industry,” vol. 44, no. 3 (1979), pp. 545-573. Journal of Air Law and Commerce,