The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 Are National Stock Markets Linked? 15 Are Econom ic Forecasts by Government Agencies Biased? Accurate? 24 The Link Between the Value o f the Dollar, U.S. TVade and M anufacturing Output: Some Becent Evidence 38 The Dubious Success o f Export Subsidies for Wheat 48 The M acroeconom ic Effects o f Deficit Spending: A Bevievv THE FEDERAL RESERVE BANK of ST. I/HIS 1 Federal Reserve Bank of St. Louis Review November/December 1988 In This Issue . . . In the first article in this Review, “Are National Stock Markets Linked?” Gerald P. Dwyer, Jr. and R. W. Hafer investigate the connection among stock price behavior in the United States, Germany, Japan and the United Kingdom. Much o f the recent com m entaiy about stock prices both before and after the October 1987 crash suggests that stock prices in different countries rose and fell as one. The authors show that, under certain as sumptions, there is no reliable relationship between the levels o f prices in various stock markets, even though changes in the different market m ea sures may be related. The empirical evidence presented is based on two data sets. One con sists o f daily stock prices for the different countries from July 1987 through January 1988. Using this data, which includes the w orldw ide decline in stock prices in October, Dwyer and Hafer find no evidence to support the notion that prices around the world “ratcheted up" before the crash, as some have argued. To put their analysis in a historical perspective, the authors examine a second data set, which consists o f monthly stock price values for each country from 1957 through 1987. Test results based on these data corroborate those using the daily data: there is no link among the levels o f stock prices across national markets. * * * The econom ic policies advocated or implemented during the Reagan presidency have elicited widespread disagreement among economists. Annual econom ic forecasts o f output growth, inflation and unemployment were a concentrated area of dispute, as economists disagreed, for example, about the expansionary effects of tax cuts and deregulation. Most years found administration supporter’s predicting strong real growth without accelerating inflation; critics — including prominent members o f the ad ministration — labelled these forecasts "rosy scenarios." Are the forecasts o f government agencies unbiased and accurate judgments about the future or are they more closely associated with advocacy and cheerleading? In the second article of this Review, Michael T. Belongia analyzes this question, comparing forecasts o f the Council o f Economic Advisers (CEA), the Congressional Budget Office (CBO) and several private-sector groups. He first tests for bias in forecasts o f real GNP growth, inflation and unem ployment and finds each group’s forecasts to be unbiased. He then tests for for ecast accuracy and finds little grounds for judging one government agency’s forecast superior to the other’s. The author also finds little to distinguish strongly between the accuracy o f private- and public-sector forecasts. Overall, "rosy scenario” and other epithets appear to be ground less rhetoric. * * * Does a decline in the international exchange value o f the dollar boost the U.S. output o f manufactured goods? In this third article in this Review, NOVEMBER/DECEMBER 1988 2 John A. Tatom examines the basis for this w idely assumed inverse rela tionship and its opposing view, that movements in the dollar’s value are positively related to changes in the U.S. capacity to produce output. As Tatom explains, exchange rates can reflect movements in underlying competitiveness. Thus, factors that reduce the ability o f U.S. producers to compete are offset, in part, by a fall in the value o f the dollar. Similarly, when those factors raise U.S. competitiveness, the associated gains in do mestic output are accompanied by a rising dollar. Examining the evidence, Tatom finds that when the dollar rose, U.S. output and capacity growth, particularly in the industries most associated with the trade deficit, w ere unusually rapid com pared w ith the past and output growth abroad. These trends evidently have reversed since the dollar began to fall. Tatom concludes that the increased competitiveness produced by eco nomic policy changes in the early 1980s has been reduced by a reversal of some o f these policies in the mid-1980s. The confusion over the link be tween the value of the dollar and U.S. output, according to Tatom, arises from a tendency to see exchange rate movements as the result of mysteri ous external influences, rather than rational responses to changes in the domestic econom ic environment. * * * One pervasive feature o f agriculture throughout the w orld is government involvement. In the case o f the European Community (EC) and the United States, such government involvement in agricultural trade policies has escalated a disagreement into a trade war. In 1985, the United States, re sponding to the rapidly expanding EC agricultural exports that allegedly were displacing U.S. exports, introduced an export subsidy program, the Export Enhancement Program (EEP). The two goals o f the U.S. export sub sidy w ere to expand U.S. agricultural exports and to persuade both the EC and other subsidizing nations to negotiate the lowering and elimination o f all production- and trade-distorting subsidies. In the fourth article in this Review, Cletus C. Coughlin and Kenneth C. Carraro examine this export program. The authors find that, w hile the EEP has increased wheat exports, the short-run costs o f doing so have been higher than other surplus disposal alternatives. In addition, the authors provide evidence suggesting that the EEP may be ineffective, and possibly counterproductive, in changing the EC's agricultural negotiating position. * * * In the final article in this Review, K. Alec Chrystal and Daniel L. Thorn ton review the m acroeconom ic effect o f deficit spending. The authors dis cuss the ways in which macroeconomists have asserted that society benefits from deficit spending, citing arguments for and against each. They observe that the once-com mon view that the market econom y is prone to extended periods o f unemployment and below-potential output because o f the private sector’s inability to generate sufficient aggregate dem and has given way to the view that deficit spending can have little permanent effect on output or employment. This means that the principal benefits from deficit spending come from using it to stabilize output around this policyinvariant path. Chrystal and Thornton note that the information policym a kers need to use deficit spending to stabilize output is extensive; m ore over, the evidence that deficit spending has succeeded in stabilizing output in weak. FEDERAL RESERVE BANK OF ST. LOUIS 3 Gerald P. Dwyer, Jr. and II. W. Hafer Gerald P. Dwyer, Jr., associate professor of economics, University of Houston, is a visiting scholar and R. W. Hafer is a research officer at the Federal Reserve Bank of St. Louis. Nancy D. Juen and Rosemarie V. Mueller provided research assistance. Are National Stock Markets Linked? A 1 M.NALYSTS generally agree that national asset markets have becom e more integrated in recent years. This process began with the relaxation of controls on capital movements in the 1950s and was followed, during the last decade or so, bv the gradual relaxation o f exchange controls. Recently, substantial improvements also have been made in computer and communication technology that have low ered the cost o f international information flows and cross-border financial transactions.1 This globalization o f financial activities has led some to argue that the behavior of stock prices in 1987 was influenced by international events to a greater extent than anyone had thought previ ously. For example, in its discussion o f 1987, the report by the Presidential Task Force on Market Mechanisms (comm only known as the Brady Commission) suggests that “[i]nvestors made com parisons o f valuations in different countries, often using higher valuations in other countries as justi fication for investing in low er valued markets. Con sequently, a process o f ratcheting up among w orldw ide stock markets began to develop.” - In ’Cooper (1986) and Bryant (1987) discuss these and related issues. Presidential Task Force on Market Mechanisms (1988), p. 10. 3lbid., Study I, p. 2. Most of the reasons offered in the report to explain the decline, however, are related to economic develop ments in the United States or changes in the dollar’s value in foreign exchange markets. The primary reasons for the break in equity prices given are: the persistent and large budget and other words, a higher level o f prices in one market increased the level in other markets. As for the fall in prices, the Brady Commission report notes that "(w]hat may have appeared strictly a ‘Wall Street' collapse was the result o f the cumulative impact of several developments occurring simultaneously in several other financial centers.” 1 There appears to be no one reason that explains the w orldw ide decline in equity values during October 1987. The timing and magnitude o f the declines differed across markets around the world. Even so, all o f the organized markets fell.4 This coincident fall suggests that changes in the markets are indeed related to one another. The behavior of stock prices since the October crash suggests that markets around the w orld do not move in tandem. As interpreted by Cowan (1988), "the first quarter o f 1988, if nothing else, dispelled the popular notion that there is one synchronized, global market.” While stock prices in the United States w ere 9.8 percent below their value on October 16, stock prices in Japan w ere trade deficits in the United States; instability in foreign ex change markets, stemming primarily from the continued fall in the dollar after the Louvre accord; the international rise in interest rates; and the threatened end to takeovers in the United States. For more, see ibid., Study I, pp. 11-13. “Roll (1988) provides an analysis of the different markets’ be havior. NOVEMBER/DECEMBER 1988 4 onlv 0.4 percent below their pre-crash level bv the end o f the first quarter. The German stock price index at the end o f March 1988, in contrast, was 28.3 percent below its pre-crash level. While some markets had recovered some or all of their October 1987 loss, others clearly had not. These disparate movements raise questions about just how the different stock markets around the world are related. Given the increase in Japa nese stock prices since the crash, should w e be surprised not to have had a similar rise in the United States? Or is it really unusual for all mar kets to move together as they did during the week o f October 19, 1987? In this article, w e examine the statistical rela tionship between the levels and movements of stock price indexes for Germany, Japan, the United Kingdom and the United States, using daily data for July 1987 through January 1988 and monthly data for the past 31 years. Thus, w e can examine the relationships both over a short period encom passing the October crash and across a longer horizon, in order to put the events o f last year into some long-run perspective. HOW SHOULD STOCK PRICES BE LINKED? In standard models o f stock price determina tion, the level o f a stock's price equals the present value o f expected future dividends. Anything that changes the fundamentals — that is, the expected future dividends or the interest rate at which those dividends are discounted — affects the price of the stock.1 One w av of thinking about linkages across na tional stock markets is to start with an extreme form o f linkage and to examine what loosens these links. Suppose that the transaction costs o f buying and selling stocks and foreign exchange anvwhere in the w orld are zero. Suppose also that investors are risk-neutral; that is, stockholders are indiffer ent to holding different stocks only if they yield the same expected return. For all stocks in the w orld to be held, the expected return in terms of any com m on currency must be the same for hold ing any stock anywhere in the world. This idea can be written as 5 The major alternative to models based on the fundamentals is known as a rational bubble model. Essentially, this model allows prices to deviate from that predicated on the fundamen tals. For a discussion of the differences between these models, see Santoni (1987) and Santoni and Dwyer (1988). FEDERAL RESERVE BANK OF ST. LOUIS Hi Eh,, = Ehn + EAe,, where Ehiit is the expected rate o f return from holding a stock in country i in terms o f i s cur rency in period t, EhMis the expected rate o f re turn from holding a stock in country j in terms of j's currency and EAe, is the expected rate of change in the price o f country j's currency in terms o f countiy i s currency. For convenience, we can call this relationship "stock return parity.” " If it existed, stock return parity w ould im ply that the expected return from holding stock o f a domestic firm is the same as the expected return from hold ing the stock o f a foreign firm. Linkages o f Price Levels Even if stock return parity holds, unexpected events will guarantee that there will be no reliable relationship between the levels o f various stocks' prices, even in the same currency. For example, suppose there is an unexpected permanent in crease in the dem and for Hondas relative to Fords which increases the expected earnings and divi dends o f Honda relative to Ford. Because o f the change in expected dividends, the price o f Honda stock will increase relative to the price o f Ford stock. This will occur even with stock return par ity. An unexpected increase in the price o f Honda stock produces a capital gain, which means that the e,v post rate of return from holding Honda stock will be higher than from holding Ford stock. This is perfectly consistent with stock return par ity, which holds that the expected rates of return are the same both now and in the future. M ore over, once the price o f Honda stock increases, w e should hardly expect that the price o f Ford stock will rise just because the price o f stock in Honda is higher.7This analysis holds both for companies in the same national market and for companies in different national markets. Internationally, even though financial markets may be increasingly inte grated, the relative levels o f stock prices lor in dexes o f stock prices) around the w orld w ill d i verge, because they represent the valuation of different firms. The preceding argument can be demonstrated more formally. In terms o f e,\ post rates o f return, equation 1 can be written as 1 1 h,,, = hi., + Ae, + 2 - eM - 6ln effect, this is uncovered interest parity applied to stocks. We ignore the second-order term Eh, ,Ae,. 7lf firms have firm-specific capital, the relative change in price levels can be permanent. 5 where e„ is the unexpected part o f the holding period return for stock i in period t, eMis the unex pected part o f the holding period return for stock j and e,., is tin? unexpected part o f the rate o f change of the exchange rate. If expectations are rational in the sense o f iVluth (1961), then the e’s are indepen dent o f the expected part o f the holding period returns.'1Assume that the variances and covari ances o f the e's are constant. If w e assume that dividends are zero, then equation 2 can be written (3i pil+l - pi., = Pj, +, - Pi, + e,+, - e, + - e,, - e,.,, where p is the logarithm o f the price o f the stock, e is the logarithm o f the exchange rate and the sub script t + 1 denotes the price one period in the future. Rearranging terms, equation 3 can be w rit ten as 1 1 Pi.,+, - Pi, +, - el+l = p,, - Pi, - e, + e,, 4 - eM - e,..,. Define x = p, — p; — e. Then equation 4 can be rewritten as (51 X, + , = X, + 6 j, - correlations between the levels o f national stock price indexes are unstable. The levels o f stock prices in different markets may rise or fall to gether, or move in opposite directions. Moreover, the size o f correlations o f the stock price levels will depend on the sample period used and the unex pected changes in the two countries’ stock prices and exchange rates in that period. Another way o f thinking about a time-series process that is a random walk is in terms o f a "unit root.” 9Although a random walk is a particu lar kind o f unit-root process, the two are not syn onymous. While its evolution may have additional components, a unit-root series wanders around in the same way that a random walk does. For exam ple, neither a random walk nor a unit-root process has a tendency to return to anv particular value over time. The algebra above has been simplified considerably bv assumptions that make the rela tive stock price indexes a random walk. Rather than maintain these assumptions (for example, constant variances o f the unexpected parts o f the returns from holding stocks and the changes in the logarithm o f the exchange rate), w e directly test for unit roots in the empirical analysis. Gj, - Equation 5 shows that relative stock prices next period simply are equal to relative stock prices this period plus the difference between the unex pected parts o f the holding period returns (e,, — e,,) and the unexpected change in the exchange rate (e,., I. In other words, even if expected rates of return are identical, relative stock prices in terms o f a com m on currency are a random walk. When the relative stock price indexes take a random step up or down, the relative stock prices show no tendency to return to anv particular value. This is important because it means that, even if the expected holding period returns o f two stocks w ere perfectly correlated, the levels o f the prices will show no stable relationship. Because relative stock prices are characterized as random walks, 8 Actually, the only implication that we need is that the expected part of the holding period return and the unexpected part are uncorrelated. Rates o f Return Stock return parity, w hile useful for making the point above, is illustrative rather than descriptive. Stock return parity implies that, since the ex pected rates o f return from holding different stocks are the same, the correlation o f expected returns is one. It is unlikely that stock return par ity holds. If stock return parity holds across na tional borders, it should hold within a countiy as well; this means that differences between the ex pected returns on domestic stock should be un predictable. This prediction, however, is inconsis tent with the data."1 Factors Decreasing the Correlations — Evi dence indicates that expected returns from hold ing stock in both the United States and other ,0See Malkiel (1985). 9 The precise definition of a unit root is based on the autoregres sive representation of a series. If the fundamental movingaverage representation of a series, say x, has an autoregres sive representation, then it can be written as [1 -t<(L)]x, = e,, where L is the lag operator such that Lx, = x, , and a(L) = Xa,L'. The polynomial in the lag operator «(L) always can be written as «(L) = (1 - (3,L)p(L). If there exists a root (J, that is equal to one, then the series x is said to have a unit root. NOVEMBER DECEMBER 1988 6 countries are related to the riskiness o f holding stock relative to other financial assets. To the ex tent that the variability o f the return from holding a stock cannot be diversified away, expected rates o f return are higher for riskier stocks." This finding suggests that stock return parity is unlikely to hold. Expected rates o f return differ across firms and industries; available evidence suggests that country risk also is important.1 2 There also are transaction costs associated with buying and selling stocks. Today, explicit transac tion costs are relatively unimportant in buying and selling large blocks of stock around the world. With improvements in communication and the ability to order trades over phone lines, the ex plicit cost to someone in London o f buying AT&T stock in N ew York is little more than the cost to someone in N ew York. Nonetheless, government restrictions are part of the costs of executing a transaction, and these restrictions have been important at times in exe cuting international transactions. Exchange con trols were one o f the ways that countries main tained the fixed-exchange-rate regime in place until 1973. By limiting access to foreign exchange, governments sought to manipulate the demand for their currency relative to foreign currency, thereby assisting their attempts to maintain a fixed exchange rate. In some cases, governments also restricted foreigners’ ability to purchase domestic financial assets. Both types o f controls have been declining gradually since the demise of fixed ex change rates." Factors Increasing the Correlations — Some forces make expected returns in different coun tries positively related even if there were no inter national financial transactions. If the demand for automobiles increases in the United States, which increases the expected earnings and dividends of domestic automobile companies, it also can in crease the expected earnings o f automobile com panies like Honda, which are headquartered in Japan and sell automobiles in the United States. Consequently, changes in stock prices in the 1 Malkiel (1985) summarizes the evidence. The riskiness of a 1 firm’s stock can be divided into its relationship with general movements in the market (market risk) and the factors that cause it to deviate from the market (non-market risk). Nonmarket risk includes those factors that influence a specific firm or industry. The idea that there are factors that cause firms or industry groupings of firms to deviate from the market portfolio applies also to the divergent movements of national stock price indexes. 1 See Solnik (1974); Cho, Eun and Senbet (1986). 2 FEDERAL RESERVE BANK OF ST. LOUIS United States and Japan can be positively corre lated even if no foreigner can buy stock in either country. This example, w hile triv ial in some re spects, points out that international trade creates a link between at least some stocks in different markets. In addition to trade, multinational operations bv firms create links through ownership o f real assets that can affect firms headquartered in different countries. For example, Ford manufactures auto mobiles in Europe. A recession in Europe would likely decrease the demand for Ford automobiles and low er Ford's earnings, dividends and stock price on the N ew York Stock Exchange. Finally, relative to data on individual firms' shares, stock index data will have a higher correla tion than the correlation of returns from randomly selected stocks on different markets. All o f the actual data that w e use below are indexes o f stock prices. Consequently, the indexes average out much o f the variation attributable to individual firms or industries. Thus, if there w ere no factors that differentially affect firms in different coun tries, the expected returns in any com m on cur rency measured by these indexes w ould be virtu ally the same. DAILY DATA SURROUNDING THE CRASH In this section, w e examine daily values o f stock price indexes for seven months surrounding the crash for evidence o f the “ratcheting u p ” in stock markets suggested by the Bradv Report. Daily val ues o f stock price indexes from Germany, Japan, the United Kingdom and the United States for July 1, 1987, through Januaiy 29, 1988, are used.1 This 4 period includes three months before the October 1987 crash and three months after it. To make the relative values o f the indexes comparable, all o f the measures are set to a base value of 100.0 on July 1, 1987. Because the markets are open in daylight hours in different time zones, the markets in our sample are not all open at the same time. We l3For an annual discussion of changes in these controls on a country-by-country basis, see any issue of the International Monetary Fund’s Exchange Arrangements and Exchange Restrictions. 1 The daily stock market indexes, both in terms of local currency 4 and U.S. dollars, are from Morgan Stanley’s Capital Interna tional Perspective. The indexes are market-weighted price averages without dividends reinvested. 7 Chart 1 Levels of Stock Price Indexes (7/1/87 100) in local currency United States United Kingdom Ratio Scale 120 I----------------------------------------------------------------------------------------------------------------------------------- Ratio Scale 71 / 81 / 91 / 1/ 01 1/ 11 1/ 21 11 / 21 / West Germany Ratio Scale ... • ?/ ... 71 / define a trading day as starting with the opening o f the European markets. The levels o f the different indexes, measured in terms o f local currency, are shown in chart 1. The behavior o f the indexes reveals some comm on movement during this period, especially around October 19. All o f the indexes decline sharply from the m iddle o f October to the end o f the month.1 '’ Before and after the crash, however, there appears to be little comm on movement in the levels o f the indexes. While the behavior o f the indexes in terms of local currency is interesting, the indexes should be measured in terms o f some com m on currency to be directly comparable. Comparing stock prices in the United States in dollars and stock prices in the United Kingdom in pounds is much like mea 1 The sizes of the decreases in stock prices in October 1987 5 differ substantially. The decline in the United States was 21.6 percent, slightly below the average decrease of 24.6 percent for a sample of 23 countries. For example, stock prices fell as i 81 / 91 / 10 1 / 1/ 11 i i 1/ 21 l 11 / 2 /1 suring the price o f apples in dollars and pounds and comparing the movements o f the two. We measure the different indexes in terms o f U.S. dollars. The dollar-denominated indexes in chart 2 show similar patterns to those in chart 1. The dif ferences in behavior of the different indexes since the crash, however, are striking. Based on the data in chart 2, the U.S. and U.K. indexes increase only slightly after the crash, w hile those in Germany continue to fall. The index for Japan, however, returns roughly to its value im m ediately fol low ing the crash. By Januaiy 29, 1988, stock prices in Germany, the United Kingdom and the United States are still below their October 19 levels. For example, stock prices in the United States at the end o f Januaiy are about 17 percent low er than on little as 5.8 percent in Austria and as much as 45.8 percent in Hong Kong (measured in U.S. dollars). For further discussion, see Roll (1988). NOVEMBER/DECEMBER 1988 8 Chart 2 Levels of Stock Market Indexes (7/1/87 = 100) in U.S. dollars United States United Kingdom Ratio Scale 1201 ------------- Ratio Scale West Germany Ratio Scale 120 80 October 12, 1987 — one week before the crash. Similarly, prices in the United Kingdom and in Germany at the end o f January are about 18 per cent and 33 percent below their October IE levels. In sharp contrast, the Japanese stock market index on January 29 is less than 1 percent low er than on October 12. Despite these different movements o f the levels o f stock prices, there is some com m on behavior in the changes in the different country indexes. A simple way to see this is to calculate the number o f days eveiy index increased or decreased. Din ing the seven months covered in chart 2, there are 20 days when all the indexes increased and 16 days when all decreased. Little significance should be attached to the greater number o f coincident increases than decreases: with the exception of Germany, increases predominate in each country during the period. Coincident increases are more likely than decreases even if the changes are unre lated. Some coincident increases and decreases in all o f the indexes are expected bv chance alone. If the probability o f an increase in one country is unrelated to events in other countries, the proba bility o f a coincident increase in all o f the indexes is about 7.7 percent, and the probability of a coin cident decrease in all o f the indexes is about 4.85 percent."’ This implies that these data w ould have about 19 days o f coincident movements due to chance alone, substantially less than the actual 36 days with increases or decreases in all four in- 1 For the data in chart 2, the indexes decrease in 52.7 percent of 6 the days in Germany, 49.0 percent of the days in Japan, 42.0 percent of the days in the United Kingdom and 44.7 percent of the days in the United States. If the changes are unrelated, the joint probability of coincident decreases is simply the product of the proportions of days with decreases to the total, which is 4.85 percent or about 7.2 days. The joint probability of the four indexes increasing or staying the same is about 7.7 percent or about 11.5 days. If the changes in the indexes are unrelated, the total number of days expected to have coincident move ments is about 18.7 days, with a standard deviation of this expected value of about 4.1 days. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 9 Table 1 Correlations of Stock Price Indexes: July 1 ,1987-January 29,1988 (logarithms in terms of dollars) Full period Country United States Germany Japan United Kingdom Germany United States 1.00 0.93* 0.22* 0.90* 1.00 0.10 0.86* Japan United Kingdom 1.00 0.16 1.00 Before October 19 Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.75* 0.40* -0 .5 6 * 1.00 0.61* -0 .3 8 * 1.00 -0 .1 4 1.00 Japan United Kingdom After week of October 19 Country United States Germany United States Germany Japan United Kingdom 1.00 -0.01 0.22 0.56* 1.00 -0 .3 8 ’ -0 .3 2 * 1.00 0.52* 1.00 ’ Significantly different from zero at 5 percent level. dexes. If the changes across stock markets were unrelated, the probability o f observing 36 coinci dent changes or more w ould be much less than 1 percent. This suggests that it is likely that changes in the indexes are related. Correlations Among the Levels o f Stock Prices There does not appear to be a stable relation ship among the levels o f stock prices (table II. Except for Japan, the evidence for the w hole pe riod suggests that the indexes are highly corre lated. If one examines the correlations o f the levels o f stock prices before and after the crash, however, the correlations change dramatically. For example, the correlation o f the U.K. index with the U.S. in dex is about 0.90 for the w hole period. Before the crash, however, the correlation is —0.56, while, after the crash, it is 0.56. Conversely, the correla ,7The test essentially consists of implementing the Dickey-Fuller test (1979) on the ratio of stock price indexes. The reported tratios are those on the lagged level of the ratio in the relevant equation. All equations include a constant term and one lagged value of the dependent variable. The critical values for the test are from Fuller (1976), p. 373. tions for the U.S. and German stock indexes are 0.93 for the w hole period, 0.75 before the crash and —0.01 afterwards. This instability is precisely what one would expect if the relative stock price indexes are ran dom walks with no long-run relationships be tween their levels. The negative correlation be tween the index for the United States and the United Kingdom before October 19, though, is not what w ould be expected if stock prices around the w orld were "ratcheting upward” before the crash. Tests fo r Unit Roots W e can test whether, as equation 5 implies, the relative stock price indexes have unit roots.1 7 l est statistics to determine whether the levels of the relative stock indexes have unit roots are pre sented in table 2. Tw o periods are analyzed: one uses data from the full period; the other examines An alternative interpretation of this test in terms of cointegra tion as defined by Granger (1986) and discussed by Engle and Granger (1987). Under this interpretation, we are testing whether two stock price indexes are cointegrated with a coeffi cient of unity in the equation relating the two indexes. NOVEMBER/DECEMBER 1988 10 Table 2 Test Statistics for Unit Roots in Relative Stock Price Indexes: July 1 ,1987-January 29,1988 (logarithms denominated in dollars) Full period Country Germany Japan United Kingdom United States Germany Japan -1 .6 7 -0 .6 5 0.53 -2 .1 6 -0 .4 3 -1 .0 6 Before October 19 Country Germany Japan United Kingdom United States Germany Japan -1 .0 3 -0 .3 2 -1 .8 6 -0 .4 0 -1 .5 7 -1 .6 1 'Significantly different from zero at 5 percent level. the relationship before the crash. A t-ratio less than about —2.89 is inconsistent with the hypoth esis that the levels o f two series have a unit root. The test statistics in table 2 are well above the 5 percent critical value, consistent with the hypoth esis that all o f the different relative stock indexes have unit roots. These results provide no reason to expect that, given an increase in the U.S. index, for example, the Japanese index also will rise, or fall. That is, there is no “normal” level o f these indexes relative to each other. This is especially important because, in contrast to the conclusion of the Brady Commission, it is inconsistent with the notion that the markets rose as one during 1987 before the crash. Furthermore, it indicates that using the levels o f the stock market indexes to judge whether there is any relationship between the markets is fallacious. Correlations o f Changes o f the Indexes The evidence indicates that there is no reliable relationship among the levels of the indexes. Sim ple correlations o f changes in daily stock prices can be used to measure the extent o f the associa tion between the rate o f increases in the indexes 1 The monthly data are from the International Financial Statistics 8 (IFS) data tape of the International Monetary Fund. The U.S. data are the monthly averages of the daily close of 400 Stan dard and Poor’s industrials on the NYSE, the figures for Ger many are the averages of daily quotations covering 95 percent of common shares of industrial companies headquartered in Germany, the Japanese data are the averages of daily closing FEDERAL RESERVE BANK OF ST. LOUIS (table 3). For the w hole period, the correlations among the changes in the U.S. index and those o f the other countries range from 0.64 for Japan to 0.32 for Germany. The correlations among the indexes for Germany, Japan and the United King dom range from 0.56 to 0.15. At the 5 percent mar ginal significance level, all but the Japan/United Kingdom correlation are different from zero. These correlations are, on average, noticeably low er when the week o f the crash in prices is ex cluded from the correlations. Correlations without the data for the week o f October 19 are presented in the low er part o f table 3. All but two are low er than those for the whole period. The only higher correlation for a subperiod is the correlation be tween changes in the Japanese and German stock indexes, a correlation o f 0.22 excluding the week of the crash and 0.21 for the w hole period. These results are consistent with the notion that m ove ments in the indexes, unlike levels o f the indexes, are indeed related. Summary o f the Short-Term Results The daily data for the period around the Octo ber 1987 crash provide little support for the notion o f prices ratcheting up or down together. Rather, they indicate that there is no constant relationship between the levels o f the indexes. There is, how ever, a positive relationship among changes in the indexes, a finding consistent with the view that either financial transactions or international trade o f goods and services affect the different indexes in the same direction. STOCK PRICE INDEXES SINCE 1957: A LONGER-RUN VIEW Investigating the link between stock markets using monthly data spanning the past 31 years provides a useful perspective on the preceding results. Chart 3 shows monthly av erage indexes of industrial share prices for each o f the four coun tries for 1957 through 1987. All stock price indexes are denominated in terms o f U.S. dollars.1 Al 8 though changes in stock prices like those in Octo ber have been quite rare during the past few de- prices for all shares traded on the first section of the Tokyo exchange and the U.K. data are the average of daily quotations of 500 industrial ordinary shares on the International Stock Exchange in London. The exchange rates used to convert the stock indexes into dollars are the monthly average rates from the IFS data tape. 11 Table 3 Correlations of Changes of Stock Price Indexes: July 1 ,1987-January 29,1988 (changes of logarithms in terms of dollars) Full period Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.32* 0.64* 0.46* 1.00 0.21* 0.56* 1.00 0.15 1.00 Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.21* 0.28* 0.46* 1.00 0.22* 0.36* 1.00 0.12 1.00 Deleting week of October 19 ‘ Significantly different from zero at 5 percent level. Chart 3 Stock Price Indexes (monthly averages, 1980 = 100) in terms of dollars United States Ratio Scale Japan United Kingdom Ratio Scale West Germany I I I l I I I NOVEMBER/DECEMBER 1988 12 cades, substantial decreases in the market indexes are not uncommon. For example, stock prices in the United States decreased by relatively large amounts in several months: the index decreased 12.0 percent in June 1962, 11.6 percent in May 1970 and 10.5 percent in September 1974. The decrease in October 1987, on a monthly average basis, was 13.3 percent. Large single-month in creases are not exactly unknown either: the index for the United States increased 12.1 percent as recently as September 1982. It also is interesting to note from chart 3 that stock price decreases in the different markets often coincide. From 1957 through 1987, stock prices declined in all four o f the markets in 31 months. Coincident increases occur more fre quently during the sample: all four stock price indexes increased in 79 months. How many of these w ould be expected by pure chance? The av erage proportion o f months with an increase is about two-thirds for each countiy. If two-thirds is the probability of an increase, the joint probability that all o f the indexes w ould increase in anv month is 19.75 percent. Given our sample of 371 months, this means that about 73 months o f coin cident increases are expected. Since one-third is the average proportion o f declines for each coun try, the expected number o f coincident decreases is about five. Because the sample contains 110 months o f coincident changes w hile only 73 w ould be expected by chance, this is longer-term evidence that coincident changes in the indexes occur more often than would be expected bv chance. Relationship Betw een the Levels o f Stock Prices Is there a long-term relationship between the levels o f stock prices during the past 31 years? To answer this, monthly data are used to test for unit roots in the relative stock price indexes. The results o f these tests are presented in table 4.1 The : 1 top panel o f the table reports the relevant test statistics for the full period. The evidence indi cates that the relative stock price indexes have unit roots. With 371 monthly changes, a t-ratio less than about -2.88 would be inconsistent with the null hypothesis o f a unit root at the 5 percent significance level. The t-ratios generally are greater than the critical value, an outcome inconsistent 1 ln order to allow for the first-order serial correlation in the 9 indexes due to the Working (1961) effect, we include one lagged change of the relative index in the regressions. We make no adjustment to the critical value for this estimated coefficient. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Table 4 Test Statistics for Unit Roots in Relative Stock Price Indexes: January 1957-December 1987 (logarithms in terms of common currency) Full period t-ratios Country Germany Japan United Kingdom United States Germany Japan -2 .9 5 * 0.73 -0 .5 1 -2 .4 2 -3 .2 7 * -0 .2 8 Flexible-rate period t-ratios Country Germany United States Germany Japan -1 .8 7 Japan 0.89 0.47 United Kingdom -1 .5 8 -1 .5 9 -1 .3 0 'Significantly different from zero at 5 percent level. with the existence o f any normal long-run level of these indexes relative to each other. The test sta tistics for Germany relative to the United States and for the United Kingdom relative to Germany are, however, less than the critical value. Unlike the others, these results are consistent with the notion that these indexes tend to some normal level. The results o f the unit root tests from the flexible-rate period, a period characterized by greater financial integration across national bor ders than the fixed-rate period, uniformly are greater than the critical value.-0Test statistics us ing the data from the flexible-rate period are pre sented in the bottom panel o f table 4. These results indicate that in eveiy instance the relative stock price indexes have a unit root. The empirical evidence from the flexible-rate period clearly is inconsistent with the notion that the lev els of stock market indexes are linked across countries over long-run periods. Correlations o f Changes o f the Indexes Correlations o f the changes in the logarithm o f the monthly stock price indexes are reported in 2 The beginning of the flexible-rate period is defined as April 0 1973. 13 Table 5 Correlations of Changes in Stock Price Indexes: January 1957-December 1987 (logarithms in terms of dollars) Full period Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.38* 0.31* 0.50* 1.00 0.31* 0.33* 1.00 0.33* 1.00 Fixed-rate period Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.35* 0.20* 0.38* 1.00 0.16* 0.26* 1.00 0.17* 1.00 Flexible-rate period Country United States Germany Japan United Kingdom United States Germany Japan United Kingdom 1.00 0.42* 0.39* 0.56* 1.00 0.48* 0.41* 1.00 0.42* 1.00 "Significantly different from zero at 5 percent level. table 5. Because the sample period incorporates both the fixed- and flexible-exchange-rate regimes, the correlations are calculated for the full 31 years and for each o f the two exchange-rate regimes. The full period correlations are relatively high across markets, and all are statistically significant. All but one o f the correlations is between 0.31 and 0.38. The outlier is the higher correlation o f 0.50 between the United States and United Kingdom. The evidence from the fixed-rate period presents a rather different picture. Although the correlation between changes in the German and U.S. indexes is about the same as the correlation for the whole period, the other correlations are much smaller. For example, the correlations be tween the stock price indexes for Germany and Japan (0.16) and between Japan and the United Kingdom (0.17) are about one-half the size of 2 Controls on financial transactions were not suddenly axed with 1 the breakdown of fixed exchange rates; instead, they have been lifted gradually with each passing year. This suggests that, if changes in these restrictions account for at least part of the increases in these correlations, the correlations should be even larger for a period beginning later than 1973. Correlations their correlations for the full period. There also is a noticeably low er correlation between the U.S. and Japanese indexes, 0.31 for the full period and only 0.20 for the fixed-rate period. The evidence from the flexible-rate period sug gests that the relationship between U.S. stock prices and the foreign markets is somewhat closer relative to the fixed-rate period. The largest in creases in the correlation are between the Japa nese and the other indexes, and the largest of these changes is between the German and Japa nese indexes, which increases from 0.16 during the fixed-rate period to 0.48 during the flexiblerate period. The correlation between stock price changes in Japan and those in the United States and the United Kingdom also increases substan tially, from 0.20 to 0.39, and 0.17 to 0.42, respec tively. This suggests that the markets are more integrated in the latter half o f the period.-1 for 1980 through 1987 provide a tentative way of examining this issue. These correlations provide modest support for this hypothesis, with two of the correlations greater for the more recent period relative to the results for the flexible-rate period in table 5. NOVEMBER/DECEMBER 1988 14 CONCLUSION Cowan, Alison Leigh. "Global-Market Notion is Dispelled in Quarter,” New York Times, April 4,1988, p. D8. Are stock markets linked across countries? The levels o f stock price indexes in different markets need not move closely together; indeed, they do not. Daily data for three months before and after the October 1987 crash and monthly data for the past 31 years show no evidence that the levels of indexes for the United States, Japan, Germany and the United Kingdom are related. This means that the levels o f indexes show no tendency to return to anv particular value relative to each other. Thus, using different levels o f indexes in various coun tries as evidence o f a link or lack thereof between the markets is unfounded. Dickey, David A., and Wayne A. Fuller. “ Distribution of the Estimators for Autoregressive Time Series with a Unit Root,” Journal of the American Statistical Association (June 1979), pp. 427-31. The changes in the stock price indexes, at least in the four markets that we examine, generally do move together. The tightness o f these links, while real, is not exceptional. For example, the correla tion o f monthly changes in stock prices in the United States and the United Kingdom is about 0.56 based on data since the beginning o f flexible exchange rates. While significantly different from zero, this correlation also is quite far from one. Engle, Robert F., and C. W. J. Granger. “ Co-integration and Error Correction: Representation, Estimation, and Testing,” Econometrica (March 1987), pp. 251-76. Fuller, Wayne A. Introduction to Statistical Time Series (John Wiley and Sons, 1976). Granger, C. W. J. “ Developments in the Study of Cointegrated Economic Variables," Oxford Bulletin of Economics and Statis tics (August 1986), pp. 213-28. Malkiel, Burton G. A Random Walk Down Wall Street, 4th ed. (W. W. Norton & Company, 1985). Muth, John. "Rational Expectations and the Theory of Price Movements,” Econometrica (July 1961), pp. 315-35. Presidential Task Force on Market Mechanisms. ary 1988). Report (Janu Roll, Richard W. “The International Crash of October 1987,” in Robert Kamphuis, Roger Kormendi, J. W. Henry Watson, eds., Black Monday and the Future of Financial Markets (Mid America Institute, October 1988), pp. 35-70. Santoni, G. J. "The Great Bull Markets 1924-29 and 1982-87: Speculative Bubbles or Economic Fundamentals?" this Review (November 1987), pp. 16-30. Bryant, Ralph C. International Financial Intermediation (Brook ings Institution, 1987). Santoni, G. J., and Gerald P. Dwyer, Jr. “ Bubbles or Funda mentals: Some Evidence from the Great Bull Markets of 1924-29 and 1982-87,” paper presented at the Conference on Crashes and Panics in Historical Perspective, Salomon Brothers Center for the Study of Financial Institutions, New York University, October 1988. Cho, D. Chinhyung, Cheol S. Eun, and Lemma W. Senbet. “International Arbitrage Pricing Theory: An Empirical Investi gation,” Journal of Finance (June 1986), pp. 313-29. Solnik, Bruno H. “ The International Pricing of Risk: An Empiri cal Investigation of the World Capital Market Structure,” Journal of Finance (May 1974), pp. 365-95. Cooper, Richard N. “ The United States as an Open Economy," in R. W. Hafer, ed., How Open is the U.S. Economy? (Lexing ton Books, 1986), pp. 3-24. Working, Holbrook. “ Note on the Correlation of First Differ ences of Averages in a Random Chain,” Econometrica (Octo ber 1960), pp. 916-18. REFERENCES FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 15 Michael T. Belongia Michael T. Belongia is a research officer at the Federal Reserve Bank of St. Louis. Anne M. Grubish provided research assistance. Are Economic Forecasts by Government Agencies Biased? Accurate? “The CBO’s analyses and forecasts, while far from flawless, have come to be viewed as the best objective evidence that economists can muster. In stark contrast, everyone knows that Executive Branch estimates pass through many political filters." Alan S. Blinder, Business Week E, I CONOMIC forecasts bv government agencies often are tainted by allegations o f political parti sanship. Forecasts by the Council o f Economic Advisers (CEA) for example, which represent the expected impacts o f the President’s econom ic policies, have been characterized as “rosy sce narios,” that are too optimistic about the pros pects for strong real growth and low er unem ploy ment. In recent years, even White House insiders have alleged that the CEA’s numbers were “ cooked” to portray favorable econom ic outcomes.' Congress has its own econom ic agency, the Congressional Budget Office (CBO), that also pro duces forecasts for real growth, unemployment and inflation on a timetable similar to that o f the CEA. In contrast to the CEA, the CBO forecasts have been w idely regarded as being accurate and objective. Still, they too have been criticized as 1 The “ rosy scenario” characterization of Reagan administration economic forecasts has been attributed to Stockman (1986). Smith (1988) reports comments from a number of observers who feel the CEA forecasts are biased. As a technical matter, it is more accurate to talk about “ Administration” forecasts in stead of “ CEA” forecasts during the Reagan years. During this biased or inaccurate, especially when the CEA and CBO outlooks have differed substantially.2 With several U.S. government agencies making econom ic forecasts and allegations being raised about the relative merits o f these alternative fore casts, a number o f obvious questions arise. The purpose o f this paper is to determine first w hether econom ic forecasts made by the CEA and CBO have been biased. Then, because allegations of bias carry the implication o f inaccuracy, the fore casts also are evaluated on this basis. Finally, to provide some apolitical benchmarks, the forecasts of several well-known private sector gr oups are examined for bias and accuracy. CEA AND CBO FORECASTS: A BRIEF HISTORY The Council o f Economic Advisers was estab lished by the Employment Act o f 1946. The Eco- period, the “troika” process involving the CEA, the Treasury Department, and the Office of Management and Budget (OMB) produced a consensus forecast not associated with the CEA independently. 2 See, for example, Meiselman and Roberts (1979). NOVEMBER/DECEMBER 1988 16 Chart 1 CEA and CBO Real GNP Forecasts vs Actual GNP Growth Percent Annual Data Percent -2 1988 nom ic Report o f the President, which the CEA publishes annually, includes a short essay bv the President and a much longer report by the CEA staff; typically, it also includes an econom ic fore cast for the year ahead. The forecasts in the Report can be regarded as true e,x; ante predictions be cause they are released in late January or Febru ary, w ell before any official econom ic data for the calendar year are reported. The Congressional Budget Office was estab lished in 1974 as part o f the new budget process created by the Congressional Budget and Im poundment Control Act. The CBO was established to provide Congress "with detailed budget infor mation and studies o f the budget impact o f alter native policies.” 1The CBO was created primarily to provide budget analyses and econom ic forecasts that are independent from those o f the CEA and Office o f Management and Budget (OMB), both of which the President and Executive Branch con 3U.S. Congress (1976), p. 1. For a detailed review of the CBO's creation and stated mission, see Meiselman and Roberts (1979) and the comments by De Leeuw, Phaup and Rivlin that follow their article. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis trol. The CBO's forecasts are reported in its E co nom ic and Rudget Outlook (or E con om ic Outlook in earlier years), which is released early in the calendar year. Annual CEA and CBO forecasts for real GNP growth, the inflation rate and the level o f unem ployment are plotted in charts 1-3 for the period 1976-87. GNP and inflation values are fourthquarter-over-fourth-quarter rates o f change. The unemployment rate showTi is the fourth quarter level. Unemployment rate forecasts are generally the fourth-quarter level but, for the last five years o f the CBO forecasts, the predictions represent the annual average unemployment rate. Although the CEA has made econom ic forecasts since the late 1940s, the data plotted in the charts begin in 1976 for two reasons. First, the CBO’s initial forecast was for the year 1976; thus, direct comparisons between the two series are limited to the post-1976 period. Second, before the early 17 Chart 2 CEA and CBO Inflation Forecasts vs Actual GNP Inflation ' Percent Annual Data 1 i---2 Percent 12 2 1976 2 77 78 79 80 1988 Li Inflation rate is measured by the GNP deflator. 1970s, the CEA forecasts often w ere couched in qualitative terms (for example, “low er inflation” or "slightly faster growth"), which cannot be ana lyzed statistically.4 An inspection o f the charts indicates that both sets o f forecasts generally move in the same direc tion; the correspondence seems particularly close for the inflation forecasts. The GNP and unem ploy ment forecasts, however, show some interesting variations. Since 1981, the CEA’s forecasts o f real growth have been generally higher than the CBO’s. For the whole period, CEA unemployment rate forecasts have been low er than the CBO’s. These figures indicate that the CEA typically has forecast stronger real econom ic activity than the CBO. Whether these forecast differences represent a systematic bias of significant magnitude, bv either the CEA or the CBO, requires further analysis.1 4Moore (1977) has constructed a CEA forecast series for the years 1962-76 based on inferences from the text of the Eco nomic Report of the President. See footnote 8 for further discus sion of these earlier forecasts. 5 Carlson (1982) also has evaluated CEA forecasts and, in the context of a monetarist model, found them to be internally inconsistent. STATISTICAL ASSESSMENT OF FORECAST BIAS Figure 1 shows a conceptual framework with which to assess the relationships that might occur if the actual values o f a specific series were plotted against the values that had been predicted. If the forecasts were perfect — that is, if the forecast errors at each point in time w ere zero — a line relating the actual to the forecast values w ould have an intercept o f zero and a slope o f one; this line, denoted LPF in the figure, is what M incer and Zarnowitz (1969) call the “line o f perfect forecasts.” Bias in a forecast m erely indicates that the mean value o f the actual series (A) is not equal to the mean o f the forecast series (P) and, therefore, that the point E, determined by the ordered pair lA,P), w ill not be on the LPF line.6The extent o f bias in 6ln more technical terms, the mathematical expectation of the actual series, E(A), is not equal to the expectation of the fore cast series, E(P). See, for example, Mincer and Zarnowitz (1969), pp. 6-9. Webb (1987) provides further discussion of what is and is not learned from tests for bias. NOVEMBER/DECEMBER 1988 18 Chart 3 CEA and CBO Unemployment Forecasts vs Actual Unemployment Percent 11 Annual Data Percent 11 5 1988 the forecast is depicted in figure 1 as the distance between a point on the LPF line and point E, which is defined by the means o f the actual and forecast series. In view o f this discussion, a standard test for bias in a forecast can be constructed by estimating a regression o f the form: (1) Y, = a + b E(Y.) + e„ where Y, is the actual value o f a variable in period t, E(Y,) is its “predicted” or “forecast" value and e, is the forecast error (actual minus predicted value).7If the forecast is unbiased, the regression's intercept, a, should not be significantly different from zero and its slope coefficient, b, should not be significantly different from one; recall that these values for a and b define the LPF line in 7 Some research in this line of work has asked which measure of the “ actual” value should be used: the first-announced (unre vised) figure or the final (revised) value? Throughout, the final, revised figure is used. This choice is defended, logically, on the basis that this value, in fact, is what people are trying to fore cast, even though it includes such unknowns as seasonal adjustments and benchmark revisions. As a practical matter, FEDERAL RESERVE BANK OF ST. LOUIS figure 1. If a = 0 and b = 1 in equation \, the ac tual and forecast values w ill differ only by random error, as represented by e,. Moreover, the error w ould equal zero, on average, over long periods o f time. The results o f estimating regressions like equa tion 1 for the CEA and CBO forecasts o f real GNP growth (y), the inflation rate (P) and unem ploy ment rate (U) over the 1976-87 period are shown in table 1. The important results for current pur poses are the F-statistics corresponding to the null hypothesis that an equation’s intercept term is equal to zero and its slope coefficient is equal to one. This hypothesis is not rejected for any o f the six equations; none o f the F-statistics is greater than 0.5 and the 5 percent critical value is 4.10. Therefore, irrespective o f forecast accuracy and estimates of equation 1 with first-announced data had no qualitative effect on any of the results. McNees (1988) also has found that the choice of measure for actual values has little impact on annual forecasts, such as these, but apparently is important for quarterly forecasts. 19 F igure 1 T h e P redictio n-R ealizatio n Diagram Table 1 Bias Tests for CEA and CBO Forecasts Realizations F-statistic for H0: a = 0 and b = 1 CEA y = 0.168 + 0.836 E(y) (0.12) (0.44) R2 = 0.27 0.27 P = -0 .8 9 9 + 1.083 E(P) (0.62) (0.37) R2 = 0.67 0.46 U = 1.035 + 0.856 E(U) (0.55) (0.57) R2 = 0.49 0.16 R2 = 0.20 0.26 P = -0 .7 0 9 + 1.054 E(P) (0.58) (0.28) R2 = 0.73 0.49 U = -0 .2 1 9 + 1.024 E(U) (0.14) (0.12) R2 = 0.68 0.02 CBO y = 0.825 + 0.727 E(y) (0.61) (0.72) i — Mean realization P — Mean prediction Ec — Corrected mean point despite assertions to the contrary, both the CEA and CBO forecasts can be called unbiased.8 ARE PRIVATE SECTOR FORECASTS BIASED? The results in table 1 indicate that the forecasts o f two government agencies are unbiased. Un biasedness, however, is not unambiguously desir able if some bias is associated with greater forecast accuracy. Zellner (1986), for example, shows that a biased forecast is the optimal predictor under certain circumstances; M incer and Zarnowitz (1969) also noted this characteristic. Still, many observers associate bias with inaccuracy in a fore cast. Is it possible instead that some other fore casts are biased, but more accurate than those o f the CEA and CBO? As a first step to investigate this possibility, the mean forecasts o f a panel surveyed by the Ameri can Statistical Association and National Bureau of 8lt also is possible to evaluate CEA performance over a longer period. Moore (1977) has constructed a CEA forecast series back to 1962 for output and inflation, inferring quantitative estimates from the qualitative forecasts presented. Although this series is subject to error from such judgmental adjust ments, the longer sample increases the power of the test statistics. The results of estimating equations such as equation 1 over the longer sample indicate bias in the inflation equation as the intercept is significantly greater than zero; this result NOTE: Absolute values of t-statistics are in parentheses. For the slope coefficients, the reported t-statistic applies to the null hypothesis b = 1. The 0.05 percent critical value for a t-statistic (two-tailed test) with 10 degrees of freedom is 2.63. The 0.05 critical value for an F-statistic with 2 and 10 degrees of freedom is 4.10. Economic Research (ASA/NBER) and the forecasts from the large econom etric models o f two wellknown consulting firms w ere evaluated by the same tests described earlier.” To make these tests comparable with those already reported in table 1 data w ere examined during the same 1976— 87 interval for the forecasts published closest to the release dates o f the CEA and CBO predictions; plots o f actual vs. forecast values are shown in charts 4— The bias tests for the private sector 6. forecasts are reported in table 2. The results do not indicate bias in the ASA/NBER forecasts for any of the three variables examined. Moreover, explan atory pow er generally is higher for these equations than for the comparable CEA or CBO equations. The forecasts from the two consulting firms also exhibit no bias in any equation and generally have explanatory pow er comparable to that o f the ASA/ NBER forecasts. Overall, the results in tables 1 and suggests that the CEA, on average, has been overly optimistic about future inflation. No bias is evident in the GNP equation. 9Stephen McNees kindly provided these data. A condition for their use, however, was that the specific firms remain anony mous. Historical data on the economic forecasts of many alternative forecasters also was available (until 1986) in the Federal Reserve Bank of Richmond’s Business Outlook. NOVEMBER/DECEMBER 1988 20 Chart 4 ASA/NBER, Firm A and Firm B Real GNP Forecasts vs Actual GNP Growth 976 77 78 79 80 81 82 83 84 85 86 87 1988 Chart 5 ASA/NBER, Firm A and Firm B Inflation Forecasts vs Actual Inflation a Percent Annual Data Percent ------- 12 rm B 2 1976 2 77 78 79 80 LI Inflation rate is measured by the G NP deflator. FEDERAL RESERVE BANK OF ST. LOUIS 81 1988 21 Chart 6 ASA/NBER, Firm A and Firm B Unemployment Forecasts vs Actual Unemployment Percent Annual Data Percent 1988 2 indicate that all five widely-cited forecasts of aggregate econom ic activity are unbiased. FORECAST ACCURACY One way to compare the accuracy o f alternative forecasts has been proposed by Fair and Shiller (1988). The test is perform ed by estimating a re gression of the form: 1 ) Y, - Y,_„ = a + b [,_„E(Ylt) —Y,_J 2 + c[,_sE( Y,,) —Y,_J + (j l , where Y, and Y,_s are the actual values o f the vari able being forecasted in periods t and t —s, respec tively, while ,_,E(Ylt) and ,_s E(Y,,) are the predic tions o f forecasters #1 and #2 at time t —s for the value o f Y in period t. In this analysis, which uses annual data and one-year-ahead forecasts, s is equal to one. If the predictions o f either forecaster embodies information beyond the estimate o f the '“That a simple extraction of past trends might be considered an alternative to “ expert” forecasts has been suggested by analy ses of forecast performance. Meltzer (1987a,b), for example, has shown that Federal Reserve forecasts were so imprecise that, predicting one quarter into the future, it was impossible to distinguish statistically between a forecast of strong real growth one-period change represented bv the regression’s intercept term, a, then one or both slope coef ficients, b and c in equation 2, should be signifi cantly different from zero. If CEA is forecaster #1 and b is significantly different from zero but c is not, one concludes that the CEA forecast contains useful information and forecast #2 has no infor mation not contained in the CEA forecast. Finding c, but not b, significant w ould cariy the opposite conclusion. Finding neither b nor c significant indicates that neither forecast has useful informa tion beyond that contained in the intercept, which is interpreted as the average s-period change in y io These tests were perform ed for all pairs o f the CEA, CBO, ASA/NBER, Firm A and Firm B forecasts o f output, inflation and unemployment. As table 3 reports, a direct comparison o f the CEA and CBO forecasts shows neither agency adds new informa tion to the other’s forecast o f real GNP growth, and recession. Another interesting result is reported by Strongin and Binkley (1988), who find that forecasts made later in the year and incorporating more information than initial forecasts were as likely to deteriorate as to improve. NOVEMBER/DECEMBER 1988 22 Table 2 Table 3 Simple Bias Tests for Private Sector Forecasts Summary of Results from Fair-Shiller Tests F-statistic for H0: a = 0and b = 1 ASA/NBER y = 0.068 + 0.999 * E(y) (0.06) (0.00) R2 = 0.41 0.01 'E(P) R2 = 0.80 0.81 + 1.010* E(U) (0.04) R2 = 0.57 0.00 y = - 1.017 + 1.156 *E(y) (0.97) (0.59) R2 = 0.62 0.65 P = -0 .3 3 5 + 1.013 * Ef!-*) (0.24) (0.06) R2 = 0.64 0.16 R2 = 0.54 Forecast pair Real GNP Inflation Neither CEA — CBO Neither CEA — ASA/NBER Neither Both CEA — Firm A Both Neither CEA — Firm B Firm B Neither CBO — ASA/NBER ASA/NBER Neither Neither CBO — Firm A Firm A Firm B Neither CBO — Firm B ASA/NBER — Firm A Firm A ASA/NBER ASA/NBER — Firm B Neither Neither Firm A — Firm B Neither Neither Unemployment rate Neither Neither Neither Firm B ASA/NBER Neither Firm B Neither Neither Neither 0.47 P = -1 .2 9 2 + 1.166 (1.16) (0.94) U = -0 .0 8 7 (0.05) Consulting Firm A U = 1.095 + 0.832 (0.64) (0.75) * E(U) pears to offer additional information to the CEA's forecasts. Consulting Firm B R2 = 0.54 0.24 * E(P) R2 = 0.74 0.19 = 0.047 + 0.968 * E(U) R2 = 0.64 0.34 y = 0.467 + 0.958 * E(y) (0.54) (0.17) P = 0.083 + 0.951 (0.08) U (0.03) NOTE: Forecaster name listed under the “ Real GNP,” "Inflation” or Unemployment rate” columns indicates it adds significant information not embodied in the other forecast. (0.30) (0.15) NOTE: Absolute values of t-statistics are in parentheses. For the slope coefficients, the reported t-statistic applies to the null hypothesis b, = 1. inflation or unemployment. Although this result is not surprising in view o f the very similar regres sion results reported in table 1, it also indicates there is no evidence to distinguish the forecasts o f either agency as a better source o f information. When CEA forecasts are evaluated against the three private sector forecasts, a different picture emerges. W hen evaluated against the ASA/NBER survey, each institution’s forecast for inflation adds to the information contained in the other and in the regression’s intercept. Neither o f their output or unemployment forecasts, however, adds to the information contained in the other. This evidence suggests that inflation forecasts can be im proved bv combining the information in the CEA and ASA/NBER forecasts. The comparisons with the two private sector firms, w hile offering a similarly mixed bag o f results, generally indicate that, for real GNP and unemployment, Firm B ap- FEDERAL RESERVE BANK OF ST. LOUIS For the CBO, the results suggest that any o f the three private sector alternatives adds information to CBO’s output forecast; two o f the three also add information to the CBO’s unem ploym ent forecast. For inflation, however, there appear to be few gains from looking at the alternative forecasts. Among the three private sector firms, none o f the results shows one to be superior to the others. Overall, the results in table 3 generally indicate that the private sector forecasts add to the infor mation in the CBO forecasts while, aside from Firm B’s contributions, the CEA and private fore casts contain similar information. CONCLUSIONS Members o f both political parties sometimes allege that econom ic forecasts by government agencies are biased. An examination o f this issue indicates that neither the CEA nor CBO forecasts exhibit any discernable bias. An evaluation o f three private sector forecasts also detected no forecast bias. Absence o f bias, however, does not necessar ily indicate that a forecast is better (specifically, more accurate). When three private sector fore casts w ere com pared with CEA and CBO forecasts, however, the private sector forecasts generally w ere more accurate than those o f the CBO; the CEA fared less w ell only relative to the forecasts of private sector Firm B. 23 REFERENCES Blinder, Alan S. “ Dithering on Hill is Crippling a Key Agency,” Business Week (September 26, 1988), p. 25. Carlson, Keith. “ A Monetary Analysis of the Administration's Budget and Economic Projections,” this Review (May 1982), pp. 3-14. Fair, Ray C., and Robert J. Shiller. “The Informational Content of Ex Ante Forecasts,” Working Paper No. 2503 (National Bureau of Economic Research, February 1988). McNees, Stephen K. “ How Accurate Are Macroeconomic Forecasts?,” New England Economic Review (July/August 1988), pp. 15-36. Meiselman, David, and Paul Craig Roberts. "The Political Economy of the Congressional Budget Office,” in Karl Brunner and Allan H. Meltzer, eds., Three Aspects of Policymaking: Knowledge, Data and Institutions, Carnegie-Rochester Conference Series on Public Policy, Vol. 10, (1979), pp. 283-333. Meltzer, Allan H. "Limits of Short-Run Stabilization Policy,” Economic Inquiry (January 1987a), pp. 1-14. _________ “ On Monetary Stability and Monetary Reform,” Bank of Japan Monetary and Economic Studies (September 1987b), pp. 13-34. Mincer, Jacob, and Victor Zarnowitz. “The Evaluation of Economic Forecasts,” in Jacob Mincer, ed., Economic Forecasts and Expectations, (National Bureau of Economic Research and Columbia University Press, 1969). Moore, Geoffrey H. “The President’s Economic Report: A Forecasting Record," NBER Reporter (April 1977), pp. 4-12. Smith, Hedrick. Stockman, David. Inc., 1986). The Power Game (Random House, 1988). The Triumph of Politics (Harper and Row, Strongin, Steven, and Paula S. Binkley. "A Policymaker’s Guide to Economic Forecasts,” Federal Reserve Bank of Chicago Economic Perspectives (May/June 1988), pp. 3-10. U.S. Congress. The Congressional Budget Office: Responsibilities and Organization (U.S. Government Printing Office, 1976). Webb, Roy H. "The Irrelevance of Tests for Bias in Series of Macroeconomic Forecasts,” Federal Reserve Bank of Richmond Economic Review (November/December 1987), pp. 3-9. Zellner, Arnold. “ Biased Predictors, Rationality and the Evaluation of Forecasts,” Economics Letters (No. 1, 1986), pp. 45-48. NOVEMBER/DECEMBER 1988 24 John A. Tatom John A. Tatom is an assistant vice president at the Federal Re sen/e Bank of St. Louis. Anne M. Grubish and Kevin L. Kliesen provided research assistance. The Link Between the Value of the Dollar, U.S. Trade and Manufacturing Output: Some Recent Evidence O BSERVERS w idely believ e that the decline in the dollar’s value against foreign currencies, which began in 1985, has boosted U.S. manufacturing output significantly. The dollar’s decline was ex pected to raise the dollar prices o f U.S. imports w hile lowering the foreign-currencv prices of U.S. exports; in response, the quantity dem anded of both U.S. exports and import-competing goods w ould rise. This demand-based linkage has been central to analyses o f both the international and domestic econom ic prospects o f this nation since early in this decade. The emphasis on an inverse relation ship between U.S. output and the value o f the dol lar became preeminent from 1980 to early 1985, when the dollar’s value rose dramatically and when a historically large trade deficit emerged.' 1 Usually this view is part of a broader analysis of economic policy; see, for example, Meyer (1986), Jonas (1986), Business Week (1987), Business Week (1988), Peterson (1987) and Summers (1987). Feldstein (for example, 1987 and 1988) has been the most vocal proponent of this view. Examples focusing on the dollar-production linkage include Berry (1986), Deutsch (1988), Revzin (1988) and Hickok, Bell and Ceglowski (1988). FEDERAL RESERVE BANK OF ST. LOUIS The relevance o f this inverse relationship, how ever, rests on a faulty assumption. Im plicit in such analyses is the view that changes in the value of the dollar are independent o f U.S. industrial devel opments rather than being caused by those very changes. When econom ic policy boosts or retards the productive capacity of the economy, the dom i nant relationship between the value o f the dollar and domestic manufacturing output should be a positive one, so that a rise lor fall! in the value o f the dollar is associated with a rise (or fall) in U.S. output. This article examines w hether the U.S. produc tion o f manufactured goods in recent years has shown an inverse relationship to movements in the v alue o f the dollar. Its principal focus is w hether the industries that are most closely asso- According to Murray (1988), former Federal Reserve Board Chairman Paul Volcker has endorsed the conventional view stating: “ We had a great consumer boom in imports that left manufacturing undernourished. Now manufacturing can carry us for the next four years.” 25 Chart 1 Index of the Nominal and Real Trade-Weighted Dollar Exchange Rate INDEX (MARCH 1973=100) 150 ANNUAL AVERAGE OF MONTHLY DATA INDEX (MARCH 1973=100) 150 70 1 1 70 1973 74 75 76 77 78 79 80 81 82 83 84 85 86 87 1988 SOURCE: FEDERAL RESERVE BOARD NOTE: The Real Exchange Rate Measure uses Trade-Weighted Consumer Price Indices. 1988 Data based on average of six months data. ciated with the growing trade deficit during the period o f the rising dollar also exhibited declining output and, in turn, w hether their output has been boosted by the general decline in the dollar’s value since then. THE EXCHANGE RATE, PRODUCTION AND TRADE The dollar’s value rose sharply from 1980 to 1985 and subsequently declined; chart 1 shows this movement for measures o f both the nominal and real exchange value o f the dollar. The real ex change rate is the nominal exchange rate (E) m ul tiplied by the ratio o f the U.S. price level (P,,s) to the foreign price level IP,,).- This rate differs from the nominal exchange rate in that it excludes m ove ments that are attributable to changes in domestic price levels like Pls or PK For example, suppose . that U.S. prices rise X percent while prices abroad 2 The Federal Reserve Board measures presented in chart 1 use a weighted-average of foreign exchange rates and price levels to construct E and PF respectively; the weights are based on , shares of trade with the United States. A host of different are unchanged. The nominal exchange value o f the dollar w ould have to fall X percent for the dol lar prices o f both U.S. and foreign goods to rise by X percent and for the foreign prices o f both U.S. and foreign goods to be unchanged. When price levels and exchange rates change for such purely monetary reasons, the real exchange rate and decisions about production, consumption and trade are unaffected. Both measures in chart 1 rose sharply from 1980 to 1985, then declined roughly as much as they had risen, however. (Ref erences to exchange rate changes below are to both nominal and real changes unless indicated otherwise.) The Conventional Analysis o f the Effects o f an Exchange Rate Change A conventional analysis o f the effects o f a change in the dollar’s value on domestic produc tion and trade is shown in figure 1. The supply measures have been developed with differing weights, baskets of currencies and price indexes, but none shows a different pattern for our purposes. NOVEMBER/DECEMBER 1988 26 and the difference — the quantity im ported — increases. Conversely, a fall in the value o f the dollar is expected to raise the dollar prices o f goods that are traded internationally, providing an incentive to raise domestic production and reduce domestic purchases; in this case, the quantity of traded goods exported w ill rise and the quantity imported will fall. Figure 1 A Rise in the Value of the Dollar Reduces Domestic Output Price of Traded Good ($ per unit) Xi Xq Quantity per Period curve (S) shows that the domestic quantity sup plied o f product (X) w ill increase if the domestic price o f the product rises, given the other factors influencing the position o f the curve. The demand curve (D) indicates the quantity dem anded do mestically at various domestic prices, given the other factors that influence its position. The world price (PJ and the domestic price are equal when measured in dollars per unit o f the product; the w orld price equals the price in a foreign countiy measured in its currency units (P*) divided by the foreign currency price of the dollar (E). The differ ence between the quantities supplied and d e manded domestically is the quantity either ex ported (when the quantity supplied exceeds that demanded) or im ported (when the quantity de manded exceeds that supplied). As drawn in figure 1, the U.S. exports product X at the initial price P ;;,. If the value o f the dollar (E) rises and the foreign price o f the good remains the same, the w orld price expressed in dollars (P*/E) is reduced; in figure 1, the price falls to P'v. As a result, the do mestic quantity supplied w ould fall and the d o mestic quantity dem anded w ould rise, reducing the quantity o f X exported. For an im ported good, the analysis is the same: when the dollar price falls, U.S. consumption rises, U.S. production falls 3Factors reflecting overall price levels both in the United States and abroad are held constant in figure 1. A given supply curve for X assumes that the dollar factor cost of resources used to produce product X are fixed, suggesting that the U.S. general level of prices is held constant; the local currency price of the FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis From the foreign perspective, the figure and results are the reverse. Thus, when the value o f the dollar rises, the w orld price measured in foreign currency rises, inducing foreigners to produce more traded goods but consume less o f them. Thus, a rise in the value o f the dollar is expected to redistribute the production o f internationally traded goods from the United States to foreign producers. Conversely, when the value o f the dol lar falls, the U.S. output o f these goods is boosted, while foreign output declines. An Alternative View o f Recent Exchange Rate Movements While there is nothing inherently w rong with the conventional analysis above, its relevance to observed exchange rate movements is question able. In the analysis in figure 1, the change in the value o f the dollar is “ exogenous,” or external to the domestic factors that influence the supply and demand curves. The dollar’s value, however, is determined in currency markets in w hich the demand for dollars in foreign exchange is m oti vated by factors influencing foreign dem and for U.S. exports and assets, w hile the supply o f dollars in foreign exchange is motivated by U.S. decisions to purchase foreign goods or assets. If changes in incentives in asset markets raise (lower) the rela tive attractiveness o f investment in the United States and raise (lower) the external value o f the dollar, the exchange rate change can only be exog enous to a U.S. market for a good if the domestic supply and demand for that product are unaf fected.3 A shift in the relative attractiveness o f U.S. vs. foreign investment that directly affects asset mar kets but not goods markets is impossible. After all, the opportunity cost o f em ploying capital in pro duction is influenced by expected rates o f return both at home and abroad. The typical rationale for product abroad also is held constant, suggesting that the price level abroad is unchanged. Thus, the change in the external value of the dollar represents a “ real” exchange rate change. 27 ignoring these related effects in the goods market is that, in the short run at least, changes in the capital stock (plant and equipment) are small rela tive to the existing stock o f such goods. Thus, the short-run output and productivity effects are also presumed to be relatively small. This simplifica tion is most questionable when the very purpose of policy actions that give rise to such a shift in investment incentives is to raise productivity and output. In the early 1980s, the real rate o f return on in vestment in the United States was raised by tax policy actions, especially those that extended the investment tax credit and accelerated deprecia tion allowances. In effect, these changes lowered the cost o f capital to U.S. firms and raised the real rates o f return that these firms were willing to bid to maintain and acquire new equity and debt financing. W hile tax changes provided a positive incentive for firms to expand capacity and output domestically, the higher rates o f return generally discouraged output and investment in activities without these new tax breaks, especially such activities abroad.4 Proponents o f the conventional view described above neglect these tax-policy-induced changes; indeed, they focus only on the supposed budgetdeficit-driven rise in (real) interest rates and its consequent effects on investment and the value o f the dollar. From an alternative supply-side view, “The hypothesis described is elaborated more fully in Tatom (1985) and (1986a). This hypothesis is not widely endorsed. Recent papers by Mutti and Grubert (1988) and, especially, Sinn (1988) address the influence of tax policy changes on international capital and trade flows and the value of the dollar; see also the comments by Gravelle (1988) and Patrick (1988). Ott (1984) and Fazzari (1987) also describe the 1981 and 1986 tax law changes for capital income and their effects. Ohmae (1988) argues that the link between the value of the dollar and U.S. competitiveness has been the opposite of that typically put forward in the popular and academic press. Boskin and Gale (1986) provide evidence on firm mobility that is consistent with Ohmae's view. Poole (1988) indicates that the 1981 tax act was the primary real disturbance in this decade and that it raised the real after-tax rate of return on investment. He also indicates the minority status of this view, however. Stockman and Svensson (1987) provide a formal model in which changes in wealth and its distribution can give rise to capital flows, current account movements and exchange rate changes that simultaneously match those described here. 5Krugman and Baldwin (1987) emphasize the importance of relative productivity developments as the factor accounting for the dollar’s decline and the growth in the trade deficit after early 1985, but do not address the possible connections of the exchange rate, trade and relative productivity developments in the 1980-85 period. however, the rise in the dollar’s value was pro duced by the same policy actions that also pro duced an increase in the supply of domestic out put."’ Similarly, econom ic policy changes that reverse investment incentives and have adverse output supply effects w ill both low er the value o f the dol lar and reduce the supply o f domestic output. From the outset o f the discussion o f tax reform in late 1985, it was clear that earlier incentives to invest, notably the investment tax credit (ITC) and accelerated depreciation for structures, w ould soon end. These changes were incorporated in the Tax Reform Act o f 1986 (and made retroactive to the beginning o f 1986 in the case o f the ITC). In response, domestic investment plum m eted from late-1985 to m id-igs?.6These tax changes should be expected to reduce both the supply o f dom es tic output and the exchange rate.7 Output and The Exchange Rate When Domestic Supply Shifts Figure 2 shows a shift in the domestic supply of product X from S to S'; such a shift can arise be cause o f a reduction in factor costs. For large countries like the United States, the increase in domestic supply w ill have an appreciable effect on the w orld supply as well: the price o f product X w ill fall as domestic output (XJ and total w orld output rise. corporate income (taxed at 51 percent) realized through re tained earnings and capital gains (taxed at 20 percent) to 59.1 percent in 1988 for corporate income (taxed at 39 percent) realized through capital gains or dividend income (taxed at 33 percent). This 1988 tax rate change excludes the end of the ITC and reductions in service lives for depreciation that further raised effective marginal tax rates, but includes a 5 percentage-point surcharge for individuals and corporations that phases out at sufficiently high incomes. From 1985 to 1987, corporate tax accruals (excluding the Federal Reserve) rose from $58.5 billion to $88.1 billion, a 50.6 percent increase. As a result, real nonresidential fixed invest ment fell from a peak of 12.6 percent of real GNP at the end of 1985 to 11.1 percent in 1/1987. This decline as a share of real GNP is the equivalent of about a $56.7 billion reduction, or 13.6 percent of investment spending, in 1/1987 alone. Canto (1988) also emphasizes the strong connection between changes in the exchange rate and tax rates, but only for personal tax rates! His explanation relies on an almost inconsequential reduction in personal tax rates in 1981 and has difficulty accounting for post-1984 exchange rate and investment developments. T here are likely other factors that could account for the decline in the dollar’s value, but the hypothesis here, explained in Tatom (1987), is that post-1984 policy developments reflect a reversal of earlier policies. 6 The maximum marginal tax rate on personal and corporate income declined only slightly, from 60.8 percent in 1985 for NOVEMBER/DECEMBER 1988 28 the price reduction abroad w ill ensure that pro duction abroad declines.9 Figure 2 An Increase in Supply Raises The Quantity Produced x0 x2 When the supply of all products in a country changes, the analysis is more complex. For exam ple, the monetary approach to the balance o f pay ments indicates that a general rise in U.S. output will low er the U.S. price level and raise the nom i nal exchange value o f the dollar. In this approach, the real exchange rate need not change, despite the increase in domestic production.1 This ap 0 proach typically does not take into account inter national capital mobility; thus, it does not em pha size the importance o f capital flows between countries as the principal factor influencing re cent exchange rate movements. X Per Period When the supply o f product X increases, its price w ill tend to fall in both domestic and foreign IP’ ) currency units to induce domestic and foreign purchasers to buy more o f it. Thus, given the value of the dollar (El, the w orld price falls to P„' and purchases o f product X rise, both in the United States and abroad. Production rises only in the United States, however. Foreign production of product X falls because its price declines and the foreign supply curve remains unchanged. For eigners w ould also consume more o f product X, so they w ould increase their imports from the United States.8 Similarly, for a good that the United States im ports, an increase in the U.S. supply o f an importcom peting good w ill raise its w orld supply and reduce its price. Just as for goods that the United States exports, the price decline will raise pur chases at home and abroad. The domestic supply increase ensures that domestic output rises, while 8 The productivity increase also explains why employment can decline despite the boost to output. Fieleke (1985) makes a similar argument about the relationship of net imports of an industry’s product and employment in that industry. He pro vides evidence showing that net import movements were uncorrelated with industry employment, which is consistent with the argument below. McKenzie (1988) has shown that productivity advances, not imports, have been the major factor behind employment losses in the textile industry. This view is explained more generally in Tatom (1986b). 9Alternatively, given P* in the analysis above, a decline in P due to an increase in domestic supply requires that E rise. Of course, the rise in the world supply of traded goods will reduce the world price of such a good measured in any currency, so FEDERAL RESERVE BANK OF ST. LOUIS In addition, real incomes will not remain con stant for such generalized output changes. The ensuing rise in U.S. incom e w ill also raise U.S. de mand for goods and services. The U.S. demand curve D in figure 2 w ill shift to the right, mitigating but normally not offsetting part o f the rise in the excess supply shown there. M ore importantly, however, the supply and dem and for product X, or products generally, w ill tend to fall abroad. U.S. policy actions that raise the real after-tax rate of return and shift domestic supply rightward from S to S', also raise the cost o f capital abroad and shift the foreign supply curve for output leftward, re ducing foreign output, incom e and demand. A decline in foreign incom e reduces foreign demand for goods and sendees, including those im ported from the United States. The effects on trade flows are ambiguous when both supply and dem and change. As long as the dominant domestic effect o f policies that raise (lower) the after-tax rate o f return in the United States is to raise (lower) the U.S. supply o f traded goods output and low er (raise) foreign dem and for traded goods, the trade flows predicted in the conventional analysis also are predicted in the supply-side analysis. That is, a rise (fall) in the that P* must decline as well. Thus, the share of domestic producers in world production will rise because of increased domestic production and reduced foreign production. >°A second approach based on the flow supply and demand for dollars, emphasizes the fall in import prices and quantities as a source of a reduced supply of dollars in international exchange and, under standard assumptions, a rise in nominal exports as a source of a rise in the demand for dollars in international exchange. The value of the dollar would rise for both reasons, although the major factor affecting the exchange rate in either view is international capital flows. 29 value o f the dollar will be associated with a rise (fall) in the quantity o f imports and a fall (rise) in the quantity o f exports. The central difference, and the focus here, is on w hether the rise o f U.S. im ports and the fall o f exports were indicators o f a “ deindustrializing” econom y or “h ollow ed'’ corpo rations, or instead were a symptom o f a redistribu tion of capital, productivity and incom e toward the United States. In the supply-side view, U.S. goods that formerly w ould have been exported are purchased at home and not abroad where incom e reductions have reduced demand; goods that formerly w ould have been produced and consumed abroad face a larger demand in the United States and a smaller demand abroad." While these outcomes are not inevitable for every traded commodity, the analy sis suggests that the conventional result — that domestic production o f exported and im ported goods varies inversely with the value o f the dollar — is a partial analysis less likely to hold if ex change rate movements arise from forces that also change domestic supply. Figure 2 illustrates h ow domestic output in creases can accompany an exchange rate appreci ation. Increases in the supply o f U.S. output gener ally w ill raise domestic output, reduce the U.S. price level and raise the nominal exchange rate. The result is a positive relationship between the exchange rate and output, contrary to the conven tional relationship. Figure 2 also challenges the notion that a rise in the value o f the dollar neces sarily redistributes production, including that of U.S. export and im port-com peting goods, away from the United States and toward our foreign competitors. These implications are examined below.1 - "Krugman and Obstfeld (1988) explain how a transfer of income from the rest of the world to the United States causes the changes in demand and trade described here. They also ex plain that such a transfer raises the demand for U.S. goods relative to those produced abroad so that the terms of trade, the price of exports relative to imports, will rise. They apply their analysis to the recent flow of financial capital, instead of an increase in current and future U.S. income. U.S. MANUFACTURING OUTPUT: A REVIEW OF SOME AGGREGATE EVIDENCE The key difference between the hypotheses above concerns the relationship between the exchange rate and domestic output. In the con ventional view, this relationship is negative; a supply-side perspective emphasizes that it can be positive. The difference centers on whether ex change rate changes are exogenous or w hether they reflect changes in domestic productivity and output. One source o f evidence on this issue is the share o f domestic manufacturing output in U.S. real GNP. Additional evidence is the experience abroad. Under the conventional view, when the value o f the dollar rose, U.S. producers should have lost market share to foreign producers as output of U.S. manufactured goods fell and foreign output rose. If, instead, the rise in the value o f the dollar reflected a decline in domestic production cost in the United States and a rise abroad — as the supply side view suggests — precisely the opposite should occur.1 Thus, examining the per : 1 formances o f U.S. manufacturing output relative to other major industrial countries is also relevant for distinguishing between these two explana tions. The Manufacturing Share o f U.S. Output The actual and cyclically adjusted shares of manufacturing output in real GNP are shown in chart 2 for the period 1/1948 to 11/1988. The actual manufacturing share is an important, but easily misinterpreted, source o f evidence bearing on the anticipate productivity improvements that follow decisions to change investment and capacity. Tatom (1986a) also provides evidence on the reallocation of investment and productivity growth toward the United States in 1980-85. ,JThe two theoretical approaches to exchange rate changes touch on a multitude of economic factors besides production, both at home and abroad, including purchases, relative prices, price levels, nominal and real trade flows. The qualitative predictions of the two theoretical analyses are the same for most of these factors under fairly standard assumptions. The critical difference involves production, and that is the focus here. ,3Tatom (1986a and 1987) shows that changes in the exchange rate occur two quarters earlier than their positively related changes in domestic manufacturing output. Exchange markets NOVEMBER/DECEMBER 1988 30 Chart 2 U.S. Manufacturing Output as a Percent of Real GNP Percent 23 Percent 23 19 1948 51 54 57 60 63 66 69 72 75 Q Based on a constant 81.9 percent capacity utilization rate in manufacturing. 78 81 84 87 19 1990 com peting hypotheses here. Since 1983, this share has been above the 1948-79 average o f 21 percent, providing no evidence that manufacturing output had weakened when compared to its previous history, despite the rise in the value o f the dollar from 1980 to 1985.1 The manufacturing share also 4 rose moderately since the value o f the dollar be gan falling in 1985. The share does not take a no ticeable jump up because o f this decline, however, just as it does not appear to have been depressed earlier when the value o f the dollar rose. ficantly stronger than w ould have been expected based on the 1948-79 record. The actual share varies cyclically because the dem and for such output is strongly cyclical; for example, periodic sharp declines in the share coincide with U.S. recessions. Transitory or cyclical incom e losses in the 1980-85 period (as indicated by a relatively low capacity utilization rate or high unemployment rate for labor) w ere relatively large, so that the actual share w ould have been expected to be low er than it actually was. The potential misinterpretation o f the behavior o f the actual share arises from the above-average level since 1983, which suggests that its perfor mance in the 1980s has not been unusual. In fact, for most o f this decade, the actual share was signi The cyclically adjusted share accounts for these cyclical variations; it surged upward to a record level in mid-1981 (after the 1981 tax act was passed) and generally remained at a historically ,4Glick and Hutchison (1988) point to the actual share of manu facturing in output as evidence against the deindustrialization hypothesis, but express agreement with the existence of an inverse relationship between the dollar's value and domestic output. FEDERAL RESERVE BANK OF ST. LOUIS 31 high level until 1985-86, when it began to decline.” Relative to its past, the adjusted share appears unusually strong in 1981-85, when the dollar was rising. Moreover, it has not surged upward since the dollar began to fall in 1985; instead, it has weakened, especially in 1987. The decline in the adjusted share in 1987 indicates that the rise in the actual share was largely due to cyclical income changes, not the belated effect o f the fall in the value o f the dollar. The pattern shown by the ad justed share is strongly at odds with the main stream view, but is consistent with the supply-side stoiy. U.S. vs. Foreign Manufacturing Output Did the U.S. share o f the w orld ’s manufacturing output rise or fall from 1980 to 1985? The Organi zation o f Economic Cooperation and Development (OECDI prepares indexes for the manufacturing output o f its 24 m em ber countries, a group that includes Europe, the United Kingdom, Canada, Japan, Australia and N ew Zealand. These indexes can be used to compute an index o f manufactur ing output for the other 23 OECD nations. Chart 3 shows this index and the index for the United States since 1960. The gap between the indexes that opens up in the 1980s indicates the unusual strength o f U.S. manufacturing in this decade. According to the OECD, the growth rate o f U.S. manufacturing output, which constituted 31.7 percent o f total OECD output in 1980, grew at a 3.2 percent rate from 1980 to 1985 w hile the value of the dollar was rising. This growth was w ell above the 1.5 percent rate for the rest o f the OECD over the same period. Moreover, the U.S. growth rate in 1980-85 was up sharply from a 2 percent rate in 1973— while production gr owth in the remain 80, der o f OECD hardly rose at all from the dismal 1.3 percent rate registered in 1973— From 1960 to 80. 1 The adjusted share is computed using the departure of the 5 capacity utilization rate in manufacturing (which captures movements in manufacturing output common to all sectors and hence is a representative business cycle measure) from the 1948-88 average of 81.9 percent. A regression of the growth rate (change in the natural logarithm) of the share on a con stant, lagged share and current and several lagged changes in the logarithm of the capacity utilization rate indicates that the lagged share and lags on the capacity utilization rate are not statistically significant at a conventional (5 percent) significance level. The equation estimated from 111/1948 to 1/1988 has an insignificant intercept 0.02 percent (t = 0.37) and a coefficient of 0.61676 (t = 27.69) for the current change in the capacity utilization rate; the adjusted R2 is 0.80 the standard error is 0.83 percent, and the Durbin-Watson statistic is 2.02. The estimate includes a significant first-order autocorrelation ad justment with p equal to 0.21 (t = 2.66). Other methods that 1973, manufacturing production in the rest o f the OECD countries had grown at a 6.4 percent rate, outstripping the 5.3 percent rate in the United States. Thus, the share o f U.S. manufacturing in the total OECD output rose markedly while the dollar rose from 1980 to 1985 — from 31.7 percent in 1980 to about 33.5 percent in 1985 — contrary to the conventional wisdom. From 1985 to 1987, the relatively faster growth in the United States eroded, despite the overall U.S. cyclical expansion. The growth o f U.S. manufactur ing output, according to the OECD, remained at a 3.2 percent rate, w hile the growth o f industrial output in the other 23 countries accelerated to a 2 percent rate. Since U.S. growth still exceeded that abroad, the U.S. share o f OECD manufacturing output continued to rise slightly, but much more slowly; it reached about 34 percent in 1986 and 1987. Thus, the comparison o f U.S. and foreign output generally supports the supply-side view that the competitive position o f U.S. manufactur ers was boosted by econom ic policy changes in the early 1980s, which subsequently w ere re versed. Thus, an inverse relationship o f produc tion and the value o f the currency does not hold for the United States or the rest o f the OECD. DISAGGREGATED EVIDENCE ON PRODUCTION AND TRADE If U.S. manufacturing production was not d e pressed by the rise in the value o f the dollar, w hy did the trade deficit balloon from 1980 to 1985? The conventional explanation emphasizes that the rise in the value o f the dollar reduced domestic output, especially the output of exported and import-competing goods.1 The supply-side view, " on the other hand, indicates that an expanded trade deficit can accompany relatively strong d o mestic output growth if domestic productivity, output and incom e rise.1. Thus, a detailed exami7 use changes in the unemployment rate or real GNP growth result in the same pattern for the adjusted share. 1 There are microeconomic arguments that emphasize other 6 sources of reduced domestic output of imported and exported goods. See Arndt (1989), Arndt and Bouton (1987) and Hooper and Mann (1989), for example. "Wharton (1986) provides evidence supporting this view. Krugman and Baldwin (1987), however, dismiss the impor tance of relative income growth in accounting for the emer gence or elimination of the trade deficit. Their argument fo cuses on an asserted difficulty of raising U.S. export volumes. It ignores the associated and currently more relevant problem (in the sense that export volume was restored to its 1980 record level as a share of real GNP in late 1987). This problem is the failure of U.S. export and import prices to rise relative to the prices of non-traded goods and services. NOVEMBER/DECEMBER 1988 32 Chart 3 Manufacturing Production in the U.S. and Other OECD Countries Index 1980=100 Index 1980=100 130 1960 63 66 69 72 75 78 SOURCE: Organization for Economic Cooperation and Developm ent (OECD). nation o f the industries most closely associated with the record trade deficit will allow us to assess whether their experience provides support for the mainstream view, despite its failure to explain overall manufacturing performance. Identifying the Deficit Industries As the value of the dollar rose from 1980 to 1985, the U.S. merchandise trade deficit also rose, climb ing from $24.2 billion to $132.5 billion1 This rise 8 was concentrated in manufacturing, where an initial trade surplus o f $20.7 billion fell to a $104.3 billion deficit. Table 1 shows the latter change and a breakdown for the two-digit standard industrial classification o f 20 industries that make up the manufacturing sector. The changes from 1980-85 and 1985-87 are indicated for each industiy. The l8Nominal trade data are used here because it is precisely the nominal deficit that is the trade-area focus of popular and macroeconomic policy discussions. Real trade data by industry are unavailable, but the output measure relevant to the hypoth eses in the text is domestic output, where data are available. The domestic industry deflators indicate that nominal prices for the deficit-related industries below have barely changed over FEDERAL RESERVE BANK OF ST. LOUIS three largest changes in trade deficits from 1980 to 1985 are in transportation equipment, non-electric m achineiy and electrical equipment. These ac count for more than half o f the total and include the largest net exporting sector in 1980, the non electrical machinery industiy. The next largest swings are in apparel products and primary metal products. These five industries account for twothirds o f the swing in the manufacturing deficit, and they are the five principal deficit-related in dustries. The changes in these industries are fol low ed by relatively small swings toward deficit in 14 o f the remaining 15 industries. Only the to bacco industry m oved toward a greater surplus (or a smaller deficit) over the period. The table also shows that the fall in the dollar's value from 1985 to 1987 was not accompanied by a the seven years; for the five-industry aggregate, nominal prices rose about 2 percent from 1980 to 1985 and fell 4.2 percent from 1985 to 1987. Changes in the real trade deficits in manu facturing and in this group, computed using domestic prices, are nearly identical to those for the nominal trade deficit. 33 Table 1 U.S. Trade Surplus in Selected Industries: 1980,1985,1987 (billions of dollars) Change 1985-87 Change 1980-85 1987 Agriculture $19.6 $ $ - 12.3 $ Mining -6 2 .4 32.6 29.8 28.6 4.0 Manufacturing Food and kindred products Tobacco manufactures Textile mill products Apparel products Lumber and wood products 20.7 1.7 1.0 0.5 - 4.9 0.1 -1 0 4 .3 2.5 1.2 2.2 - 14.7 2.4 -1 2 5 .0 4.2 0.2 2.7 9.8 2.5 -1 3 2 .7 0.9 2.2 2.8 - 20.0 2.0 -2 8 .4 1.6 1.0 - 0.6 - 5.3 0.4 - - - 2.1 2.9 0.4 5.1 2.2 - 4.0 3.8 0.1 11.7 8.6 - - 2.7 3.6 0.1 9.0 12.8 1.3 0.2 0.2 2.7 4.2 1.9 4.3 2.2 9.2 4.5 - 2.6 9.2 3.5 13.4 3.4 - 0.6 2.0 1.0 2.3 1.4 ; - 7.3 Furniture and fixtures Paper and allied products Printing and publishing Chemical products Petroleum and coal products 0.6 0.7 0.5 14.1 -1 0 .6 Rubber and plastic products Leather products Stone, clay and glass Primary metal products Fabricated metal products - 0.1 2.9 0.3 6.5 2.5 - 2.0 7.2 2.5 15.7 2.0 - Machinery, except electrical Electrical equipment Transportation equipment Instruments Miscellaneous 23.9 0.9 - 0.7 3.0 - 2.5 - 6.2 19.0 27.9 0.2 9.4 - 17.7 - 19.9 - 27.2 3.2 6.9 - 3.3 3.8 - 22.2 - 42.0 1.4 - 11.9 I 1985 - O •S t 1980 Industry -1 0 .0 - 3.2 -14 .1 - 1.2 - 2.5 NOTE: Deficits are negative entries. SOURCES: Data for 1980 and 1985 are from Arndt (1989) and Bureau of the Census, U.S. Department of Commerce, Statistical Abstract of the United States, 1988; 1987 data are in Bureau of the Census, Foreign Trade Division, Highlights of U.S. Export and Import Trade, FT 990, December 1987. decline in the trade deficit, the deficit in manufac turing, or the deficit for the five deficit-related industries. A decline in the deficit in mining was offset bv a decline in the surplus for agriculture. The manufacturing trade deficit grew by $28.4 billion over the period. Only sLx industries showed positive movements in their trade surplus, and this group included only one o f the deficit-related industries, primary metals. Positive changes also were recorded for chemicals, tobacco, food, lum ber and petroleum. For the other four major de ficit industries, the total deficit rose $32.6 billion: when primary metals are included, the trade de ficit o f the five principal deficit-related industries rose $30.3 billion, slightly more than the $28.4 billion increase in the total manufacturing deficit. Thus, these five industries account for all of the 1985-87 rise in the manufacturing trade deficit. Comparative Output Performance fo r the Deficit-Related Industries The top panel o f table 2 shows the growth rates in manufacturing output for the five deficit-related industries and the other 15 industries for the pe riod o f the rising value o f the dollar, the period of the falling value o f the dollar, and the earlier seven-year period that is roughly a comparable cycle-peak-to-cvcle-peak period for the United States. Over this earlier period, the value o f the dollar declined somewhat (chart 1). The data in the table show that the five deficit-related indus tries boom ed during the period o f the rising dol lar; indeed, they w ere the sectors that pushed the overall manufacturing growth rate up to a 3.4 per cent rate. The other 15 industries, as a group, showed much less acceleration in output growth NOVEMBER/DECEMBER 1988 34 Table 2 U.S. Output and Productivity Growth Rates (compounded annual rates of change)___________________ 1973-80 1980-851985-87 Output Growth Rates Manufacturing 1.0% Five deficit-related industries' 1.0 Other 15 manufacturing industries 1.0 Real GNP 2.2 3.4% 4.9 2.2 2.6 3.3% 3.6 3.1 3.1 Productivity Growth Rates2 Manufacturing 0.9 Five deficit-related industries 0.7 Other 15 manufacturing industries 1.1 Real GNP 0.0 4.4 6.3 2.9 1.1 3.8 5.3 2.7 0.7 'Includes primary metals, nonelectric machinery, electric equipment, transportation equipment, and apparel and other textile products, 2The output growth rate minus the employment growth rate; civilian employment is used for the real GNP productivity measure. SOURCE: National Income and Product Accounts, Table 6.2, U.S. Department of Commerce. value o f the dollar declined, but in a direction opposite to that predicted by the mainstream view. The further increase in the trade deficit was associated with a switch to slower domestic pro duction growth, both overall and in the five princi pal industries.| Manufacturing output growth 1 J slowed slightly, led by a substantial slowing in output growth in the five deficit-related indus tries.2 This reduction in output growth o f the five " deficit-related industries, both absolutely and relative to the other 15 industries, is inconsistent with the conventional view, but is consistent with the view that earlier incentives for domestic pro ductivity growth had been reduced. The bottom panel o f table 2 shows labor pro ductivity growth, measured by the difference be tween output and em ployment growth rates, for the five industries and other manufacturing firms. The sharp acceleration in productivity in manu facturing in 1980-85, led by the five deficit-related industries, clearly stands out, as does the relative decline for these same industries since 1985. in 1980-85, compared with the 1973-80 period. The five-industiy group includes the steel and automobile industries which are often viewed as mature or declining, but it also includes the com puter industry where rapid growth has led the expansion o f the non-electric machinery industry and o f overall manufacturing in this decade. The comparisons made here using the data in table 2 are not reversed bv omitting non-electric machin ery from the measures, however. The evidence in table 2 confirms and strength ens the aggregate evidence. The aggregate data are not obscuring a negative relationship between the value o f the dollar and output in the deficit-related industries. In fact, the positive relationship is even more apparent for the five industries.-1The results are strongly at odds with the view that the expan sion in the trade deficit in 1980-85 came at the expense o f domestic production. Instead, declin ing net exports reflected increased domestic pur chases that outstripped the relatively rapid growth of domestic production.- The top panel o f table 2 also shows that the pattern of production changed in 1985— as the 87 Moreover, as developm ents since 1985 suggest, the declining dollar and the nascent rev ersal in 1 The decline in output growth is much more pronounced when 9 non-electric machinery is omitted from the five-industry and manufacturing measures. The growth rate for the four-industry total slowed from 2.6 percent in 1980-85 to zero in 1985-87, leading the decline in the growth rate of manufacturing which fell from a 2.3 percent to a 2.0 percent rate for the same peri ods. In 1973-80, the four-industry growth rate was 0.2 percent, and manufacturing less non-electric machinery grew at a 0.7 percent rate. 2,The 1981 tax act generally provided a subsidy to structures and to equipment that increased with its durability; these subsi dies were reversed in the Tax Reform Act of 1986. If the five deficit-related industries are relatively more equipment intensive in production, then their supply is relatively more affected by the changed taxation of capital income. Also, the rise in the real after-tax rate of return domestically will raise the cost of capital abroad, changing investment patterns so as to reduce foreign productivity growth relative to what it would otherwise have been, reinforcing the positive relationships abroad. “ The cyclically adjusted output growth rate for the five deficitrelated industries rose 2.5 percentage points from 1973-80 to 1980-85; this increase is statistically significant, t = 2.47, according to a pooled t-test. The adjusted growth rate fell by a statistically significant 2.9 percentage points in 1985-87 from its 1980-85 rate (t = -5 .5 0 ). The cyclically adjusted growth rate is found using the regression of the actual growth rate of the five industries’ output for the period 1967-80 on the real GNP growth rate. When the rate of change of the real ex change rate is regressed on the cyclically adjusted growth rate for 1967 to 1987, its coefficient is positive, 0.088, but not sig nificantly so at conventional levels, t = 1.49. FEDERAL RESERVE BANK OF ST. LOUIS 2 When domestic industry price deflators are used to adjust 2 nominal trade deficits, the resulting real net imports can be added to real output to obtain real domestic purchases. This procedure indicates a 9.4 percent annual rate of growth of real purchases in the five deficit-related industries from 1980 to 1985 and a slowing to a 6.7 percent rate from 1985 to 1987. Comparable figures for real manufacturing purchases are 6.8 percent and 4.4 percent, respectively. For the other 15 indus tries, the figures are a 4.6 percent rate in 1980-85 and a 2.2 percent rate in 1985-87. 35 Table 3 U.S. Manufacturing Capacity Growth and Deficit-Related Industries Industry1 Manufacturing 3-industry 4-industry 5-industry 1973-80 3.4% 4.2 3.6 3.3 1980-85 2.9% 4.4 3.6 3.6 Change 1985-87 Change -0 .5 % 0.2 0.0 0.3 2.6% 2.4 1.8 2.3 -0 .3 % -2 .0 -1 .8 -1 .3 ’The three-industry group includes electrical equipment, nonelectrical machinery and transportation equipment; the four-industry group adds the primary metals industry; the five-industry group adds other nondurables. Other nondurables includes apparel and other textile products, tobacco products, printing and publishing, and leather products. SOURCE: Computed from data published by the Federal Reserve Board. the trade deficit look to be the results of policy actions that have reversed the earlier productivity boom in these key industries. Thus, bv reducing their competitiveness internationally, these policy actions have allowed weaker sectors in the United States and abroad to expand. On net, these changes w ill reallocate w orld consumption and production away from the United States. Were Current Production Changes at Odds with Longer-Term Output Plans? As noted earlier, manufacturing output is strongly influenced by cyclical factors. One way to avoid the influence o f such temporary factors at the industry level is to examine capacity output measures. The long-run choices o f capacity and its optimal output are based on expected prices and costs. The capacity choice is more forwardlooking and is based on more permanent consid erations than the current output choice. If a rise in the value o f the dollar w ill reduce the optimal domestic output o f an industry, then, regardless of current output developments, firms w ill cut back on the growth o f capacity.-3 facturing capacity slowed in 1980-85, then slowed further in 1985-87, two o f the three measures of capacity growth in deficit-related industries accel erated in 1980-85, then slowed in 1985-87.2 This is 4 precisely the same pattern follow ed by actual out put in table 2. The exception is the four-industry measure, where the decline in primary metals capacity growth held the group’s rate to the same pace in 1980-85 as in 1973-80. When the dollar was rising, capacity growth in the deficit-related industries, by all three mea sures, exceeded the overall average for manufac turing capacity growth and accelerated relative to the average for manufacturing. Thus, the share of manufacturing capacity in these industries was expanding and expanding faster than it had ear lier. For the five-industiy measure, capacity growth was slightly below the overall manufactur ing growth rate in 1973-80, but it jumped to about a 24 percent faster growth rate than in manufac turing in 1980-85. Table 3 shows the growth rates for manufactur ing capacity and several industry groupings o f the principal deficit-related industries over the same periods as in table 2. While the growth o f manu When the dollar fell from 1985-87, these devel opments w ere reversed. The capacity growth rates in the deficit-related industries declined and were slower than for manufacturing as a whole. The share o f capacity in the deficit-related industries began to decline slightly. This result is inconsis tent with the view that international com petitive ness in these industries had im proved since 1985. 2 The Federal Reserve Board compiles data on industrial capac 3 ity for the sectors in table 1, except that apparel products are lumped into “other nondurable manufacturing” which also includes tobacco products, printing and publishing, and leather and products. a 0.5 percent rate in 1985-87. The growth rate of manufactur ing capacity per worker, measured by the difference in the growth rate of capacity and employment, accelerated from 3.3 percent in 1973-80 to 3.9 percent in 1980-85, then declined to 3.1 percent in 1985-87. ^Manufacturing employment expanded at a 0.1 percent rate in 1973-80, then declined at a 1.0 percent rate in 1980-85 and at NOVEMBER/DECEMBER 1988 36 CONCLUSION The mainstream view that the dollar’s decline has im proved U.S. competitiveness is based on a partial and misleading econom ic analysis. This view mistakenly focuses on the effects o f exoge nous exchange rate movements on trade and out put. In a broader analysis, the exchange rate is determined precisely by those factors that drive econom ic competitiveness. Thus, a rise in the value o f the dollar can reflect an improvement in competitiveness, rather than a cause o f its decline. Similarly, a fall in the dollar’s value can reflect a decline in competitiveness; it is not necessarily a factor that w ill improve it. While there are firms and even industries within the groups examined here — for example, the primary metals sector — in which relative productivity changes have not been significant so that an inverse relationship between the dollar exchange rate and production and em ployment is observed, they are not typical. For the U.S. manufacturing sector as a whole and the industries most closely connected to the U.S. trade deficit, the relationship between movements in the value of the dollar and output growth dur ing this decade has been a positive one. In the early 1980s, U.S. manufacturing output, especially when adjusted for the effect of the U.S. business cycle, was unusually strong relative to both its own past experience and output growth abroad. The industries most closely related to a $125 billion surge in the manufacturing trade de ficit were the leading sectors in this strong growth; these industries showed a sharp acceleration in capacity growth over the same period that rein forced their growing dominance in econom ic per formance. From 1985 to 1987, these trends, like the value of the dollar, reversed. Only the trade deficit contin ued its previous pattern, growing somewhat larger over the period, and this increase was fully ac counted for by the same key industries. Over the period, the share o f manufacturing output in U.S. production, on a cyclically-adjusted basis, did not increase. Meanwhile, actual manufacturing output growth abroad accelerated both absolutely and relative to its counterpart in the United States. In the United States, at least, this pattern was dom i nated by the slowing o f output growth in the key deficit-related industries. From 1985 to 1987, ca pacity growth slowed in the deficit-related indus tries to a pace below that for manufacturing. A central lesson of this evidence is that the ef fects o f changes in the dollar exchange rate on FEDERAL RESERVE BANK OF ST. LOUIS domestic production, at least during the 1980s, are dominated by the effects o f the econom ic policy changes that also have produced the exchange rate movements. The evidence suggests that the increased U.S. manufacturing competitiveness produced bv econom ic policy changes in the early 1980s has been reduced by reversals o f some of these policies in the mid-1980s. REFERENCES Arndt, Sven W. "An Evaluation of Policies to Resolve the Trade Deficit," in Albert E. Burger, ed., The U.S. Trade Deficit: Causes, Consequences and Cures (Kluwer Academic Publishing Co., forthcoming 1989). Arndt, Sven W. and Lawrence Bouton. Competitiveness: The United States in World Trade (American Enterprise Institute for Public Policy Research, 1987). Berry, John M. "Narrower Trade Gap Will Require Cut in Spending," Washington Post, November 2, 1986. Boskin, Michael J., and William G. Gale. “ New Results on the Effects of Tax Policy on the International Location of Investment,” National Bureau of Economic Research Working Paper No. 1862 (March 1986). Business Week. pp. 56— 67. “America After Reagan” (February 1,1988), ________ . “ It’s Time for America to Wake Up” (November 16, 1987), pp. 158-77. Canto, Victor A. “Tax Rates Move Currencies, and There’s No Easy Fix,” Wall Street Journal, September 22, 1988. Deutsch, Claudia H. “ U.S. Industry’s Unfinished Struggle,” New York Times, February 21, 1988. Feldstein, Martin. “ Unwanted Advice From Abroad,” Wall Street Journal, March 3, 1988. ________ _ “ Correcting The Trade Deficit,” Foreign Affairs (Spring 1987), pp. 795-806. Fazzari, Steven M. “ Tax Reform and Investment: Blessing or Curse?" this Review (June/July, 1987), pp. 23-33. Fieleke, Norman S. “ The Foreign Trade Deficit and American Industry,” New England Economic Review (July/August 1985), pp. 43-52. Glick, Reuven, and Michael Hutchison. “The Dollar and Manufacturing Output,” Federal Reserve Bank of San Francisco Weekly Letter (August 26, 1988). Gravelle, Jane G. "Discussion of ‘U.S. Taxes and Trade Performance’,” National Tax Journal (September 1988), pp. 341-2. Hickok, Susan, Linda A. Bell, and Janet Ceglowski. “The Competitiveness of U.S. Manufactured Goods: Recent Changes and Prospects," Federal Reserve Bank of New York Quarterly Review (Spring 1988), pp. 7-22. Hooper, Peter H., and Catherine L. Mann. “ The U.S. External Deficit: Its Causes and Persistence,” in Albert E. Burger, ed., The U.S. Trade Deficit: Causes Consequences and Cures (Kluwer Academic Publishers, forthcoming). Jonas, Norman. “The Hollow Corporation,” Business Week (March 3, 1986), pp. 57-59. Krugman, Paul R., and Richard E. Baldwin. “ The Persistence of the U.S. Trade Deficit,” Brookings Papers on Economic Activity (1987:1), pp. 1-43. 37 Krugman, Paul R. and Maurice Obstfeld. International Economics (Scott, Foresman and Company, 1988). McKenzie, Richard B. “ Textile Gripes Are Made of Whole Cloth,” Wall Street Journal, April 8, 1988. Meyer, Stephen A. “Trade Deficits and the Dollar: A Macroeconomic Perspective,” Federal Reserve Bank of Philadelphia Business Review (September/October 1986), pp. 15-25. Murray, Alan. 11, 1988. "Improbable Growth," Wall Street Journal, July Mutti, John and Harry Grubert. “ U.S. Taxes and Trade Performance,” National Tax Journal (September 1988), pp. 317-25. Ohmae, Kenichi. “ No Manufacturing Exodus, No Great Comeback,” Wall Street Journal, April 25, 1988. Organization for Economic Cooperation and Development. Main Economic Indicators (June 1988). Ott, Mack. “ Depreciation, Inflation and Investment Incentives: The Effects of the Tax Acts of 1981 and 1982,” this Review (November 1984), pp. 17-30. Patrick, Robert J., Jr. "Comments On U.S. Tax Structures and Competitiveness,” National Tax Journal (September 1988), pp. 343-5. Peterson, Peter G. “The Morning After,” The Atlantic Monthly (October 1987), pp. 43-69. Poole, William. “ Monetary Policy Lessons of Recent Inflation and Disinflation,” Journal of Economic Perspectives (Summer 1988), pp. 73-100. Revzin, Philip. 17, 1988. “ Rude Awakening,” Wall Street Journal, March Sinn, Hans-Werner. “ U.S. Tax Reform 1981 and 1986: Impact on International Capital Flows,” National Tax Journal (September 1988), pp. 327-40. Stockman, Alan C., and L. E. O. Svensson. "Capital Flows, Investment, and Exchange Rates,” Journal of Monetary Economics (March 1987), pp. 171-201. Summers, Lawrence H. "The Results Are in on the Deficit Experiment,” Wall Street Journal, February 20,1987. Tatom, John A. “Will a Weaker Dollar Mean a Stronger Economy?” Journal of International Money and Finance (December 1987), pp. 433-47. ________ . “Why Has Manufacturing Employment Declined?" this Review (December 1986b), pp. 15-25. _________ “ Domestic vs. International Explanations of Recent U.S. Manufacturing Developments,” this Review (April 1986a), pp. 5-18. _________ “Two Views of the Effects of Government Budget Deficits in the 1980s,” this Review (October 1985), pp. 5-16. Wharton Economics Forecasting Associates. Outlook (November 1986). Quarterly Model NOVEMBER/DECEMBER 1988 38 Cletus C. Coughlin and Kenneth C. Carraro Cletus C. Coughlin is a senior economist and Kenneth C. Carraro is an economist at the Federal Reserve Bank of St. Louis. Thomas A. Pollmann provided research assistance. The authors also would like to thank Kenneth W. Bailey for providing data and helpful suggestions. The Dubious Success of Export Subsidies for Wheat O N MAY 15, 1985, the U.S. Department o f Agri culture introduced an export subsidy program called the Export Enhancement Program (EEP). The program's main goal is to increase U.S. agri cultural exports.' The program also is intended to induce European Community (EC) reductions in agricultural subsidies during the current round of multilateral trade negotiations under the General Agreement on Tariffs and Trade (GATT). Although the EEP focuses on exports and trade policies, it is a direct outgrowth o f the domestic farm policies o f the United States and the Euro pean Community. The above-market price guaran tees o f these policies have resulted in surplus production. To dispose o f these surpluses, many governments have chosen to subsidize agricultural exports.- 'U.S. Department of Agriculture (May 1988). 2 The General Agreement on Tariffs and Trade contains the international rules governing export subsidies. Article XVI:4 prohibits export subsidies on industrial products that lead to export sales at lower prices than domestic sales; however, this does not apply to agricultural goods. U.S. farm interests were sufficiently powerful to prevent the prohibition on export subsi dies from encompassing agricultural goods. Article XVI :3 recommends that export subsidies on agricultural goods be avoided, but, if they are used, the subsidizing country should not garner a “ more-than-equitable” share of trade for the good. While European farm interests prevented the extension of the FEDERAL RESERVE BANK OF ST. LOUIS This paper examines the EEP’s primary goal o f expanding exports in the context o f disposing of governm ent-owned wheat. Other research is used to compare the cost o f reducing wheat stocks via the EEP with the cost o f simply destroying the wheat. Although this paper focuses on wheat, chiefly because wheat has accounted for the bulk o f EEP activity, the econom ic principles used here can be generalized to similar programs for other commodities. The secondary goal o f influencing EC farm pol icy is analyzed in the fi-amework o f the current round o f GATT talks. In 1985, the United States proposed that all trade- and production-distorting subsidies be eliminated over a 10-year period.3 Other groups, including the EC and the Cairns export subsidy prohibition to agricultural goods in the Tokyo Round of Multilateral Trade Negotiations (1974-79), the mean ing of equitable was clarified. For example, a more-thanequitable share includes the displacement of another country’s exports. A reference period of the three most recent years in which normal market conditions existed is to be used in deter mining an equitable share. See Hufbauer and Erb (1984) for additional details. 3Roningen, Sullivan and Wainio (1987) estimate annual welfare gains for the United States of slightly less than $4 billion from a multilateral removal of these measures. 39 Group, have offered alternative proposals.4The prim aiy reason for changing the current agricul tural policies that benefit domestic farmers at the expense o f consumers, taxpayers and others is the cost of such programs. For example, the ministers o f the Organization for Economic Cooperation and Development recently stated: The cost of agricultural policies is considerable, for government budgets, for consumers and for the economy as a whole. Moreover, excessive support policies entail an increasing distortion of competi tion on world markets; run counter to the princi ple of comparative advantage which is at the root of international trade; and severely damage the situation of many developing countries.’ This paper examines EEP's role in encouraging successful negotiations to liberalize agricultural trade in the G A IT process. Before examining the issues o f export expansion and trade negotiations, w e describe the export subsidy programs o f the United States and Euro pean Community and, in the process, provide historical background necessary to understand the EEP's objectives. THE FARM POLICY ENVIRONMENT UNDERLYING THE EXPORT ENHANCEMENT PROGRAM The stage for this export subsidy program was set by steady losses in the share o f w orld agricul tural trade held by the United States and by paral lel EC gains in export shares. From 1977 to 1985, the U.S. share o f the w orld ’s net wheat exports declined from 41.9 percent to 28.8 percent, while the EC’s share rose from —1.6 percent to 15.1 per cent.'* These changing market shares can be linked to EC export subsidies. Chart 1 shows that the U.S. wheat export price generally exceeded the subsi dized EC export price between 1978 and 1987.7 Since 1983, the gap has tended to widen. EC ex port subsidies are responsible for this gap because internal EC prices are far above U.S. market prices. This EC subsidy is the catalyst for the U.S. export subsidy program, which is targeted at those coun “See Rossmiller (1988) for an outline of the major features of the proposals by the United States, the European Community, the Cairns Group, Japan and Canada. The Cairns Group consists of Canada, Australia, New Zealand, Indonesia, Malay sia, the Philippines, Thailand, Brazil, Uruguay, Argentina, Columbia, Hungary and Chile. tries where EC-subsidized exports have displaced U.S. exports. Surplus Production and Government Wheat Stocks To understand the EEP goal o f export expan sion, one must examine the U.S. farm programs used to support the production o f most crops. The most important consequence o f these pro grams is that they generate surpluses because price guarantees, with the exception o f the earlv 1970s, have been above market-clearing levels. There are two main instruments o f the crop pro grams: loan rates and target prices. Both are price guarantees that are announced well before farmers make planting decisions. To participate in these programs, farmers generally have been re quired to reduce crop acreage. For example, wheat farmers must set aside 10 percent o f their 1989 wheat acreage base to qualify for the wheat price support program. T he loan rate, set at $2.06 per bushel for the 1989 wheat crop, serves as a price floor. If the mar ket price is low er than $2.06, a farmer pledges the wheat crop to the government as collateral in ex change for a “loan” o f $2.06 per bushel. If the price o f wheat rises above the loan rate, the farmer can repay the loan with interest, recover the crop and sell at the higher market price. If the market price does not recover, the fanner defaults on the “loan,” thus ceding the crop to the government. Bv law, the government keeps the acquired surpluses off the market until the price reaches a higher level, known as the release price, at which time the surpluses can be sold on the market. While the loan rate acts as an explicit price sup port, the target price functions as an explicit in come support device and is the final price that farmers receive for their crop. At the end o f a crop year, farmers receive “ deficiency payments” equal to the difference between the target price and either the market price or the loan rate, depending EC was a net importer of wheat in that year. EC data have been calculated for the EC-12 and exclude intra-EC trade. T hese data are not adjusted for wheat quality and transporta tion differentials. Such differentials, however, are relatively constant and, therefore, do not distort significantly the rising price gap trend. 5 See Wilson (1988), p. 5. 6Net wheat exports represent exports minus imports. The nega tive market share figure for the EC in 1977 indicates that the NOVEMBER DECEMBER 1988 40 Chart 1 U.S. and EC W heat Price D iffe re n tia ld Dollars per ton 50 Dollars per ton 50 _ 20L-------------------------------------------------------------------------------------------------------------------------------------- 1-20 JU LY 78 JU LY 79 JULY 80 JULY 81 JULY 82 JULY 83 JULY 84 JU LY 85 JU LY 86 SOURCE: International Wheat Council □.(U.S. Gulf - EC Rouen fob Price) on which is higher." Table 1 shows the target, loan rate and market prices since the introduction of the target price mechanism in the crop year end ing June 30, 1975. For 1988, the target price w7 as $4.38 per bushel. Since the higher of the market price and loan rate was the market price o f $2.60 per bushel, the deficiency payment was $1.78 per bushel. Until 1988, the deficiency payment rose throughout the 1980s. As suggested above, as the market price declines relative to the loan rate, farmers find it more pro fitable to surrender their crops to the government for the loan rate price rather than sell on the mar ket. These surplus stocks are accumulated by the Com m odity Credit Corporation, the agency of the U.S. Department o f Argiculture (USDA) charged with the administration o f the price support pro 8ln some years, farmers have received a portion of this pay ment, an “ advance deficiency payment,” at the beginning of the crop year. FEDERAL RESERVE BANK OF ST. LOUIS grams. Chart 2 shows the inverse relationship between the accumulation o f wheat stocks and the price gap measured by the market price minus the loan rate. When the price gap increases be cause o f crop shortages or strong demand such as in the early 1970s, stocks are reduced. When the price gap narrows, and especially if the gap is negative, the accumulation o f stocks occurs. The large increase in wheat stocks in the 1980s reflects the relatively small price gap. Disposing o f Government Wheat Stocks Various programs are used to dispose o f the stocks that are ow ned by the government. Some of the surplus disposal is directed to the domestic market, but most is directed to foreign markets 41 Table 1 U.S. Wheat Prices (dollars per bushel) Year' Target price Loan rate Market2 price Deficiency payment 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 $2.05 2.05 2.29 2.90 3.40 3.40 3.63 3.81 4.05 4.30 4.38 4.38 4.38 4.38 4.10 $1.37 1.37 2.25 2.25 2.35 2.50 3.00 3.20 3.55 3.65 3.30 3.30 2.40 2.28 2.06 $4.09 3.56 2.73 2.33 2.97 3.80 3.99 3.69 3.45 3.51 3.39 3.08 2.42 2.60 NA $0.00 0.00 0.00 0.65 0.52 0.00 0.00 0.15 0.50 0.65 1.00 1.08 1.98 1.78 NA 'The year ending June 30 2 Average price received by farmers SOURCE: Wheat Situation and Outlook (February 1988), p. 20. Chart 2 Surplus Wheat Stocks and the Price Gap of the Wheat Market Price minus Wheat Loan Rate Dollars per bushel Millions of bushels 1500 1250 2.00 Price Gap 1965 67 69 71 73 75 77 Stocks = CCC Inventories and Farmer-Owned Reserves 1000 1987 NOVEMBER/DECEMBER 1988 42 Table 2 Export Subsidies by the United States Under the EEP and Export Restitutions by the European Community (millions of U.S. dollars) United States' Year 1982 1983 1984 1985 1986 1987 1988 Total EEP European Community2 Wheat EEP3 Total export refunds Wheat export refunds $209 701 744 $ 4,668 4,647 4,895 5,028 7,121 10,245 12,861 $550 720 388 365 681 NA NA $280 937 995 'Value of EEP bonus awards at market value. Data are for fiscal years. Fiscal year 1988 covers the period from October 1, 1987 to September 30,1988. 2EC refunds for calendar years. Data for 1988 are estimates. 3 Wheat subsidy values are estimated by taking wheat's share of the total bonus value, 74.8 percent, and applying this percent to total subsidy value. through export.1Obviously, the EEP belongs to the 1 latter category."1 More than 100 initiatives, targeting more than 60 countries and 11 comm odities have been an nounced under the EEP since June 1985. The EEP functions by giving governm ent-owned surplus comm odities at no cost to private U.S. exporters. This allows them to sell U.S. comm odities at prices that are below U.S. market prices in order to be competitive with other export-subsidizing coun tries. An EEP/sales initiative states the targeted country and the quantity of a specific com m odity to be sold. Knowing that a subsidy is available, private U.S. exporters can offer to sell the com m odity at prices below the market cost o f acquir ing it in the United States. These bids are contin gent upon receiving the necessary subsidy from the USDA. The foreign buyer may accept bids made by numerous U.S. exporters. The U.S. exporters then bid against each other to receive the USDA’s sur plus stocks as a payment for the export subsidy. During this process, each exporter states how 9 One of the most notable examples of domestic surplus dis posal was the 1983 Payment-in-Kind (PIK) program that gave surplus commodities to farmers who agreed to limit their pro duction. The school milk program, which sells milk at belowmarket prices, and programs to distribute other dairy products to food-stamp recipients are other domestic examples of sur plus disposal policy. FEDERAL RESERVE BANK OF ST. LOUIS large a subsidy is required to make the export sale. For example, if one exporter requests a subsidy of $30 per ton and another requests $35 per ton for sale o f the same com m odity to the same country, the USDA w ould award the subsidy payment to the low er bidder. Thus, the bid process helps the USDA get a larger volum e o f exports per dollar of subsidy. If the exporter s bid for the subsidy is successful, the com m odity sale to the foreign country is made; otherwise, the sale to the foreign country is voided. Upon proof o f shipment and landing o f the com m odity in the targeted market, the exporter is paid by the USDA with a generic com m odity certificate in the amount o f the bonus. The certificate can be exchanged for its value in any of the surplus stocks held by the USDA. The exchange o f certificates for most comm odities is made at the “Posted County Price,” which is rep resentative o f an average local market price. The exchange o f certificates for wheat is accomplished through a USDA auction. As shown in table 2, EEP subsidies have in creased since their inception in 1985. Total EEP '“Another example of surplus disposal through export is the Food for Peace Program, also known as P.L. 480. The pro gram provides surpluses either at no cost or at below-market prices to low-income countries. In an analysis of P.L. 480, Luttrell (1982) concluded that the food shipments were largely a gift that reduced the incentive for food production in the recipient nations. 43 subsidies in fiscal year 1986 amounted to $280 million and grew to $995 million in 1988. During this time, the value o f wheat subsidies grew from $209 million to $744 million Had it not been slowed bv the severe drought in the United States that reduced the availability of wheat surpluses, EEP growth in 1988 would have been even more substantial. The European Community’s Export Refunds While the EEP is o f recent origin, the EC’s export subsidy program has been in effect since the founding o f the Common Agricultural Policy (CAP) in 1962. The program, however, was not a source of trade friction with other agricultural exporting nations because the EC was an im porter of most agricultural commodities. The CAP originally was designed to encourage domestic production in Europe following the food shortages during and afterW orld War II. It encourages com m odity pro duction by offering a guaranteed price that often has been significantly higher than the w orld price. Because domestic prices generally have been higher than w orld prices, the CAP uses a variable levy to protect EC farmers from lower-priced im ports." Over time, European farmers responded to the high price guarantees with greatly increased pro duction, resulting in large surplus stocks. To dis pose o f the surpluses, the EC makes payments to exporters, known as export restitutions or refunds, to allow them to sell the higher-priced EC com modities at the low er w orld price. As CAP price guarantees have remained above w orld market prices, export subsidies have expanded further to dispose o f the mounting surpluses. Export re funds by the EC have grown from $4.7 billion in 1982 to a projected $12.9 billion in 1988. Bailey (1988a) states that EC export subsidies for wheat rose from $365 million in 1985 to an estimated $1.8 billion in 1988 and that the EEP probably ac counted for 35 percent to 40 percent o f the increase.1 2 "The variable levy taxes imports at the rate of the difference between the world price and the EC threshold price. For exam ple, in March 1987, the EC threshold price for wheat was $8.53 per bushel while the world price was $1.95 per bushel. Im porters would have been required to pay a levy of $6.58 ($8.53 - $1.95). These payments represented a large income source for the EC when it was a major importer. 1 The EEP is only one device that the United States allegedly 2 has used to influence the EC. The Farm Bill of 1985 sharply cut the crop loan rate which allowed the market price to plunge while maintaining a high level of income support for farmers. ANALYSIS OF THE EEP — ATTAINMENT OF GOALS AND COSTS The EEP will be judged on the basis o f the costs associated with expanding exports and inducing negotiations to liberalize agriculture throughout the world. First, w e examine the effect o f the EEP on exports and assess its costs relative to simply destroying the surplus production. Second, we examine the effect o f the EEP on the EC’s w illing ness to reduce governmental involvement in agri cultural production and trade. EEP and the Goal o f Export Expansion The primary goal o f the EEP is to increase the volume of U.S. agricultural exports. Wheat exports have increased sharply since 1985, the first year of the EEP, growing about 60 percent in 1987. Not only has the level o f exports expanded, but the U.S. share o f the w orld's wheat market increased from 28.8 percent in 1985 to an estimated 41.6 percent in 1988. To what extent can the rise in exports be attributed to the EEP? The following discussion highlights many of the empirical dif ficulties involved in answering this question and discusses one study that has addressed this ques tion. The EEP, as an export subsidy program, will increase the quantity o f exports by driving down the price o f exports. As Belongia (1986) has noted, however, export revenues w ill not necessarily rise as the quantity o f exports increases.1 If the world 1 demand for wheat is inelastic, then the EEP would cause a reduction in export revenues. Therefore, the price elasticity o f export demand for U.S. wheat is a crucial variable for determining the overall effects o f the EEP. Estimates o f the price elasticity of export de mand for U.S. wheat cover a w ide range o f values. Gardiner and Dixit (1987) summarized studies over the past two decades that estimated this elasticity. This cut in market prices led to higher export subsidy costs for the EC. In addition, the 1985 Farm Bill introduced the practice of marketing loans for cotton and rice. The marketing loans also led to lower world prices while maintaining farmers’ in come. Cotton and rice, however, are not exported in any signifi cant quantities by the EC. ,3See Belongia (1986) for a discussion of the profitability of farming and the pitfalls of using export volume as an indicator of the farm sector’s economic health. NOVEMBER DECEMBER 1988 44 The short-run (that is, one year or less) price elas ticity ranged from —0.14 to —3.13 with an average o f —0.72, while the long-run (that is, more than one year) price elasticity ranged from —0.23 to — 6.72 with an average o f —1.93. The lack o f a con sensus estimate precludes a definitive assessment about the desirability of EEP; however, some sug gestive evidence can be assembled. With 14 o f the 17 estimates o f the short-run price elasticity in the inelastic range, evidence suggests that increasing export subsidies will de crease export revenues in the short-run. A one time, across-the-board subsidy is clearly unwar ranted in this case.'4 If the export subsidy continues, then the longrun price elasticity is relevant. A definitive conclu sion is no longer possible. The studies suggest that the long-run price elasticity is likely to be elastic. If so, then export revenues w ill increase in the longrun due to the export subsidies. With export reve nues likely decreasing in the short-run and in creasing in the long-run, additional information about the magnitudes o f the export revenues and subsidy costs over time and the appropriate dis count rate is required before a definitive conclu sion can be reached. In fact, information requirements are even greater. An implicit assumption of the elasticity discussion is that the EC, as well as other countries, do not attempt to counteract the EEP. The parallel rise in U.S. and EC export subsidies, as well as an ecdotal evidence presented later, reveals this as sumption is not appropriate. In addition, the EEP is targeted to specific markets where U.S. exports have been displaced by the EC. Thus, information is required about the price elasticity of export de mand for specific markets. Consensus estimates concerning specific markets are simply not avail1 Using a model of international wheat markets, Sharpies (1984) 4 simulated the 1983 effects of an across-the-board $34 per ton ($.93/bushel) U.S. subsidy on its wheat exports. A specific goal was to compare the costs of using export subsidies to reduce surplus stocks with using the payment-in-kind acreagereduction program. Assuming the EC counter-subsidized to maintain its existing volume of wheat exports, the U.S. subsidy would have caused a 300 million bushel increase in U.S. ex ports, which represents a 20 percent increase above the level of unsubsidized exports. The direct budget cost would have totaled $1.6 billion or approximately $5.30 for each additional bushel of wheat exported. A less costly alternative would have been for the government to buy the additional 300 million bushels at the existing $3.65 loan rate and then destroy it. 1 The lower foreign-exchange value of the dollar, which might be 5 expected to have price effects similar to an across-the-board subsidy, accounted for little of the increase in wheat exports. 1 The cost of the wheat subsidy for the two crop years 1986/87 6 FEDERAL RESERVE BANK OF ST. LOUIS able. Finally, other factors that influence the level of U.S. wheat exports must be accounted for. Despite the difficulty o f estimating EEP’s effect on export revenues, the EEP clearly has boosted the volume o f wheat exports by eliminating the EC’s export price advantage. Since 1985, the U.S. export price has been $30-$40 per ton higher than the subsidized EC export price, a difference offset by the average EEP subsidy o f approximately $33. The effect o f this subsidy, along with four other factors that influenced U.S. wheat exports over the past three years, w ere analyzed by Bailey (1988b). These other factors w ere the low er price support loan rates for wheat, reductions in the yields o f competing exporters, increased imports by the Soviet Union and the Peoples Republic o f China not attributable to the EEP and finally the low er value o f the dollar. Other factors that influence wheat exports, such as w orld econom ic health and production in im porting countries, w ere not evaluated. Bailey found that the EEP was responsi ble for about one-third o f the increase in wheat exports from 1985 to 1987 attributable to the five factors.1 The EEP was responsible for roughly a 3 305 m illion bushel increase in wheat exports over 1986/87 and 1987/88. Over this same period, the cost o f the EEP subsidy given to exporters for wheat sales was approximately $1.24 billion.1 0 These estimates translate to an approximate cost o f $4.08 for eveiy bushel o f increased exports. The average U.S. Gulf export price for wheat over these two years was only $3.16. In terms o f its pri mary goal, the EEP increased exports, but it did so at a high cost in the short-run. Destroying the governm ent-owned stocks, which entails an o p portunity cost o f approximately $3.16 per bushel, w ould be a more cost-effective form o f surplus removal than the EEP with a cost o f $4.08 per bushel.1 7 and 1987/88 was estimated because year by year cost data were not available. The share of total EEP wheat sales (through August 4,1988) made in the two years was approxi mately 85 percent. The total market value of the wheat subsi dies given to exporters through August 4,1988 was $1.46 billion. The two-year cost of wheat subsidies therefore was estimated at $1.24 billion ($1.46 billion times 85 percent). ,7The availability of surplus commodities is another important factor in the EEP which became apparent in the drought year of 1988. When government stocks of wheat are depleted by drought or by other factors, the EEP program would be forced to reduce its activity. Such irregularity makes the program less reliable from the perspective of importing countries. The reac tion of importers likely would be to diversify sources and reduce reliance on a single export source. In addition, the changes in the EEP possibly prevent the full impact of the rising exports from the higher price elasticities of demand in the long run from occurring. 45 EEP and the Goal o f Liberalizing Agriculture Worldwide In addition to increasing exports, the EEP is being used to pressure the EC to liberalize its agri cultural production and trade policies. By increas ing the costs o f the EC's agricultural support pro grams, the United States hopes to induce the European Community to negotiate major reduc tions in these programs.1 The political nature o f 8 the agricultural programs o f both the United States and the European Community preclude any definitive conclusions about the response o f the EC to the EEP in the long run; however, insights from strategic trade theory and the observed initial EC responses that are identified below suggest the EEP has been ineffective."1 In a strategic environment, a small number o f econom ic agents make interdependent decisions.-0 A decision by one agent can alter the costs and benefits facing another agent. Thus, agents at tempt to judge the response o f their rivals before determining the best course o f action. Contrary to a w orld o f perfect com petition with many agents each too small to influence the market outcome, agricultural trade policy can be viewed as a strate gic environment that can be altered by govern mental decisions. Obviously, the United States and the EC are major decisionmakers in this environ ment. Subsidies play an important role in strategic trade policy. Export subsidies have been recom m ended for strategic industries that are expected to earn additional returns sufficient to exceed the total cost o f the subsidy. Strategic trade policy is controversial for a number o f reasons, one of which is that strategic trade policy tends to create an adversarial situation between countries.2 ' Countries affected adversely are inclined to re 1 The rising costs of the EC’s agricultural programs have already 8 lead to some reductions in price supports. In 1984, the EC imposed dairy quotas and began charging farmers who ex ceeded their quotas. More recently, the EC has stated its willingness to reduce grain support prices if grain production exceeds 160 million metric tons. The relationship of these cuts to the EEP, however, is unknown. 1 Strategic trade theory has become popular because of recent 9 developments that have focused on the importance of econo mies of scale, production experience and technological change as determinants of trade patterns. These determinants raise the possibility that productive resources such as labor and capital can earn higher returns in some industries than others and that certain sectors generate benefits that accrue to other sectors. Both possibilities can be used to justify an activist use of trade policies to influence domestic as well as foreign activity spond with their own subsidies. Mutually destruc tive trade wars are a distinct possibility. In fact, recent developments in w orld agricultural trade are characterized as part o f a trade war.2 Without 2 question, both U.S. and EC export subsidies for wheat have increased rapidly in recent years. Paarlberg’s (1988) recounting o f the trade war between the United States and the EC provides a number of incidents that tend to reinforce the preceding discussion. In Januaiy 1983, the United States subsidized the sale o f 1 million tons of wheat flour to Egypt to undercut subsidized offer ings by the EC. To prevent the EC from buying back the market, the United States forced Egypt to agree not to import wheat flour from anv non-U.S. supplier until June 1984. In the short run, the United States displaced the EC sales o f wheat flour to Egypt. The EC, however, responded by subsidiz ing exports o f 320,000 tons o f unmilled wheat to Egypt in spring 1983 and new subsidized wheat sales to Iran, Syria, Libya and Algeria. The EC also began com peting in China and Latin America. In addition, it reached an agreement with Egypt in October 1983 on future subsidized sales. One o f the major problems o f the EEP in liberal izing agriculture is that m ixed signals are being sent to the EC and other agricultural nations. The U.S. proposals are a highly publicized initiative to stimulate a cooperative search to reform agricul ture through GATT. At the same time, the retalia tory challenges to European-subsidized export sales can be term ed "non-cooperative activism.” Tangermann (1985) argues that the U.S. export subsidies w ill be counterproductive in achieving a reduction o f EC agricultural subsidies. His reasons are both politically and econom ically based. First, the EC’s costs o f matching the U.S. subsidies are relatively small. If U.S. subsidies had reduced w orld grain prices bv 10 percent in 1982, the EC and increase income domestically. Strategic trade theory combines international trade theory and political theory to explain the dynamics of trade policies and assist in designing policies that are in a nation’s best interest. 20 See Richardson (1986) for a more lengthy discussion of strate gic trade policy. 2 Subsidies for research and development have been recom 1 mended for strategic industries whose competitive positions depend on generating technological advances. In addition to creating an adversarial situation between countries, there are concerns that special interest groups will capture the benefits from the subsidies at the expense of the nation. 2 See Lochhead (1988) for a characterization of this trade war. 2 NOVEMBER/DECEMBER 1988 46 could have maintained its export volume by an increase in its agricultural budget o f only 0.8 per cent. Second, since the United States can be por trayed as the enemy, there will be much political support for expenditures to counteract the U.S. subsidies. Additional doubts about the effectiveness and wisdom o f waging full-scale trade war have been raised by Paarlberg (1988). Since the United States has much larger foreign markets to defend, Paarlberg estimates that it w ould have to outspend the European Community by 50 percent in a fullscale war simply to retain its market share. In ad dition, since some o f the U.S. major foreign com petitors are also large importers, a full-scale war w ould likely lead to foreign retaliation in the form of import restrictions that w ould be costly to both the United States and the other countries. As a major im porter from the United States, the EC is in a position to make a full-scale war more costly for the United States. In addition, the U.S. negotiating position in GATT is weakened because the United States is doing the same thing that is the source o f its irrita tion with the EC.2 For example, the Cairns Group, 3 a negotiating coalition o f 13 agriculturally oriented nations, has objected strenuously to the contin ued use o f agricultural export subsidies.-4Much irritation stems directly from the econom ic conse quences o f increasing subsidies by the United States and the EC. Oleson (1987) has noted that the U.S. and EC policies have caused the price of wheat to fall, im posing major losses on such grain exporters as Canada, Australia and Argentina. Strategic trade theoiy suggests that the lack of clarity about U.S. policy will inhibit the desired foreign response. The United States is willing to subsidize exports; however, it maintains that a liberalized agricultural system is a goal. A basic question, which focuses on the credibility o f the U.S. proposal, is w hether the goal o f a complete liberalization o f agricultural production and trade in the United States is feasible politically. Producers o f agricultural products in the United States have been beneficiaries o f price support programs since the 1930s.-3These programs pro vide substantial benefits to the farm sector. 2 The EEP also has sparked a diplomatic controversy. The 3 program was written to exclude subsidized sales to the Soviet Union. The Soviet Union used this discrimination as a basis for reneging on its commitment to purchase a minimum of 4 million tons of U.S. wheat each year. Under pressure from domestic producers and exporters, the exclusion was reversed and the Soviet Union became eligible for subsidies on 4 million tons. FEDERAL RESERVE BANK OF ST. LOUIS Roningen, Sullivan and Wainio (1987) estimate that a multilateral liberalization w ould cause a loss of U.S. producers' surplus o f slightly less than $10 billion. Thus, w hether the U.S. Congress w ould actually support legislation for the complete liber alization o f agricultural production and trade is uncertain. This uncertainty about the true U.S. position likely deters the EC from agreeing to the stated U.S. position. CONCLUSION The initial evidence from the Export Enhance ment Program, although far from conclusive, raises doubts about the wisdom o f U.S. agricultural export subsidies for wheat and, by implication, all commodities. While the EEP has contributed to increased U.S. wheat exports, the cost o f disposing o f the resultant surplus, even if one ignores the escalating U.S.-European Community trade war, is higher than the cost o f simply destroying the wheat in the short-run. In addition, the U.S. has im posed costs on agricultural exporters through out the world. Strategic trade theory and the EC’s initial re sponse to the export program suggest that the EC w ill be more likely to escalate the trade w ar than agree to U.S. proposals for eliminating all production- and trade-distorting agricultural p o li cies. This, however, does not rule out the possibil ity that negotiations to liberalize agriculture are doomed, but rather suggests that the EEP w ill be ineffective, and possibly counterproductive, in affecting the EC’s position. REFERENCES Bailey, Kenneth W. “ What Explains Wheat Export Rise?” Agricultural Outlook (July 1988a), pp. 22-25. ________ “ The Impact of the Food Security Act of 1985 on U.S. Wheat Exports: An Econometric Analysis.” Ph D. Dissertation, Department of Agricultural and Applied Economics, University of Minnesota, September 1988b. Belongia, Michael T. “The Farm Sector in the 1980s: Sudden Collapse or Steady Downturn?” this Review (November 1986), pp. 17-25. Gardiner, Walter H., and Praveen M. Dixit. Price Elasticity of Export Demand: Concepts and Estimates, U.S. Department of Agriculture, Economic Research Service (February 1987). This reversal embarrassed U.S. diplomats and angered major farm competitors such as Canada and Australia. 2 See Houck (1988) for further details. 4 2 See Gardner (1987) for a study using the public choice ap 5 proach to explain producer protection across agricultural com modities since the 1930s. 47 Gardner, Bruce L. "Causes of U.S. Farm Commodity Pro grams,” Journal of Political Economy (April 1987), pp. 290310. Haley, Stephen L. Targeting of U.S. Agricultural Export Subsi dies: A Theoretical Analysis, U.S. Department of Agriculture, Economic Research Service (June 1988). Houck, James P. “ What in the World is the Cairns Group?” CHOICES (Second Quarter 1988), p. 34. Hufbauer, Gary C., and Joanna S. Erb. Subsidies in Interna tional Trade (Institute for International Economics, 1984). Lochhead, Carolyn. “ Disarmament Negotiations in the Agricul ture Trade War,” Insight (November 28, 1988), pp. 42-44. Luttrell, Clifton B. “ Good Intentions, Cheap Food and Counter part Funds,” this Review (November 1982), pp. 11-18. Oleson, Brian T. “World Grain Trade: An Economic Perspec tive of the Current Price War,” Canadian Journal of Agricul tural Economics (November 1987), pp. 501-14. Paarlberq, Robert L. Company, 1988). Fixing Farm Trade (Ballinger Publishing Richardson, J. David. “The New Political Economy of Trade Policy," in Paul R. Krugman, ed., Strategic Trade Policy and the New International Economics (MIT Press, 1986), pp. 2 5 782. Roningen, Vernon, John Sullivan, and John Wainio. “The Impact of Removal of Support to Agriculture in Developed Countries," paper presented at the American Agricultural Economics Association meetings, August 1987. Rossmiller, Ed. “ Agricultural Proposals in the GATT,” CHOICES (Second Quarter 1988), pp. 30-31. Sharpies, Jerry. “Are Export Subsidies the Answer to U.S. Grain Surpluses?” U.S. Department of Agriculture, Economic Research Service (October 1984). Tangermann, Stefan. “The Repercussions of U.S. Agricultural Policies for the European Community,” in Bruce L. Gardner, ed., U.S. Agricultural Policy: The 1985 Farm Legislation (Amer ican Enterprise Institute, 1985), pp. 329-44. Wilson, Ewen M. "A CHOICES Debate: Export Subsidies on Value-Added Products — Effects May Differ From Policy Objectives,” CHOICES (Second Quarter 1988), pp. 5 and 7. U.S. Department of Agriculture, Economic Research Service Wheat Situation and Outlook, various dates. U.S. Department of Agriculture, Foreign Agricultural Service. FAS Fact Sheet — Export Enhancement Program, revised May 1988. _________World Production and Trade: Weekly Roundups, various dates. NOVEMBER/DECEMBER 1988 48 K. Alec Chrystal and Daniel L. Thornton K. Alec Chrystal is the National Westminster Bank Professor of Personal Finance at City University, London, and Daniel L. Thorn ton is a research officer at the Federal Reserve Bank of St. Louis. Dawn M. Peterson provided research assistance. The Macroeconomic Effects of Deficit Spending: A Review J . OLLOWING the Kevnesian Revolution in macroeconomics, a large number of economists argued that deficit spending was required to achieve two of the stated national econom ic objectives: full employment and a high rate o f econom ic growth.1 Society was thought to benefit from deficit spend ing because o f the reduction in lost output and because the econom y w ould achieve a higher rate of growth. This view o f deficit spending has been chal lenged increasingly over the years. A sizable num ber o f economists now believe that deficit spend ing has little effect on em ployment and output, especially in the long run, and that it primarily results in a redistribution o f output, either within the private sector or as a transfer o f resources from the private to the public sector.- Support for this viewpoint has produced a growing concern about the potentially harmful effects o f deficit spending and the size o f the public debt.3 'One of Keynes' initial arguments was that saving would exceed investment at a level of output consistent with the full employ ment of labor. That is, the U.S. savings rate was too high. The view that the budget should be in persistent deficit was termed the “ new fiscal policy.” To see how opinions about deficit spending have changed in two decades, compare the deficit discussions in Levy (1963) with those in Levy, et. al. (1984). 2 The once-common view that the market economy cannot sustain full-employment equilibrium has given way to the FEDERAL RESERVE BANK OF ST. LOUIS The existence and magnitude o f the benefits from deficit spending have important implications for the public policy debate. Presumably, the deci sion to incur deficits is affected by the public's belief about whether deficits provide benefits to some individuals at little or no cost to others, or w hether they merely redistribute income. Hence, a central issue in the debate over deficit spending is whether, and to what degree, it can be used to produce net benefits for society as a whole. The purpose o f this paper is to examine some o f the arguments and evidence on whether deficit spending yields net benefits to society. DEFICIT SPENDING: SOME KEY TERMS The phrases "deficit spending” and "fiscal pol icy” are not necessarily synonymous. While deficit spending is a particular fiscal policy action, not all concept of the natural rate of unemployment. For a discussion of these issues, see Modigliani (1986b), Blinder (1986) and Laidler (1988). ’ For a discussion of the potential harmful effects of the public debt, see Bruce and Purvis (1986), Barro (1987) and Levy, et. al. (1984). 49 fiscal policy actions produce or involve deficits.4 For example, the government could devise a policy whereby expenditures and taxes are changed bv the same amount. This well-known "balanced budget’’ operation affects aggregate demand, be cause the change in government expenditures affects aggregate demand more than the change in taxes, but does not affect the deficit." Despite the balanced-budget multiplier, the stance o f fiscal policy today is often associated with, and frequently measured by, the size o f the federal budget deficit.1Thus, in this article, deficit 1 spending and the stance of fiscal policy w ill be treated as synonymous. Furthermore, since they both produce the same qualitative shift in aggre gate demand, no distinction will be made between deficits that arise from increases in government spending and those that result from tax reduc tions. Cyclical and Structural Deficits and Discretionary Fiscal Policy It is important to differentiate between "cycli cal” and '"structural’' deficits when examining the effects o f policy changes on the economy. Tax revenues rise during the expansion phase o f the business cycle and fall during the contraction phase; in contrast, certain government expendi tures (e.g., unemployment compensation) fall dur ing expansions and rise during contractions. These counter-cyclical components o f the “There is a well-known caveat to this statement. Government tax rate changes are not neutral. The government may change certain marginal tax rates and simultaneously alter government expenditures to produce no net effect on aggregate demand, all other things constant. The ultimate effect on aggregate output, however, need not be neutral; the non-neutrality of the tax rate change could produce changes in aggregate supply. Such analysis underlies much of the recent work by Auer bach and Kotlikoff (1987) and Kotlikoff (1988). Consequently, they have challenged the usual convention of associating deficit spending with fiscal policy. For example, Kotlikoff (1988), pp. 489-90, states that “ .. . fiscal policies can matter a lot, but deficits may nonetheless tell us nothing useful about the true stance of fiscal policy.” They argue that, within their life cycle model, the labels “taxes" and “spending” are arbitrary. For them, a tight fiscal policy occurs when a larger burden of “ government consumption” is borne by current rather than future generations. Aggregate demand increases because the marginal propensity to spend of the public sector (1) is greater than the marginal propensity to spend of the private sector (< 1 ). If the private sector’s marginal propensity to spend is large, the difference between the marginal propensities will be small and so, too, will be the effect of tax-financed expenditures on aggregate de mand. deficit— the so-called automatic stabilizers— are intended to smooth cyclical swings in income. The structural deficit, on the other hand, reflects discretionaiv fiscal policy actions.7It is the part o f the deficit that is invariant to the phase o f the business cycle. Chart 1 presents measures o f the actual and cyclically adjusted budget deficit. Although these measures depart substantially at times, generally they move together. While the analysis in this paper applies equally w ell to cycli cal and structural deficits, from now on the dis cussion will focus solely on structural deficits. THE NET BENEFITS FROM DEFICIT SPENDING The effectiveness o f deficit spending depends on two factors: the slope o f the aggregate supply curve and the extent to which deficit spending shifts the aggregate demand curve. These factors are discussed in detail in latter sections o f the paper. In this section, w e present some general notions underlying the view that society can be a net beneficiary from deficit spending. The initial popularity o f using deficit spending to increase output was based on the belief that the market econom y is unable to sustain aggregate demand at a level consistent with full-employment output. This idea o f persistent unem ploy ment is illustrated in chart 2 which shows a gap between actual and “ potential" real output.* The 7See de Leeuw and Holloway (1983) for a detailed discussion of these concepts and Fellner (1982) for a critique of these mea sures. For a discussion of these concepts and a breakdown of the deficit, see Erceg and Bernard (1988). 8 There is an issue, not taken up here, about the extent to which such unemployment is “ involuntary." According to the usual textbook definition, involuntary unemployment occurs when individuals are willing to work at the market wage but are unable to find employment; that is, when there is an excess supply of labor at the market wage rate. If the market is com petitive, the wage rate should fall to eliminate the involuntary unemployment. Hence, nearly all theories of involuntary unem ployment require some form of nominal or real wage rigidity. In early Keynesian models, involuntary unemployment was due to nominal rigidities in wages. This explanation requires real wages to fall when output rises. Empirical evidence, how ever, suggests that real wages are pro-cyclical. Recently, research by “ New Keynesian Economists” suggests that persistent under-employment equilibria and involuntary unem ployment can result from nominal price rigidities in the output market because of monopolistically competitive firms, and because of rigidities in real wages due to “ efficiency wages.” See Blinder (1988), Mankiw (1988), Rotemburg (1987), Pres cott (1987), The New Keynesian Microfoundations (1987) and the cited references. 6lt is common to measure fiscal action by the full-employment budget surplus or deficit. For a discussion of this, see Carlson (1987) and Seater (1985). NOVEMBER/DECEMBER 1988 50 Chart 1 Actual and Cyclically Adjusted Budget Surplus/Deficit Billions of dollars Billions of dollars 50 ----------------------------------------------------------------------------------------------------------------------------------- 50 -1 0 0 -1 5 0 -1 5 0 -200 -200 -2 5 0 1987 -2 5 0 1955 Chart 2 Actual and Potential GNP Billions of 1982 dollars 4000 Billions of 1982 dollars 4000 3500 3500 3000 3000 2500 2500 2000 1500 1500 1000 1955 1000 59 63 FEDERAL RESERVE BANK OF ST. LOUIS 75 79 83 1987 51 potential path o f real output usually is associated with some full-employment rate o f unem ploy ment. Periods in which real output falls below its potential represent episodes o f persistent exces sive unemployment. If the econom y is prone to periods o f prolonged unemployment due to de ficient aggregate dem and for goods and services, the government could run a sustained deficit to make up for the deficiency. If successful, this de ficit would keep output closer to its fullem ployment potential. Moreover, on average, real output growth would exceed the rate that would otherwise occur. Deficit Spending and Capital Accumulation Deficit spending could have a secondary effect on the rate o f econom ic growth. Production o f real output Iv) is related to factor inputs, labor IN) and capital (K), via a production function, that is, v = f(N,K). The marginal products o f both labor and capital are positive: for any quantity of capital (labor), output increases as more labor (capital) is used. The growth o f the labor force is often con sidered synonymous with population growth, w hich is determined in part by factors that are independent o f econom ic considerations. The size o f the capital stock, on the other hand, is usually assumed to be related to econom ic factors. The higher the rate o f capital formation (investment), the higher the rate o f econom ic growth. Firms determine the most profitable level o f output and, simultaneously, the optimal capital/ labor ratio. Because o f the nature o f capital goods, the decision to acquire capital is based (among other things) on expectations o f output growth. If the market econom y is subject to prolonged peri ods o f unemployment and slow growth because of insufficient demand, expectations for output growth and investment will be low er than if these periods did not occur. If deficit spending raises the path o f real output over what it w ould achieve otherwise, investment and, thereby, potential real output growth should rise even higher. Thus, de 9 Achieving a higher rate of economic growth was part of the fiscal policy agenda during the 1960s. See Levy (1963). '“Recently, Sickel (1988) has investigated the asymmetry of the business cycles. He tests for both the “steepness’’ and “ deep ness” of post-World War II cycles and finds evidence that cyclical troughs are deeper than cyclical peaks. "This discussion implicity assumes that deficit spending does not alter the path of y*, i.e., that deficit spending merely dampens the cycle. 1 Many authors merely assert that there are benefits from more 2 stable output growth without identifying these gains, e.g., ficit spending could produce a higher rate o f ac tual and potential growth because o f increased capital formation.1 1 Deficits and Symmetric Business Cycles The gains in output discussed so far are predi cated on the assumption that cyclical swings in output around its potential path are asymmetric: cyclical downturns are longer and more pro nounced than cyclical upturns. Since w e are as suming that cyclical swings are due to variation in the demand for goods and services, this means that increases in the demand for goods and ser vices are less frequent and smaller than decreases. If, on the other hand, fluctuations in aggregate demand around potential output are symmetric, periods during which output is above or below the potential path also will be symmetric."’ This is illustrated by path 1 in figure 1 and by the aggre gate demand and supply curves in figure 2. Given the slope o f the aggregate supply curve, symmetric variation in aggregate dem and produces symmet ric movements in output about the potential level, v*. On average, there are no "net output” gains to be achieved from deficit spending over the cycle. Periods o f deficit spending w hen the econom y is below the full-employment path w ould be matched by periods o f budget surplus when out put is above the path, so the budget w ould be balanced over the cycle and the average output level w ould be the same as with no fiscal action. Society still may benefit, however, if the govern ment runs deficits during the contraction phase o f the cycle and surpluses during expansions. A cy clically balanced budget could stabilize aggregate demand and reduce the variability in output; this is illustrated by path 2 in figure 1." The Benefits From Stable Output More stable output could reduce the risk associ ated with capital investment and, as a result, in crease investment.'- Consequently, the capital Modigliani (1986a), (1986b) and Bossons (1986). At other times explanations of these gains sound hollow. For example, Bruce and Purvis (1986), pp. 60-61, argue for the benefits of avoiding a cyclical downturn by stating that “ a government deficit will provide some stimulus to the economy and hence help reduce the dead-weight costs of unemployment that would have occurred in the absence of the deficit." In the case where the government runs a surplus in order to prevent an economic boom, they argue that the surplus helps “ avoid the dead weight costs that again arise because the economy is away from its long-run equilibrium.” (Italics added.) NOVEMBER DECEMBER 1988 52 Figure 1 Symmetric Swings in Output stock w ould increase, as would the level o f poten tial output .1 The econom y w ould then achieve a 3 higher rate o f growth than otherwise. Additional benefits could arise if more stable output growth results in more stable consum p tion. Economists usually argue that people maxi m ize the utility o f their consumption over some planning horizon and that the utility gains from increased consumption are smaller than the losses from equally probable decreases in con sumption.1 Even if the distribution o f shocks to 4 incom e and, therefore, consumption are symmet ric, the distribution o f utility gains and losses will be asymmetric. Consequently, the expected utility o f consumption rises as incom e is stabilized. The Benefits from Stabilizing Nominal GNP There are additional benefits from stabilizing aggregate dem and if cyclical movements in nomi- 1 The issue is whether the growth rate of real output is made 3 permanently higher. Certainly, if economic stabilization policy merely causes the level of real output to be higher but does not affect the rate of real output growth permanently, there would still be a period immediately following the enactment of stabili zation policy in which the observed rate of real output growth would exceed the full-employment growth rate. 1 That is, the utility function is concave. Such gains from eco 4 nomic stabilization have been suggested by New Keynesian economics. See Rotemburg (1987), p. 83. To illustrate this point, assume that consumption is a random variable that is uniformly distributed on the closed interval 1 to 2, and let the utility of consumption be the simple concaved function, u = C 5. In this case, the expected value of utility is 1.22. Now assume that income and, hence, consumption are more variable, but FEDERAL RESERVE BANK OF ST. LOUIS Figure 2 Symmetric Swings in Output and Aggregate Demand and Supply nal GNP are symmetric, but cyclical movements in real output are asymmetric. That is, the aggregate supply curve is more steeply sloped above poten tial output as in figure 3. In this case, random vari ation in aggregate dem and w ould produce larger changes in real output below the potential output level than above it. Of course, the change in nom i nal spending above and below potential output must be the same if variations in aggregate de mand are symmetric about the natural rate. Stabi lizing discretionary fiscal policy reduces both in flation and unemployment over the cycle and, thus, the cost o f lost output associated with un em ployment and the cost o f inflation.” Finally, deficit spending could yield net benefits if it merely offsets downward shifts in aggregate demand. For example, assume that cyclical swings in real output are symmetric so that there are no output gains on average over the cycle from stabi lizing aggregate demand. Deficit spending still could result in net output gains for society, if de- with the same expected value. Specifically, assume that con sumption is now uniformly distributed on the closed interval 0 to 3. In this case, the expected value of utility of consumption is reduced to 1.15. Hence, reducing the variability of consumption increases the expected (average) utility of consumption. Of course, consumption may fluctuate much less than output over the business cycle if the life-cycle or permanent income theo ries of consumption are correct. 1 The costs of expected inflation are in terms of its effects on 5 long-term bond markets, the misallocation of productive re sources and its effects on regulations. The casts of unexpected inflation are primarily in terms of its redistribution of wealth. For a discussion of these costs, see Leijonhufvud (1987) and the references cited there. 53 Figure 3 Asymmetric Swings in Output but Symmetric Swings in Nominal GNP AS ficits w ere incurred when aggregate demand was weak, but surpluses were not incurred when ag gregate demand was strong. Of course, in this case, the level o f government debt w ould rise, both over the cycle and over time. CRITICISMS OF THE ALLEGED BENEFITS OF DEFICIT SPENDING As we have seen, the gains from deficit spending consist o f reducing “lost” output due to reduced employment, increasing the growth rate o f real output or stabilizing output and consumption. To achieve these gains, deficit spending must shift the aggregate demand schedule and the aggregate supply curve must be upward-sloping, at least in the short run. If the aggregate supply curve were vertical, shifts in the aggregate demand schedule w ould not affect output. Consequently, there could be no output gains from offsetting shifts in aggregate demand. Of course, if the aggregate sup ply curve were positively sloped, deficit spending w ould be effective only if it succeeds in shifting the aggregate demand curve. Attacks on the ef ficacy o f fiscal policy have focused, therefore, on ,6This applies to monetary policy as well. 1 For an interesting discussion of Keynesian and classical eco 7 nomics, see Blinder (1986), Laidler (1988), Eisner (1986) and Niehans (1987). 1 There is a problem in defining ‘'persistent” unemployment and 8 establishing if and when it differs from cyclical unemployment. Many economists argue that there is no such thing as persist ent unemployment because the market economy eventually the slope o f the aggregate supply cuive and the ability of deficit spending to shift aggregate de mand.1 " Asymmetric Cyclical Variation in Output Both the Great Depression o f the 1930s and the rise o f Keynesian economics, with its emphasis on underem ploym ent equilibrium, led to the accept ance o f the notion that the market econom y is neither able to sustain a full-employment level of output nor able to move back to it quickly when aggregate dem and failures occur.1 Prior to Keynes, 7 it was com m only believed that output w ould natu rally move to the level consistent with no involuntaiy unemployment. While shocks to either aggre gate demand or supply might cause temporary periods o f unemployment, resources were thought to be sufficiently mobile and wages and prices sufficiently flexible that the econom y w ould return to its full-employment equilibrium fairly quickly. Keynes argued that the econom y might remain permanently below its full-emplovment level be cause o f insufficient aggregate demand and mar ket imperfections.'" This below-full-emplovment equilibrium requires an upward-sloping aggregate supply curve. Typically, it was also argued that the aggregate supply curve w ould becom e steeper around the full-employment level o f output, like the aggregate supply curve in figure 3. The Phillips Cun e The Keynesian view was strengthened by the discoveiy o f what appeared to be a stable long-run empirical relationship between the rate o f in flation and the unemployment rate; this relation ship was called the Phillips Curve."1If unem ploy ment was too high (relative to the full-employment rate), policymakers could achieve a permanent increase in output by increasing aggregate de mand through deficit spending. The cost w ould be a permanent increase in inflation. The extent of the cost is determined by the slope o f the Phillips Cuive. The closer incom e was to its full- will adjust to the point at which the labor market clears. Keynes himself almost certainly believed this to be true in the long run; however, he regarded the long run to be too long for the adjust ment to be left to market forces alone. His much-quoted de fense of his view was that “ . . . in the long run we are all dead.” ,9This apparent empirical regularity was first discovered by Phillips (1958) who used wages and unemployment. NOVEMBER/DECEMBER 1988 54 em ployment level, the steeper the slope and, con sequently, the higher the inflation rate. Presum ably, without deficit spending, the econom y w ould be stuck permanently below the fullem plovm ent level o f output. The Natural Rate Hypothesis and Rational Expectations: A Counter View to the Phillips Curve The view that the econom y could remain per manently at underem ploym ent equilibrium was challenged by the Natural Rate Hypothesis.2 It 1 1 reintroduced the once-prevalent argument that the econom y eventually w ill return to its fullem ployment equilibrium. That is, the Natural Rate Hypothesis im plied that the long-run Phillips Cuive is vertical at the natural rate o f unem ploy ment. The implications o f the Natural Rate Hypothesis were enhanced bv the rational expectations revo lution, which argued for the same conclusions, albeit along different theoretical lines. Rational expectations models o f the business cycle showed that systematic stabilization policies could not affect real output permanently in markets popu lated by 'rational” individuals."1 Both theories argue that the em ployment rate w ill tend toward its natural rate; consequently, demand management policies w ill be unable to keep the unem ploym ent rate below the natural rate in the long run. The natural rate o f output, y,„ is determ ined solely by the level of em ployment N,„ consistent with the natural rate o f unem ploy ment, given the stock of capital K. That is, y„ = f(N„, Ki. Since dem and management policies have no last ing effect on em ployment or the capital stock, they 2 See Friedman (1968) and Phelps (1967). 0 2 Neither the Natural Rate Hypothesis nor many rational expec 1 tations models give rise to involuntary unemployment as de fined in footnote 8. Many rational expectations models, how ever, give rise to cyclical movements in the natural rate of unemployment. See Fischer (1977), Taylor (1988) and McCallum (1986). For a list of other factors that could cause the unemployment rate to change without involuntary unemploy ment, see Blinder (1988). 22ln chart 2, “ potential” output is defined arbitrarily. Conse quently, persistent unemployment can exist by definition. This applies to estimates of “ potential” GNP as well as cyclicallyadjusted deficits, etc. See Fellner (1982) and de Leeuw and Holloway (1982) for a discussion of this point. 2 Also, it does not say explicitly what the level of the natural rate 3 is. See Carlson (1988) for a discussion of the level of the natu ral rate. FEDERAL RESERVE BANK OF ST. LOUIS have no effect on the natural rate o f output. In effect, these theories make it less likely that there w ill be asymmetries in the business cycle, thus, eliminating the possibility o f permanent gains in net output from deficit spending. Unless shocks to demand or supply are asymmetric, on average, cyclical downturns need be no more pronounced nor of longer duration than cyclical upturns ~ The Natural Rate Hypothesis asserts that the long-run aggregate supply curve is vertical at an output level consistent w ith the natural rate o f unemployment. It does not assert, however, that the short-run aggregate supply curve w ill be verti cal at this level o f output.2 Hence, accepting the 3 Natural Rate Hypothesis does not im ply that soci ety cannot benefit from appropriately timed and im plemented deficit spending; however, it limits significantly the benefits that society can receive from deficit spending. As discussed previously, society benefits only if deficit spending reduces cyclical swings in output or nominal GNP.2 4 CAN DEFICIT SPENDING SHIFT THE AGGREGATE DEMAND SCHEDULE? Even when the aggregate supply curve (short- or long-run) is upward-sloping, deficit spending will have little effect on output or prices if the increase in aggregate dem and that it produces is largely offset by a deficit-induced decrease in private spending, that is, if deficit spending fails to change aggregate demand. Competition f o r Credit— Indirect Crowding Out Through Interest Rates When the government runs a deficit, it issues government debt.2 Thus, the demand for credit 1 increases relative to the supply. All other things 2 Actually, in such models, deficits can provide benefits in the 4 absence of stabilizing output. These benefits come from smoothing taxes over the cycle. Public finance theory asserts that variation in tax rates across goods or activities results in welfare losses under most conditions. Consequently, it would be more efficient to run deficits and surpluses over the busi ness cycle rather than balance the budget annually by altering tax rates. See Bossons (1986) and the references cited there. 25ln models with a government budget constraint deficits are often financed directly through money creation. Given the current institutional structure, however, the government must initially issue debt even if it is subsequently monetized. See Thornton (1984a). See Thornton (1984b) for a discussion of and evidence on debt monetization. 55 unchanged, this causes interest rates to rise, re ducing private expenditures in interest-sensitive sectors o f the economy. Hence, the increase in aggregate demand associated with the deficit could crowd-out private expenditures indirectly by affecting interest rates.-" Since investment spending is one o f the most interest-sensitive components o f spending, analysts often argue that deficit spending might retard the rate o f capital formation and, hence, econom ic growth.-7 Deficit Spending and the Trade Deficit Assuming that deficit spending increases the demand for credit, its effect on interest rates d e pends on whether the econom y is “ op en ” or “ closed.” In the preceding example, w e im plicitly assumed that the econom y was closed so that the government ran a deficit by borrowing from the private sector. In an open econom y with a floating exchange rate and perfect capital flows, the results w ould be somewhat different.2 8 An increase in the budget deficit puts upward pressure on domestic interest rates. This leads to inflows o f financial capital and an appreciation of the exchange rate. This appreciation, together with the higher domestic demand, is associated with a current account deficit in the balance of payments. In effect, the government deficit is 2 This problem cannot be solved by monetizing the debt. The 6 increased rate of money growth will result merely in a higher rate of inflation and, hence, higher nominal interest rates. Many advocates of countercyclical fiscal policy view this as one of the most serious drawbacks to deficit spending. See Modiqliani (1986b). 2 This argument ignores how the deficits are spent. Recently, 7 Heilbroner (1988) has argued that deficit spending is neces sary to finance the purchase of public capital, that is, infrastruc ture. Other economist (for example, see Sturrock and Idan (1988)) argue that the real burden of deficits comes only when they are used to finance current consumption. This does not establish the desirability of deficit spending; it merely asserts that spending for infrastructure capital may increase the rate of economic growth, depending primarily on the relative produc tivity of the factor resources in the two sectors and on the productivity of public versus private capital. The idea that such expenditures should be financed by deficits rests largely on the long-lived nature of capital goods. Since these capital goods provide services over a number of years, it is argued that public sector capital goods should be financed by borrowing just as businesses or households fi nance their acquisition of durable goods. In the case of busi nesses, however, debt service is financed out of the increased earnings that the capital goods are expected to provide. In the case of households, deficit financing is used to better match the desired consumption with expected future income. Hence, households, too, expect to service the debt through higher incomes. No similar increased earnings necessarily accrues from the acquisition of public capital. Income will increase only if the marginal product of public capital is larger than that of private capital. This is a difficult point to establish. Proponents of this view point to the productivity gains that could accrue financed bv a larger trade deficit.2 The econom y ” may gain in terms o f higher short-term consum p tion, but at a cost o f an increase in external debt. The decline in private expenditures is affected through higher interest rates, a larger trade deficit or both. In any event, the result is the same: the group that gains directly from deficit expenditures does so at the expense o f those w ho lose, with little or no net increase in aggregate demand. The only difference is that those w ho gain directly are more readily identified than those w ho suffer indi rect losses through higher interest rates or in creased foreign claims on U.S. assets.1 " Ricardian Equivalence Another argument, referred to as the “ Ricardian Equivalence Hypothesis,” holds that deficit spend ing cannot shift the aggregate demand curve.'1The closed-econom y conclusion that deficit spending does not crowd-out private spending directly im plies that government debt is net wealth to soci ety. In other words, when the government issues debt to purchase goods and services, the holder of the debt views it as an asset; but the taxpayer does not view it as a liability (or, at least, views it as a smaller liability). That is, individuals believe that they w ill not have to pay current or future taxes to service or retire the debt. from public expenditures on education and the like; however, these services could be provided by the private sector. Hence, this argument is about the appropriate role for government and public goods. See Aschauer and Greenwood and Aschauer (1988a, b and c) for a discussion of the benefits from social infrastructure expenditures. Hence, the only real argument for deficit financing of such expenditures is that it would equalize their costs and benefits across generations. This implies, however, that the increased indebtedness that such expendi tures necessitate will eventually be retired through increased taxes unless the infrastructure acquired is infinitely lived. 2 The assumption of perfect capital flow s m eans that domestic 8 real interest rates could not rise above world levels without inducing an inflow of financial capital from overseas. For a situation in which there is no expectation of exchange rate changes, this means that domestic and foreign nominal interest rates must be equal. ^See Mundell (1963). This result assumes no change in mone tary policy to accommodate the defict. ^In this model, the real market value of government debt is part of society’s net wealth. In the closed economy model, at the natural rate of unemployment, the increase in wealth resulting from the increase in nominal debt due to deficit spending is just offset by a decline in wealth due to higher prices, interest rates or both. In the open economy model, it is offset by a reduced stock of national wealth due to increased claims by foreigners on U.S. assets. 3,Technically, Ricardian Equivalence argues that, for a given level of government expenditures, aggregate demand will not change as the government switches from tax to bond financing. As O’Driscoll (1977) points out, Ricardo was merely offering this as a theoretical possibility and did not himself believe it. NOVEMBER/DECEMBER 1988 56 Ricardian Equivalence, on the other hand, as serts that public and private debt are perfect sub stitutes. Individuals believe that they or their heirs will have to pay taxes equal to the deficit-financed expenditures, so an increase in present value o f the expected future taxes just equals the current deficit. At the macroeconom ic level, Ricardian Equiva lence implies that deficit spending w ill not be associated with increases in real interest rates, output, prices or the trade deficit:* Consequently, the Ricardian view yields a radically different no tion o f the national debt. For those w ho believe in the benefits o f deficit spending, the national debt, which is the accumulated deficits, should be viewed as a blessing, not a curse. For those who believe in Ricardian Equivalence, deficit spending merely results in a redistribution o f income and the national debt represents the cumulative amount o f this net transfer. Can Discretionary Fiscal Policy Be Successfully Implemented? There is also an argument against the useful ness o f deficit spending that is independent o f its ability to shift aggregate demand. It is critically dependent, however, on the Natural Rate Hypoth esis and on whether shifts in aggregate demand caused by other factors are temporary or perma nent. It has been suggested that policymakers do not have the information needed to offset shifts in aggregate dem and to stabilize output.” This argu ment is usually couched in a discussion o f the lags in econom ic policymaking. For fiscal policy, the most important o f these are the "recognition” and "implementation” lags. The recognition lag is the time between when a need for corrective action arises (an exogenous shift in aggregate de mand) and when policymakers recognize the need. The issue is simply w hether policymakers know where the econom y is in the business cycle at any particular point in time. The implementation lag is the time between when the need for corrective action is recognized and when policymakers take action. Thus, even if policymakers are quick to recognize that the de mand has shifted, by the time they react to the situation, it may have changed and the need for corrective action may have vanished. 3 Analysts frequently argue that Ricardian Equivalence must be 2 invalid because the necessary microeconomic conditions for its validity are so stringent that they cannot possibly be satisfied. For example, see Buiter (1985). Also, see McCallum (1984). FEDERAL RESERVE BANK OF ST. LOUIS This argument is presented graphically in figure 4a. Assume that the Natural Rate Hypothesis holds and that the short-run aggregate supply curve is symmetric around the level o f output consistent with the natural rate o f unemployment. Assume further an exogenous decrease in aggregate d e mand, shifting it from AD to AD'. N ow if policym a kers did not react to the shift in demand im m edi ately, the process o f adjustment toward the natural rate w ould begin; the price level w ould decline and the quantity o f output dem anded w ould increase. Once policymakers reacted to the problem bv increasing deficit spending, they w ould shift the aggregate demand curve upward, bringing output back to its natural-rate level. If the shift in aggregate dem and were temporaiy, a delay in policy might actually exacerbate the situation if deficit spending coincided closely with the return o f aggregate dem and to its former level. This is illustrated in figure 4b, where the simultaneous increase in deficit spending and the return o f aggregate demand to its form er level shift aggregate demand to AD". Of course, if the decline in aggregate demand were permanent, the timing o f policy w ould be less important. Deficit spending eventually w ould move the econom y back to the natural rate; the timing o f the policy action w ould determine only how quickly deficit spending m oved the econom y back to its full-employment potential. Of course, the econom y w ould move back eventually to full em ployment even without deficit spending. Demand or Supply Disturbances Another problem is that policymakers must be able to differentiate between demand- and supplvside disturbances. Recently, some have suggested that business cycles can be explained solely by supply-side disturbances. Indeed, some "real busi ness cycle” models have successfully produced cyclical swings in output that mimic real w orld data. W hether all cyclical swings in econom ic activity can be explained by such models is the subject o f intense debate. Nevertheless, to the extent that some cyclical swings are the result of supplv-side shocks, fiscal policy can succeed in stabilizing output only by exacerbating m ove ments in prices (or it can help stabilize the price level only bv exacerbating movements in output). 3 lt is argued that inappropriately timed policy might destabilize 3 the economy. See Friedman (1968). 57 Figure 4 The Timing of Changes in Fiscal Policy Consequently, policymakers must know not only where in the business cycle the econom y is at any point in time, but whether its position was caused by a shift in aggregate demand, aggregate supply or, perhaps, simply the cyclical dynamics o f the economy, unrelated to exogenous distur bances in either aggregate dem and or supply. In short, some would argue that the information required to use discretionary fiscal policy effec tively is simply too great. WHAT IS THE EVIDENCE? Assessing the evidence on discretionary fiscal policy is difficult. Effective discretionary fiscal policy implies that output should be more stable and suggests that perhaps the rate o f real output growth should be higher on average when fiscal policy was used aggressively. It also suggests that deficit spending should be positively correlated with interest rates, prices (or inflation) or trade deficits. 3 0ne of the earliest of these was the Andersen-Jordan equa 4 tion. See Andersen and Jordan (1968). 3 See Barro (1987), Bernheim (1987) and Aschauer (1988a). For 5 more recent studies which report results consistent with Ricar dian Equivalence, see Evans (1988), Koray and Hill (1988) and Leiderman and Razin (1988). A number o f large-scale econom etric models suggest that fiscal policy has significant short-run and, in some cases, long-run effects. Estimates of reduced-form models, however, typically show no long-run effects o f deficit spending and, often, only weak short-run effects.3 Hence, such models 4 essentially substantiate the Natural Rate Hypothe sis. These studies are subject to considerable con troversy because o f the difficulty in finding com m only accepted variables that reflect discretionary changes in fiscal policy and the continued contro versy over reduced-form estimation. The greatest challenge to the orthodox view of deficit spending comes from the Ricardian Equiva lence H y p o th e sisM a c ro ec o n o m ic evidence from three recent surveys is largely consistent with the Ricardian view.3* In general, there is no statistically 1 significant relationship between structural deficits and interest rates or inflation, or between the budget and trade deficits.3 These results are bol 7 stered by work that shows a high negative correla3 The microeconomic evidence yields mixed results. 6 3 Barro (1987) reports that he finds a statistically significant 7 correlation between government deficits and the trade deficit only if 1983 is included. NOVEMBER DECEMBER 1988 58 tion between public and private savings.’" The Evidence on Stabilization One com m only cited piece o f evidence that demand management can stabilize the econom y is a comparison o f the volatility o f U.S. output, unemployment and industrial production, before and after W orld War II. The fact that the pre-war series are more volatile than the post-war series has been cited as evidence o f both the inherent instability o f unmanaged capitalism and the suc cess o f demand management policies in stabiliz ing the economy. There are several criticisms o f this evidence. First, pre- and post-war data vary in terms o f a quality and uniformity. Indeed, some argue that the excessive pre-war volatility o f the comm only used series on unemployment, GNP and industrial production is due to various quirks in their con struction." Second, even if the post-war econom y is more stable, this may be due to other changes in eco nomic fundamentals, not to discretionary fiscal policy per se.4 Furthermore, even if fiscal policy is 0 responsible for the apparently more stable post war economy, this may be the result o f increased relevance on the automatic stabilizers, not to dis cretionary fiscal policy. Also, post-war real output growth in the United States is below its pre-war growth. The discrep ancy is even larger if the Depression years are omitted.4 Moreover, there has been a secular rise 1 in the u n e m p l o y m e n t rate. These adverse m ove ments roughly coincide with a secular rise in the U.S. structural deficit.4 Hence, if the more stable post-war econom y is used as evidence on the suc cess of fiscal policy, the associated slower output growth and higher unemployment must be con sidered the costs o f stability. 3 0 f course, in a closed economy with output unchanged, the 8 public sector deficit must equal the private sector surplus. Other studies of consumption have tried to determine whether government debt is net wealth, e.g., Tanner (1979) and Kochin (1974). Again, the results are consistent with the Ricardian Equivalence Hypothesis. 3 See Romer (1986a, 1986b, 1986c and 1988). Romer's evi 9 dence has been challenged by Weir (1986) and Lebergott (1986). 40Pre- and post-war real output series for the United Kingdom, Germany and Italy show significant decreases in the variability of real output of a similar order of magnitude as that of the United States. The pre-war standard deviations of annual output growth for the United States, United Kingdom, Germany FEDERAL RESERVE BANK OF ST. LOUIS CONCLUSION This paper has examined the theoretical argu ments about the wisdom o f deficit spending. The once-prevalent Keynesian approach, which con cludes that such gains clearly exist, has come under attack. Increasingly, both theoretical inno vations and empirical evidence suggest that m od ern economies are not w ell characterized by the Keynesian view. Support for the Natural Rate Hy pothesis, which argues that deficit spending has no effect on the equilibrium level o f output and employment in the long run has grown. If this hypothesis is valid, the gains from deficit spending result from stabilizing output around the level consistent with the natural rate o f unemployment. Such an effective use o f deficit spending, however, imposes information requirements on policym a kers that are unlikely to be attained. In general, empirical evidence on the effects of deficit spending is sparse and, for the most part, ambiguous. Most persuasive is the growing macroeconom ic evidence, consistent with Ricardian Equivalence, that deficit spending has no long-run effect. The challenge for those w ho argue that deficit spending merely redistributes incom e and that stabilization policy w ill likely hurt is to ex plain phenomena like the Great Depression. Through adherents to both extreme Keynesian and extreme rational expectations views (and ev erything between) usually are able to rationalize historical events on their own terms, the Great Depression is as likely to be seen as an example o f what bad policy can create as it is o f what good policy can eradicate. REFERENCES Andersen, Leonall C., and Jerry L. Jordan. “ Monetary and Fiscal Actions: A Test of Their Relative Importance in Eco nomic Stabilization,” this Review (November 1968), pp. 1124. and Italy were 6.61, 3.98, 6.10 and 4.79, respectively. The post-war standard deviations were 2.83, 2.00, 2.45 and 3.49. In all cases, the decline in variability was statistically significant at the 5 percent level. The data were obtained from Liesner (1985). 4,The growth rate of real output from 1869 to 1938 was 3.1 percent, from 1945 to 1983, 2.7 percent, and from 1965 to 1983, 3.7 percent. These growth rates were calculated from data in Gordon (1986). 4 The unemployment rate averaged 4.5 percent in the 1950s, 4.8 2 percent in the 1960s, 6.2 percent in the 1970s and 7.7 percent in the 1980s, respectively. 59 Aschauer, David. “ The Equilibrium Approach to Fiscal Policy," Journal of Money, Credit and Banking (February 1988a), pp. 41-62. Kochin, Levis A. "Are Future Taxes Anticipated by Con sumers?" Journal of Money, Credit and Banking (August 1974), pp. 385-94. _________“ Is Government Spending Stimulative?" Federal Reserve Bank of Chicago Staff Memoranda SM 88-3 (1988b). Koray, Faik, and R. Carter Hill. “ Money, Debt, and Economic Activity," Journal of Macroeconomics (Summer 1988), pp. 351-70. ________ “ Public Spending and the Return to Capital," Fed eral Reserve Bank of Chicago Staff Memoranda SM 88-2, (1988c). Aschauer, David A., and Jeremy Greenwood. “ Macroeconomic Effects of Fiscal Policy,” Camegie-Rochester Series on Public Policy (Autumn 1985), pp. 91-138. Kotlikoff, Laurence J. “What Macroeconomics Teaches Us about the Dynamic Macro Effects of Fiscal Policy,” Journal of Money, Credit and Banking (August 1988, Part 2), pp. 47995. Laidler, David. “Taking Money Seriously,” Canadian Economic Journal (forthcoming). Auerbach, Alan J., and Laurence J. Kotlikoff. "Evaluating Fiscal Policy with a Dynamic Simulation Model," The Ameri can Economic Review (May 1987), pp. 49-55. Lebergott, Stanley. “ Discussion,” The Journal of Economic History (June 1986), pp. 367-71. Barro, Robert J. “The Ricardian Approach to Budget Deficits,” Henry Thornton Lecture, City University Business School, London, November 1987. Leiderman, Leonardo, and Assaf Razin. “Testing Ricardian Neutrality with an Intertemporal Stochastic Model,” Journal of Money, Credit and Banking (February 1988), pp. 1-21. Bernheim, B. Douglas. “ Ricardian Equivalence: An Evaluation of the Theory and Evidence," NBER Working Paper No. 2330 (July 1987). Leijonhufvud, Axel. "Constitutional Constraints on the Mone tary Powers of Governments,” in James A. Dorn and Anna J. Schwartz, ed., The Search for Stable Money: Essays on Mone tary Reform (University of Chicago Press, 1987). Blinder, Alan S. “The Challenge of High Unemployment,” The American Economic Review (May 1988), pp. 1-15. _________“ Keynes After Lucas," Eastern Economic Journal (July-September 1986), pp. 209-16. Levy, Michael E. Fiscal Policy, Cycles and Growth (The Confer ence Board, 1963). Levy, Michael E., et. al. Federal Budget Deficits and The U.S. Economy (The Conference Board, 1984). Bossons, John. “ Issues in the Analysis of Government Defi cits,” in John Sargent, ed., Fiscal and Monetary Policy (Uni versity Toronto Press, 1986), pp. 85-112. Liesner, Thelma. Economic Statistics: 1900-1983 (The Econo mist Publications Ltd., 1985). Bruce, Neil and Douglas D. Purvis. “ Consequences of Govern ment Budget Deficits,” in John Sargent, ed., Fiscal and Mone tary Policy (University of Toronto Press, 1986), pp. 43-84. Mankiw, N. Gregory. “ Recent Developments in Macroeconom ics: A Very Quick Refresher Course,” Journal of Money, Credit and Banking (August 1988, Part 2), pp. 436-49, Buiter, William H. “ A Guide to Public Sector Debt and Deficits,” Economic Policy: A European Forum (November 1985), pp. 14-79. McCallum, Bennett T. “ Are Bond-financed Deficits Inflation ary?: A Ricardian Analysis,” Journal of Political Economy (February 1984), pp. 123-35. Carlson, Keith M. “ How Much Lower Can the Unemployment Rate Go?," this Review (July/August 1988), pp. 44-57. ---------------"On Real’ and Sticky-Price’ Theories of the Busi ness Cycle,” Journal of Money, Credit and Banking (Novem ber 1986), pp. 397-414. _________“ Federal Fiscal Policy Since the Employment Act of 1946,” this Review (December 1987), pp. 14-30. de Leeuw, Frank, and Thomas M. Holloway. “The HighEmployment Budget and Potential Output: A Response,” Survey of Current Business (November 1982), pp. 33-35. Eisner, Robert. “The Revolution Restored: Keynesian Unem ployment, Inflation and Budget Deficits,” Eastern Economic Journal (July-September 1986), pp. 217-21. Erceg, John J., and Theodore G. Bernard. “ Federal Budget Deficits: Sources and Forecasts,” Federal Reserve Bank of Cleveland, Economic Commentary (March 15, 1988). Evans, Paul. “ Are Consumers Ricardian? Evidence for the United States,” Journal of Political Economy (October 1988), pp. 983-1004. Fellner, William. “The High-Employment Budget and Potential Output: A Critique," Survey of Current Business (November 1982), pp. 26-33. Fischer, Stanley. “ Long Term Contracts, Rational Expecta tions, and the Optimal Money Supply Rule,” Journal of Politi cal Economy (February 1977), pp. 191-205. Friedman, Milton. “The Role of Monetary Policy,” The Ameri can Economic Review (March 1968), pp. 1-17. Gordon, Robert J., ed. The American Business Cycle: Continu ity and Change, NBER (University of Chicago Press, 1986). Heilbroner, Robert L. “ The Importance of Red Ink: How I Learned to Love the Deficit," The New York Times, September 4,1988. Modigliani, Franco. “ Comment on R. J. Barro, U.S. Deficits Since World War I,’ ” The Scandinavian Journal of Economics, Vol. 88, No. 1 (1986a), pp. 223-34. _________The Debate over Stabilization Policy (Cambridge University Press, 1986b). Mundell, R. A. “Capital Mobility and Stabilization Policy under Fixed and Flexible Exchange Rates,” The Canadian Journal of Economics and Political Science (November 1963), pp. 47585. The New Keynesian Microfoundations: “ Discussion,” in NBER Macroeconomics Annual 1987 (1987), pp. 114-16. Niehans, Jurg. “ Classical Monetary Theory, New and Old,” Journal of Money, Credit and Banking (November 1987), pp. 409-24. O’Driscoll, Gerald P., Jr. “The Ricardian Nonequivalance Theorem,” Journal of Political Economy (February 1977), pp. 207-10. Phelps, Edmund S. “ Phillips Curves, Expectations of Inflation, and Optimal Unemployment Over Time,” Economica (August 1967), pp. 254-81. Phillips, A. W. “ The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United King dom, 1861-1957,” Economica (November 1958), pp. 283-99. Prescott, Edward C. "Comment,” NBER Macroeconomics Annual 1987, pp. 110-14. NOVEMBER DECEMBER 1988 60 Romer, Christina D. “ Is the Stabilization of the Postwar Econ omy a Figment of the Data?” The American Economic Review (June 1986a), pp. 314-34. _________“ Spurious Volatility in Historical Unemployment Data," Journal of Political Economy (February 1986b), pp. 1 37. ________ “ New Estimates of Prewar Gross National Product and Unemployment," The Journal of Economic History (June 1986c), pp. 341-52. _________“World War I and the Post War Depression: A Reinterpretation Based on Alternative Estimates of GNP,” Journal of Monetary Economics (July 1988), pp. 91-115. Rotemberg, Julio J. "The New Keynesian Microfoundations,” NBER Macroeconomics Annual 1987, pp. 69-104. Seater, John J. "Does Government Debt Matter?: A Review," Journal of Monetary Economics (July 1985), pp. 121-32. Sickel, Daniel E. “ Business Cycle Asymmetry: A Deeper Look,” unpublished manuscript, Board of Governors of the Federal Reserve System (1988). FEDERAL RESERVE BANK OF ST. LOUIS Sturrock, John, and George Iden. “ Deficits and Interest Rates: Theoretical Issues and Empirical Evidence," Congressional Budget Office Staff Working Paper (1988). Taylor, John B. “ Aggregate Dynamics and Staggered Con tracts,” Journal of Political Economy (February 1980), pp. 1 23. Tanner, J. Ernest. “An Empirical Investigation of Tax Discount ing,” Journal of Money, Credit and Banking (May 1979), pp. 214-18. Thornton, Daniel L. “The Government Budget Constraint with Endogenous Money,” Journal of Macroeconomics (Winter 1984), pp. 57-67. ---------------“ Monetizing the Debt," this Review (December 1984b), pp. 30-43. Weir, David R. “The Reliability of Historical Macroeconomic Data for Comparing Cyclical Stability,” The Journal of Eco nomic History (June 1986), pp. 353-65. FEDERAL RESERVE BANK OF ST. LOUIS REVIEW INDEX 1988 JANUARY/FEBRUARY JULY/AUGUST K. Alec Chrystal and Geoffrey E. Wood, “Are Trade Deficits a Problem?” William G. Dewald, “ Monetarism Is Dead; Long Live the Quantity Theory” Cletus C. Coughlin, K. Alec Chrystal and Geoffrey E. Wood, “ Protectionist Trade Policies: A Survey of Theory, Evidence and Rationale” Michael T. Belongia, “ Prospects for International Policy Coordination: Some Lessons from the EMS" Daniel L. Thornton, “The Borrowed-Reserves Oper ating Procedure: Theory and Evidence” Kenneth C. Carraro, “ Farm Policy and Mandatory Supply Controls — The Case of Tobacco” Clemens J. M. Kool and John A. Tatom, “ Interna tional Linkages in the Term Structure of Interest Rates” Keith M. Carlson, "How Much Lower Can the Unem ployment Rate Go?” R. Alton Gilbert, “ A Comparison of Proposals to Restructure the U.S. Financial System” MARCH/APRIL R. W. Hafer and Joseph H. Haslag, “The FOMC in 1987: The Effects of a Falling Dollar and the Stock Market Collapse” Thomas B. Mandelbaum, “The District Business Economy in 1987: The Expansion Continues” Kenneth C. Carraro, “The 1987 Agricultural Recov ery: A District Perspective” SEPTEMBER/OCTOBER Mack Ott, “ Have U.S. Exports Been Larger Than Reported?” Cletus C. Coughlin and Thomas B. Mandelbaum, “Why Have State Per Capita Incomes Diverged Recently?” Lynn M. Barry, “ District Bank Performance in 1987: Bigger Is Not Necessarily Better” Michael T. Belongia and Kees G. Koedijk, “ Testing the Expectations Model of the Term Structure: Some Conjectures on the Effects of Institutional Changes” MAY/JUNE Albert E. Burger, “The Puzzling Growth of the Mone tary Aggregates in the 1980s" Gerald P. Dwyer, Jr., and R. W. Hafer, “ Is Money Irrelevant?” G. J. Santoni, “The October Crash: Some Evidence on the Cascade Theory” Cletus C. Coughlin, “ The Competitive Nature of State Spending on the Promotion of Manufacturing Exports” Tobias F. Rotheli, “ Money Demand and Inflation in Switzerland: An Application of the Pascal Lag Tech nique" Daniel L. Thornton, “The Effect of Monetary Policy on Short-Term Interest Rates” NOVEMBER/DECEMBER Gerald P. Dwyer, Jr., and R. W. Hafer, “Are National Stock Markets Linked?” Michael T. Belongia, “Are Economic Forecasts by Government Agencies Biased? Accurate?” John A. Tatom, “The Link Between the Value of the Dollar, U.S. Trade and Manufacturing Output: Some Recent Evidence” Cletus C. Coughlin and Kenneth C. Carraro, “The Dubious Success of Export Subsidies for Wheat” K. Alec Chrystal and Daniel L. Thornton, “The Mac roeconomic Effects of Deficit Spending: A Review” NOVEMBER/DECEMBER 1988