The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
July/august FEDERAL FlE SffiFtV IC BANK P M 3 5 1 A The Battle for Energy Independence: How Much of a Good Thing? FEDERAL RESERVE BANK The Auto Industry: Slowdown in Sales, Stall in Jobs Forecasting the Economy with Mathematical Models: Is It Worth the Effort? business m h r FEDERAL RESERVE BANK of PHILADELPHIA IN THIS ISSUE . . . The Battle for Energy Independence: How Much of a Good Thing? . . . America's vulnerability to periodic oil embargoes can be lessened, but the goal of achieving energy self-sufficiency in the next few years could cost more than it is worth. The Auto Industry: Slowdown in Sales, Stall in Jobs . . . As consumers curtail their purchases of big-ticket items, a major casualty is the automobile industry where the jobless rate exceeds the overall national level. Forecasting the Economy With Mathematical Models: Is It Worth the Effort? . . . Econometric forecasting is a long way from being an exact science, but prognos ticators can continue improving their models by incorporating judgmental data and cor recting past errors. On our cover: Of the historical houses in Philadelphia's Fairmount Park that are opened to the public, Strawberry Mansion is the largest. It was once the home of U. S. District Court Judge William Lewis, a friend of George Washington. In 1798 Judge Lewis built the center section, naming the house Summerville. The wings were added in the mid-1820s by a subsequent owner. The present name stems from the early 1840s when a resident sold strawberries and cream to visitors. The furnishings of the mansion are Federal, Regency, and Empire. BUSINESS REVIEW > produced in the Department of Research. Editorial assistance is pro s vided by Robert Ritchie, Associate Editor. Ronald B. Williams is Art Director and Manager, Graphic Services. The authors will be glad to receive comments on their articles. Requests for additional copies should be addressed to Public Information, Federal Reserve http://fraser.stlouisfed.org/ of Philadelphia, Philadelphia, Pennsylvania 19105. Phone: (215) 574-6115. Bank Federal Reserve Bank of St. Louis The Battle for Energy Independence: How Much of a Good Thing?* By Tim othy H. Hannan Abundant low-cost energy has been fun damental to the American way of life for a long time. It's hard indeed to imagine Americans without their climate-controlled houses, aluminum cans, and large gasoline burning automobiles. Yet, as anyone who cooled his heels in a gasoline line last year can testify, a stable source of abundant lowcost energy can no longer be taken for granted. Domestic demand for energy has in creased rapidly in recent years; domestic supply has not. To help fill this widening gap, Uncle Sam has relied increasingly on imports from the Middle East, where a volatile mixture of oil and politics has already re sulted in one serious embargo and poses an ever-present threat of future embargoes. As the recent gasoline lines and closed fac tories so dramatically demonstrated, a sud den curtailment of foreign oil can cause con siderable economic disruption in a nation grown accustomed to relative energy abun dance. To reduce the threat of similar economic disruptions in the future, the na tion has embarked on a policy of energy selfsufficiency. Government funds are being allocated to stimulate research and develop ment of alternative sources of energy, volun tary conservation efforts are being pro moted, and— just to help voluntary conser vation along— tariffs are being imposed on imported oil. All of this brings up the question of the desirability of these efforts and the degree to which they should be pursued to bring about energy self-sufficiency. As economists never tire of proclaiming, resources are not limit less. The economy cannot at the same time satisfy all desires for more goods and ser vices, higher quality environment, and great er reliance on domestic production of energy. In the area of energy policy, this means that hard choices must be made not only among the various methods of reducing *This article deals primarily with the economic issues involved in seeking energy independence. Political or diplomatic considerations also may be important in de termining the degree of energy self-sufficiency appro priate for the United States. 3 BUSINESS REVIEW JULY-AUGUST 1975 Voluntary Conservation. In addition to ef forts designed to increase domestic energy production, a reduction in dependence on foreign sources of energy can also be achieved by policies designed to reduce domestic demand. Voluntary conservation is a currently practiced example of such a poli cy, and it has met with at least limited suc cess. However, often self-interest and the goals of voluntary conservation don't jibe. An individual who believes his neighbors will adequately conserve energy may find it in his self-interest not to do so. Because of this "free-rider problem," as economists often call it, conservation on a voluntary basis is generally recognized as having significant limitations. For this reason, policymakers have increasingly called for mandatory, and perhaps less palatable, means of reducing energy dependence. energy dependence but also among the vari ous levels to which energy dependence should ideally be reduced. Because re sources are scarce, complete energy selfsufficiency in the near future may come at a very high price indeed. ENERGY DEPENDENCE: HOW CAN IT BE REDUCED? Uncle Sam's arsenal contains many weapons to combat the energy problem. Most are designed to cut U. S. consumption of energy, boost domestic production of energy, or perhaps achieve some combina tion of thetwo. But as the current debate over energy policy serves to emphasize, the vari ous methods of reducing energy depen dence are not identical, and much con troversy remains concerning the appropriate path to follow. Consider a few of the more important alternatives available. Rationing. Mandatory conservation through rationing is one such policy and has in fact been proposed by a number of na tional leaders. The problems involved in de veloping an equitable rationing system, however, are simply enormous. Decisions would have to be made on how to allocate gasoline, fuel oils, jet fuel, diesel fuel, and many other refinery products to the thousands of categories of consumers—a function which, according to Treasury Sec retary William E. Simon, would require 15,000 to 20,000 full-time employees, incur $2 billion in Federal costs, and require 3000 state and local boards to handle the exceptions.2 Perhaps more important, rationing does not provide the needed incentives for suppliers Research and Development. Governmentfunded research designed to accelerate development of alternative sources of energy can play an important role in en hancing the nation's domestic production of energy, particularly in the long run.1 The future availability of low-cost energy from nuclear, solar, and geothermal sources, or from synthetic fuels and oil shale deposits, may require substantial investments in re search and development. Although the return to such investments may prove quite significant, so too may be the time required for these investments to pay off in the form of abundant low-cost energy. Thus, research and development of new technologies is generally viewed as having only long-run significance. velopment may provide a large gain to the economy as a whole, but there may be little opportunity for any one firm to derive a large enough part of this gain to warrant undertaking the research. Hence, Government partici pation in such efforts is needed. ’Although the private sector must be counted on to undertake most of the energy research and develop ment, Government-funded research may prove to be quite important. Development of new energy tech nologies often involves expanding basic knowledge of fundamental processes. In such cases, research and de- Statement of the Hon. William E. Simon, Secretary of the Treasury, before the Ways and Means Committee of the U.S. House of Representatives, January 22, 1974, Department o f Treasury News, pp. 9-10. 4 FEDERAL RESERVE BANK OF PHILADELPHIA and six-inch insulation remarkably “good buys." Second, unlike a policy of voluntary con servation or mandatory conservation through rationing, the impact of the tariff in reducing energy dependence is not limited to that of simply discouraging consumption. This is because a price rise brought on by the tariff will also increase the incentives of domestic producers to bring more energy to the market. Economic rewards are important. Faced with a rise in the price of energy, pro ducers of coal, oil, and other sources of energy can be expected to search for and develop additional sources. Energy deposits identified by geologists but previously too costly to work— such as the vast oil shale deposits in Colorado and Wyoming— may now be tapped simply because higher prices make doing so profitable. And efforts to de velop new technologies in the production of energy may be stimulated for the same reason. Thus, by raising the prices we must pay for energy, a tariff on imported oil both reduces domestic consumption of energy and in creases dorhestic production— making the nation less dependent on foreign sources of energy. of domestic energy to increase domestic production. Without new energy produc tion, rationing would continue to be needed many years into the future. The Tariff. Imposing a tariff on imported oil is another tool available to policymakers. A tariff is simply a tax placed on each unitor the value of each unit of an imported good, and its imposition on oil is designed to increase the price paid for imported oil. Of major sig nificance is the tariff's effect on the price of domestic oil. With the imposition of a tariff, domestic oil becomes relatively more attrac tive to consumers of energy. As long as the price of foreign oil exceeds that of domestic oil, users will try to buy from domestic pro ducers. When this happens (and as long as at least some domestic oil is not subject to Gov ernment price controls), the average price of domestic oil will be bid up to a higher level.3 Because of the dual role of prices in dis couraging consumption and promoting pro duction, this whole process results in less dependence on foreign energy sources. First, the rise in the price of oil, both foreign and domestic, will cause domestic purchas ers of energy to review their expenditures and cut down on the more easily avoided uses of energy. In the industrial sector, for example, firms that did not consider energy conservation measures worthwhile when energy prices were low will now find it profit able to eliminate heat leaks, switch to less energy-intensive technologies, or improve waste-heat recovery systems. Consumers who once drove large automobiles 30 miles to work and failed to insulate their homes will now find public transportation, small cars, The Quota. Unlike the tariff, the quota re stricts imports in terms of quantities, rather than in terms of a tax on each unit or on the value of each unit. Its impact, however, is quite similar. Like the tariff, the quota (by directly reducing the supply of imported oil, rather than by directly increasing its price) causes an increase in demand for domestic energy. Since a significant portion of domes tic energy production is not subject to price controls, this means that the average price of domestic energy will rise, performing the dual function of discouraging domestic con sumption and encouraging long-run domes tic production. Thus, the quota, like the tariff, provides policymakers with a doublebarreled weapon that can be used to make the nation more self-sufficient in energy. Governm ent price controls are currently in effect on only a portion of domestically produced crude oil. In applying price controls, a distinction has been made between “ old oil" and “ new oil.” New oil isdefined as all oil produced on a property in excess of output in the same month of 1972. New oil and oil from wells produc ing less than ten barrels per day are not subject to price controls. Domestic “ old oil," however, is currently held at a price of $5.25 per barrel. 5 JULY-AUGUST 1975 BUSINESS REVIEW tariff arrangements are designed to reduce imports by either reducing domestic con sumption of energy, increasing domestic production, or achieving some combination of the two. But as some economists have The tariff and the quota can differ in terms of the revenue that they generate for the Government or in terms of the predictability of their economic impact (see Box 1). In gen eral, however, the similarities are more strikBOX 1 TARIFFS AND QUOTAS: THE SIMILARITIES AND THE DIFFERENCES The economic impact of tariffs and quotas can be quite similar. In fact, for any given tariff, there is a theoretically equivalent quota. If supply and demand responses to price changes are known with certainty, it is possible to predict the level of imports that will result under a certain tariff and simply impose that quota to achieve the same result. There are, however, some potential differences between the two means of restricting imports. One potential difference is the revenue that they generate for Uncle Sam's coffers. Since a tariff is a tax, it provides revenue for the Treasury as long as it doesn't discourage all imports. But a quota is not a tax. It simply sets the level of imports allowed into the country and therefore does not generally provide revenue to the Government. Both means of restricting oil imports cause the domestic price to rise above the world price, but the difference goes to the Government in the case of the tariff and usually to the oil importers in the case of the quota. However, even this distinction can be eliminated if, under a quota, the Government chooses to auction off import licenses. By pursuing such a scheme, the Government could obtain roughly the same funds from selling import licenses under a quota as could be collected under a tariff. With the right conditions, both approaches can generate the same revenue. A potentially more important difference between a tariff and a quota stems from the fact that it is often not possible to predict future changes in supply and demand conditions. Under these circumstances, tariffs and quotas thought to be the same can have divergent results. For example, if world oil prices decline unexpectedly, a tariff will result in an unexpected increase in the percentage of the domestic market supplied by foreign oil, while a quota will not. Also, the failure of domestic supply to expand as expected will lead under a tariff to an increase in imports, but under a quota it will cause an unanticipated increase in the price of domestic oil. Because of uncertainty, the tariff and quota can lead to unexpected and different results. been pointing out, there are also ways to soften those periodic blows from the Middle East without significantly reducing overall imports of oil, and a policy of oil storage is perhaps the most frequently mentioned example. Storage performs the function of being an alternate source of supply when the going gets rough. By stockpiling oil bought from ing than the differences. Both provide an in centive for domestic production, both dis courage domestic consumption, and, to bring about these results, both require that we pay higher prices for energy. Oil Storage. Policies such as Governmentfunded research and development, volun tary conservation, rationing, and quota or 6 FEDERAL RESERVE BANK OF PHILADELPHIA oil can be so severe in the short run, there may be a positive gain from policies designed to discourage imports gradually in the long run. These long-run policies can cause the economy to make adjustments without the major disruptions associated with sudden embargoes. By cutting consumption, increasing pro duction, or stockpiling reserves, the country can help protect itself from future embar goes. Of particular importance, the nation's foreign and domestic policies do not have to be unduly influenced by foreign producers of oil. But while there's something to be gained from such policies, there are also significant costs. Because resources are indeed scarce, reducing the nation's vulnerability to foreign oil embargoes requires sacrifice. If it is to be achieved through increased domestic pro duction, large expenditures maybe required for further exploration and for research and development of alternate sources of energy. If it is to be achieved by reducing domestic consumption, money will have to be spent on better insulation, more efficient engines, and improved heat-recovery systems. More over, we will have to get along on less energy consumption even when embargoes are not underway. Tariffs and quotas also impose these kinds of costs since they are simply tools designed to increase production and decrease consumption. And because they do so by raising the price of energy, they also bring about higher gas prices, higher heating fuel costs, and higher prices of goods whose production requires large amounts of energy. Even an oil storage policy, which is not de signed specifically to reduce consumption or increase production, may require consider able sacrifice in the form of large expenditures on oil storage facilities. foreign sources or by storing domestic oil in the ground in the form of reserve capacity, sudden shortages of imported oil can be par tially or totally filled by dipping into a stockpile accumulated for just such a rainy day. Oil storage, then, is another of the many potentially useful steps that can be taken to ensure a steady supply of energy. REDUCING ENERGY DEPENDENCE: THE GAINS AND THE COSTS Clearly, there is a potential gain to all such efforts designed to reduce the nation's vul nerability to oil embargoes.4 When the spigots are turned off temporarily in the Middle East the resulting economic disrup tions can cause considerable hardships. This is because domestic supply patterns and domestic consumption patterns cannot be changed readily at a moment's notice. It takes time to expand domestic energy pro duction and introduce expensive production technologies which are not required when Middle East oil is flowing freely. And on the consumption side, it takes time to change over to more energy-efficient applicances, smaller automobiles, better-insulated build ings, and less energy-intensive technologies in commerce and industry. Because of this short-run inability to adjust to less energy, sudden embargoes can mean production bottlenecks, factory layoffs, cold homes, and other hardships. Therefore, the advantage of policies designed to avoid or reduce their impact can be large. This can be true even of policies such as a tariff or a quota, which are designed to replace temporary curtailments in imported oil with a permanent one. Be cause periodic sharp reductions in imported 4ln addition to avoiding or reducing the impact of embargoes, policies designed to make the nation more self-sufficient in energy can also help the balance of payments problem, {However, since fluctuating ex change rates tend to correct imbalances in the balance of payments, this advantage may not be a very significant one. THE QUESTION OF POLICY As is the case with so many economic prob lems, hard choices must be made among competing ends. To protect the nation from 7 JULY-AUGUST 1975 BUSINESS REVIEW sufficiency even in the near future if we are willing to pay the price for it. Imports of foreign energy can be prohibited by quota, extreme conservation measures can be im posed, or tariffs can be set high enough to discourage all imports of oil, causing the price of energy to rise until the domestic supply of energy satisfies domestic demand. (See Box 2.) All of this can be done, but is a policy of energy self-sufficiency, carried to this extreme, worth the costs? There are a number of reasons to suggest that striving for total self-sufficiency, at least in the near fu ture, may not be worth the sacrifice. future oil embargoes, substantial sums may have to be expended and hardships may have to be endured. This means that the benefits of reducing the country's vulnerability to foreign oil embargoes must be weighed against the costs of bringing about such a result. In such circumstances, economists often apply a simple rule: increase the activity so long as the additional gain that results ex ceeds the additional cost. In the present case, this means that it is worthwhile to in crease activities such as research and de velopment efforts, oil storage programs, tariffs or quotas, and conservation programs only to the point where the additional gain associated with insulation from embargoes equals the increased costs of such efforts. Beyond such a point, devoting more re sources to the effort simply will not pay. Where this point lies is always difficult to determine without further information.5This framework, however, does establish the probability that a number of policies de signed to reduce our vulnerability to foreign embargoes— tariffs, research and develop ment, and oil storage, for example— may in deed be justified up to a point. But perhaps more important, it can prove useful in analyz ing the desirability of a much publicized goal—that of achieving complete energy self-sufficiency. Those Last Steps toward Self-Sufficiency. One reason is that as the U. S. approaches energy self-sufficiency, the cost of taking such additional steps may increase, while the advantage of making an already relatively self-sufficient nation still more sufficient may not be great. The additional costs are particu larly important. The nation moves toward energy self-sufficiency by expanding domes tic production and reducing domestic de mand, but the further that either of these activities are pursued, the greater will be the sacrifice required. Expanding domestic sup ply in the near future will require that we turn to increasingly costly methods of energy production, and reducing domestic con sumption will require that increasingly high valued uses of energy be abandoned. The sacrifice required to change the thermostat from 75 to 65 degrees may not be great, but that required by an additional 10-degree twist of the dial may be substantial. It is for these reasons that total energy self-suf ficiency, at least in the near future, may be too much of a good thing. Put simply, the gain from making those last steps toward energy self-sufficiency may not be worth the higher costs required to complete the trip. It may be better to settle for something less. COMPLETE ENERGY SELF-SUFFICIENCY? To reduce the nation's dependence on un stable sources of foreign energy is one thing; to eliminate it is another. This difference in degree can be extremely important. It is no doubt possible to achieve total energy self 5On the one hand, if the probability of a recurrence of last year's embargo is low, as many believe, then the fruits of even the smallest efforts to reduce the nation's vulnerability to foreign oil embargoes may not be worth the cost. On the other hand, if the probability of recur ring embargoes is high, then substantial efforts may be justified. Risk-Free Sources of Foreign Energy. Not all of the oil currently being imported into this country comes from the politically volatile 8 FEDERAL RESERVE BANK OF PHILADELPHIA BOX 2 THE "PRICE'' OF ENERGY SELF-SUFFICIENCY A rough idea of the energy prices required to achieve energy self-sufficiency by 1980 can be obtained from a number of supply and demand estimates presented below. ENERGY EQUILIBRIUM IN 1980 Millions of Barrels of Oil per Day Equivalent, at Prices Per Barrel* $7 $9 $11 Fuel Domestic Supply Crude oil and natural gas liquids (including Alaskan) Natural gas Coal Uranium and hydroelectric New technology Total Supply Domestic Demand Net imports 10.6 (2.0) 14.7 6.1 5.2 0.0 36.6 44.2 7.6 10.7 (2.0) 14.5 8.0 5.2 0.0 38.4 42.4 4.0 10.9 (2.0) 14.4 8.0 5.2 0.1 38.6 40.6 2.0 SOURCE: Energy Self-Sufficiency, An Economic Evaluation (Washington: American Enterprise Institute for Public Policy Research, 1974), p. 8. *A fuel is made "oil equivalent" by finding the number of barrels of oil which has the same heating value as a given quantity of that fuel. These estimates, which were derived from a number of statistical studies, indicate the supply of different fuels and the total domestic demand for energy that can be expected at the prices of $7, $9, and $11 per barrel (in constant 1973 dollars). As economic theory would suggest, higher prices mean more energy will be produced domestically and less of it will be consumed. But here is where part of the problem of energy self-sufficiency emerges. As should be noted from the Table, the expected supply of various types of energy in 1980 is relatively unresponsive to price increases. In addition, the reduction in domestic demand for energy that can be expected to result from a price increase is estimated to be quite small. This means that in order to reach the point at which domestic supply equals domestic demand, which is required if no energy is to be imported, we may have to pay prices significantly higher than $11 per barrel (in constant 1973 dollars). As can be seen, this is significantly higher than the price of energy that would be required if we relied on some imports. 9 BUSINESS REVIEW IULY-AUGUST 1975 Middle East. Much comes from countries that are less likely to institute embargoes. A suffi ciently restrictive policy can eliminate im ports from relatively secure sources just as well as it can eliminate those from insecure sources. But why bear the cost if little is to come out of it? The primary gain from reduc ing imports is the reduction in periodic dis ruptions resulting from embargoes, but if a source of supply is relatively secure, there is little reason to incur the higher costs re quired to eliminate such imports. This means that policies should be less restrictive toward secure sources of foreign energy than those required by insecure sources— yet another reason to question the advisability of total energy self-sufficiency. part of the gain from reducing imports is the resulting reduction in the economic impact of embargoes. But if a storage policy is insti tuted, embargoes become less serious, thus reducing the gain to be obtained by eliminat ing all oil imports. This does not necessarily mean that all efforts to increase energy selfsufficiency should be abandoned in the pres ence of a storage policy. Some movement toward self-sufficiency may still be justified. However, it does provide yet another reason to question the goal of independence from all sources of foreign energy. CONCLUSION Uncle Sam's arsenal contains many weapons that can be used to reduce the na tion's vulnerability to periodic oil embar goes. Some, such as voluntary conservation programs and mandatory conservation through rationing, are designed to reduce domestic consumption. Others, such as ef forts to develop alternative sources of ener gy, are designed to increase domestic pro duction. Still others, such as oil storage policies, are designed to soften the blow of periodic embargoes without significantly re ducing overall imports. Because all are cost ly, however, a proper balance must be struck between the gains and costs resulting from their use. Reducing the nation's vulnerability to a sudden oil embargo is important, but so too are the substantial sacrifices required to do it. Since periodic oil embargoes can cause serious economic disruptions, it may well pay to reduce our dependence on foreign sources of energy, at least to a degree. But running the full distance to achieve total self-sufficiency in the next few years may simply not be worth the cost required. S Oil Storage. If the goal of complete energy self-sufficiency means eliminating all oil im ports, then the advantage of oil storage policies is another reason why the goal may not be desirable. If the cost of storing oil and using it during embargoes is not excessive, it may well pay to store at least some oil to smooth out the disruptions when they oc cur.6 But if a policy of oil storage is undertaken, what does this mean for the goal of selfsufficiency? Simply stated, it reduces the need to eliminate all imports. A substantial 6Storage can take the form of either increasing domes tic reserve capacity or stockpiling oil purchased abroad. The question of whether reserve capacity or storage from foreign sources is better is a simple cost calcula tion. If the landed price of foreign oil plus storage is less than the incremental cost of developing domestic capac ity, then storage of foreign oil is preferable, and vice versa. 10 The Auto Industry: Slowdown in Sales, Stall in Jobs By Clara Prevo GHART 1 CONSUMERS HAVE RECENTLY SEEN THEIR INFLATION-ADJUSTED SPENDING POWER DROP . . . Billions of Constant (1958) Dollars Disposable Personal Income 625 600 575 550 — i * i SO U RC E: II III IV I II III 1973 1974 (Annual Rate Seasonally Adjusted) U. S. Department of Commerce. 11 IV I 1975 BUSINESS REVIEW JULY-AUGUST 1975 CHART 2 AND THEY RESPONDED IN PART BY CUTTING BACK ON THEIR PURCHASES OF DURABLES. 12 FEDERAL RESERVE BANK OF PHILADELPHIA 13 BUSINESS REVIEW JULY-AUGUST 1975 CHART 4 WHICH HELPED PUSH THE JOBLESS RATE IN THAT INDUSTRY ABOVE THE OVERALL RATE. Percent J F MA MJ J A S O N D J 1973 SO U RC E: F M A M J J A 1974 S O N D J U. S. Department of Labor, Bureau of Labor Statistics. 14 FM 1975 Forecasting the Economy with Mathematical Models: Is It Worth the Effort? By Nariman Behravesh The months following the Arab oil em bargo of 1973-74 could well go down in his tory as the nadir of the art and science of economic forecasting. The embargo, oil price increases, and the ensuing recession jarred the U. S. economy, leaving economists with forecasts that were in many cases em barrassingly wrong. For example, errors as sociated with price level and real GNP predic tions as much as tripled after mid-1973.1 Quite a comedown for those who in earlier years had earned high marks for forecasting! On average, forecasters who keyed their predictions only to mathematical or econo metric models were proved less accurate than those who relied on pure judgment or a combination of judgment and econo metrics.2 The quality of the forecasters' judgment helped to determine the relative accuracy of economic predictions duringthis period. Less clear-cut, though, is the degree to which econometric models helped or hin dered those who used them. Some skepticism about econometric fore casting is clearly justified. Mathematical models are still in their formative stages. When used to forecast the economy, they tend to underestimate the peaks (high points) and troughs (low points) in business cycles and to miss the timing of these busi ness cycle turns. Yet, most forecasters using econometric models can compensate for 1See Stephen K. McNees, "How Accurate Are Economic Forecasts?" New England Economic Review of the Federal Reserve Bank of Boston, November/ December 1974, pp. 2-19. 2lbid. A judgmental forecast is formulated without the help of an econometric model but depends on a variety of inputs including the forecaster's intuition, trend pro jections, and the use of leading indicators. 15 JULY-AUGUST 1975 BUSINESS REVIEW large extent, these internal variables may in fluence each other. For example, GNP is di rectly related to national income, which in fluences consumers' expenditures on goods and services, which in turn helps to deter mine GNP. However, these internal variables also depend on other variables such as Gov ernment expenditures, exports, tax rates, and lending rates of central banks— some of which may not be determined purely by economic forces. These variables can be called external variables because they are not explicitly determined by the model.3 A forecaster intending to use a model to pre dict economic activity must supply the pre dicted values for these external variables. weaknesses inherent in the models. These models are invaluable for zeroing in on the effects of policy changes on the economy. Moreover, since considerable research in empirical economics is being directed at refining these models, forecasters will probably find them increasingly useful aids for prognostication. INSIDE A PANDORA'S BOX An econometric model used by a forecast er is a set of mathematical and statistical rela tionships that purports to describe economic behavior. These models are based on eco nomic theory and, in the process of model building, the relationships in the models are estimated and tested using the historical data (see Box). Most econometric models used in predict ing the status of the economy are quite large (40 to 400 equations). These so-called macroeconometric models are designed to pre dict economic variables such as the Gross National Product, the price level, the un employment rate, and interest rates. Such variables, which are determined within the model, can be called internal variables. To a determination of whether a variable is internal or external to the model depends on its builder. For exam ple, some model builders may designate Government expenditures as an external variable since these expen ditures are determined by a number of noneconomic forces that the model cannot consider. Other model builders may feel that Government spending depends primarily on economic activity and, therefore, should be included among the internal variables and described explicitly by the model. Econometric models must al ways have some external variables; otherwise, the forecaster faces an everything-depends-on-everythingelse situation. ANATOMY OF AN ECONOMETRIC FORECAST Building a Model. If we were interested in building an econometric model our im mediate questions would be: What economic variables do we want to describe? What does economic theory have to say about these variables? What does the data show about these variables? Here is how these questions may be answered. Suppose, for example, we want an overall description of consumption behavior in the U. S. economy. A review of relevant economic theories might turn up this assertion: Aggregate consumption is related to disposable or after-tax income. If the data for consumption and disposable income were graphed (see Diagram), the scatter of points would lie nearly on a straight line with a slope of about nine-tenths. Then it could be said that on average in the U. S., nine-tenths of disposable income is used for consumption expenditures.* In mathematical terms, this relationship would be: Consumption = .9 x Disposable income. *The consumption relationship being described is a long-term one. The distinction between long- and short-term consumption will not be made in the interests of simplicity. 16 FEDERAL RESERVE BANK OF PHILADELPHIA RELATIONSHIP BETWEEN CONSUMPTION AND DISPOSABLE INCOME. Consumption 600 Billions of Dollars (1958) 500 400 300 200 ’35 100 '« ^?30 | 200 300 400 500 600 Disposable Income Notice that this simple relationship is not an exact one. For example, in the Depres sion and war years, consumption was less than nine-tenths of disposable income (that is, in the Diagram, the observations for these years fall below the line). The opposite is true for the '60s. The inexactness of this simple model can be traced to factors such as changes in wealth, depressions, and wars that have not been taken into account. The model builder can rewrite the consumption equation to account for the approximate nature of the model: Consumption = .9 x Disposable income + Error "Error" refers to all the factors that affect consumption which the modef builder has not taken into account. By includingsomeof these factors in the consumption equation, the size of the error can be reduced.** If this consumption model were used for forecasting, **lf more han one explanatory variable is used to describe consumption, plotting the data and fitting a line as we have done in Chart 2 would be difficult. However, there are statistical methods that can do the same thing. 1 17 BUSINESS REVIEW JULY-AUGUST 1975 the reduction of this error would be a step toward more accurate forecasts. Another salient characteristic about the scatter of points in Chart 2 is that if consump tion is below average (that is, below the line) in a particular year, then it is likely that it will be below average for a few years (the war years). The same is true when consump tion is above average (the '60s). This tells us that consumption patterns vary slowly in response to changes in the economy— that is, the "error" or unexplained portion of the consumption model is not random. In fact, this error is systematic and correlated with its past and future values (econometricians referto this type oferroras serially correlated). Systematic or serially correlated errors are common in macroeconometric models and should be taken into account when these models are used for forecasting. Forecasting with a Model. An econometric forecast is obtained by projecting the estimated model to include the year or years of interest. Suppose we were interested in predicting consumption expenditures in the United States in 1974 and 1975. If it were known that disposable income in those years was $650 billion and $750 billion (mea sured in 1958 dollars), respectively, then the simple model introduced above could be used to forecast consumption. This model would predict consumption in 1974 and 1975 to be $580 billion and $630 billion, respectively (also measured in 1958 dollars). Such forecasts are approximate, since by ignoring the other factors that affect con sumption in the simple model, these factors are ignored when this model is used to predict consumption. Sharp-eyed forecasters would have to decide if there were any factors that would induce more or less consumption in 1974 or 1975. For example, if it were expected that economic activity was slower than usual in these years, then con sumption would also be subpar; therefore, we would want to adjust the predicted consumption levels downward. In this way we would be able to consider the "other factors" which affect consumption and which the simple model does not take into account. A more sophisticated forecaster would weigh the possibility that if consump tion fell below average in any one year it may remain there in the following years (that is, economic variables may move slowly through time). To compensate, we would adjust consumption downward for a greater time. Thus, an econometric model tempered by the forecaster's judgment can yield better forecasts. Multiequation Models. The model presented above has a number of shortcomings. From a behavioral point of view, it is a simplistic model of consumption. From a forecasting point of view, this single-equation model depends on forecasts of dispos able income, which may be just as difficult to predict as consumption expenditure. Furthermore, disposable income is influenced by the level of consumption in the economy (since consumption contributes to GNP, which is directly related to dispos able income). These types of problems are usually solved by adding more equations to the model. Just as consumption forecasts required us to supply predictions of disposable income in the above model, forecasts of the internal variables of a large econometric model (such as GNP, prices, and unemployment) require predictions of the external variables (such as Government expenditures, taxes, and the money supply). Furthermore, in the same way that the consumption forecasts above could be modified to account for information not already included in the models, adjustments can be made to the forecasts of large econometric models. 18 FEDERAL RESERVE BANK OF PHILADELPHIA THE CYCLES PRODUCED BY AN ECONOMETRIC MODEL The value and reliability of macroeconometric models in forecasting business cycles can be studied in two ways. The first method compares the actual historical values of key internal variables such as real GNP with the values a forecaster would have obtained from the model. This method provides insight into the model's ability to duplicate the economic conditions which occurred, when it is sup plied with the actual historical values of the external variables. The second method compares the size and duration of fluctua tions for a predicted variable, such as real GNP, with the actual business cycle fluctua tions of that variable. Such a comparison would allow the forecaster to judge the reasonableness of the business cycles pro duced by the model when he has to rely on forecasts of the external variables. The model under scrutiny here represents the state of econometric model-building in the late 1960s. Chart 1 compares the actual values of real GNP from 1956 to 1965 with a historical fore cast of real GNP by an econometric model.4 The predicted values rise and fall at about the right time but don't trace out the cycles in real GNP very well. In fact, these forecast values underestimate both the peaks and the troughs in the actual series. One explana tion for this difference may be that the peaks and the troughs in the actual series were caused by unanticipated occurrences that the model was not "smart" enough to capture. These unanticipated events or "shocks" may have consisted of major strikes, changes in international markets, or shifts in Government policies that were not explicitly built into the model. The second method of analyzing the "tracking" record of an econometric model — looking at the long-run forecasts it gen erates— yields similar conclusions. Fore casters wishing to make long-run predic tions must begin by predicting the long-run changes in the external variables.5As a result, these long-term forecasts are no more accu rate than the predictions of the external vari ables supplied by the forecaster. For lack of better information, long-run forecasters usually assume that external variables will change slowly and with virtually no fluctua tions. However, this implies thatthe long-run forecasts generated by an econometric model may also be fluctuation-free (Chart 2— dashed line). Clearly, such forecasts do not trace out anything resembling a business cycle. More realistic cycles can be traced by econometric models if the modeler tries to account for the occurrence and impact on the economy of events such as wars, strikes, and embargoes. One way of doing this is to impose random shocks on the models (Chart 2 — dotted line). But these cycles are too fre quent and short-lived compared to an actual series such as in Chart 1. These cycles are too short because the model moves the economy back to a "normal" position immediately after the shock is felt. However, in reality, the economy often takes more time to adjust to such disruptions. If the model user spreads the impact of these shocks over a number of “ •Historical, or after-the-fact, forecasts used in the first method of analyzing the tracking record of econometric models require that the user provide values of the exter nal variables of the models (such as Government expen ditures, taxes, and exports). In these forecasts the external variables are set at their actual historical values. Data for these external variables are fed into an econo metric model which then predicts the values of internal variables such as real GNP, prices, and unemployment. Unfortunately, usable forecasts of the external vari ables may be as difficult to get as predictions of the internal variables. Short-term forecasts of variables such as Government spending, taxes, and money supply growth may be easily obtained through Government budget estimates and other sources. However, getting accurate long-run forecasts of such external variables is a tougher undertaking. This, in turn, undermines the ac curacy of all long-term forecasts, both econometric and judgmental. 19 BUSINESS REVIEW JULY-AUGUST 1975 CHART 1 MODELS TEND TO UNDERESTIMATE THE ACTUAL PEAKS AND TROUGHS OF A BUSINESS CYCLE. Billions (1958 Dollars) SO U RCE: B. G. Hickman, e<±, Econometric Models of Cyclical Behavior (New York: National Bureau of Economic Research, 1972). periods, the fluctuations in the predicted series are smoother and begin to resemble the actual fluctuations in the series. (Com pare the solid line in Chart 2 with the actual series in Chart 1.) Accordingly, model users must be wary of the fact that business cycles produced by econometric forecasts are less pronounced than those the economy normally experi ences.6 In part, this may be a result of the inability of the models to foresee and, there fore, cope with the impact of unanticipated events, especially those whose impacts are spread over a number of periods. Fortunate ly, judicious use of judgmental information can at least partially compensate for such model weaknesses. fore, cannot duplicate business cycle behavior. On the other hand, econometric models may be good represen tations of the economy if, indeed, business cycles are a result of shocks to an economy which would otherwise be stable. It can then be argued that no matter how good a model is, it will inevitably fail to predict some unantici pated shocks and, consequently, miss some business cycle fluctuations. 6 The smoothness of econometric forecasts relative to economic time series may be explained in two different ways. On the one hand, econometric models may not be good representations of economic structure and, there- 20 FEDERAL RESERVE BANK OF PHILADELPHIA CHART 2 LONG-RUN FORECASTS OF REAL GNP WITH AND WITHOUT SHOCKS. Size of the Variable 1975 SO U RC E: 1980 1985 B. G. Hickman, ed., Econometric Models of Cyclical Behavior (New York: National Bureau of Economic Research, 1972). THE TRACKING PERFORMANCE OF SOME MODELS clues as to their ability to track past business cycles.8 Although these models have * changed significantly since 1969, the state of the art has probably not changed enough to make the types of results presented here ob solete. How accurately have forecasters relying solely on some of the major econometric models been able to spot the timing and magnitude of business cycle turns?7A look at the 1969 versions of three models which make quarterly forecasts provides some “VictorZarnowitz, Charlotte Boschan, and Geoffrey H . Moore, “ Business Cycle Analysis of Econometric Model Simulation," in B. G. Hickman, e d ., Econometric Models o f Cyclical Behavior (New York: National Bureau of Economic Research, 1972), pp. 311-541. The models considered in this study are the Wharton Econometric Forecasting Unit m odel, the O ffice of Business Economics model, and the MIT-Penn-Fed model. 7Aturning pointinthe business cycle occurs when the economy shifts from a positive growth period to a nega tive growth period and vice versa. The former points are called peaks and the latter troughs in the reference-cycle terminology of the National Bureau of Economic Re search. 21 BUSINESS REVIEW JULY-AUGUST 1975 Spotting the Turning Points. Table 1 sum marizes the accuracy with which these three models were able to predict the timing of turning points for six-quarter historical fore casts.9 On average the historical forecasts spotted a turning point two-thirds of the time when the economy actually peaked or bot tomed out. There did not seem to be a ten dency on the part of the models to predict a turning point when one did not occur. The models tended to predict turns too soon. This is especially true for historical forecasts that preceded the turning point by three quarters. The closer the turning point to the start of the forecast period, the better the chance of calling the turn. These results did not differ for upturns or downturns. In order to correct such errors, it would help if the forecaster could pinpoint some of their sources. The forecasting mechanism of business cycles in many quarterly models is linked to investment and inventory cycles, both of which are leading indicators in business cycles.1 However, investment and 0 inventory cycles are not the only factors that account for business cycles in the economy. It is entirely possible that model builders haven't fully accounted for the complex linkages between such leading indicators and the economy. Generally, the closer the turning point, the more useful and reliable the information that signals the turn will be to the model. So, the closer the forecast is to the turning point the greater is the likeli hood that the model will correctly spot the cycle peaks and troughs. In general, a 9A six-quarter historical forecast starting, for example, three quarters ahead of the turning point, would begin nine months before the quarter in which the turn occur red and would end six months after the quarter of the turn. It should be remembered that for a historical fore cast the external variables are set to their actual historical values. 10Leading indicators are economic variables that will usually peak before the economy peaks and bottom out before the end of a recession. These indicators are iden tified and classified by the National Bureau of Economic Research. TABLE 1 HOW THREE MODELS* SPOTTED TURNING POINTS: 1957-61** Too Soon Too Late On Time 43% 26% 31% 37 28 35 28 36 Average of Forecasts Starting 3 Quarters Ahead of Turning Point Average of Forecasts Starting 2 Quarters Ahead of Turning Point Average of Forecasts Starting 1 Quarter Ahead of Turning Point Average of All Forecasts 33 29 39 35 *The three models in question are the 1969 versions of the Wharton, Bureau of Economic Analysis, and MIT-Penn-Fed models. **Victor Zarnowitz, Charlotte Boschan, and Geoffrey H. Moore, “ Business Cycle Analysis of Econometric Model Simulations," in B. G. Hickman, ed., Econometric Models o f Cyclical Behavior (New York: National Bureau of Economic Research, 1972), pp. 311-541. 22 FEDERAL RESERVE BANK OF PHILADELPHIA TABLE 2 HOW THREE MODELS* FARED IN PREDICTING THE SIZE OF PEAKS AND TROUGHS: 1957-61** Too Large Average of Forecasts Starting 3 Quarters Ahead of Turning Point Average of Forecasts Starting 2 Quarters Ahead of Turning Point Average of Forecasts Starting 1 Quarter Ahead of Turning Point Average of All Forecasts Average of Forecasts during Contractions Average of Forecasts during Expansions Too Small Correct 21% 54% 25% 15 62 23 15 17 14 21 55 57 57 56 30 26 29 23 *The three models in question are the 1969 versions of the Wharton, Bureau of Economic Analysis, and MIT-Penn-Fed models. **Zarnowitz, Boschan, and Moore, “ Business Cycle Analysis of Econometric Model Simulations/' in Hickman, ed., op. cit., pp. 311-541. modeler must assume that short-run fore casts are more accurate than longer-run ones. predict a turning point too soon. If the mod els called a peak or a trough too early, then at the peak or trough the predicted series would underestimate the actual rise or de cline that occurred. Undershoots can also result because the models ignore the cumulative effect of the "other factors" that are overlooked in the model structure. Here again, the closer the starting point of the forecast to the actual turning point, the bet ter and more plentiful the information signal ing the turn, and so the more accurate the forecasts. Predicting the Size of Peaks and Troughs. The Achilles heel of many macroeconometric models is their proclivity to smooth out busi ness cycles and, in so doing, undershoot the size of both peaks and troughs. The three models under consideration did, in fact, smooth over past cycles (see Table 2). These models tended to underestimate both peaks and troughs. The closer the beginning of the forecast was to the actual turn, the better the chance the models had of correctly predict ing the size of a peak or trough. On average, the models were better at foretelling the depth of the slide during a recession than they were at gauging the peak to which the economy rose before experiencing a con traction. Models undershoot the size of the peaks and troughs for several reasons. In part, this may be a result of the model's tendency to SHARPENING THE FORECASTS On the whole, this evidence suggests that, without adjustments by the forecaster, the tracking record of econometric models leaves some room for improvement. There are two general ways to hone the tracking and predictive abilities of econometric mod els. The first is numerically adjusting an exist ing model prediction to correct for past 23 JULY-AUGUST 1975 BUSINESS REVIEW misses and to impose the forecaster's judg ment. The second strategy is refining and im proving the model itself. Forecasters can improve their results by anticipating and mathematically correcting the tendencies of the models to smooth out economic fluctuations. This can be ac complished by looking at past error patterns (that is, the difference between the actual and the predicted series, such as in Chart 1), and adjusting the forecast to compensate for these errors. If, for example, a model tends to understate GNPgrowth during expansions and to overstate GNP growth during contrac tions, the model user can adjust GNP growth predicted by a model upward or downward to counteract this tendency. A great deal was learned about this process and about econometric models from the larger than usual forecasting errors made in the months right after the Arab oil embargo. Most econometric forecasters will also use their judgment to anticipate the impact on the economy of events they expect to occur. This information is then used for the neces sary adjustments to the forecast. For exam ple, during the Arab oil embargo economet ric forecasters tried to estimate the effect of the boycott on production and consumption activities and to fine-tune their models cor respondingly.1 The virtue of econometric 1 models is that these adjustments are fed through the model so that an embargo's im pact on the economy can be measured. Thus, correction of past error patterns and imposi tion of informal judgment on econometric models should, in general, yield better fore casts. The second method of improving econometric forecasts, which entails chang ing the structure of the model and updating it, could also result in improved forecasts. Econometric forecasts can be refined by try ing to incorporate other types of predictive information, such as anticipatory data, into the models. For example, a recent study has shown that incorporating the plant and equipment investment anticipations of the Bureau of Economic Analysis into a model can reduce the forecasting errors of business-fixed investment.1 To a lesser de 2 gree, incorporation of the University of Michigan's consumer sentiment index into a model will improve consumer expenditure forecasts. Including this anticipatory data also improves the ability of models to predict turning points. Still another way of upgrading the overall performance of econometric models entails "reestimating" the models continuously by adding new observations to the data base and recalculating the equations used for predic tion. Most macroeconometric models that are used commercially are reestimated every three to five years. Given their size, rees timating them more often is costly and im practical. Nevertheless, within a three- to five-year period institutional and behavioral changes in the economy could possibly in validate part of the model. For example, the high rates of inflation in 1974 may have al tered economic behavior. Econometric models which were estimated before then would have missed this change. Small mac roeconometric models can be reestimated every quarter when national income data are released. However, this type of reestimation alone is not sufficient to reduce significantly the forecasting errors of the models. Up graded econometric forecasting requires ad justing the model by employing judgment and the analysis of past errors. Finally, some research in economics is being directed at improving the structure of the models and at using economic data more efficiently in estimating and quantifying "See Donald L. Raiff, "Forecasting in a 'Shortage' Economy," Federal Reserve Bank of Philadelphia, 1974 (unpublished paper). 12F. Gerard Adams and Vijaya G. Duggal, "Anticipa tions Variables in an Econometric Model: Performance of the Anticipations Versions of Wharton Mark III," In ternational Economic Review 15 (1974): 267 - 83. 24 FEDERAL RESERVE BANK OF PHILADELPHIA policy menus for economic policymakers. It is relatively easy for an econometric model to provide a range of forecasts made under a variety of policy assumptions. As the impact of changes in Government expenditures and the growth in the money supply are traced through the model, the policymaker can de termine the effect of various policies on the economy. Finally, once a large econometric model has been built, it can be employed for pre dicting a multitude of economic variables with a small expenditure of time and effort. For example, some current models regularly predict as many as 400 variables. The judgmental forecasting of the same number of variables, on a regular basis, may be very time-consuming. these models. It is likely that functional rela tionships can be discovered and refined which will allow modelers to predict specific internal variables more precisely. WHY USE ECONOMETRIC MODELS AT ALL? Although econometric models, on their own, cannot track business cycles very well, they do provide an explicit and wellorganized framework within which the forecaster can apply judgment to improve their predictive ability. Judgmental forecast ers have some implicit model of economic behavior in mind to rely on in formulating their predictions. However, such models are rarely made public along with the judgmental forecasts. The advantage of econometric models is that one can readily pinpoint and, therefore, try to correct weaknesses in the model structure and the assumptions under lying the forecast. Another important advantage of econo metric models is the way in which adjust ments feed through the entire model to provide forecasts that are, at all times, con sonant with forecasters' theories of how the economy is structured. Obtaining consistent forecasts under a variety of assumptions is more difficult for a judgmental forecaster be cause the relationships between economic variables in a judgmental "model" are not as clearly defined as those in an econometric model. Econometric models also help serve up CONCLUSION Pure econometric forecasting does not provide very accurate predictions of the tim ing, size, and duration of business cycles. This is especially true for longer-run econo metric forecasts. Nevertheless, forecasters who adjust these models to impose judg mental information and to correct model errors can substantially improve their accuracy. Furthermore, flexibility and con tinued improvements of econometric fore casting relative to judgmental forecasting do make the efforts channeled into econometric model-building and predicting worthwhile. 25 BUSINESS REVIEW JULY-AUGUST 1975 ECONOMICS INFLATION Inflation is currently a major problem facing the U.S. Can policymakers curtail it? If so, how much will their actions "cost" society? Is inflation "bad," and if so, why? Are there ways of "living with inflation" that cushion its negative impact on the individual and society? Six articles reprinted from the Philadelphia Fed's Business Review these questions in det£ seek to promote an understanding of the problem for both policymakers and the general public. Copies are available free of charge. Please address all requests to Public Information, Federal Reserve Bank of Philadelphia, Philadelphia, PA 19105 FEDERAL RESERVE BANK OF PHILADELPHIA On December 31, 1974, Americans were permitted to buy and sell gold for the first time in some 40 years. Since then questions have been raised about the once-hallowed, almighty metal's worth and importance. For example, has its status in the United States and in the international monetary system changed? If so, in what manner? A pamphlet recently produced by the Philadelphia Fed's Department of Public Information con siders the role of gold— past, present, future. Copies are available free of charge. Please address all requests to Public Services, Federal Reserve Bank of Philadelphia, Philadelphia, PA 19105. riDKRAL RESERVE B N A 3 riDIRAL RESERVE BANK FEDERAL RESERVE BANK of PHILADELPHIA PHILADELPHIA, PENNSYLVANIA 19105 business review FEDERAL RESERVE BANK OF PHILADELPHIA PHILADELPHIA, PA. 19105