The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Foreword This This volume contains the papers presented at the October 7, 1977 Productive Capacity: Estimating the 1977 conference on "U. “U. S. Productive Utilization Gap," Utilization Gap,” held in St. Louis, Missouri. The conference, cosponsored by the Center for the Study of American Business the Federal Reserve Bank of St. Louis, examined the current and the current and projected estimates of U.S. productive capacity and utilization. Appropriate measures of of the gap between productive productive capacity utilized and full productive capacity are important to the development of appropriate appropriate economic policy. The three basic papers were presented by George Perry of the Brookings Institution, Peter Clark of the staff of the Council of Economic Advisers, and by Robert Rasche and Jack Tatom of Reserve Bank of St. Michigan State University and the Federal Reserve Louis. An economic policy address was given by William William Nordhaus, aa member of the Council of Economic Advisers. A paper by James Ragan of Kansas State University on projections of capacity utilization is also included in the Supplemental Supplemental Papers section of this volume. Discussion papers were presented by Pham Chi Chi Thanh of American University, Laurence Meyer of Washington University, and Frank deLeeuw deleeuw of of the Congressional Budget Office. • I POTENTIAL OUTPUT: RECENT ISSUES AND PRESENT TRENDS George L. Perry would be associated associated Potential output measures the real GNP that would with operating the economy at some specified level of labor utilization. The concept offers answers to two basic types of questions: be the level of GNP if unemployment was at a specified level? what would (Or what would unemployment be if GNP were at some specified level?) And what will unemployment be at some point in the future if GNP grows at some specified rate? (Or what will GNP be if some specified unemployment target is achieved at some point in the future?) In the process of developing the concept of potential and providing providing the needed estimates for answering these questions, we have gained aa number of insights into the cyclical characteristics of the economy. Okun's law, which summaOkun’s summa- rizes many of these characteristics in linking marginal output to marmarginal changes in unemployment rates, is probably the most robust macromacroeconomic relationship yet developed. Despite the general success of the original potential concept and related relationships such as Okun’s Okun's law, several developments of the l97Os have cast doubt on traditional methods of measuring the nation’s 197Os nation's Dr. Perry is Senior Fellow at the Brookings Institution. The author is grateful to Jesse Abraham for his research assistance and to Evelyn Taylor for her typing. The views expressed in this paper are the author’s and not necessarily those of the officers or trustees of the author's Brookings Institution, Institution. • I economic potential. First, the changing composition of the labor force, and more dramatically, of the unemployed impinge on potential output measures in two distinct ways: Labor input measured in efficiency units has been diverging from labor measured by a head count; and aa tradiona11y used as a benchmark for for measurmeasurconstant unemployment rate, tradionally coning potnetial output, has moved noticeably away from measuring a con- stant degree of labor utilization measured in efficiency units. In the unemploypast, I have addressed both these issues of labor force and unemployment composition, and they are both incorporated in recent official analyses of potential by the Council of Economic Advisers. Second, the slowdown in the growth of the capital stock -- which has been particpartic-- ularly marked since 1973 1973 once an allowance is made for investment that is going to meet environmental requirements -- has raised anew the -- question of whether explicit attention to the size and growth rate of in estimating potential output. The most recent capital is needed In official CEA estimates are based on analysis by Peter Clark~that Clarklf that takes account of variations such as these in capital stock growth. And the recent slowdown in the growth of the capital stock, measured after deducting an estimate of investments going to pollution abatement, is an important source of the slowdown in potential output growth estiesti- mated by CEA. Third, the dramatic rise in energy prices has caused some analysts to make estimates of potential that are seriously afaffected by this this energy price explosion. The most notable examples of this Rasche and John Tatom that this new wrinkle are the papers by Robert Rasche have been published in theFederal the Federal Reserve Bank of St. Louis Review.~ Review ..?! 22 Finally, my colleague, Edward Denison has called my attention to the importance for potential output measures of changes in the environment businesses operate -- including the rise in crime, pollution in which businesses —— abatement regulations, and regulations covering safety and health practices. His research in this area is still underway. In this paper, II will report on some work that deals explicitly with the first of these departures from tradition and that indirectly supports the last of these as well -- that is, the demographic issues —- and the Denison issues. But before getting into this analysis, I want But to turn to why I am ignoring explicit attention to the capital stock, although ideally II would like to integrate it into the analysis. And why I am ignoring the impact of energy prices, and think that giving that development aa prominent role in modifying potential output meameasurements is mistaken. The Case Against Using Capital It is hard to argue that capital should not be included in estimating potential output because everyone knows it belongs in the calculation. l96Os, the same CEA that introduced Back in the 1960s, introduced potential introduced output into the mainstream of policymaking and debate also introduced the investment credit in order to stimulate capital formation. capital is ignored, it is for a simple pragmatic reason: If one cannot fpr capital in a find an important or statistically significant role 15cr freely estimated aggregate production production function or any equivalent relation that one might use in estimating potential output. Although this negative result is well known, II thought II would would try again using 33 the newly—developed newly-developed data on the capital stock from the Bureau of Economic Economic Analysis. I tried, in turn, several versions of the capital stock including the total stock, equipment separately, the stock with estimated expenditures for pollution abatement subtracted, and the nonfarm business stock for manufacturing alone and for the total nonfann sector. None of these worked. This left me with a choice of research strategies. Constrain the capital stock to play some specified role in determining potential exoutput. Or see how well the trend in labor productivity can be ex- plained by taking account of cyclical factors and changes in labor force composition. Several considerations led me to opt for the latter approach. reAny capital stock series must rest on assumptions about re- tirements of physical capital from the stock. We probably do not know enough about these retirements and about whether they proceed smoothly or whether they, in turn, depend on current rates of investment. The degree to which the capital stock is utilized at any point in time is not only hard to measure but is a very uncertain concept at bottom. proNobody can ever explain how we had enough of a capital stock to pro- duce the output we did during World War II. generally, since it More generally, is the flow of productive services from from the capital stock that we prepresumably want to measure, we have to deal with the fact that the flow of services from aa given stock can be expanded simply by expanding the double-shift operations operations hours over which we utilize it. Thus moving to double—shift doubles the effective capital stock without any new investment taking place. AA department store that starts staying open from 66 to 9 in the 4 evening adds a third to the effective capital stock of that operation. ooeration. This consideration is particularly troublesome when we try to measure potential output since we are then interested in the effect of the capital stock at some relatively high rate of production -- precisely -- the situation in which a more intense utilization of the existing stock might be expected. When the economy reached aa 4 percent unemployment rate during 1966, the level then defining its potential, available measures of capacity utilization in manufacturing reached levels subsubstantially higher than we have observed since then, despite achieving even lower unemployment rates. meanIf these utilization figures are mean- from the capital ingful, they indicate that the available services from stock are quite expandable and are not closely linked to the level of the unemployment rate. Finally, even if we were ingenious enough to integrate these comcomcapital and Its its relation to potential plexities into our concept of of.capital output, we would still have to deal with the fact that the capital today's, but the stock that will exist stock that interests us is not today’s, systemat the time potential is achieved. The measured capital stock system- atically grows faster as the economy expands toward potential and more when actual output recedes from potential. slowly during recessions when To decide what potential output will be in 1981, we would have to forefore- cast the levels of investment that would take place each year in a move analysis. After recession, to potential and integrate these into the analysii. when the economy is well below beiow its potential, the capital stock always looks low relative to aa trend line estimate of potential output. 5 Of course, against these complications that arise from trying to utilize the capital stock in potential calculations, we have to weigh the possible improvement we could get if we correctly measured the relevant stock and its effect on labor productivity at potential. Clark has made a careful attempt at doing this and we may have to wait until until potential is approached to know which research strategy gives more accurate answers. ~seAainst acts The Case Against Hih High Estimates of Ener Energy Price Im Impacts The sharp increases in energy prices of recent years led to assumplarge reductions in potential output under certain restrictive assumptions about how energy and output are related. The simplest of these assumptions, and the one utilized by Rasche and Tatom, is that output in the business sector is governed by a Cobb-Douglas production funcfunction with capital (K), energy (E), and labor (L) as inputs and a disdisembodied productivity trend growing at the rate rater: r: (1) V = Aert La Kb Ec, a+b+c = In this assumed production relationship, the output elasticity of energy is one and its price elasticity is minus one, one. That is, doubling output will double energy demand while doubling the relative price of energy will cut the amount demanded in half. plausible. propositions seerrsentirely see~sentirelyplausible. The first of these these The second very unlikely. it lies behind the proposition proposi on that the rise in energy prices has substantially substantially reduced potential output. 66 Yet Measured by the wholesale price of energy to users, the relative price of energy rose 57 percent between 1973 and 1976. If we believe in the country, the elasticity of -1.0, and apply it to all energy used In our energy use should be only 64 percent of its former level (relative to trend and adjusting for output effects, both of which are relatively small by comparison). We oil-import problem and should should have no oil—import probably be exporting oil to the rest of the world -- unless they too -- -1.0 for their demands. had price elasticities of —1.0 Statistics on energy used in production, as opposed to use by government consumers, are hard to get and Rasche and Tatom resorted to using price data rather than quantity data in their production function. Cobb-Douglas assumptions, energy use is Since, under the Cobb—Douglas given by (2) EcVP~ where PE is the relative price of energy, their price series proxied for the unobserved quantities of fuel use. Up to 1973, there was not much variation in the relative price of fuel, so it was probably hard to view the resulting estimates very critically. But what has happened to energy use since 1973 shows the model is wildly unrealistic. And therefore so are its implications for potential output. I have made some estimates of how much business has curtailed energy use in response to the increase in fuel prices since 1973. An accurate total of business fuel consumption is hard to get, but I was able to assemble time series covering about 60 percent of the total. 7 The main omissions were coninercial commercial uses of petroleum for heating and transportation. I estimated the following follo\11ing relationship over the period 1949—1973, 1949-1973, (3) lnft = A+p1t+bflnU where EE Is is my series on BTUs used by business, QQ is gross business potenti a1 product, u U is utilization measured as the ratio of QQ to potential ft o,V is an estimate of the annual trend in IQ This equation says Is that, through time, energy per unit of output has displayed a trend of and p r1 pp . While at a point in time, energy use will be (l+b11)) perpercent greater for every one percent additional output that is produced. 1 per year. equaTable 11 shows three sets of coefficient estimates for this equa- tion, the estimates differing according to whether the utilization term is included and whether the equation is adjusted for serial correlation. time period used for estimating since data on They also differ in the time. UU were not available before 1954. The estimates indicate aa trend dee] ine of 1. decline 1.33 to 1. 1.66 percent per year in energy per unit of output. In the t,10 two equations that have it, the coefficient on the utilization term indicates aa cyclical elasticity of energy use with respect to Hov1ever, the estimate of bb11 has aa low t—statis— t-statisoutput of 1.3 to 1.4. However, tic in both equations 4a and 4b and equation 4c may be more reliable. The relative price of energy trended down through most of the estimation period and its effect cannot be separated from the trend in energy per unit unit of output. But the post-sample prediction errors from the Table 11 equations provide estimates of how much energy use has been 88 Table 1. Estimated Equations for Energy Per Unit of Ouput, 0uput, Private Business Sector~/ Sectors” Coefficients Equation number Constant AA Time p1 Percent errors Data Utilization Period b1 bl R 1973 1976 forecast -3.0% -3,0% -5.9% -3.8% —3.8% —7.0% -7.0% —2.1% -2. 1% -7. O;i, —7.0% D.W. S.E. S.E. rho prediction ‘C (4a) 2~9l 2. 91 (9.82) —0,01278 -0.01278 (—8.6) (-8.6) 0,4413 0.4413 ((1.43) l . 43) 1954-73 0.805 0.9 0.034 (4b) 3.04 (9. 4) (9.4) -0.01266 —0.01266 0. 3111 0.3111 (0.93) 1954-73 0.877 1954—73 (-5.9) (4c) 3.46 (115.8) —0.01557 -0.01557 (—8.0) (-8.0) — 1.3 - 0.027 0.561 o. 561 929 1.9 0.031 1949—73 0, 0.929 1949-73 0.641 o. 641 reduced as a result of the post-1973 price changes. The ratio of energy to output in the business sector declined by 10.2 percent between 1973 and 1976. Equations 4a, 4b, and 4c in Table 11 predicted declines of 7.3, 7.0, and 5.3 percent respectively. Thus my measure declined by 2.9 to 4.9 percent more than predicted, of energy use decli given the behavior of output over this period. Thissis is the response facone can attribute to higher energy prices or other, unspecified fac- tors, using the equations of Table 1. Since the relative price of energy rose by 57 percent over this period, the indicated price elasticity for business use of energy is betwqen betw~en 0.05 and 0.085. Low as these estimates are, they probably still overstate the true amount of energy saving that has occurred thus far. The Table 1l equations assume a constant trend through 1973 in energy per unit of business output. In fact, the nearer to 1973 one starts to estimate the trend, the steeper the estimated decline rate is. If the decline that would have occurred without the price explosion was greater than attrithe Table 1l equations indicate, the extra decline that can be attri- buted to the price explosion is correspondingly smaller. Other studies, based on long time series and on cross-sections, have estimated higher price elasticities than these. The absence of any spectacular change in the relative ative price of energy before 1973 would make any statistical estimates from time series up to that time uncertain; while cross-sections may reflect differences other than just the price of energy, and may not be useful for predicting the response to a change in price in the U.S. The present estimates come from the 10 first opportunity to observe the response of use to aa large change in It may be that three years is much too short a period to price. observe long-term effects; with more time, energy use may respond further. And it is probably true that total use of energy, as opposed to business use, is more price elastic. But it is business use that is relevant for potential output calculations. And it is the response to stabidate and over the next few years that is relevant, at least for stabilization problems, not the response that may eventually occur occur over a period of many years when the capital stock -- and eventually the —- geographical distribution of the population and life styles -- have all —— had a chance to evolve. Effects on Potential Output Knowledge of the quantity of energy conserved permits some guess at the decline in labor productivity and potential output that resulted from higher energy prices without making restrictive assumpassumptions about the form of an aggregate production function. Valued at 1976 prices, the estimated energy saving of 2.9 to 4.9 percent was worth $2.8 billion to $4.8 billion. Reducing this to 1973 relative prices, the range is $2.0 to $3.4 billion. Since the most profitable at the old price and the substitution of other factors are marginal at substitution least profitable profitable are marginal at the new, the midpoint is appropriate, providing aa range of $2.4 billion to $4.1 billion for the value of other inputs substituted substituted for energy, energy. Even if business is assumed to substituting labor for energy, have accomplished all this saving by substituting not much extra labor could have been used in this process. 11 11 $4.1 billion sector. is 0.5 percent of employee compensation in the business sector, billion is 0.3 percent. $2.4 substitution Since an unknown amount of the substitution must involve capital as well as labor, the added labor input would be smaller still. still. If we assume the substitution proportional to the substitution is proportional two-thirds, one—third one-third split usual two—thirds, split of shares between capital and labor, substituted falls to 0.2 percent to 0.33 percent. the estimate of labor substituted Finally, some part of the energy saving must have involved no substituFinally, tion of other inputs at all: lowering thermostats to 68 degrees in winter and raising them to 75 degrees in summer or turning out every other light in hallways are obvious examples, but there must have been many less obvious examples of "waste" waste that were eliminated only after the OPEC crisis made firms more energy conscious, conscious. The The amount of energy saving that involved labor substitution must be smaller than all this "costless" conservathe total energy saving by the amount of all ‘costless conservaI know of no way to pin down the answer more accurately; but on tion. the basis of the evidence here, it seems unlikely that higher energy prices have caused more than aa 0.2 percent loss of labor productivity and potential output between 1973 and 1976. It seems likely that there will be more energy saved in the longer run. But it also seems plausible that any growing conservation 11 come disproportionately from substituting capital rather of energy wi will than labor for energy. If the price elasticity of energy use after ten years is substantially greater than after three years, it is presumably because the capital stock will be changed much more over the longer period. Thus, in response to higher energy prices, we would predict an 12 unusually large amount of investment with greater confidence than we would predict an unusually slow growth in labor productivity. productivity. Don't Higher Hurt? her Prices Hurt The apparent paradox in all this is that the inability to substitute labor for energy has kept potenti potential output from being being prices. affected by the increase in energy prices, If the substitutability assumed by Rasche and Tatom were in fact available, our potential potential output would have fallen just as they described. our economic welfare has been reduced by OPEC. The answer is that Our consumption possipossi- bilities for other goods and services are smaller by the amount that our fuel bills are larger. Substitution would help us reduce the size of this fuel bill, but the possibilities for substitution are slight, at least in the short—run, short-run, as the low response of energy use to date has shown. shown, Since much of the added revenue from higher energy prices has gone to U.S. producers, it gets complicated to figure out exactly who is worse off and by how much. Furthermore, our exchange rate may have been affected by OPEC's OPEC’s price increase and the subsequent spending and trade investment decisions of oil producers, thus altering our terms of trade with the rest of the world and further muddying the full calculation of us. what it has cost us, such a measure. measure, But for the present purpose, we are not after For calculating potential output effects, we need to know how much labor productivity has been affected. the effect, thus far at least, is negligible. 13 And the answer is Potential GNP Estimated For the reasons I have just described, my own recent estimates of attentionto potential output are made without explicit attention to the capital stock or energy prices.W prices ..Y An examination of the residuals from my estimating equations should indicate whether these, or other, omissions are inappropriate. II define the potential path of the economy as the trend line of GNP passing through actual real GNP in mid-1955 real GHP mid—1955 and growing at aa rate that would hold the weighted unemployment rate at its mid—1955 mid-1955 level. This is similar to the long—established long-established benchmark for potential originally presented by Okun, except that the path is defined by concon- stant weighted unemployment rather than the conventional aggregate unemployment rate. Weighted unemployment measures underutilized labor in efficiency units rather than bodies. As such it is a better sunnary summary measure of labor market tightness than conventional unemployment. But it is not intended to define a "noninflationary" utili“noninflationary” level of labor utili- zation. That is another, and more complicated, matter. Over the past 15 years, the conventional unemployment rate along the potential path has drifted upward, from 4.1 percent to 4.9 percent in 1976. The main reasons for this are the declining proportion of high-weight adult males in the work force and the rising relative unemployment rates of low-weight young workers of both sexes. The drift in conventional unemployment rates along the potential path is very close to the drift along the CEA’s CEA's path, which is based on a similar treatment of the labor force. 14 The economy's economy’s potential labor force is calculated using participation rate equations for each of 14 total l4 demographic groups in the total labor force. These equations account for cyclical variability in the labor force of women and younger men. They also provide estimates of participation rates along c:1ong the potential path. the trends in participation Labor input is measured by weighting ,veighting the employment and labor force in each of the 14 age-sex age—sex categories by relative wages and summing them. Potential weighted labor force and weighted employment are obtained by adjusting particithe actuals to potential using the cyclical components of the participation rate equations. An equation estimating average hours worked per year, again with a cyclical and trend component, provides estimates of potential average hours. Multiplying this by potential weighted employment each year gives potential weighted total hours, the basic measure of potential labor input in the analysis. analysis. Weighted labor productivity is defined as labor input divided by output. Since the labor input measure is already weighted to take account of average productivity productivity differences among workers in different different demographic groups, weighted productivity is already cleansed of this productivity source of cyclical variaticn variation and trend in conventional productivity measures. business The relationship between labor input and output in the business sector provides the basis for examining the behavior of weighted proproeconomy, ductivity and estimating aa potential output path for the economy. The productbasic model starts out with the proposition that weighted labor product- ivity grows exponentially along the potential path: it -9- (5) C) - where Q where Q is output in the business sector, r is the annual growth rate, tis indicate potential t is aa time index, and the bars over variables indicate values. This can be modified to allow for aa break in the growth trend: (6) = Be (r1t1+r2t2) H Cyclical deviations of productivity from its trend are expressed by (7) (7) __~i_ Q _[_ Q where sS > > = = iTh13 ~H’ 11 if, as expected, productivity is higher the higher the potential. level of actual hours or actual output relative to potential. Previous work has shown that some lags exist in this cyclical relation, and they are allowed for by modifying 7 7 to (8) H H H H = = (Q)6IQ/Q Q ° Combining equations equations 66 and 8 to eliminate ~ Q leads to the basic equation Combining used for statistical statistical estimation. estimation, In the original analysis, analysis, some alternative alternative specifications specifications were tried and residuals were examined to determine whether a break in the time trend was important in modelling potential. potential. The evidence only slightly favored the hypothesis of a trend break and two alternative slightly alternative 16 16 estimates of potential output were made, one based on a constant trend 1954-1976 interval, the other with aa in weighted productivity over the 1954—1976 break in that trend in 1969. Now, however, on the basis of the analysis that Edward Denison compelis currently conducting, the case for a break in the trend seems compelling. Denison is measuring the effect on productivity of business costs or expenditures -- those associated with dishonesty and crime, —— with compliance with health and safety requirements, and with pollution control -- he finds productivity growth has been eroded since the late -- 1960s. When Denison’s Denison's final estimates are available, it will be possipossil960s. ble to integrate them carefully into an analysis such as the present one.&/ nmv, they lend 1end independent support to the estimates that one.~I But for now, allow for a break in the weighted productivity trend. The equation estimated with a break in the productivity trend in 1969 is (9) log (~) t = -5.28 + O.0179T54 0.0179T54 + 0.651 log —5.28 (13.5) (16.5) Q l,og (li)t 1-0.0237074 7-0.0237074 -- 0.0019169, o.0019T69, -0.093 ~og (E) - log (~) 1 (-3.3) (-2.3) '-j (-1.5) Q t QQ tlJ(,33) SEE= SEE = 0.0062, D.W. = = - 1.77, estimation period is 1954-1976. 1954-1916. T54 and T69 are the annual tim! time trend duninies dummies startstartIn this equation, 154 ing in 1954 and 1969 respectively. They indicate an annual trend in weighted productivity of 2.75 percent through 1968 and 2.46 percent thereafter. D74 is a duniny dummy for the year 1974 when productivity 17 experienced its largest residual. While a variety of observations about business behavior that year led me to use the duuaBy, dummy, its only noticeable effect is to reduce the size and importance of the lagged adjustment term. The residuals from the equation equa.tion for 1973, 1975, and 1976 are only 0.2, 0.3, and 0.3 respectively, indicating only a slight underprediction of productivity and no trend toward a growing error. The final estimate of potential output that arises from combining my potential labor input estimates with the trendin trend in potential weighted productivity are sumarized summarized in Table 2. In calculating these estimates, labor input and output outside the business sector are assumed the same at actual and potential. Actual GNP in 1976 was estimated to be 8.3 percent below potential. 1976-81 period, potential is projected to increase at an average In the 1976—81 rate of 3.88 percent, just slightly slower than in the 1970-76 1970—76 period and noticeably faster than in the previous intervals covered by this in the labor force is The projected 2.1 percent annual growth In study. noticeably slower than the 1970-76 1970—76 average and slightly slower than in the 1955-70 period. 1965—70 perIod. However, as a comparison of the last two lines in producthe table show, the difference between conventionally measured productivity and weighted v1eighted productivity narrows sharply in the projection projection period. v/here potential output growth was slowed slov1ed over the past decade Where by the changing demographic composition of the work force, in the period ahead, as a result of the maturing of the baby boom, it is not. vJhile the impact of energy prices on potential has been shown to While be slight, a very modest adjustment to the point estimates presented 18 Table 2. Profile of Changes in the Economy at Potential, Selected Intervals, Intervals, 1955-81 (annual rate of growth in percent) Sector and economic measure 1955-60 1960-65 1965-70 1970-76 Projected 1976-81 Total economy Labor force l. 01 1.01 1.29 2. 17 2.17 2.39 2.08 2.08 Employment 0.97 1. 23 1.23 2. 14 2.14 2.31 2.07 Real GNP 3.49 3,49 3.49 3.53 91 3. 3.91 3.88 3.88 Employment 1.21 1.11 2.30 2.87 2.07 Total hours 0.95 0.84 1.41 1.41 2.41 2,41 1.53 1.53 Output 3.62 3.42 3.64 4.34 3.92 Output per hour 2.65 2.56 2.19 2. 19 1,96 1 . 96 2.35 Output per weighted hour 2.79 2.79 2.67 2.48 2.49 Business Sector 19 here can be made to allow for it. Reducing the estimate of the 1976 1976 output gap by 0.2 percent and the annual growth rate of potential potential in the 1976-1981 1976—1981 interval by 0.1 percentage point is about all the adjustment that seems appropriate. This brings the current annual growth rate of potential to 3-3/4 3—3/4 percent. The main implication of this analysis for the capital stock is not that its present size calls for aa downward adjustment of potential estimates, estimates, but that we should expect strong business investment demand and a rapid expansion of the stock if the economy grows enough to approach its its potential level over the next few years. Footnotes 1’ Peter K. Clark, “A New Estimate of Potential GNP (Council 1/ (Council of Economic Advisers, 1977; processed). Economic 2/ Robert H. Rasche and John A. Tatom, “The "The Effects of the New tnergy Regime 011 on Economic Capacity, Production, and Prices,’ Energy Prices," 2—12 and Federal Reserve Bank of St. Louis Review, May 1977, pp. 2-12 GNP,” Robert H. Rasche and John A. Tatom, “Energy "Energy Resources and Potential Potential GNP," Federal Reserve Bank of St. Louis Review, June 1977, pp. 10-23. 10—23. 3/ Potential QQ is taken from the analysis presented later in this 1/ paper. It could just as well v1ell be taken from the CEA estimates as there is 1little ittl e difference between the two tv10 over the relevant re 1evant period. 4/ The estimated coefficients correspond to Equation (3) in the text. text. t—statistics are given in parentheses. I-statistics 5/ The analysis is presented in greater detail in George L. Perry, “Potential Output and Productivity," Productivity,” ~Brookings Pa ers on Economic "Potential Papers ~~vit, Activity, 1:1977. i:1977. 6/ The analysis anal is will be presented in a forthcoming issue of the ~eof0urrentBusiness. Survey of Current Business. 20 POTENTIAL GNP IN THE UNITED STATES, 1948-1980 Peter K. Clark Introduction Introduction The concept of the output attainable attainable by the economy if resources were fully utilized utilized has interested interested economists for many years. This measure of maximum sustainable output, usually called ''potential “potential GNP,'' GNP,” has been aa useful tool for analyzing policies designed to bring about utilization of labor and capital resources. the full utilization The potential GNP measure that gained the widest recognition was first proposed by the Council of Economic Advisers in 1962 ..!J After first 1962.1/ making a number of calculations unemploycalculations relating the overall rate of unemployment to constant-dollar constant-dollar GNP, it it was determined that a reasonable estimate of the GNP attainable at 4% unemployment equaled actual GNP in thereafter. mid-1955 and grew at aa 3.5% annual rate thereafter. Between 1962 and 1976, CEA revised its potential GNP estimates aa number of times; the annual growth rate for potential output was finally pegged at 4.0% for the period 1968-1975 when trend output was thought to be rising rapidly relative to the unemployment rate. Still, Still, in 1976, a judgmental variant of CEA’s CEA's original procedure?! procedureY was still still being used to determine potential President's Council of Economic Dr. Clark is aa staff member of the President’s Advisers. However, the views expressed are solely the author's author’s and are not necessarily those of either the current or past Council of Economic Advisers. 21 output. Potential and actual GNP were still defined to be equal in mid-1955, and the benchmark unemployment rate was still 4.0%. Research on potential GNP from 1964 to 1974 produced a number of different views on the best estimation technique, but very little 1” All of the results were themselves)/ disagreement about the estimates themselves. similar to the CEA estimates or even somewhat higher. Perry [12], (12], for compensate for the example, used a weighted labor input measure to compensate outchanging composition of the labor force, and found that potential out1970's, or 0.3% higher put was growing at 4.3% per year in the early 1970’s, than the CEA estimate of 4.0%. economy's performance between However, several aspects of the economy’s 1973 and 1976 indicated that the maximum sustainable output might be significantly lower than the CEA estimates. First, in 1973 a number of bottlenecks occurred both in primary materials industries and in labor markets which indicated that the economy might have even exceeded the non-inflationary leveL level, rather than being below potential by 2.4% as CEA estimated. Second, shifts in the composition of the labor force toward demographic groups (particularly those aged 16-24) with relatively high unemployment rates indicated that the labor market in 1976 would be much tighter with a 4.0% unemployment rate than it was In in 1955. In other words, if a 4.0% unemployment rate was consistent with a stable inflainflation rate in 1955, 1955, it would produce accelerating inflation in 1976. 1976. Third, the productivity slowdown slo,1down evident in the BLS statistics since the mid-1960’s mid-l960's did not seem to be adequately included In in the CEA estimate of 4.0% potential growth from 1968 to 1975. Although higher 22 labor force growth offset the poor productivity performance to some extent it it was not clear that the sum of these two effects should result CEA's origorigin growth of potential a full one-half percent higher than CEA’s inal estimate estimate for the 1950s. l950s, The productivity decline in 1974 1974 was so extraordinary compared with declines observed in earlier earlier recessions that it demanded special attention. attention. The persistence of the low level The of productivity in 1975 and 1976 indicated that a permanent setback may have taken place. Such an occurrence occurrence would have lowered maximum sussus- tainable output still further. And finally, the Commerce Department revised its real output series in 1976, shifting from 1958 to 1972 prices. Such a shift shift could patterns of growth and to normally be expected to change the observed patterns lower measured growth rates. The new data needed to be incorporated into the potential output measure. For these reasons, in 1976 the Council of Economic Advisers decided to undertake aa comprehensive review of the official official potential GNP series and the methodology used to derive it.~’ it.:!! Using aa technique that incorporated the contribution of capital formation to output growth, together with a variable unemployment benchmark rising from 4.0% in 1955 calculated.!?./ This to 4.9% in 1976, aa new estimate of potential was calculated.~! paper refines the statistical statistical methodology used to determine CEA’s CEA's new potential UMP GNP series and updates the estimates with data through the second quarter of 1977. The new CEA CEA estimate of potential GNP grows 3.5% per year between between 1968 and 1976, significantly significantly lower than previous estimates. 23 result This result makes the formulation of economic policy harder in one sense, but easier in another. The problem is that lower potential output implies that high employment will generate generate less output than previously estimated. For the government, this means lower revenues and a smaller budget marmar- reductions, gin for new programs or tax reductions. On the other hand, lower propro- ductivity implies that a smaller increase in output will achieve the same unemployment and capacity utilization utilization targets. If increases in aggregate demand are constrained by low investment or a climate of fiscal conservatism, poor productivity performance is not unambiguously bad. Disaggregation of GNP hisThe crucial determinant of the difference between any two his- torical potential torical potential GNP estimates is the rate of growth of productivity. The main question is then: productivity How much has the rate of productivj~y growth slowed down? The The answer answer is not easy to obtain obtain because because producproduc9r2~sloweddo~p? is not easy to tivity varies widely with the business cycle, growing rapidly in expanexpansions in economic activity, and growing more slowly or even falling during recessions. Most of the research reported in this paper was productivity growth devoted to obtaining good estimates of the trend in productivity by careful adjustment for cyclical factors, factors. These cyclically adjusted productivity growth figures are then combined with estimates of factor input to obtain potential output. The first step in estimating the trend in productivity growth was the division of GNP into four components: 1. Gross Output Originating in the Rest of the World 24 2. Compensation of Government Employees 3. Gross Housing Output 4. Private Nonresidential Nonresidential Output GOP, Gross output originating originating in the rest of the world, or GNP minus GDP, was an obvious candidate for exclusion from the productivity estimates because this contribution contribution to GNP is generated by by investments outside the U.S., and should not respond to domestic inputs of labor or capital. Compensation of government employees is the only measure of Compensation government output in the national income and product accounts. This component is deflated by an index of salaries of government workers, which implies that real output of the government sector is a weighted average of government employment. Therefore, productivity productivity for the weiqhted employment divided by employment, and government sector is weighted productivity growth is defined as zero in the National Income and Product Accounts. When measuring productivity growth, it it is reasonable to exclude government output so that variations variations in the ratio of governgovernment to total employment do not affect the productivity calculations. Segregation of housing output into a separate category was based on the possibility possibility that the real return from residential residential capital and nonresidential capital might be different. different. In theory, such aa disparity disparity should be only temporary, but in the actual actual analysis, analysis, it it was thought that the fixed nonresidential capital stock measured by the Commerce Department was only a proxy for non—labor non-labor inputs to private private production. Since housing was easy to exclude, it seemed worthwhile to do so. 25 Another important reason for excluding these three sectors is that their output is not related to the domestic business cycle. If unemployment is high and capacity utilization low, government output, the imputation to the residential capital stock, and gross product originating in the rest of the world are not necessarily low. ThereThere- in these sectors. fore, potential and actual output can be assumed equal In Private nonresidential output, the residual in GNP after (GNP-GDP), compensation of government employees, and output attributable to the residential capital stock have been subtracted, corresponds economists' preconception of private sector output, closely to many economists’ produced by capital and labor. It is this output which Is is most closely made.§! studied, and for which productivity estimates will be made.~ Potential and Actual Capital Input to the Private Nonresidential Sector Productivity is a ratio of output to input; real output for the private nonresidential sector may be derived by subtraction, as described in the last section, but the corresponding capital and labor inputs must be estimated. Capital input was taken to be an estimate of Inputs the effective private private fixed nonresidential capital stock multiplied by an estimate of capacity utilization. The effective capital stock meameasure used was the B.E.A. gross stock of private nonresidential capital, adjusted for investment in pollution abatement equipment.ZI equipment.1” Quarterly data were linearly interpolated from annual data; projections of capital stock were derived from an investment forecast in which the ratio of nonresidential fixed investment to real GNP rises to ten 26 percent by 1980. Six percent of fixed investment was assumed to be for pollution abatement throughout the forecast period. Annual averages of the capital stock series are given in Table A-1. A—l The newly—revised newly-revised Federal Reserve Board manufacturing capacity utilization rate-W utilization rate~ was taken as the starting point for estimating the degree of capacity utilization private nonresidential sector. utilization for the private However, since output in manufacturing is much more cyclical cyclical than private sector output as a whole, the cyclical variation in the FRB manufacturing index must be reduced. This was accomplished by multimulti- plying the difference between 87.5 and the Fed index by 0.5, the approximate ratio of the percentage standard deviation around trend for private nonresidential output to the percentage standard deviation around trend for manufacturing output. If the ratio of utilized utilized capital to output is fixed in the short run, such an approximation is reasonable. reasonab 1e. The potential capacity utilization utilization rate of 87.5% was chosen because it it was this rate that was reached in mid-1955, mid—1955, early 1968, and all all of 1973, all periods when it it is generally considered that output was near its potential level. If there were a close relationship between changes in the rate of inflation inflation for private nonresidential output (or the profit rate) and measured capacity utilization, utilization, it it would be appropriate to estimate the relationship, relationship, and define ''potential'' “potential” capacity utilization utilization as that rate which resulted in non-accelerating prices. However, in the absence of such a ''Phillips “Phillips curve'' curve” for capital, 87.5% is a reasonable benchmark. 27 Potential and Actual Labor Input to the Private Nonresidential Sector emoloyment and labor force require a much Extensive data on employment more elaborate set of calculations for the estimation of private labor input. The labor input measure that was constructed tried to adjust for the productivity of different groups of workers by dividing di vi ding the (16-19, 20-24, 25—64, 25-64, 65+) and labor force into four age categories (16—19, also disaggregating by sex. in each of these 88 Private employment In governcategories was obtained by subtracting an estimate of civilian govern- ment employment from total civilian employment. Private employment for each group was then weighted by mean weekly earnings for that group In in May 1973)1 1973.2/ Use of the weekly earnings weights approximates the contricontriincluding bution to production of an employee in each demographic group, Including both average hourly earnings and average weekly hours. It would be better to have weights that vary over time rather than one fixed set of weights, but data are not available to construct variable weights. Therefore, the effect of changes in the age-sex weights representing changes in average weekly hours and average hourly earnings is included in In the estimated trend terms described later. Rates of growth of l. Although the weighted and unweighted labor input are shown in Table 1. growth rate of weighted employment is less than the rate of unweighted employment, it is only the change in this difference that explains part of the productivity slowdown since 1966. Determination of the potential level of labor input requires two extensive extensIve calculations. determined. First, the potential labor force must be Then a benchmark unemployment rate is calculated, and used 28 Table 11 Rates of Growth of Weighted and Unweighted Private Employment, 1948-1976 Private Employment, 1948-1976 Time Time Period Private Private Emp l oymenta Employmenta Private Employment b Private Employment b (weighted by 1973 earnings) 1948-1955 1948—1955 .58 .61 1955-1966 1.13 l. 13 .75 1966-1973 l. 78 1.78 1.38 aa b b Civilian employment employment minus government employment Civilian minus civilian civilian government employment from from the Current Population Survey. Civilian employment minus civilian government government employment by Civilian employment minus civilian employment by eight age-sex (16-19, 20-24, 25—64, 25-64, 65+; M, F) weighted by age—sex groups (16—19, May 1973 mean weekly earnings, all from Population Survey. May 1973 mean weekly earnings, all from Current Current Population Survey. 29 to translate potential labor force into potential employment. Since labor input is a weighted sum of employment from eight age-sex groups, levels for potential labor force and the full employment unemployment rate must be determined for each group. Full-employment labor input governis then potential employment for each age-sex group reduced by govern- ment employment, employment, and then weighted by mean average weekly earnings in 1973. Potential Labor Force Potential labor force for each age—sex age-sex group Is is calculated by estimating a cyclical adjustment to labor force participation for that group, and then adjusting actual labor force to full employment labor force using the adjustment. The general form of the labor force participation equation is: (4) (4) L~t it = = aa. +b.~U 1 +c1 t+d1 Tlt+e1 ~MILt+f. •5(~ + g. · AGt+ AG + tit 1 t Where L. Li l = = civilian labor force in group i POP.1 POP 1 = = in group Ii civilian noninstitutional population In UU = = unemployment rate of men 25-54 t = = ti me time Tl = = dur.llTiy which vihi ch equals equa 1s 0O until 1966:4, 1966; 4, and then trend duniny increases 1, 2, 3, 4... 4... MIL = = military employment divided by the civilian 16-24 noninstitutional population of men aged 16—24 30 SCC = = degree credit enrollment in higher education as a percent of population aged 16-24 AG = = civilian employment in agriculture. proportion of civilian Estimation results for equation (4) (4) are given in Table 2. Using the lagged unemployment rate gives the largest largest estimate of cyclical variation in the labor force, even though the estimates are smaller than might have been expected. unemployment Use of a contemporaneous unemployment rate or a distributed lag on the adult adult unemployment rate generates lower estimates of cyclical variation. By using the unemployment rate of men 25-54 as aa cyclical variable for all groups, the problem of upward simultaneous equation bias is avoided for all groups except men 25-64, where cyclical variation in labor force participation is very small. The cyclical adjustment for each group was the estimate in coefficient was insignificant coefficient Table 2, except for men 25-64, where an insignificant estimated even though substantial upward bias due to simultaneity simultaneity was suspected, and for women 65+, where the cyclical coefficient coefficient was insiginsignificant and the wrong sign. literature on pretest estimators The literature suggests that some of the other cyclical coefficients coefficients should be set to zero,J..Q/ but this was not done. zero)-~’but bit high. be a Thus, potential labor force may be Of course, if the reaction of labor force participation participation to long periods of low unemployment is much stronger than its its average reaction over the cycle, potential labor force could be underestimated. civilian In 1976, the estimates in Table 22 imply a potential civilian labor force 1.1 million workers larger than the actual labor force. 31 Table Table 22 Cyclical Variation Variation in Labor Force Participation Estimates of Cyclical Rates by Age and Sex Rates by Age and Sex Estimation 1953:1 to 1976:4 Estimation interval 1953:l (standard errors in parentheses) (standard errors in parentheses) gg p R2 (.00043) ( .00043) .00277 1.95 l.95 ((.52) .52) .60 .897 1.81 .00270 (.0005) 1.27 1.27 (.59) (. 59) .57 .57 .934 . 934 1.97 1.97 1.11 (.35) (. 35) .45 .45 .810 1.89 .00237 (.00026) .57 .988 .988 1.87 —.00024 -.00024 (.00005) —.00066 -.00066 ((.0001) .0001) .64 .979 .979 1.86 —.00025 - .00025 ( .00074) (.00074) .00151 ((.00013) .00013) .00043 ( .00022) (.00022) .78 .995 1.91 .481 (.015) (. 015) -.00183 —.00183 ((.0013) .0013) -.00267 —.00267 (.00024) (.00024) .00138 (.00042) (,00042) .80 991 ..991 l.68 1.68 .108 ((.006) .006) .0003 (.0008) -.00007 -.00007 ( .00009) (.00009) -.00045 ( .00017) (.00017) .61 .61 .817 .817 2.15 2,15 Group a b cC Men 16-19 16—19 .411 ((.072) .072) -.0096 —.0096 (.0016) ( .0016) .00040 (.00058) (.00058) 16—19 Women 16-19 .260 (.080) (. 080) -.0059 (.0018) .00091 (.0005) (.0006) Men 20-24 20—24 .877 (.059) -.0009 —.0009 (.0013) .00080 ((.00036) .00036) Women 20—24 20-24 Women .427 (.001) (.001) -.0019 —.0019 ( .0012) (.0012) .00111 (.00013) Men 25.64 25-64 ..963 963 ( . 004) (.004) —.00017 -.00017 (.0004) Women Women 25—64 25-64 .333 ((.008) .008) Men 65+ Women 65+ dd ee -196.3 —196.3 (44.1) (44.l) f —.0039 - . 0039 (.0016) d.w w w N ~ This figure is only slightly slightly higher than the approximately .9 million “discouraged workers” estimated by BLS for 1976.ll/ 1976.]-]-! "discouraged workers" Projections of labor force by by group were made with the estimated labor force participation equations. Since they include aa cyclical adjustment, the projections are slightly higher than those made recently by BLS. Annual totals for potential labor force are given in Table A—2. A-2. Full Employment Unemployment Rates The establishment of a benchmark unemployment rate for use in estimating potential output is aa difficult problem. If there were aa good statistical relationship between unemployment rates and the ininflation rate, the vector of unemployment rates by age and sex that yields aa constant rate of inflation could could be determined directly. Unfortunately, there seems to be no unique relationship between “Phillips curve" curve” method of unemployment and inflation, so this simple "Phillips estimating an appropriate unemployment benchmark is not available. The picture is further complicated by increases in the proportion of the labor force comprised of young people (aged 16-24) and of adult (aged 25-64) 25-64) women, which seems seems to have changed the relationship between the unemployment rates of different age-sex groups. The significant significant change in the unemployment survey in 1967 also tends tends to make the determination of an unemployment benchmark which is consistent over time somewhat arbitrary. The procedure actually used makes the assumption that aa 4.0% overall unemployment rate represented full employment in 1955. By looking at the relationship of unemployment rates between age and sex 33 groups in 1955, the eight age-sex unemployment rates that would have yielded aa 4.0% overall unemployment rate in 1955 may be determined. It is further assumed that the unemployment rate for men aged 25-54 has remained aa stationary indicator of the state of the labor labor market. The increase or decrease in each group’s increa•;e group's unemployment rate is estimated using an equation of the form (5) u.l U. = = 1 Cl ~. • 1l Lp. * ( L~ i ) + * u + y. + ~ Si * U + y.l1 *(~LP I LP 1 I ±) where U. of age-sex age-sex group group ii unemployment rate = = unemployment rate of U u of men 25-54 as before, = unemployment rate of l = A LP. 1l A Cp LP = = (I~ LiL· equation (4) times times Pop. from equation Popi ·.. Po~.1 1 ~ (Li =L. l1 \ 0P~i (Pop~) The inclusion of the the/ L:i) term, the relative proportion of group ii in Iftp LP ~ the labor force (purged of short-term short—term variations) was based on the idea of partial segregation of labor markets. A A relatively high proportion of the labor force in aa particular group may make it difficult for members of that group to find satisfactory employment. The coefficient yi estimates the change in relationship between the unemployment rate of group i and the unemployment rate of men 25-64. 25-64. The data used in estimation of equation (5) (and equation (4)) have been adjusted for by multiplying the change in sampling procedure starting in 1967 by by BLS using employment and labor force by ratios obtained in 1966 by 34 both sampling techniques.1Y techniques.i?! While this adjustment is reasonable for there is no evidence on its accuracy during high-employment years, there periods of low economic activity. Estimation results for equation (5) are given given in Table 33 for all eight demographic groups. The unemployment rates for women 25-64, 25—64, women 65+, and men 25-64 25—64 did not exhibit significant change relative to ..Jl/ The sign of the rate for men 25-54 25_54.i~! y was negative for men 65+ indicating indicating the operation of other forces such as Social Security in the labor market for these workers. However, aa downward trend was evident, so the equation was re-estimated with aa time trend, as shown. The changes in benchmark unemployment rates rates by demographic group are illustrated by the results in Table 4. The relative labor force proportion of younger workers (ages 16-24) 16—24) has risen sharply; equation (5) hypothesizes that this shift in proportions was responsible for the observed change in relative unemployment rates. When combined with the high high employment labor force, these benchmark benchmark with estimates estimates of of the employment labor force, these unemployment rates rates yeld aa benchmark for the overall unemployment rate, also shown in Table 4. The overall benchmark unemployment rate equivequiv- 1977. alent to 4,0% 4.0% in 1955 is 5.1% in 1977. It would be an abuse of the term "full employment" to call 5.1:s full employment” 5.1% the full employment unemployment rate in 1977, given the high benchmark rates for teenagers, and the fact that the burden of this joblessness is distributed unequally across races and demographic groups. Rather, the estimates in Table 33 are a strong reminder an overall unemployment rate of 33 or 4% would be be characterized by by aa very tight labor labor market market for for adults. characterized very tight adults. 35 Tab 1e 33 Table Estimates of the Relationship of Age-Sex Unemployment Rates to the Unemployment Rate of Men 25-54 (Equation 5) of Estimation (standard Group IInterval nterva 1 1948:2-1977:2 1948:2—1977:2 errors errors in parentheses)a Time Trend ii:-2 a. Si S; Men 16-19 .103 -1. 03 (2,33) 2.33) 1.87 1.87 (.152) . 152) 208.7 208. 7 (51.5) (51. 5 .77 . 77 .939 Women 16-19 -7. 99 —7.99 1. 31 1.31 (.189) (. 189) 493.8 (60.2) .76 .936 1.92 l. 92 (.116) (.l1~J 36.5 19. 1 (19.1) .70 .947 (1.23) .42 ( . 97) 1.08 . 106) 87 .4 87.4 18.9 .73 .919 .95 2.7 ( .49 .46 .990 .60 .925 .49 .786 .42 .559 1 (2.15) Men 20-24 Women 20-24 20—24 ~106~_ - .63 - Men 25-64 25-64 - l. 15 -1.15 ( .26) ~(.02)(,49 Women 25-64 25-64 Men 65+ Women 65+ ~0.5 aa 2.33 2,33 (.02) .77 .93) .05) 2.37 ~07~ ( .29) .70 (. 07) —1.59 -1. 59 .82) .09) ..36 36 y. 1 .11 3.7 -.0072 ((.0028) .0028) 227,4 227.4 60.5 Data adjusted for 1967 CPS survey change. Data adjusted for 1967 CPS survey change. 36 pp Table 4 High Employment Benchmark Unemployment Rates 1955 1955 and and 1977 Rates 1977 (percent) (percent) Demographic Group 1955 1977 Women Women Women Women 10.8 6.2 3.8 2.7 17.7 9. l 9.1 4.3 l.8 1.8 l l. 7 11.7 6.5 3.0 3.6 14.8 7. l 7.1 2.4 3.2 4.0 5. l 5.1 Men Men Men Men 16-19 16-19 20-24 25-64 25—64 65+ 16-19 20-24 25—64 25-64 65+ Total 16+) (Both Sexes, 16+} Note: Unemployment rates assume the survey technique actually used in that year. 37 5.1% 4.0% in 1955 and 5. 1% in 1977 are in no sense estimates of the lowest overall unemployment rate that does not cause inflation to accelerate. Rather, the time series of unemployment rates generated by the equations in in Table 33 is aa consistent set of unemployment rates over time generated by the assumption that the unemployment rate of men aged 25-54 is aa stationary measure of labor labor market tightness. The non-accelerating inflation rate of unemployment, or NAIRU, was probably about 0.4 percentage points higher in 1955, and 0.6 to 0.9 percentage points higher today. The high employment level of labor input is calculated in three steps. First, employment in each age-sex group group is estimated by multimulti- plying the the potential potential labor labor force force by by one one minus minus the benchmark unemployment unemployment rate. Second, civilian government employment is subtracted from these potential employment estimates to obtain potential employment in the private nonresidential sector. Third, potential private nonresidential employment in each age-sex group is weighted by mean average earnings in May 1973 and aggregated to obtain weighted potential labor input. input. Cyclical Adjustment of Productivity and the Calculation of Potential GNP The crucial step in the estimation of potential GNP is the determination of good estimates of productivity at benchmark input levels. If an equation explaining the variation of productivity with the rate of input utilization can be obtained, then aa benchmark benchmark level proof input can be entered into the equation to obtain the level of pro- ductivity associated with that benchmark over time. 38 The basic specification of the variation of productivity with utilization rates used in this study is: (6) yt ~t 7~/P ‘tt = = ( It’\~ t TP \ ‘~ I / \~ (It~i I p \I t—l where yt = Y~ = = It = = = real output of the private nonresidential sector real output of the private nonresidential sector in in quarter quarter tt potential value of Yt in quarter t. weighted combination of labor and capital input in the weighted combination of labor and capital input in the t. private nonresidential sector in quarter t. 2 = (K * CU )l/3 · (Lt) 2 /3. t t t (Kt * CU~)”3 (Lt) /3. = of It in quarter t. = potential value of I = Kt = = nonresidential adjusted for for nonresidential fixed fixed capital capital stock stock adjusted po 11 ut·i on abatement investment. investment. pollution CUt cut = = Adjusted Federal Reserve Board manufacturing index. capacity utilization index. Lt Earnings-weighted private employment. 141 6) ‘t Js~J holds, holds,1~’ ((6) If the invertibility condition ~sot lsol >> ls1I = = can be be can ‘Pt past yt expressed as aa convergent series of past yP ((7) 7) 1t - II f = tt {Yt-s) (Vt_s\ '\ ~ s= \ YPt-s s-Uo \y~t-sJ Equation (7) may be familiar to many readers as aa specification of the lagged response of of inputs to output that has been discussed extensively extensively 39 in the literature iiteraturei-~i .0 Equations (6) and (7) ( 7) say that in the long run, the percentage gap between potential and actual input is aa constant fraction i/(s~+~~) 1/( Bo +si) oof f the percentage gap between potential and actual In the short run, this fraction is smaller, due to the lagged output. response of input to outputJ~’ output.ill An alternative specification for the cyclical relationship between output and input is equation (7) with aa one-period lag: ( 8) (8) It I~ (Vt\ao - = \V~) - y t-l)cq (_V~~\ai ( \V~-11 Yt-1 Equation is aa variant variant of is sometimes called "Okun's Law." Equation (8) is of what what is sometimes called ‘Okun’s Law.” If we let JG t and yG VG t = = = = p p ~t It It -It p It I~ p p yt _yt VtJt percentage input input gap gap = percentage = = percentage output gap = lt then (8) becomes: log (1 - ItG) = ~ log (1 - V~) + ~1log (1 - V~1) The approximation log (1 ++ x) ~ xx for small small x implies: G G G G GG ~o Vt It = ao Yt ++ a~ a1 Vt_i Yt-l = which gives a percentage input gap as a function of current and lagged output gaps, in much the same way Okun’s Okun's Law relates an unemployment gap to current and lagged output gaps. 40 One further assumption besides either (6) or (8) is needed: aa specification for the growth in cyclically adjusted total factor productivity: (9) (9) 1_V~ log log ( y~ ) IP It t \ ) = f(t)+u f(t) + ut , = t where f(t) describes how productivity productivity has grown over time. The specification of f(t) was made on an ad hoc basis; basis; namely, total factor productivity was assumed to grow at aa constant rate from 1948 to 1966, 1966, and at a different rate from 1967 to the present, to correspond with the productivity slowdown that has been widely observed. Additional "kinks" “kinks” in f(t) are necessary to help explain the extraordinarily bad productivity performance observed in late 1973 and all of 1974. Three variants of f(t) were used: (lOA) (bA) f(t) = = aa+bt+cTi + bt + cTl (108) (lOB) f(t) = = a ++ bt ++ c(Tl) c(T1) ++ d(T2) (lOC) (bC) f(t) = = c(Tl) ++ d(T3) aa++ bt ++ c(Ti) where t ·= Ti Tl = T2 TZ = T3 = time trend 1,2,3,4 ....thereafter thereafter 4, then 1,2,3,4.. 1966:4, o.....Oo until 1966: = 0. l.O thereafter . 75 and 1.0 . 5, .75 ....0o until 1973:4, then .25, .5, = 00.. o....00 until 1973:3, then 1,2,3,4,5,4,3,2,1, and 0O thereafter. = 0.. IIA'l variant gives additi ona 1 consideration to plummeting The “A” gives no additional productivity in 1974, and just treats it as another set of observations on the cyclical variability of productivity. "B" variant implies a The “B” once-and-for-all downward shift in the trend level of productivity in once—and-for-all 1974, due to shift in in the relative price of energy 1974, possibly possibly due to the the shift the relative price of energy or or 41 underestimation of real output in an inflationary environment. The "C" “C” variant explains the lower productivity in 1974 as an extraordinary cyclical movement that disappears by the end of 1975. 1975. Equations (6), (9), and (10 A-C) can be combined to yield aa regression equation for total factor productivity in the private nonnonresidential sector which can then be used for estimating potential GNP. ~ Equation (6) implies: log H )+ bo~(~ = (~-1)bog Sibo~(~i) Substituting in (9) and (10 A—C) A-C) yields: a ++ bt ++ cTl Nt\ + k’t / ( llA) (hA) log (Vt~ ((11B) llB) bog(Vt\ ==a+ a + bt ++ cTl ++ dT2 + + (s (so-h) bog(It ~+ B1log sflog 0 -l) log(:.!_)+ ~I ,/ :. Ip = IT + (se-i) bog sibog(’t-l u ' t S 1 ‘~ t—i \+ u I~ / \ (11C) (llC) log = a + bt (se-i) 0 -l) log ( :; ) =a+ bt ++ cTl cTl ++ eT3 eT3 ++ (s It’ bog(~t\+s bog log(:!) +s11log \ (‘ti I t-1 \ ++ uut . ( rL1 / \‘tiJ The regression equations derived from the alternative specification (8) A-C) are: along with equations (9) and (10 A—C) (i2A) (12A) bog(Vt log~::): “ It ~ (ao - +a1)a +a1)a + + ~ (a 0 (a~-1) bogflt\ ki~1 42 +a +a11)) - bt bt cL~ al + (cto ~1) cTl + (ao + +a1) cTl bog log ( Vti\ yt-1 .) ++ u~ I~-1) It-1 ( 12B) (128) (a0+aj) a + (ao+ai ) bt + (cto+czz) cli + (a +a~ ) d12 0 lo~(”t): ‘ t — - (a0—l) 1o9(t) —a~ 109( t_i + + Ut u t \ t—iI (12C) iog(~t~: (a0+a1) a I’ / + - (aO + (a0 +a1) cli (czo-1) log + ~ (a0 +~2)c13 (v~\-aj ‘ bt t logP’t-l\+ rn U It should be noted that the algebraic manipulations required to derive (12) from (8) and (10) imply that the disturbances In in (12) will exhibit second-order serial correlation. Thus it was not surprising when secondsecond— order serial correlation was found in the estimation of (12) (and Cochrane-Orcutt two handled by by a second-order second—order generalization of the Cochrane-Orcutt stage procedure). Estimates of the parameters In A-C) and (12 A—c) A-C) in equations (11 A-c) Cochrane-Orcutt two-stage procedure are given in Table obtained by the cochrane-Orcutt 5 below. Standard errors are not given for the parameters of f(t) In in the estimates of equation(12), since these are obtained by dividing least squares coefficients, implying that they have infinite variance. It is reasonable to assume that equations hA-c llA-C give more reliable estimates, for two reasons. mation errors for estiFirst, the division problem allows esti- and a11 in equation (1.2) (12) to contaminate the growth parameters a, b, and c. Second, the longer lag specification (equation a 0 (7)) seems more appropriate than the short 1-period lag in equation (8). 43 Table 55 Parameter Estimates Estimates and Implied Implied Potential GNP for Equations (llA—C)and (llA-C)and (l2A—C): (l2A-C): Cyclical Variation in Total Factor Productivity Long-Run Growth Growth and Cyclical 1977:2) (quarterly data; estimation estimation interval 1948:3 1948:3 or 44 to 1977:2) Equation +> +> a bb llA -3.70 —3.70 (.0095) .00456 (.00017) —.00155 -.00155 ((.00038) .00038) 11B 118 -3.70 —3.70 ((.0083) .0083) .00443 (.00016) -.00058 —.00058 (.00042) .00455 (.00018) -.00139 —.00139 (.00039) C C llC 11C -3.70 —3.70 (.0096) 12A —3.69 -3.69 ..00442 00442 —.00099 -.00099 12B 128 -3.69 —3.69 .00436 -.00044 l12C 2C —3.69 -3.69 .00441 --.00084 .00084 in parentheses.) parentheses.) (Standard errors in d e (so—l) (So··l) .878 (.114) -.042 —.042 ( .011) (.011) -.00523 (1—m0) (1-ao) c~ "1 p; Pl p P2 2 d—w d-w _ 1955 1955 Potential GNP 1977 1977 Potential GNP GNP ~2 R2 (billions of 1972 1972 dollars) —.422 -.422 (.115) (. 115) .78 1,70 1. 70 .997 656.1 656. l 1392.4 -.414 .822 —.414 (.111) (.110) ( .110) .76 .76 1,80 1.80 .997 653.7 1378.1 1378. l .79 .79 1.86 .997 655.9 1399.4 1399 .4 -.00659 —.00659 .806 (.0022) (.112) (.112) -.028 Si B1 -.352 —.352 (.113) (. 113) .656 ( .022) (.022) —.148 -.148 (.022) 1.02 l.02 —.34 -.34 1,62 1.62 .9998 658.8 658. 8 1413.8 .644 (.023) -.163 (.022) ( .022) .97 -.'32 -:32 1.59 1,59 .9998 657 .7 657.7 1402.6 .649 (.022) -.144 (.021) ( .021) 1.08 -.39 -.39 1.80 .9998 658.6 1420.4 1420.4 It is difficult to discriminate between the two hypotheses about about the 1973-74 1973—74 "productivity “productivity disaster" disaster” implicit in the BBand and CC variants of the equations. If the B B variant is the correct specification, and the level of productivity shifted downwards in 1974, high inflation infbation rates could be the cause. If such high inflation rates caused price increases proto be overestimated, real output has been underestimated, and the productivity loss exaggerated. de nee for this view vi ev1 can be found Some evi evidence in the Federal Reserve Board Board Industrial Production Production index, Index, which fell less than real GNP over the 1973-75 period. One also suspects the rapid rise in the relative price of energy, although the mechanism for loss in productivity due to the high price of oil is not obvious. In a theoretical model with homogeneous capital, even if the elasticity of substitution were zero, potential GNP measured in 1972 dollars would not fall at all. Cobb-Douglas formulation generates implausible AA Cobb—Douglas reductions in energy usage of 40 to 50%. AA vintage model for capital could explain the drop only if U.S. capital is more energy intensive than foreign capital. In this case, production using the most energy— energy- intensive capital in the U.S. might not cover cover variable costs at world output prices. The with the the "C" specification, is is that that The other other view, view, consistent consistent with “C” specification, the cyclical movement in in productivity productivity was was just just much much stronger stronger in the the cyclical movement in the 1973-75 recession than in previous downturns. Probably the truth lies once-and-for-a 11 drop in tota producsomewhere in between; aa once-and-for-all total1 factor productivity of about 2% combined with some extra cyclical loss may be close to correct. 45 The estimates of potential GNP shown in Table 5 5 are derived by eliminating the cyclical components in each equation and setting labor and capital inputs to their potential values. This yields aa potential for private nonresidential GNP, which then is added to the non-cyclical components (compensation of government employees, imputation to the residential capital stock, and income from investment abroad) to obtain potential GNP. 61W, The BB variant shows the lowest potential for 1977, productiviey in reflecting the pessimistic assumption that the drop in productivity 1974 not explainable by by normal cyclical factors was permanent. The equation (12) estimates are higher than those from equation (11); this difference may be caused by incorrect specification of of the lag between changes in output and input gaps in equation (12). For example, the long-run elasticiey elasticity of the input gap with respect to the output gap is 1/(Go equation (11B), (118), while the same elasticity elasticity is 1/(13o + 0~) = .716 in equation 1) = (a 0 ++ a 1 ) == a~) .562 in equation .562 in equation (128). (12B). By allowing the lag lag to to be be longer By allowing the longer in (11), the sum of the coefficients coefficients is larger; the larger long-run elasticity implies aa smaller output gap for aa given input differential. Estimates Using Only Labor Input ~atesUsin0n1LaborInut Discussions of potential output are usually based on labor input only, largely because the measurement of the capital stock is based based on a number of arbitrary (but necessary) assumptions, and because the weight of capital in total input is the subject of some controversy. controversy. It is instructive, then, to investigate the effect of the capital stock estimates on the calculation of potential output by performing the analysis using labor input input only. All the same equations ((llA-C) ((hA-C) and and 46 (12A-C)) may be estimated by replacing I~,the It, the combination labor and capital input, with Lt~the Lt, the labor input component only. The basic equations are then (6’) v -w‘tt (8’) Lt t and 1L ~o1 =it t ~t P-t—i~ plLp \L,~ j \Lt_l =I~t\ct’o ( t~ni ~\ \ti (9’) logl/YPt \I (PA \t / ____ = f(t) t-11 + u t The analysis is exactly the same, but the basic productivity concept is 1abor productivity instead of total factor productivity. labor capita 1 If the capital input measure is sufficiently poor, ignoring capital will produce better estimates. The results of the “labor "labor input only” only" regressions are given in Table 6. The results are virtually identical; estimated potential GNP is about 1% lower in 1955 and about 1% higher in 1977. The difference is primarily due to somewhat higher capital utilization rate relative to in the mid-1970s, compared to 20 years earlier. the unemployment rate In The elasticity of the output gap with respect to the labor input gap is higher than the elasticity of the output gap with respect to the weighted gap for labor and capital. 1/(s~ l/(s 0 +s~ +s 1 )) = = (llB') while (a 0 + '"i) == .400 for equation (12B’). (12B'). (lhB’) + ‘~) .569 for equation This is not sursur- prising, for capital utilization utiiization should adjust more rapidly to output 47 Table 6 Parameter Estimates and Implied Potential GNP for Equations (11*-C’) (llA .. C 1 ) and (12A-C’): (12A-»C: '): Long-Run Growth and cyclical Cycl ica1 Variation Variation in Labor Productivity (quarterly data: estimation interval 1948:3 to 1977:2) (quarterly Equation ‘HA’ l lA' +> bb CC —3.11 -3. 11 ( .011) .00683 —.00179 -.00179 ( .00020) (.00020) (.00044) -3.11 —3.11 ( .010) ..00671 00671 (.00020) -.00081 —.00081 1K’ !lC' -3.11 —3.11 . 00682 .00682 (.00020) ( .00020) -.00162 —.00162 (.00045) 12A’ 12A' -3.10 —3.10 .00663 -.00117 128 12B' —3.10 -3.10 .00654 -.00053 —.00053 12C’ !2C' —3.10 -3.10 ..00663 00663 .•. 00109 —.00109 llB' 118’ 0) a ‘ ( .011) (Standard errors in parentheses) (.00052) dd ee -.045 —.045 (.014) —.00690 -.00690 ((.00265) .00265) —.029 -.029 —.00679 -.00679 (Go—i) (Bo-!) Gx S,1 (i—ao) ( 1-ao) aj 01 P1 Pl d—w d-w R2 1955 Potential 1955 1977 Potential (billions of 1972 dollars) 1.79 1. 79 ( .20) (.20) —.98 -.98 (. 20) (.20) .77 . 77 1.75 1. 75 .998 650.2 1405.5 1.67 (.20) -.99 —.99 (.20) (. 20) .77 . 77 1.86 .998 648.0 648,0 1389.6 1.67 (.20) ( .20) —.86 - .86 .78 1.87 .998 650.0 650.0 1413.1 1413.l (.20) .784 .784 ( .022) (.022) —.190 - . l 90 (. 021) (.021) .85 1.15 .9998 .9998 651.8 1421.9 .773 . 773 ( .022) (.022) —.197 - . 197 ( .020) (.020) :86 :s6 1.19 1. 19 .9998 650.2 1410.1 .771 . 771 ( . 021 ) (.021) —.188 -.188 (. 020) (.020) .86 1.27 .9998 651.8 651.8 1426.2 than labor utilization. It is also not surprising that the sum of the coefficients ~o a0 + a ct~ “Okun’s 1 is very close to the sum reported in "Okun's Law'' (12') is essentially Okun's Law” equations, given that (12’) Okun’s Law, as explained earlier. Unlike the total factor productivity estimates, second order serial correlation was not significant significant in (12A’ (12A' - 12C’), 12C'), implying that - some other form of specification error is responsible for the low Durbi n—Watson statistic statistic after the first-order first-order serial correl ati on Durbin-Watson correlation correction. The range of the 12 estimates of potential GNP derived from the regression equations are given qiven in Table 77 and shown pictorially in Figure 1. Projections of the labor force, capital stock and the comcom- ponents of noncyclical output given in Tables A-1, A-6 were A-h, A-2, and A—6 used to obtain potential GNP projections to 1980. The large increase in the range of potential since 1973 reflects the uncertainty generated by the precipitous productivity decline in 1974. By 1980, the range of estimates is almost 4% of potential GNP, aa figure that does not overoverunemployestimate our ignorance about the level of output in 1980, when unemployment and capacity utilization may be nearer their benchmark levels, levels. Most of the productivity decline since s i nee the late 1ate 1960s 1960s cannot cannot be explained by by the changing age-sex age-sex composition of the labor labor force, the the changing industrial composition of labor-hours, or changes in the rate of qrowth of the capital/labor ratio.ill growth of ratio.2~’ Instead, the slowdown must be allocated to aa residual residual category, or ‘technical "technical progress.” progress." Since reareaSince sons for apparent changes changes in in the of productivity growth are sons for apparent the trend trend rate rate of productivity growth are not well well understood, understood, it may may be be the the case case that that the the trenc' trend productivity 49 49 Table 7 Estimates of Potential GNP (billions of 1972 dollars) (billions of 1972 dollars) Year 1948 1950 1955 1960 1965 1970 1975 1980 Minimum Estimate Maximum MaxImum Estimate 494.5 517. 1 517.1 535. 7 535.7 560.0 580.8 604.0 625. 625.11 648.0 676.0 501. 8 501.8 524.6 543.6 571.0 593.7 615.9 636.5 658.8 685.7 696.0 708.1 721.9 743.5 772.0 801.0 819.9 848. 848.66 880.8 913.9 944.6 981.9 1017. 1017.44 1053.8 l 090. 8 1090.8 1124. 1124.99 1165. 1165.33 1208. 1208.88 1249. 1249.22 1278. 1278.66 1331 .7 1331.7 1378. 1378.11 1426.0 1477.l 1477.1 1531 .6 1531.6 732.0 752.9 779.6 807. 1 807.1 828.5 856.5 888.3 923.3 957.7 997.2 1034. 1034.11 1074. l 1074.1 1114.4 1152. 2 1152.2 1194. 1194.22 1240.2 1240.2 1282. 0 1282.0 1324. 1324.99 1373.7 1426.2 1477.7 1530. 1530.99 1587. 1587.55 Unofficial ** Unofficial 50 1977 CEA Estimate 492.8* 514.4* 537.0* 560. 5* 560.5* 584.9 608.2 627.7 651 .4 651.4 673.9 697.2 721. 721.33 746.2 771 • 9 771.9 798.6 826.4 857. i 857.1 890.3 925.0 925.0 960.8 996.3 1031. 1031.77 l 068. 3 1068.3 1106. 2 1106.2 1145. 1145.55 1186. 1186.1l 1228. 1228.22 1271 . 7 1271.7 1316. 1316.99 1363.6 1412.0* 1462.1* 1513.9* 1567.7* Figure 1 ·Jange 1968-1~8 and ~iange Es~irnate and CEA Estimate c;·1\JP: CEA Potential 1963—19800 Potentia l CNP: dollurs )~----- ---of -197F liori~ of ------- -~--('lifl fbiUion>~ 1972 dollars) -— 1600 1690 Maximum Esti1nate Maximum Estimate Estimate CEA Estimate CEA Estimate Minimum ~inimum Estimate 1500 - --, 1400 01 1300 -. 1200 - - - - - 1100 1100 1000 — ____ 63 68 69 713 71 72 73 74 75 76 76 77 7? 78 ?8 79 79 89 80 growth rate will be higher from 1977 to 1980 than it was from 1966 to to 1973. 1974, Even if there was a permanent 2% loss in productivitv productivity in 1974, altered relative prices may nay generate the incentive for productivity at low energy energy prices. increases that were not particularly profitable at Therefore. Therefore, cautious optimism either in the form of assuming that the 1974 productivity decline was temporary, or in the assumption that the productivity growth trend will be higher from 1977 to 1980 than it GlIP in 1980 of about $1560 billion was in 1966-73, generates potential GNP 1972 dollars. The estimates for the years 1952-1968 1952—1968 conform very closely to previous estimates of potential output for the U.S. economy. Potential GNP is calculated to be very close to actual GNP ($654.8 billion 1972 rate of potential potential is is very very close close to to dollars) in 1955, and the growth rate 18/ l 952-1962 . - The Okun's Okun’s original estimate of 3.5% per year for 1952-l962.~ growth of potential in 1962-68 is also very close to the 3.75% per year that had previously been estimated by the Council of Economic Advisers.W Advisers.-~” Since 1968, however, the growth in potential output has estimated. been much lower than was previously estimated. Potential output growth 3,5% per year instead of 4.0% for 1968-75 is estimated here at about 3.5% for that period estimated by the Council in 1976. 201 for that period estimated by the Council in l976.~’ Although part of the difference between the previous 4% growth rate and the new CEA 3.5% rate can be explained by the t.he increase in the unemployment benchmark from 44 percent to 4.9% in 1976, by far the largest of the decrease is is due to slow slow productivity growth. largest part part of the decrease due to productivity growth. CalCal- culations of the trend in total factor productivity using 4% unemployunemploy- 52 52 estimated potential GNP would be ment as a benchmark indicate that estimated .3 to 1.1 percent higher with this standard, depending on how the reduced unemployment is distributed among demographic groups. Unemp1o~entand Unemployment and Real GNP Change: Checking the Results As a rough check on the potential GNP estimate of 1332-1374 billion 1972 dollars for 1976, the relationship between changes in the overall unemployment rate and changes in real GNP was estimated using equations 13A and 13B. h3B. (13A) {13A) AU LIUt ~ (13B) (h3B) LIU AU .38 .24* = .38 - .24* (.04) ( .04) ((.02) .02) = — A LI %GNP .h8* A %GNP %GNP - .18* LI %GNP ~ t1 t (.02) (.02) t-1 - = .77 d-w == 1.89 data: = quarterly 1953:2-1976:4 .45 - .25k %GNP - .10* ALI %GNPt %GNP l .25* LIA %GNPt (.04) (.02) (.01) (.04) (.02) ~ (ofl Q3* ALI IGNPt_ %GNP 4 - .03* - .08* LIA %GNPt 2 - .05* LIA %GNP (.01) (.005) t-3 (.003) (.01) -2 (.005) t-3 (.003) (last four regression coefficients constrained to lie on a straight ine) straight 1line) R2 = d-w=l.89 = .75 d-w=h.89 data: quarterly 1953:2-1976:4 1953:2-1976:4 t = = — - - — - = Ut_U = Ut-Ut == percentage point point change in the overall 1 unemployment rate U. tl unemployment rate U. A%GNPtt = = 100*(GNPt_GNPt ll%GNP l00*(GNPt-GNPt-l)/GNPt-l 1)/GNPt1 , == percentage change in Gross National Product measured at 1972 prices. AUt , Equation 13A implies that aa one percentage percentage point reduction reduction in the overoverall unemployment rate will be associated with a 2.4 percent increase in real GNP in the long run, while equation 13B l3B implies an eventual 2.0 percent increase. A unemployA 2.8 percentage point decrease in the unemploy— 53 ment rate in 1976 from from the realized 7,7% 7. 7% to the 4. 9% CEA benchmark, benchmark, ment 4.9% rea 1 GtJP 5. 6 to 6. 7%. Since Si nee GNP at 1972 prices would therefore increase real GUP 5.6 6.7%. was $1264.7 billion in 1976, these increases imply of 1336 to 1350 billion 1972 dollars. aa potential output These figures are below the middle of the potential GNP range for 1976. It should be noted that the result of 2.5 or less for an estimate of% of % 6 A GNP/6 GNP/A U U is lower than the 3.0 or greater used by some economists. The confusion here probably lies in the distinction between the shortrun and long-run responses of unemployment to output. In the short short run (one quarter), it takes an additional additional increase of 4% 4% in real output to reduce unemployment by an additional one percentage point. However, additional unemployment reductions are forthcoming in future quarters, even if there are no additional marginal increases in real output. Since attention is focused on real growth and unemployment during economic activity, activity, it is natural to estimate the growth periods of periods of slack slack economic it is natural to estimate the growth in output that would give an acceptable decline in the unemployment rate. At the beginning of aa recovery, this “required "required real growth” growth" may be very high. high. Conclusion Concl us ion The new CEA estimates of potential output are confirmed by the updated results presented in this paper. The changes generated by an additional year's data very small, small, with with the the biggest adjustment additional year’s data are are very biggest adjustment being the the increase increase in in the the unemployment unemployment benchmark 4.9% to being benchmark from from 4.9% to 5.1%. 5.1%. The atypical productivity decline experienced in 1974 has not been 54 reduced in the past past year, adding more weight to the argument that that 22 to 33 percent was permanently lost from the trend level of productivity. The CEA potential potential GNP estimates in the 1977 Economic Rqpprtof Report of the President are, if anything, optimistic about the gains in output resulting from a reduced level of unemployment. In the second quarter of 1977, CEA estimated that a reduction in the unemployment rate from the observed 7.0% to 4.9% would have increased real GNP from $1330.7 to $1405.8 billion, or 5.6%. The 1977:2 potential estimates range from 1372.3 to 1419.6 billion, using aa slightly slightly higher 5.1% unemployment benchmark. "Okun's Law,” Law," with aa multiplier of 2.0 to 2.5, yields a “Okun’s range of $1386.6 to $1400.6 billion, using the 4.9% unemployment benchmark. Thus, the results reported here indicate an "output “output gap” gap" which is generally smaller than the official CEA gap of 5.6%. Estimates of the current output gap which are significantly significantly larger than 5.6% must be based on assumptions about large cyclical variations in the labor force and productivity which are unsupported by the data. a1 growth rate" Results on the "potenti “potential rate” for the economy over the next five years are much less precise. Structural models (as opposed to the empirical trend-fitting equations used in this study) of growth in labor force participation participation and productivity have not been developed to the point that they can be used to make good conditional predictions. Therefore, any projection of potential output must be an extrapolation of past trends. A growth rate in potential output of 3.5% per year is A consistent with the growth rates of the labor force and output per worker which have been observed since the late 1960s. 55 However, high labor force growth, coupled with aa return to the pre-1966 trend in tctal total factor productivity, strong capital stock growth and lower relative youth unemployment rates could generate spectacular economic growth over the next five years. Alternatively, sluggish performance in all these areas could result in aa very low 101-1 real growth growth rate. Erratic behavior of of productivity, coupled with recent changes in labor force parti ci pa ti on trends and unstable unstable prices prices make make any any projection projection of future future participation growth grov1th rates subject to wide variability. 56 Footnotes ‘I! AJ2p ort of the Council of Economic Advisers, January 1962, 1962, l/ AnnualjRe Report p. p. y2/ 49ff. See Okun (11) for an explanation of the various methodologies used to relate unemployment and real output. Black and Russell (1) 3/ For example: Kuh (8), Thurow and Taylor (20), Black l! and Perry (12), {12). 4/ CEA was not alone in its concern about its "old" “old’ estimates of potential output. Data Resources voiced its concern over the potential output output estimates in early 1976 (Brinner (3)). (3)). Publication Economic Re Report of the 1977 ~g~pmic ort, generated additional studies, including Perry (13), and Rasche and Tatom (15). (15). 5/ 'ii See the 1977 Economic Report of the President for aa non—technical non-technical discussion of the issues involved, and Clark (4) for some of the statistical results results used used in in the the re-estimation re-estimation process. §j 6/ This concept of the private sector sector is is close to what Denison calls "nonresidential business sector." the “nonresidential sector.” See Denison, (6), p. 2lff. 21ff. It is is also very very close to to the the Bureau of Labor Labor Statistics' "private Statistics’ “private sector." business sector.” (17). 7/ See Musgrave, (9), and Segel and Rutledge, (17), ?J 8/ §j See Raddock and Forest, (14). 9/ '}j 1973-1976. Data are available by age and sex for May of the years 1973—1976. 1973 was chosen because it is closest to aa cyclical peak. Such an adjustment is sometimes called “Perry-weighting” "Perry-weighting" since aa similar weighting scheme was used by George Perry in in adjusting adjusting the weighting scheme was used by George Perry the unemployment rate: (12). 10/ See, for example, Sclove, Sciove, (16). lQ/ and Earnings, various issues. 11/ Bureau of Labor Statistics, Employment ipjp~ep~,~ndE~nins, ll/ 12/ .!Y See Stein, adjustment subject to (19). Since Since the parallel surveys in 1966 used for this are only half the size of the CPS, these ratios are considerable sampling variability. 13/ As mentioned earlier, this result indicates that the unemployment DI — rate for all persons 25-64 could be used as aa cyclical indicator in place of the unemployment unemployment rate for men 25-64. Observations of the adult womens’ womens' unemployment rate relative to that of adult men 57 point increase in the differential between them from shows a .5 point 1962-1966, 1967 (as predicted by the BLS 1962—1966, a .7 point increase in 1967 1967 to 1968. partial samples) and then aa .6 point decrease from 1967 This strange behavior of the women's women’s unemployment rate influenced the decision to use the rate for adult men, although results using either rate are virtually identical. 14/ See G.E.P. Box and G.M. Jenkins, (2), p. 67ff, for aa discussion of l'±I the conditions under which aa moving average process such as (6) can be converted to aa one-sided autoregressive scheme. called “short-term ''short-term increasing 15/ This lagged response is sometimes called Ji! returns to labor.” labor." See Sims, (18). 16/ One possibility possibility is that capital input response is instantaneous, while labor input response is lagged. This implies aa cyclical adjustment that treats labor and capital inputs differently. Experiments with such a specification yielded results insignificantly different than those reported below. 17/ See Norsworthy and Fulco (7) for aa discussion of the reasons for the productivity slowdown. Embodied technical progress and investinvestment in research and development may have contributed to the slowdovm, but these factors were not analyzed. slowdown, }W 18/ Okun, (8). ]2/ 19/ Business Conditions Digest, August 1976, p. p. 95. 20/ Ibid. 58 References 1/ ]j Black, S.W. and R.R. Russell, “An "An Alternative Estimate of Potential GNP,” Review of Economics and Statistics, February 1969, 70—76. GNP," 1969, p. 70-76. Y Time Series Analysis, Holden-Day, 2/ Box, G.E.P., and G.M. Jenkins, Lime San Francisco, 1970. ]/ 3/ Brinner, R.E., “The "The Growth of Potential GNP," Resources GNP,’ The Data Resources U.S.~~-TermBul1eti,~,, U.S. Long-Term Bulletin, Winter \~inter 1976, p. 91-110. 4/ 1lf Clark, P.K., "A “A New Estimate of Potential GNP," GNP,” in ~~Conrefl, U.S. Congress, Joint Economic Committee, Hearings on the Economic Report of the President, January 19, 1977. 1977. 5/ !if Council of Economic Advisers, "The “The Annual Report of the Council of Economic Advisers,” Advisers," in The Economic Report of the President, various annual issues. §! 6/ Denison, E.F., Accounting for United States Economic Growth, Growth, 1929-1969, Brookings Institution, Hashington, Washington, D.C., 1974. 1974. Po ten ti a 1 Output in the U.S. Economy, Joint 7/ Knowles, J.K., Potential ?J Economic Committee, 1959. 8/ §.I Kuh, E. "Measurement of Potential Output," E.,, “Measurement Output,” American Economic Review, September 1966, p. 758-76. 758—76. 2J Residential 9/ Musgrave, J.A., "Fixed “Fixed Nonresidential Business and Residential 1925-1975," Survey of Current Capital in the United States, 1925-1975,” Business, April 1976. 10/ Norsworthy, J.R., J.R., and L. Fuico, Fulco, "Productivity Costs in the “Productivity and Costs Private Economy, 1973," Monthly Labor Review, June 1974. 1973,” ~p,~hl 11/ ill Okun, A.M,, “Potential Okun, A.M., "Potential GNP: Its Measurement and Significance,” Significance," in American Statistical Association, Proceedings of the Business and Economics Section (1962). 12/ Perry, G.L., “Labor "Labor Force Structure, Structure, Potential Output, and Productivity," Papers on Economic Activity, (1:1971), (1 :1971), Productivity,” Brookings Papers p. 11-47. — ill 13/ Perry, G.L., G. L., "Potential Productivity," Brookings Papers “Potential Output and Productivity,” g~i~jP~2!~ gn Economic Activity, Actiyjj~y, (1977:1), p. 11-47. on 59 14/ Raddock, R.D., R.D., and L.R. Forest, "New “New Estimates of Capacity Utilization: Manufacturing and Materials," Materials,” Federal Reserve 1976, p. 892-905. Bulletin, November 1976, 15/ Rasche, R.H., and J.A. Tatom, “Energy "Energy Resources and Potential GNP,” GNP," Federal Reserve Bank of St. Louis Review, June 1977, p. 10-24. 16/ Sclove, S.L., “Improved "Improved Estimators for Coefficients in Linear Regression," Journal of the American Statistical Regression,” Stati sti cal Association, June 1968. ]]j 17/ Segel, F.W., F.l✓., and G.L. Rutledge, “Capital "Capital Expenditures by Business for Air, Water, l✓ ater, and Sol id ,Jaste and Solid Waste Pollution Abatement, 1975 and Planned 1976," Survey of Current Business, July 1976. 1976,” ~ 1976. 18/ Sims, C.A., “Output "Output and Labor Input in in Manufacturing,” Manufacturing," ~r~~ins Brookings Papers on Economic Activity, (3:1974). ~ 19/ ill Stein, R.L., “New "New Definitions for Employment and Unemployment,” Unemployment," Monthly Labor Review, February 1967. ~ 20/ Thurow, L.C., and L.D. Taylor, "The Interaction Between Actual and and “The Interaction Potential Rates of Growth,” Growth," Review of Economics and Statistics, Statistics, November 1966, p. 351-60. ~1/ 21/ U.S. Department of of Commerce, ~ Business Conditions Digest, August 1976. 1976. inessCopciItionsDiest,, 60 Table A-i A-1 Fixed Nonresidential Capital Stock at 1972 i972 Prices Excluding Pollution Abatement Capital (billions of 1972 dollars) 1948 1950 1955 1960 1965 Note: 632.8 632.8 658.0 681 . l 681.1 707.5 734.2 761 .l 761.1 787.9 815.4 845. 7 845.7 876.3 902.2 925.0 950.5 976.4 1003.6 1033.5 1067.2 1110.4 1966 1970 , 1975 1980 1164. 1164.22 1219. 3 1219.3 1273.3 1331.1 1387 .6 1387.6 1437.2 1437.2 1486. 1 1486.1 541 . 7 1 1541.1 1600.6 1649.0 1687.1 1729. 9 1729.9 1783.5 1845. 6 1845.6 1913. 2 1913.2 Figures are average values of capital stock during the given year. 61 Table A—2 A-2 Table Potential Civilian Force 1948-1980 Potential Civilian Labor Labor Force 1948-1980 (millions of persons) (millions of persons) 1948 1950 1955 1960 1965 60.6 61.5 61.5 62.4 62.0 62. l 62,1 62.9 62.9 63.8 65.0 65.0 66 .5 66.5 67.0 67.0 68.0 68.0 68.6 68.6 69.8 69.8 70.8 70.8 70.8 72.0 72.0 73.2 74.4 1966 1966 1970 1970 1975 1975 1980 1980 62 62 75.6 75.6 77 .2 77.2 78.6 78.6 80.5 80.5 82. 82.77 84. 84.33 86.7 86,7 88.7 88.7 91. 91.1l 93.3 93.3 95.9 95.9 98.0 98.0 99. 7 99.7 l101.4 01 . 4 103.3 103.3 A-3 Table A-3 Table 75 Employment, 1948-19 Private Employment, Private 1948—1975 ) persons of (millio ns (millions of persons) 1948 1950 1955 53. 1 53.1 52.3 52.3 53. 53.1l 53.9 53.9 53.8 54.7 53.5 53.5 55.3 55.3 56.9 56.9 56.9 56.9 55.6 55.6 56.9 56.9 1960 1960 1965 1965 1970 1970 1975 1975 1976 1976 63 63 57.8 57.8 57.6 57.6 58.0 58.0 58.7 58.7 60.0 60.0 61. 5 61.5 62.6 62.6 63.2 63.2 64.3 64.3 65.9 65.9 66. 2 66.2 66.3 66.3 68.4 68.4 70.8 70.8 71. 9 71.9 70.3 70.3 72.5 72.5 Table A-4 Table A-4 Estimated Capacity Utilization Rate for the Private Sector, l948~lg76~ Sector, l948-1976a (percent) 1948 1950 1955 1960 85.0 80.8 85.2 86.7 86.4 88.4 83.8 87. 3 87.3 86.8 85.6 81. 3 81.3 84.6 83.8 1961 1965 1970 1975 1976 a 82.4 84.5 85.5 86.6 88.5 89.3 87,2 87.2 87.3 86.9 83.4 82.8 85.3 87.5 85.8 80.5 83.8 Annual average rate. rate, Quarterly Quarterly series from the Annu~l average series (Rt) (Rt) is is derived derived from the FRB manufacturing utilization rate (Ft) (see Raddock and and Forest (10)) according to the following following formula: Rt = 1/2 (87.5 + Ft) Rt= 64 Table A-S A-5 Full Employment Unemployment Benchmark Equivalent to 4.0 Percent Unemployment in 1955 ( percent l (percent) 1948 1950 1955 1960 1965 4.4 4.3 4.3 4.2 4. l 4.1 4.0 4.0 4.0 4. l 4.1 4. 1 4.1 4.1 4. 1 4.1 4.2 4.2 4.2 4.3 4.3 4.5 1966 1970 1975 1980 4.6 4.5 4.5 4.6 4.7 4.7 4.9 5.0 5.0 5. l 5.1 5. l 5.1 5. 5.1l 5.2 5.2 5.2 Note: Unemployment rates are computed relative to the sampling procedure actually used in In a given time period. The CPS in the benchmark survey change in 1967 causes the shift In unemployment rate from 1966 to 1967. 65 Table A-6 Projections of Noncyclical GNP Components 1976-1980 (billions of 1972 dollars) (billions of 1972 dollars) Compensation of Federal Employees Year 1976a 1977 1978 1979 1980 Compensation of State and Local Government Employees 48.4 97.3 98.7 l 01 . 9 101.9 l 05. 2 105.2 108.7 108.7 48.6 48.6 48.7 48.7 Gross Ouput Attributed to Residential Housing Stock Originating Gross Ouput Originating in Rest of World (GNP-GDP) (G NP GD P) - 1976a 1977 1978 1979 1980 a 6.7 8.0 9.2 9.2 10.3 11. 11.33 111. 111.66 114.88 114. 118.4 118.4 122.2 126. 126.22 1976 figures are actual, included for comparison. 1976 figures are actual, included for comparison. 66 POTE/lTIAL GROi/T!{ RATE POTENTIAL OUTPUT NHJ AND ITS GROWTH — THE OOfclI!!ANCE DOMINANCE OF HIGiiEf; HIGHER rnrnGY ENERGY COSTS IN I:i THE TliE 1970’S l970'S Robert H. 1-!, Rnsche John A, Tatori R11sche and ,John A. Tator,1 l96Os, the level Since the early early 1960s, ]eve l and the rate of growth grm'lth of potential output have become subjects, potenti a 1 outrut becor,1e increasingly i ncreas i n!J1Y important ir1rortant subjects. While \/hi 1e policymaker’s concepts has become po 1i cymaker' s and the public’s rub l i c' s acceptance of these concerts becorie widespread, since the early 1videsrrend, e11rly 1970s there t!:enc has been heen considerable controversy concerning measurement of potential concernina the rneasurer,1ent rotential output and its its growth. By fly 1973 lS173 it it had become hecorce clear to many nany observers that the Council of Economic Advisers (CEA) measure of potential output was too high. That measure shov1ed showed slack in the economy e~ual equal to $30 billion (1972 dollars) while many mc1ny observers thought thou9ht the economy was 1,1as operating at or above its potential, at least in the early part of the year. In irid-1973, mid—1973, Business Week Heek summarized the “Debate "Debate Over Gauging the GilP GHP Gap," Gap, pointing out the ir1portance importance of the issue for assessing stabilization policy, particularly prospects. ]J recession prospects, for near term inflation and La1·irence Klein, iiAlan 1an Greenspan, Geoffrey Moore Lawrence economy was much closer to full utilization and others argued that the econoriy potential of resources than the CEA CEf1 rotenti al output measures then revealed. Nevertheless, llevertheless, Arthur Okun and George Georqe Perry were said to remain renain defenders of the slack economy view, econol"y vie1•1. sayino: "I “I am Perry is quoted as saying: Ors. Rasche and latom Tatom co—authored co~authored this paper while \lhiie both were employed enployed Drs, by the Federal Reserve Hank Bank of St. Louis. 67 that we have the the industrial to bring bring the the not persuaded that we do do not not have industrial capacity capacity to unemployment rate do1~n down to 44 percent. 1looks ooks 1like i ke a firm number,” number." bi 11 ion gap still sti 11 To me, the $20 billion Y V Since 1973, energy price developments, the inflation experience, 1abor s 1uggi sh capital growth in the recovery, and unusual labor a recession, sluggish productivity changes have brought the potential debate to a turning point. Within the past year, Business Conditions Conditions Digest ceased publication of the CEA series. Peter K. Clark’s Clark's study, “A "A New Estimate of Potential Potenti a1 GNP," GNP,” circulated, and the CEA reported a new series for potential output in the 1977 Economic Report of the President, President. While the revised estimates reduce the previous measures of potential output, the ~ Report points to evidence of the need for further revisions due to a productivity loss since 1974. The ~Repo1't also suggests that the growth rate of potential is about 3,5 prior 3.5 percent -- lower than the prior —- official view. Early this year a study by Data Resources, Inc., reached a similar conclusion. lf May of this year, the theoretical In our Review article in i1ay <lue to foundations for aa loss in economic capacity and potential output due the 1974 energy price change are derived. Our results support the capacity loss hypothesis which has been discussed in numerous Review capacity v1hich receives indirect support articles in the past, an hypothesis which monetarist price price equation. 1/ Our Review from an investigation of a monetarist article in June provides further empirical support for the loss in economic capacity capacity due to the change in the relative price price of energy using aa production function approach to measure potential output. contention that downward revisions are results support the CEA contention 68 The necessary in their revised potential output series for recent years. Nevertheless, our potential output measure appears consistent with the lfovertheless, old CEA series until the early seventies, especially in 1955 1955 and 1969-70. 1969—70. §I Perry's, “Potential "Potential Output and Productivity,” Productivity," appeared in George Perry’s, the July Brookings Papers. §j Brog jns Paers. His potential output measures also dmvnv1ard revisions from the old official series, especially indicate downward for the late l9GOs 1960s and early seventies seven ti es when his potential potenti a 1 output measures are actually below the new CEA measures. measures, Perry's Perry’s measure for 1973 shows the economy operating slightly potential, a point of slightly above potential, agreement with vtith Clark. Perry's measure of potential output Clark, 7J Li’ Ho~1ever, However, Perry’s begins begins to grow more rapidly in about 1970, and it it grows more rapidly than our measure or the new CEA measure of potential output. In fact, Perry's o 1d CEA Perry’s growth rate for 1970-76 1970—76 is about as fast as that of the old period. measure which he has repudiated for the earlier period, The three studies share a general conclusion that, at least through 1973, 1 d CE/1 potenti a 1 output of the 1973, the oold CEA series overstated the potential U.S. econo111)1, economy. There is some difference in the pattern of the downward case, revisions in each case. Interestingly, both Clark and Perry reduce while our measure is only $15 1973 potential output by over $30 billion 1°1hile billion belm1 below that in the old CEA series. differences, serious differences. After After 1973, there are more Our measure shovis shows a substantial substantial effect effect of energy developments, the others do not, not. Also, Perry's Perry’s measures indicate a sharp acceleration in the rate of potential output growth in the l970s 1970s the nevi new CEA estir.1ate. estimate. with potential growing much faster than tile In this paper we review briefly the theoretical and empirical 69 basis of our earlier potential output results. To provide a context for this review, quarterly potential output reviews the discussion focuses upon quarterly measures nEasures instead of the annual measures presented in the June article. article, Our measures of potential output in the June Review and below rely Clark's vmrk rates. heavily upon Clark’s work on the labor force and participation rates, developments Ignoring energy deve 1 or errors due to assumptions concerning capi ta1 growth or growth of non-private non-pr'ivate business sector output output and capital production employment, the growth rate of potential output from our production function analysis should be his. he about equal to his, The only remaining difference would be that our estimated labor, capital, and trend coefficients deviate very slightly from his assumed labor coefficient of two-thirds, and estimated trend term of about 1.55 percent. Thus, in our earlier work we found it convenient to follow Clark's Clark’s analysis 1976. and assume a 3.5 percent rate of growth of potential output after 1976, Since there is a 1 arge gap between the CEA and Perry estimates of the large outlook for potential output growth, we examine the growth issue as well. During the recovery (since 1/1976), I/1975), investment in plant and equipment has been a continuing concern to economic analysts, not only because it reflects the business outlook of investors and affects current employment, emp 1oyment, but also a 1so because it affects the future growth of actual output (or, implicitly, impli t]y, the growth rate of potential output). Our research offers an explanation exp 1anati on of both the ssluggish 1 uggi sh growth of investment and slower slower than expected growth in potential output since the recession, Finally, we offer some comments on the prospects for decade, potential output growth for the remainder of the decade. 70 70 Quarterly Potential Output: 1949-1977 1949—1977 Until recently, the conventional method for measuring potential output focused upon the relationship between output and labor force growth. The growth rate of potential output depended upon the growth rate of the labor force, secular changes in hours per worker, and and labor productivity trends. trends. In several papers, dating back to the original potential output studies, the importance ,of of accounting accountin<J explicitly for for growth in the capital stock is emphasized. The use of an aggregate production function relating potential resource employment to potential output is the obvious obvious solution and one which has been followed most recently by Clark and Brinner. However, they simply employ conventional assumptions: a labor share of income of two—thirds two-thirds and a residual or one-third. capital share of one—third. Our theoretical work on capacity output suggests that energy price rroductivity of domestic labor and capital carita1 changes have an effect on productivity resources. Moreover, we are not content to fix factor share coefficients, especially since data on energy use is not collected in a a11ows ready computation of its factor factor share in cost. form which allows Thus, our work v/Ork begins with a production function, but the coefficients of the three resources —- labor, capital, and energy —- are estimated. From this Froni production function, potential output is measurable given assumptions concerning potential resource employment. AA Quarterly Aggreqate Aggregate Production Function measurina potential output The fundamental relationship used for for measuring is a production function for private business sector output. 71 Output (L), (Y) is hypothesized to be a function of hours of all persons (1), capital (K), energy (E), and disembodied technological progress. The Cobb—Douglas and r is the constant trend rate of production function is Cobb-Douglas growth, growth. ( 1) (1) y_MrtLaKsEY The demand for energy may be derived from the production function and eneruy usage found by equating the supply of energy to the the rate of energy econorr,y is a price price taker in the energy demand for energy, assuming the economy market. market, Substituting the equilibrium quantity of energy in the yie 1ds production function functi 011 yields (2) (A* elt ert ~ L a(j, KKfl8 PP -y)l/1-y (2) Vy =~ (A* ·js the relative price of energy, measured by deflating the where PP is wholesale price index for fuel, related products, and power by the implicit price deflator for for private business sector output. 81 Hours of for the private business sector are all persons data and output for prepared by the Bureau of Labor Statistics of the U.S. Department of Labor. The capital stock data is based upon interpolation of the structures end-of-year net stock of fixed nonresidential equipment and structures. 91 The interpolation uses prepared by the U.S. Department of Coimnerce. Commerce. 2,’ fixed investment in quarterly rates of constant dollar nonresidential fixed GNP accounts as weights in finding end-of-quar,ter net capital capi ta1 the SIP finding end—of-quarter stocks. The flow of capital services is computed by multiplying the previous end-of-quarter capital stock by its utilization rate as measured by the Federal Reserve Board index of capacity utilization. end-of-year capital stock is Since a consistent measure of the end—of-year 72 only available with ~1ith a lag, laf), estimates of the quarterly capital stock from the prior rrior (II/19481V/1975) (II/l948IV/1975) relationship after 1975 had to be found from of quarterly changes in the net stock to the quarterly rate of investr:ient and, to account for depreciation, the nonresidential fixed investment lagged net capital stock. The equation is: (3) Kt — ~t—l= 1.012 (4.5) (4.5) + .2457 It — (29.2) = .98 .0252 Kt1 (-21.Cf) (—21.4) thU. == 2.10 2. 10 S.E. == .37 ,37 = p .49 v1here Kt is the constant dollar net stock of equipment equipr.1ent and structures where at the end of quarter tt and It is constant dollar nonresidential fixed fixed investment in quarter t. The quarterly production production function, estimated for the period Il/1948 hor10geneity constraint, is: 11/1948 —- IV/1975 1•:ith with a linear homogeneity Y = 1.5380 L + .2774 ln 1n • 7226 ln ln Vt 1.5380++ .7226 In L+ in Kr,, ( 4) (4) (13. 77) (21.24) (13.77) — (8. 15) (8.15) .HJ401n .1040 in pp ++ .0046 tt (-5.05) (—5.05) ( 15. 35) (15.35) R2 = .98 D.td. D. ~!" == 1.93 .0076 S.E. == ,0076 pp = .80 The indirect es estimates ti mates of the production function parameters are (standard errors in parentheses) 25. ]5; (3.09%), ~ y = 9.4% {1.86%), a = 65.5% (3.09%), 8B == 25.1% (1.86%), ' r —= .4% . 4"' , J oI(.03%) , . 03'L) ,,~ 73 outrut elasticities are not The estimated equation and the output significantly different from the annual estimates or the quarterly estimates reported in the June article. article . 1.2/ .lQ/ As earlier, the estimated elasticity of energy eneroy is lower lov1er than that found using annual data. Therefore, the energy price effect will appear smaller in the potential output series. The quarterly production function above was also estimated with the output elasticity elasticity of energy constrained to be 12 percent, the estimate found using annual annual data. An F test of the constraint indicated that (y (; =~ 12%) 12;;) could not be rejected (F ,, 1 106 106 Nontheless, the more conservative estimate of y is used below. .66) •.!1! == .66).!!! Some statistical statis 1•1e reported in the June properties which we repentinr;. article bear repeatinq. First, the production function is stable when 1 l or 197S. \Jhen energy is omitted 01,1itted from the estimated through 1973 1975. t4hen aldin<J the observations o:Jservatfons for 1974-75 results in a production function, a4ding th,, estimate e,;tinat.:? of the output elasticity '=lasticity of labor and a sharp decline in the significant rise in the standard error of the equation. AA Chow test on s tructura 1 change when energy is the additional observations indicates structural whe:1 it is included. omitted from from the equation, equation, but not when Second, in our discussion (Appendix (Appenrlix II) of potential biases in the estimation estimation due to Cobb-D011'.)1 as production function, we vie noted that if the the assumed Cobb—Douglas own—price own-price elasticity of demand demund for for energy is not unity, our assumption imparts a downward bias b-ia,; to the estimate of the output elasticity of energy (y) and an upward up,nrd bias to the output elasticity of labor (a). The consistency of th2 the estimate of a with the labor labor share data indicates that that this bias, if present, is not substantial. In any bfos were v1ere present, it would not bias the estimated event, even if this bias 74 regression coefficients upon which viiiich the potential potent-ial output measures neasures are regression based. In estimating estinating the annual production function for the June article we account for other other factors article vie attempted attempted to account factors which have been cited as Influencing inf1uencin9 productivity in the last decade and which 11hich are believed by many to have lowered loviered potential output in recent years. Such adjustments involve the labor force, capital and trend measures. rneasures. An attempt to adjust for the quality of hours in the tile production function by accounting for the labor force share of young youno people proved to be statistically insiunificant. insignificant. i\11 adjusticnt adjustr:ent to the gross 9ross capital stock An rer,iove pollution abatement abaterilent capital does not affect the coefficients to reuove or improve the standard error of the production function and does not appreciably affoct neasure of potential output either. affect the measure AA break in the trend rate of gro\'lth growth was also allowed because of an observed slowdown in productivity growth 1967. slov1dmm orov1th since 1967. The slowing slo11inr1 of the trend was statistically statisticolly significant, but, lacking lackinu a explanation for it, we vie have chosen to ignore i rJnore it. i::1portant It may appear that the break in the trend has important implications i1;1plications for the prospective growth 9rov1th rate of potential output. thc major effect is to raise potential output measures rieasures in the Instead, the mid- l 960s. Trend terms terr.is with 1/i th a break in 1967 1%7 show silo\'/ trend growth 9rov1th to be at mid—l960s. a 2.02 percent rate prior to 1967 and 1.55 percent since then. The trend growth ~rowth in the production function above ilbove of 1.6 l .G percent riercent is not markedly greater than the current trend rate where a trend growth 9ro11th s 1ov1dovm is allowed. a1101·1ed. slowdown 75 Hours Per Worker Potential hours per 1:orker worker in the private business sector in the vJOrker to a June paper is ·'ound ?ound from an equation relating hours per worker trend and a cyclical '!ariable the civilian variable ~~ the unemployment rate of the labor force estir;1ated for the period from the second quarter of 1948 estimated through 1975. ·,\'io changes have been made to measure potential hours l975~ Iwo per worker. First, for consistency the tile cyclical variable has been changed to the difference difterence between the actual unemployment rate and the fuli~employm2ntunempioyr;ient fu11·•emp1oyment une";ployr.1ent rate. the regression equation, ·121 This change has little little effect on The second change cnange is to allow for the unusual behavior of hours 1961-67 period by using a dummy variable. per worker in the l96l~67period Inspection Inspection of the earlier results reveals that the equation has systematically large errors (more than one standard deviation) for every quarter from mid-1961 mid~l96l through the first quarter of 1967. More importantly, the actual hours per 11arker worker exceed the estimated potential levels throughout the period by relatively large amounts. Various specifications specifications were tested, including alternative time intervals for a temporary or permanent shift and changes in the trend rate of decline of hours per worker. worker, The time interval chosen and a temporary (versus up1,ard in hours per worker fit the observed error permanent) shift upward pattern most closely and yielded the lowest standard error. Tests for a break in the trend rate of decline in hours per worker failed to support such an hypothesis once the temporary shift upward was taken into account, Also a comparison of equations allowing only a break in the trend or only the temporary shift in the level of hours per worker 76 revealed the te"1porary temporary shift yielded the lowest standard error. error, 121 The equation for hours per 1'/orker worker is: + ,0136 ln !!Pk/ • 7983 - •.3229 3229 u:1 IIPW == .7983 UN - ,0010 .0010 t ± .0136 D D (1004.6) (-36. l) (13.1} (1004,6) (-10.7) (—10,7) (—86.1) (13,1) — R2 = 99 = • .99 ll.ll. == .74 • 74 [LW, Sample: II/1948-IV/1975 iI/l948—IV/1975 Saraple: S,E, = • .004 S. E. = 004 v1here HP\'/ where HPW is hours per worker, llll UN is the excess of the unemployment rate over its ful1-employ1,1ent full—employment rate, tt is time, and UD is a dummy v1hich rises rises to one in steps steps of one-fourth variable which one—fourth beginning in the third quarter of 1961, 1961, and phases out in the same way reaching zero in the second quarter of 1967. ill B’ Potenti a1 hours per worker are found by setting UI, equa 1 to Potential UN equal zero. It should be noted that potential hours per worker in the mid-1960s are higher hi1her than our previous esti111ates mid—1960s estimates so that the difference betv1een our measure of potential output and the old CEA measures appear between smaller than in Chart IV of the June dune paper. ~erl’Potentia10utut Quarterly Potential Output To measure potential output, potential resource er;1ployment employment in the private business sector and the tl1e potenti a 1 output in the remainder rernai nder potential of the economy must be he measured. measured, Actual output is assumed to be econor;iy which are not included included in potential output for sectors of the economy the private business sector (general ( genera 1 government, governrnent, rest of world, vmrl d, imputed output of housing, and output of households and nonprofit institutions). institutions). C<1pital Capital ernploynent employment is assumed to be the services of the 77 capital stock at an 87.5 percent FRB ms capital utilization rate. This ut i1 i zat ion rate was chosen as a full—employment fu 11-e1;1p 1oy,nent measure r,1easure because It it was utilization preva"ilino rate during the tile benchmark bench11:ark year of 1955 and in other the prevailing fol"l-employment periods. peacetime full—employment Energy employment ernployment is asswned assumed to be that demanded at potential output, given the actual relative price of energy in each quarter. Potential hours of all persons are measured by combining emp 1oyment and potential measures of potential private business sector employment ,/Orker. Potential private business sector employment Is is hours per worker. found by subtracting the 11rn,1ber unemr,loy2d at full—employment fo11-emp1oyrnent from number of unemployed Clark's Clark’s measure measure of the potential civilian labor force to obtain emp 1oyr1ent and then subtracting subtracti n9 actual actua 1 employment ernp l oyment potential civilian employment 151 . ~ business tJ11s1ness . t . ' t' ou_s,ae rie private pr1vaLe sec t or. -El outside the sector. 11hich the potential output 1.1easures The assumptions upon which measures are optil.listic and und may 11ay result in overestimates based are, if anything, very optimistic of potential output, especially in recent years. First, by using the estimate of the output elasticity of energy employment of 9.4 percent instead of the 12 percent estimate estitnate from fror: the annual regressions, the the impact of energy price increases is Is lowered. Second, Clark’s Clark's series for the full-employment ful1-e,1ployi;1ent unernployr:ent rat1, may may lead to a significant unemployment rate overstatement of potential employment. er,1ployr,1ent. ,!achter (DPEA, 1:1976 1 :1976 and BPEA, Wachter (SPEA, l :1977, pp. 4851) has made a strong stron9 case for the th1c full—employr;ent ful1-employr.1ent 1:1977, uner.1ployment rate being higher in recent years than thnn the Clark and Perry unemployment estimate. {His work also implies ir:1plies a s1oHer (His slower rate of growth of the force)$ potential labor force). paper by Ronald Talley. Sir.:ilar measures have been derived in a recent Similar 161 Finally, the measures may nay be overly 78 optimistic because they do not adjust for the tr,e factors discussed earlier which would ,iou1 d all a l1 tend tend to reduce potential notenti a 1 output In in recent years such as pollution abatement investment, investr.;ent, the changing chan~ing composition of the labor force and slowdown slowdovm in the trend growth 9rowth of factor fac:or productivity. !leasures (]Uarter1y potential output ou-:µ:;t from 11/1948 !I/1948 through Measures of quarterly II/1977 are presented in Table 1. 1~ 11/1977 These measures are compared to others for selected years in Table 2 where v1here annual averages of quarterly measures are shown for selected years. years~ The other measures are those those of t:1e CEA until January 1977, called “old "old CEA,” CEi\,'' the tr1e CEA CE/\ measures reported the in January 1977, called “new nnew CEA,” CEJ\,n and Perry’s Perry's second series which does not have a break in trend growth. grm,1th. Using the the old CEA CE/\ series as a standard, it nay ueasure is fairly close in 1955 and may be seen that our measure 1:170 (closer than either eiV1er the new CEA CE/\ measure masJre or Perry II). II), Comparison 1970 to measures i;ieasures In in 1960 and 196S indicates that our estimate of the growth 1965 indIcates rate of potential potentic1l In in the late 1960s is higher hir;,-,er than the others 1·1hi1e while our 9rov;th Is is slower than the others othc:rs during the estimate of potential growth earlier period, especially from fron 1955—60. 1955-60, Fron 1970 until 1973, our From CEf\ series, liut measure is iJe1m1 below that in the old CEA but it is closer than the measures,, other two measures. Gy 1973, both the t(w CEA and anJ Perry reduce the old CEA CEJ\ By respectively)~ measure hy by sizable amounts amounts ($37 and $32 billion, respectively). In ccntrast, rieasure is only or,y about $16 billion below belovi the old CEA CE,q contrast, our measure i;;casure,, Figure 1 shm-;s the tile: differences betv-,een measure. 1 shows between three of the measures actua 1 output through throu~1h mid—1977. c,i d-1 ~J77 ~ from 1972-HO 1972—80 and actual fdter 11973, ::J73, our measure J.i1easure shows sh0;1s the impact ir.:pact After of the tne loss 1 oss of of potential output due to the large increase increa~:;c in the relative re·iative price of energy. By 1976, our measure is over $30 billion below the old ~EA 79 TABLE 11 QUARTERLY ~UARTERLY POTENTIAL GNP (Bi 11 •ions of 1972 Dollars at Annual Annua 1 Rates) (Billions I 1948 49 50 r•~ 51 52 53 54 ;)j fi 13 ~ 'I 513.1 ~)33,,0 533.0 562,,9 562.9 592.1 615. 2 615.2 629.2 56 6Ll6,. 7 646.7 669.7 57 h8G,,8 686.8 58 59 60 70B.3 708.3 S5 55 6'i 61 r~ 62 Of- 63 64 65 66 ,... ·-7 o, 67 68 69 70 71 70 , 72 ~ 73 74 75 765 77l III II 486, 486.11 520.7 537.7 572. 4 572.4 593.2 593"2 616,4 616.4 632.6 651. 6 651.6 676.3 676.8 692.3 720.5 740.1 731,,4 731.4 757 ,,6 757.6 IQ? ?: U,-1,,,' 783.6 8HL6 810.6 764,,0 764.0 739~0 789.0 819,5 819.5 ! 500.4 526. 1 526.1 543,6 543.6 578.9 601.9 618.0 639,3 639.3 1307. 1307.77 134 7. 9 1347.9 1353.8 798, l 798.1 824.0 829&9 829.9 335,,9 835.9 348.5 848.5 873,,8 873.8 902,9 902.9 950. 0 950.0 9"" ') 982.2 1027.0 1027"0 879.3 912.7 959.3 996,1 996.1 1035. 2 1035.2 1056,,9 1056.9 1071,3 1071.3 111 L9 1111.9 1151. 0 1151.0 1084.8 1100.9 1100" 9 1137, 4 1137.4 1178. 0 1178.0 1,-....J ?....,"!1,,u C: 1231.6 1259?6 1259.6 1283.0 1324. B 1324.8 1361.0 l 361 "0 L,. 1194, 8 1194.8 1244.4 1259,'.) 1259.0 1294, 1 1294.1 "1335,,9 1335.9 1366,9 1366.9 80 501.8 526.8 552.2 584.0 611.2 621.2 641.5 669.7 683.9 704.8 732.9 754.2 775.2 803,2 822.3 856.7 880.5 921.6 967.7 1006,9 1006.9 1045. 3 1045.3 1092.4 1125. 3 1125.3 1167 .o 1167.0 1222.5 1265,4 1265.4 1279.7 1315. 9 1315.9 661.3 661.8 680.5 702.9 730,7 730.7 743,7 743.7 767. 1 767.1 866,,2 866.2 890,,6 890.6 935,,0 935.0 978.0 1012,,4 1012.4 Qt:.," IV 1121.3 1155.7 1209.4 1257. 0 1257.0 1255.9 1265.9 TABLE TM:LE 2'~ r iras urf's of Potential Pot2ntiill Output Alternative Ileasures (Sill ions of 1972 Dollars) rlol lars) (Billions Oldd CEA C[/\ 01 :~e\'l CEA CEA !levi Perry II Perry 657.5 756.0 766.0 9n7 .. o 907.0 1114. 9 1114.9 65G.6 656.6 779.9 932. 1 932.1 1124.4 651.4 771. 771.99 92 E-i., 0 925.0 1106. 1106.22 657.8 775.l 775.1 918.0 1091. 7 1091.7 1971 1152,3 1152.8 1972 1973 1201. 1201.33 11 fi(). 9 1169.9 i21G.7 1216.7 12f)S,,4 1265.4 131,i.tJ 1315.9 1368.6 1421.2 1145.5 ll 8fi. 1 1186.1 1228 .. 2 1228.2 1271. 1271.77 1316.9 13fi3. 6 1363.6 1136. 1136.00 1184. 1184.88 1233. 1233.1l 1283.6 Rr1schr.-Taton1 Rasche—Tatom 1955 1960 1965 1970 1974 1249.77 1240. l 2GG,) l 1266.1 1975 1976 1300 .. 2 1300.2 134n,5 1340.5 8l 81 1334.9 1388. 1 1388.1 FIGURE 11 Actual Output and Ahernative Alternative Actua~Output Potentia~Output Potential Output Measures Ratio Stole Scale Scale Ratio Scale Trillions of Dollars Trillions of Dollars Quorterty Doto Dollars 7 1.7 , - - - - , - - - - - - , - - - - , - - - - - - - , - ~ - - - , - - - - , - - - - - , - - - - - - , - - - - - , 1 .Li 1.6 1.6 1.5 c:x~ 00 ro N 1.4 t~ 1.3 1.3 I l 1.2 - - - - t - - - - t 1.2 Actual Output 1.1 1972 1973 1914 1975 1976 1976 1917 1918 1979 1980 Sources, U.S. tJ.S. Deportment ~ Economic Department of Commerce, Council of Economic Economic Advisers, and George George Perry Perry, Brookings Papers on Economic 3 ~Activity \!:1977\. 1d9771. L!.. George George Perry’s Perry's series on annual annual measure. measure. U. series is is on [1 Potential output grows at 3.5"1;ofter 2nd quarter quarter 1977. ~,Potential 3.5% after plotted for Actual Output: OutpuE 2nd Quarter 1977 Latest data plotted Prepared Prepared by Federal Federal Reserve Bank of St. Louis bfi1icn below Df!lov1 Perry's tJl11ion below belo,1 the measure, S43 $48 billion Perry’s measure and S23 $23 billion Comparina real OIIP a::P to potential output yields Comparing new CE!\ tii:wsure, CEA measure. differe 1t measures Measures of the C44P Gr!P gap !'JIIP in 1976. The new CEA gap oap of markedly different 1 G,5 parcent, Perry’s Perry's gap of 8.2 percent percont and the t,;e old CEA CE/\ gap of 10.3 6.5 percent, percont outr,ut i11ply tl1at economic performance perfonnance was worse percent of potential output imply that pnivi ous pas tv;a r year except 1975. during 1976 than in any previous postwar In contrast, our measure rooa,ure of the thn gap (lap is below he1(M all ~11 postwar postwar recession years except 197041. Tt1us, ile11cr' s recent claims claius that the tho difference between Thus, Holler’s alternative measures nili!asures of potential rotentfa1 output is sr~ialland s,,.all and that there is con:.iderilble $1ack in the t:w econony seriously misstate r.Jisstate the case. )1j considerable slack Ile He apparently converts the revised CEA CEi\ measure mensure of 1412,0 and Perry’s Perry's Potential I measure r,e11sure of 1436.7 1436, 7 for for 1977 F.!77 to current dollars and concludes conc1 udes that in current dollars the gap is $116 btllion billion to $148 billion. The highest potential rotential measure, 11(,asm·e, the o1d CC/\ estimate estirmte would, if allowed a11m1ed to old CEA aro.; 3,75 percent percent-~ rate~irip1y ari gap gup of $200 billion. bi11ion, grow at 3.75 —e its 1976 rate —- imply esti~ate yields a gap of $165 ~165 billion. At the Perry’s Potential II estimate otiler extrct1e, I,1easure other extreme, the rm1,1 new CEA measure ~~ —— adjusted using their conservative productiv1ty decline dec1ino due to energy developments developments-· measure of the productivity —. would do11ur gap of 0f abo:rc ~69 billion. i;i1lio11. Our second quarter imply a current dollar about $69 potential measure, on the same basis, implies an even smaller gap of about $57 billion. One ~ayquestion whether the $200 billion measure tOk!'n seriously since the t!le CEA CEi1 apparently does not. not, However, should be taken s cudy for the tne Joint ,Joint Economic [conoP1ic Comittee Co1 irn ttr.e suggests sugoes ts the gap is a recent study even larger than the old CEA measure implies. ‘1W ~onetheless, 1a range Nonetheless, of !57 ""mn $165 billion biliion in alternative a1tr!rnntive measures of the $57 to $200 or even 83 current dollar GflP GIIP gap does not seem very close and the difference in gaps of 3 percent of potential output versus 10 percent is staggering in iitself tse 1f as we 11 as for what it might suggest to activist well policy-makers. policy—makers. Table 22 also indicates that in recent years our measure of potential output has been growing more mare slowly, slowly. This must be the case when 1974 is in the interval over which the growth rate is computed since our measure includes the potential output loss due to the energy not, price change while others do not. But, even for 1975 to 1976 our growth rate of 3 percent is markedly below belov1 the new nevi CEA's CEA’s 3.5 percent, the old CEA's Perry's 3.9 percent. CEA’s 3.75 percent or Perry’s important for While an assessment of the size of the GNP gap is important understanding the recent performance perforr.1ance of the economy, a measure of the prospective growth rate of potential output, while more difficult to pin down, is equally important for policy—making policy-making purposes. The Growth Rate of Potential Output: 1975 - 80 — ti a 1 output from 1975 to 1977 has been The growth rate of paten potential below both the CEA and Perry estimates. estinates. It is easy to understand why this is the case for Perry’s Perry's estimate since it follows so closely the old method of estimating estimating potential growth which concentrated on potential growth in labor employment and trends in labor productivity. As Otto Eckstein has noted, this method may have yielded plausible results in the past, but tao too many studies show that its results are implausible in the seventies seventies because it does not examine the changing factors determining labor productivity. 84 W 12/ The rise in energy costs costs can explain the slow growth of potential output over the last two years, years. Hhile While the increase in the relative price of energy has been less dramatic over the last two years, from the fourth quarter of 1974 through the first quarter of this year the relative price has increased 10 percent. Our earlier 1"1ork, work, it will be recalled, concerned the effect of a 35 percent increase in the prior year. year. The energy and energy price coefficient esti"iates estimates in the production function above indicate that, spread over two years, a 10 percent increase in the relative price of energy reduces the growth rate of private business sector potential output by half a percent: the difference between Clark's Clark’s estimate of potential output growth (3.5%) and our measure of the gro1"1th growth of potential (3.0%). Moreover, economic economic theory suggests a short-term short-term effect on the future gro11th growth rate of potential output due to the large 1974 increase in the relative price of energy. In particular, a rise in the relative price of energy depresses the demand for existing supplies of capital capita 1 resources and, if new capital is energy intensive relative to the remainder of the economy, raises the relative supply price of those goods, Thus, investment investr,rent falls below belol'I v1hat othervtise would have have been what it otherwise for some period until a desired capital output ratio is restored. Such "sluggish" grol'lth has been observed over the last two years and “sluggish” capital growth has had a retarding effect on the 9r0v1th growth rate of potential. The effect of a higher relative price of energy on the growth of potential resources is more fully discussed in the next section. Then we turn to the outlook for potential output grov1th growth in the remainder of the decade. Since Since this outlook depends on prospective energy price 85 well, developments, some sor,ie attention is devoted to this issue as well. non-observable economic variable such as Speculation on how fast aa non—observable po ten ti a1 GNP will wi 11 grow should be considered "second order potential “second order rretaphysics", practitioners of the subject. metaphysics”, with apologies to practitioners Nonetheless, it is useful for understanding the near-term growth, employrrent possibilities of the economy econofl\Y and for policy employment and inflation possibilities formulation purposes to examine the question. The Implications implications of the 1974 Capacity LossflGro!hRateofPot!QtialOa~ut Los 2 for the Growth Rate of Potential Output Our analysis of the 1974 rise in the relative price of energy shows that the productivity of existing labor and capital resources fell. fell. The production function estimates bear out the direction and magnitude of the productivity loss. We did not explore the impact of the loss in potential output on the future rate of growth of potential output. However, the analysis which yields the loss in potential output also suggests aa decline in the rate of growth of potential for some period in the future. In particular, the supply of plant, equipment and labor resources can be affected due to a rise in the cost of energy. With Hi th given supplies of potential capital capita 1 and 1labor, abor, the demand for each falls when the relative price price of energy rises, rises. These shifts in the demand for resources measured in terms of decreases in their rental prices may lower the growth of labor and capital resources and reduce the future future potential output rate. ~ The conventional analysis of the labor supply decision suggests that there are two major impacts of the energy price increase. 86 The shift in demand for labor services tends to reduce real wages. Such aa reduction in real 1;ages wages induces both a substitution effect and an full—employment supply of labor income effect which v1hich tend to reduce the full-employment soi;1ev1hat. somewhat. '?:!JI The second impact i111pact arises from a change in the value of non-human wealth. non—human The price level impact i~1pact of the higher relative price of energy reduces the real rea 1 va 1 ue of net monetary rionetary wealth. Hea 1th. value l\t At the same tirne, time, the lower productivity of existing capital assets reduces the present value of those assets. Given that leisure is a normal good, suc,1 a11 reduction in real 1·1eilltil tile full—employment full-er,1ployment such wealth tends to increase the supply of hours of all persons. possible,!a priori, to sign It is not possible, the effect of the energy price rri ce increase; the wealth ,1ea 1th and c1.nd labor 1abor iincome ncorne effects tend to increase the labor supply while a substitution effect tends to reduce it. We know of no evi de nee that there is a net effect evidence on the fu11-emr1oyment full-er.iployment supply of labor in either direction. W ZJJ In cari ta 1 resources, tile result is much clearer. In the case of capital the resUlt carital due to a left11ard The reduction in the rental price of existing capital leftward demand for the stock or flow of services of capital also shift in the dernand shifts the demand for ne11 new capital goods. Other thinrJS Other things equal, investment steady—state stock of capitai,~Y investr.,ent tends to fall, as does the steady-state capital.W Oti1er things are, of course, not equal. Other In particular, the replacement replacerr.ent cost of capital, or the supply price of flew ncM capital goods, may be expected to change as v1el1. well, If capital goods are more energy intensive than aggregate output, as one vJOul re 1ati ve supply price wouldd expect, the relative 93/ of capital goods 0oods v1ould rise. 231 would tend to rise, ~— Since Since both the demand and capitill investrient falls a forti fortiori supply of nevi new capi tal goods tend to fall, investment on capi tal stock, as does the long—run lona-run equilibrium carital stock. 87 ce AA simple model relating the rental price of capital, the price of new capital goods, the stock of capital, and and the investment Investnent rate 24 / Figure illustrates these points. &5.~/ Fiaure 22 illustrates the steady state relationship between the variables. In Quadrant I the flow demand for the services of capital, Q, is Is shown as a function of its rental price, R, and parameter a. The services of capital are assumed to be proportional to the stock of capital, K. In Quadrant II, Il, the price of a unit of the stock of capital is related to the rental price as a discounted perpetual gross Income income stream where r is the real rate of interest and w w is the depreciation rate. Quadrant quadrant III shows the supply Interest of new capital goods, I, in terms of the price of a unit of capital and s. Finally, Quadrant IV shows the steady state the shift parameter, ~. relationship between gross investment, I, and the stock of capital, K, which is proportional to Q. 0 0 , At Q I0 , P R0 and the Implied implied K’ K0 Q°,i°, pa,, and R° is in an initial steady state equilibrium. the economy Is An increase in the relative price of energy shifts the demand for the services of capital downward and to the left. Given the K and services existing capital stock K’, 0 , Q , the (~°, 0 ne~, capital goods. does the demand price of new replacement replacement so the capital stock declines. rental price falls as is less than Investment Is The new steady state ution occurs at a lower rental price, price of new capital, and with th solution a smaller capital stock and flow of capital services. An increase in the supply price of new capital shifts the supply curve In in Quadrant III upward and to the left. The process of returning to the steady state through temporary negative net investment is easily is a higher steady state price of traced through through the graph. The result Is 88 FIGURE 2 2 ., R R I R u.S It I I I I 'o ~pO ,p F’ DI q ID fl I 8~ capital, a lower lo,ier stock and rate of replacevaent replacer.1ent of capital and a higher higner rental price. Finaliy, combining both shifts yields a smaller capital Finally, contining stock in -- achieved through a temporary declIne decln1e in In the steady state —— tile net investment fovestrnent rate. rate, the \.Jhether the steady state price of capital and and Uhether rental price µrice of its services is higher depends on the dominance of the init·ial reduction in the supply of capital goods over the reduction in initial der:ar.d for the services of capita 1. 25 / demand capital. F1Dure sl10,15 the ti1e GUP G,lP price deflator for plant and FIgure 3 shows and equipment rel0tive to the PBS Pii5 deflator. cteflator. Table 3 shows the slow slm1 rate of growth of relative capital in the last two years cor1pared investment since compared to the rate of lnvestment t:,is measure of the 1949, and to subperiods since then. The rise in this wen as recent Investment investment relative pr1ce price of capital goods in 1974, as well behavio,·, are consistent with this theoretical analysIs analysis and the behavior, 26 increased replacement cost hypothesis. hypothesIs. / grm1th of capital in the recent past ;:,ast is is summary, the tile sluggish growth In suninary, consistent with the lower productivity of existing capital resources as 11el welll as the increased relative price of capital goods, both of which arc consequent to the ttie large Increase.’in increase in the relative price of energy. are S,.c:1 a sfo,:~ng Such slowing in capital growth is merely transitional so that, in a srn1; in~ economy econo,;1y with ,1i th the absence of further resource supply shocks, growing 9ro,'ltn of ;:;f capita its normal norma ·1 growth capital1 resources eventually approaches Its ni ·i ;.1 t :On shh to the growth grm-1th of 1 abor resources and anu potential potenti a 1 output. relationship labor Ti112 energy 12ner-gy price change not only reduces the potential poter:tial output yielded The by a particular ;;articu1J.r rate of use of services of labor and capital, it may by also leave ieai,,: the economy econociy with v1ith fewer capital carital services than ,vould have nave than would othen--d se been the tho case, after son1e adjustment,, otherwise some period of adjustment. 90 FIGURE 33 FIGURE Relative Price of Capitcil Capital Good? Goods* 1947=100 120 •~0 1947~100 1947=100 Quarterly Doto ----,120 120 118 118 116 116 114 114 112 112 110 110 110 108 · 108 106 106 106 -S 104 102 100 98 --'--'--'--'----'--"---'-.J.--'---'---"'"""-.J.--'--..__. 98 98 ss st s1 ss 59 60 61 62 63 64 65 ~6 i1 is M 10 n 12 Mn 1, 111s '--'---'-..J-JJ -'---'--'- 1947 ~e 49 so s1 s2 53 s~ n n Sources: U.S. Deportment of Commerce and U.S. Department of labor Shaded oreos represent periods of business recessions. Note, First quarter 1947=100 *Ratio of pri-.:c indo -::,f ncn residential fi~ed iflvestment to price in.:lo: ·of v;;tp.;: :11 th0 p;:C, ::.:.:: ~usin,ess sector. Latest doto plotted: 2nd quarter Prepared Bank of at St. Louis Prep a red by Federal federal Reserve Bonk TACLE 33 TABLE The Rate of Grov1th Equipnent Growth of Plant and Equipment (Constant Dollar 1!8t Capital Stock) (Constant Dollar Net Capital Stock) Period I/1950 1/1950 I/1955 1/1955 I/1960 1/1960 I/1965 1/1965 I/1970 1/1970 (Growth Rate) I/1955 1/1955 I/1960 1/1960 I/1965 1/1965 -— 1/1970 I/1970 - 1/1975 I/1975 — 1i.2% 4.2% 3,6 3.6 3.3 5.5 3.7 -— — — I/1975 —- 1/1977 I/1977 1/1975 1. 8 1.8 I/1950 —- 1/1977 I/1977 1/1950 3,9 3.9 92 The Prospects Prospr"cts for the Relative r:el~tive Cost of Energy [ner9y Ener0y huve been heen heavily influenced by Federal Energy prices have :1ror;ra;1 announced in August, Auoust, 197i. 1971. regulations since the price control progran This has been especially true for 11etroleur:1 petroktsn markets since 1973 1·1hen when rui sed the 11orl prier> of crude oil oi 1 above the protected protected OPEC actions raised worldd price market price in the United States. '.>tates. In order to insulate the U.S. econo111y fro:1 ver/ large larse increases increu,;es In in the world V1orld price of oil, economy from the very regul dor:estic crude oil owners mmers from regulations were putt in place to prevent domestic receivinJ "vriHlfol1 ant: to avoid uvoid the tho recessionary impact ir,1pact of receiving “windfall profits” 1Jrofits" and ') 7 / • ncrct1se d petro l cuL1 prices. 1increased rr1• ces ~ -t. petroleum 1 / Tl10 oxistinq regulation ro~ulation is the tl1e crude oil oil The centerpioco centerpiece of existing entit·iemcmts pro,irar.:, a method nethod for allocating allocatin9 controlled domestic crude entitlements prograra, oi arnmq competing competi n~ refiners. re nors. oil1 among The essence ossonce of the program is to provide an "equal" c1ail7 on controlled oil to all refiners based on their total “equal” claim inp11ts. The T,10 effect c ct of the progran pro0r<1r1 is to equalize equa 1 i ze the price of oil inputs. ,1hici1 is a 11eighted avera~e of the crude oil to all refiners at a level which weighted average controlied price and the world v1orld price, set by OPEC, DPEC, 1·1here ,1ei~hts controlled where the weights are based on the basod on thr, share of iuports iqiorts and doriestic oil in total oil inputs. inputs. Tiius, the entitlement en tlen;ent progran progrilu provides a means noans for holding holdin9 the domestic Thus, price of energy ~rice r:nol"'.7Y below bolm·1 the world price and ilnd a means mr:,ms for distributing the rents which i'Jhi ci1 would ,-:nu1 d otherwise otherv,i se accrue accru2 to domestic crude producers, producers. “rents” 1 l 11 1 [xistin<J regulation rcqulation hRs 1mrkr1J to hold the price of crude oil to Existing has worked domestic buyers L,e]m1 the tire world 1•1orld price, but to allow al1rn·1 it to gradually f)raJually below rise toward the world price. ir;1p1icitly implicitly subsidized oil owners. m·mers~ riroc2ss, of course, imports ir,1ports are In the process, with the rents expropriated cxpropri ated from frou domestic crude Not only are cheaper than they would dot arc refined rcf·ined products dieurer vrnuld be he 93 TAGLE 4 TABLE The Composite and Irirorted Imported Refiner Acquisition Cost of Crude Oil Oi 1 Import Imrort Cost (Dollars/Barrel) (Donars/flarrel) 1974 1974 I II II III IV IV 1975 1975 II . II III IV 1976 I II IIII II IV 11.59 11. 59 12.93 12. 65 12,65 12. 6:1 12,60 Composite Comrosite Cost (Dollar/Barrel) (Do11ar/8Rrre1) 8,24 fl.24 9,34 9.34 9.20 9,20 9.30 13.03 13.03 13.56 13. 56 14. 11 14.11 14.84 14,84 10.96 10,96 13.35 13.43 13,43 10,58 1o. 58 10. 72 10.72 9,83 9,98 9.98 10. 72 10,72 110,94 o. 94 11,26 11 .26 13,52 13. 52 13,59 13. 59 Percent Difference Difference 34,1% 34. 1% 32.5 31.8 31,8 30.4 28,2 28.2 30.7 27.5 30.3 30,3 23,3 23.3 22.5 21,2 21.2 18,8 18.8 Uational Energy Infomation Information Center Source: Based on data from the ilational ifonthly Energy ~~thiv Ener Reviev1 Review (April, 1977), p. 73. 94 in reoulations because of the incentive to produce In the absence of the regulations competinq energy sources tend to have lower prices as well, more, but competing due to their smaller demand. de11iand. Table 4 shows the “Refiner "Refiner Acquisition Cost” Cost" of Imported imported oil and the composite cost which 11h i is the weighted average of the price of frorn 1974 through 1976. domestic and imported oil from in These costs are, are, In ces, but they reflect the impact imp act of the control contro 1 effect, delivered prices, v1ilich the world program. The last column shows the percentage by which acquisition cost of imports, exceeds the price, measured by the refiner aèquisition composite, or domestic price. ce. Over time, this excess has fallen due to both the actions of the Federal Energy Administration Adr,1inistration (FEA) (FE/\) In in Its its attenpt to gradually remove rermve the difference, and due to the increasing attempt share of high cost iniports ir1ports in total crude usage and the falling share and rate of production of “cheap” "cheap" domestic donesti c crude oil due to the dm,1estic price controls. doraestic Ex·isting ne1·1 energy policy Existing regulations and major proposals for a new 11 en s ion the “rationalization” ionalization" of the domestic donestic petroleum petroleurn market so that envision dor,estic prices and enorjy usage are based upon social socia1 costs or reflect domestic enet~yusage j ,\ • This Is . economic imposed). econor:lic scarcity (even if artificially artifi a l '1y 1r1pose, is evident, for example, examplu, in the Energy Ener0y Conservation and Production Act of 1915 1975 (ECPA) al1ov1s controlled domestic prices to rise over tine and which allows termi nutes do;estic dor,1es tic price control in 1979. 1979. The crude oil tax of the tl1e terminates Administration's cne icy proposal as well 1-1el 1 as crude oil decontrol Administration’s energy policy sililre this titis desired proposals share desired result, Thus, it is very likely that the dispa ty in retroleur:1 prices, shown shmm In in Table IV, of about 20 percent disparity In petroleum 1930. at the end of 1976 1ii1l will be eliminated before the end of 1980. 95 AA 20 percent rise in the cost of crude oil to domestic refiners will not only raise the price of refined products, but also, through substitution effects on energy users and direct and Indirect indirect cost effects an competing energy producers, raise the price of other energy resources. To assess the impact of raising domestic crude oil costs to the 1/1orld world price on the relative price of energy resources and and on potential output, we have examined the relationship between the relative price of crude oil and of energy prior to the deluge of controls on primary, intermediate, and retail markets which began In in August, 1911. 1971 • The relative price of energy is that used In in our aggregate production analysis and the wholesale price of crude oil Is is used as a measure of the domestic cost cost of crude oil prior to August, 1971. The relative cost of crude oil is found by deflating by the implicit price deflator for the private business sector. The simple linear regression of first differences in the logs of the relative price of energy (P) on the relative price of (P) from 11/1948 Il/1948 through of domestic crude oil D’c) C II/1971 11/1971 is: (6) 2l+.4 3S4AlnPc c~lnP=e.OO ln P = - .0021 + .4354 c ln Pc (—1.68) (-1.68) (5.18) R2s.23 D.\J. = 1. 77 D.W.=L77 S.E. —= .010 ,010 This simple regression may be used to obtain information on the increase in the relative price of energy resources occasioned by the expected rise In in domestic dor;1estic crude prices and the average avera9e cost of crude in oil to domestic consumers. 281 Given the 18.8 percent disparity In 96 prices at the end of 1976, the equation Indicates indicates an 8.2 percent rise in the relative price of energy as the disparity is removed sometime over the next three years. The 8.2 percent rise in the relative price ~rice of energy in the United States assumes no change d1ange In in the relative price of energy or crude oil in the world market. To the extent that our imports of is some possibility that the energy resources would 1✓ould be reduced, there is 11ea1th "dominant firm”, firm", the OPEC producers, wealth maximizing price of the “dominant might change. The fundamental question is the effect of such a U.S. elasticity of world v10rld demand for OPEC oil. policy change on the elasticity AA simple reduction in demand is not likely to lower the relative price of OPEC oiL oil. The relative price of oil and other energy resources would tend 1•mrld market only If if demand del1and became more elastic and to decline in In the world this would not necessarily occur simply because of a reduction in U.S. imports. An increased responsiveness of domestic supplies to the world tile elasticity of demand for OPEC oil and price would tend to reduce the other energy Imports and, thus, tend to reduce the cartelized ;vorld world other irnports and, oi L However, only decontrol decontrol of domestic energy markets would price of oil. ensure such responsiveness of domestic suppliers and such a policy does does not appear likely over the next three years. Taxing existing supplies to iriplies little or no to raise their cost to the world level implies responsiveness of domestic supplies to world ,1orld prices and, to the extent such actua1ly reduces that responsiveness, provides a suth a policy change actually case for an even higher higher domestic and world 1-1orld price of energy resources. Thus, an 8.2 percent rise in the relative price of energy resources sometime over the next three years yenrs appears arrears to be a reasonable prospect. 97 ~ntial0utut The Future Gro11th Rate of Potential Output In order to find the growth rate of potential output, assumptions concerning the gro11th growth of potential resources are necessary. He briefly describe the assumptions assur,1ptions which 11e We we use below, In each case 11e nost optimistic ai;mng we have tried to choose the most among alternative assumptions. assumptions~ The ilon-Private-Business Non-Pri vate-Business Sector and Employment Em yment Growth -- Hhile While - - output outside the private business sector {PBS) (PBS) has grown with time, it is not significantly affected by hy employment. employrnent. AA quarterly regression of non-PBS 1oy1"ent, the unemployment uner.1p 1oyment rate of the civilian non—PBS output on emp employment, II/1977 indicates that labor force, force, and time for the period II/1948 11/1948 - 11/1977 — only the time trend of 3.24 percent per year is significant (t=12,2). (t=12.2). ,25. The t-statistics t—statistics of the insignificant variables are less than .25, equation is adjusted for autore~ression autoregression and has an R R22 of .997 and rercent. standard error of 1.6 percent. The <Jrov1th rate of non—PBS non-PBS output The growth during the past tvm ii.c1s also been 3.24 percent, uhile vias two years has while it was lower (2,7%) (2.7%) in the prior five years (11/1970 (II/1970 - 1/1975), I/1975). — Thus, it appears arpears reasonable or perhaps rerhilrs slightly optimistic optfoi sti c to assume the trend rate will continue. Employrnmt important because it Employment growth in the non-PCS non—PBS sector is important grov1th of PBS potential e1:1ploy1.1ent limits the growth employment and output. Er,1ployment Employment growth in the private rrivate business sector contributes more to total output er.,ploy,.1ent in the non—PBS non-P1lS sector. than an equivalent increase in employment In the post—war post-war period, non-PBS en;p 1oy1:1ent has grown 0rown more r,ore rapidly than non—PBS employment potentiill or actual PBS PCS employnent. potential employment. tionethelcss, riaintain an Nonetheless, to maintain growth rate of potential optimistic bias in the 9ro11th potential output, it it is assumed 98 t~at the future growth groc,c:h rate of potential enployr:1ent that employment in In the private DJ5iness sector s,,ctor is the ti1e same sarae as the growth rate ,·ate of potential potentia1 employnent.~ employment. 292/ f business 1 us,i·no Clark’s Clurk s estimates csti01ates of Potential employment growth is found using civi-!ian labor force and full—employment ful1-emp1oyr:;ent unemployment unemp1oyt1ent the potential civilian 9rm1th of pi,tccnti a 1 employment er•p i:,yment is 1.65 percent per rate. The rate of growth of potential yeLJ.r while tJhi le hours growth 9ro~-1ti'! is "is 1.25 1 ~25 percent per year In in the private year business sector from 11/1977 through 1980. Hat1;; of the Net i{et Stock of Plant Pl ant and Equipment The Growth Rate -- The —— .,·:.st difficult Groblera •n assessing future potential output growth Is 1s most problem in n,:,ng the growth rate o• ti1e capital stock. finding of the As Table 3 indicates, ti:e growth of capital has nas been Deen relatively slow in recent years, but the sin over fow five year yer.r intervals i over a few In the past, has been at relatively ;'ap1d rates. rates, rapid detemine whether 1·1i1ether the Since it is difficult to determine trac1sitional adjustment adjust;;ient of the tt~,e capital stock to a higher relative price transitional of energy is complete, complete: 11 and also since future increases are likely whIch which rnay not may not have been anticipated by investors, a continued 1ow low rate of investment should be allowed allm,,,ed for as aa possible outcome. outcome" To do this, this, we include 2a low 10,1 esti1,1ate 9rov1th based on capital Include estimate of potential output growth gcov:th of 11 .8 ~3 percent per year, the rate of increase of the past two growth _years~ years. optir,1istic measure of capital growth may be found from AA r10re more optimistic 9rov1th to potential output growth grovJth before the relationship of capital growth 1973. (a1.nuu1 rate) of capital exceeds that of The mean rate of growth (annual Il/194S potential output for the period 11/1948 ~ IV/l973hy L'/1'-173 by .55 percent. percent, — rna.y be used to estimate estir:1utc a rate of capital growth .55 "55 relationship may uercent faster than the growth orm,, th rate of potential output. output., percent 1 99 This Gro,1th —— -- Trend growth Is is allowed to remain at Energy and Trend Growth tile private business sector. The impact of 1.6 percent per year in the energy developments may be seen more clearly by measuring the growth rate of potential output assuming no change In in the relative price of betvieen now and the end of 1980. Coithlnlng Combining the asstanptions assumptions energy between above concerning potential resource employment with the production function (4) yIelds yields an annual rate of growth of potential output of 3.8 functIon percent. if Investment investment continues to yield the low rates of increase In in If pl ant and equi pr,1ent of the last 1ast two years, the growth rate of potential potenti a1 plant equipment would be reduced to 3.2 percent. The higher rate Is is predicated upon a much larger larger rate of growth grov1th of the capital stock (4.4 percent per year). such Only one of the five year intervals shown shm·m in Table 3 shows growth of the capital stock of 4.~, reriod 1965 to 4.4 percent per year or above, the period 1970. However, capital stock growth arowth has attained this rate during durina perfon1ance, such as the mld—1950s riid-1950s and durina other periods of peak performance, and during 1973. The assur.1e no The potential output growth rates of 3.2 —- 3.8 percent assume in the relative price of energy. eneruy. change In Accountin<J for an increase In in Accounting tfoc over the next the relative price of energy of 3.2 8.2 percent some time grov1th from frrn,i the present to three years noticeably reduces the rate of growth 1930. The the end of 1980. The additional energy price change will very likely ter,1porary shock with v1ith i;1uch its effect occurring over a tend to be a teMporary much of Its short period of tir.1e. time. nonetheless, lionetheless, since the tinlng ti11ing of the change is inpact on currently unknmm, unknown, the best that can be done is to show its Impact the gro1-1th in the tl1e growth rate over the longer period. Such an increase In relative price of energy eneray reduces the naxiinum uaxi1,1urn expected growth grm-1th rate of 100 3. 3.88 percent to 3. 3.55 percent. The i1,1p 1i cit rate of growth grov1th of the capita inplicit capital1 stock to achieve this result is 4,1 4.1 percent, essentially the mean annual rate of growth of capital fror.1 fron II/191\8 11/1948 through 1973. If capital grows at the rate of the last two years, 1.8 percent, the rate of growth of potential potentiul output will he be only 3.0 3,9 percent per year, the rate achieved so far since 1974. 1974, The results frame the alternatives quite well. Perry's Perry’s estimated grm1th growth rate of potential output for the next few years is rougilly equal to our highest estimate. roughly Cut that estimate estimate requires But unusually rapid capital accur:iulation, accumulation, consistent v1ith with our estir.1ates estimates of hut probably not his -- and, more importantly it the recent gap -- but — —— changes and their effect. ignores the prospects of further energy cost chanaes Accounting for energy price developrients developments and assuming capital growth to remain the same as the last two years results in a grov,th growth rate of potential output 1,1hich which is the sane as that we have observed for the last two years, 3,0 percent. Finally, allowing allol'ling for energy price developments and a more historically normal nomal pace of capital growth of estimate equal to Clark's 4.1 percent under peak conditions yields an estimate Clark’s of 3.5 percent per year. year. Ue regard a 3.5 ,Je re9ard 3. 5 percent growth rate of output to be a reasonably optimistic estimate potential outrut estimate of the potential growth rate when the recent response of investment to energy cost changes is considered. Since the current G:IP GNP gap is quite small srnll compared to alternative esti1nates, and since our investigation of the growth rate of potential estimates, suggests it will grow at a niaxir:ium about 3,5 3.5 percent, we vie conclude maximum of about that the econorny vlill achieve full—employment fu1l-ernploy1,1ent and peak operating economy will 101 performance perfonnance within a year if the actual growth rate of real output since 1974 continues. Unlike other studies of potential output, output. we conclude that more stimulative monetary or fiscal policies policies are neither necessary or desirable. Canel us ion Concl Our n~asures measures show the econor.iy rercent of economy to be producing over 97 percent 1977. its potential at mid-year mid~year1977. In addition, our measures show potential output to have grown grmm at about aa three percent rate during the recovery, recovery. The rate of growth of potential GNP for the remainder of the most, decade is about 3.5 percent, at most. contrast to the mainstream view. These findings stand in stark Several recent studies have shown the be seriously flawed, basis of this conventional view to he Nonetheless, most observers are reluctant to alter their views viel'ls on U.S. economic perfonnance or the potential output growth grov1th rate after 1973, apparently apparently performance due to the power of historical extrapolation, extrapolation. Our conclusions follow from a theoretical analysis of the role of energy resources and the relative re 1ati ve price of energy in the production process of the U.S. tJ,S, economy, economy. The empirical analysis of the er,1ploy1,1ent supports relationship of aggregate production to resource employment the theoretical conclusions. conclusions, The anaiysis analysis provides empirical estimates production function parameters which allow allo\'/ the quantification of of production effects of changes in the supply of potential resources on output possibilities. These estimates go well v1e1l beyond the tile specificity allowed by other studies, which either fail to take resources such as capital or energy into account, or 1·1hich which fall back on standard assumptions 102 about some of the relevant coefficients. Our potential output series reflects our earlier conclusions and those added here. In particular, the large increase in the relative price of energy led to a change in the pattern of resource use which constitutes efficient production, changing the demand for aall 11 resources, but, most importantly, permanently reducing the productivity abor and capita cost of of existing 1labor capital1 resources. Increases in the cost energy over the last two yenrs, years, and further increases yet to come during the remainder of the decade, continue the negative neg11tive energy cost effect on potential output but to a lesser extent, extent. The direct productivity effect of the higher cost of energy is compounded by an indirect effect temporarily reducing the rate of capital accumulation. accur,1ulation, The reduced incentive to invest was shovm shown to be due to both the reduced productivity of the services of existing capital and its increased replacement cost, cost. These conclusions are supported by the unusually sluggish growth of capital since 1974. ,fo stinulative denand r1anagement policies are We have argued that stimulative demand management both unnecessary and inflationary, inf111tionary, and that at potential output the federal budget shows a very large deficit. The economy will will very likely achieve its potential output rate within a year with only moderate grm1th. growth. Carlson 1977, Review) has verified that, that, Garison (August 1977, rather than aa high employment bnlanced be the balanced budget in 1977, as would he 1d CEA measures neasures were correct, the high—employment hi gh-erir 1oyrnent deficit case if if the oold is currently about $20 billion. Thus, within a year it will 11ill become 3iear it virtually impossible to postpone critical fiscal decisions concerning permanent financing financinCJ of the existina the means of permanent existing and/or desired role of the federal government in a markedly changed American economy. 103 Footnotes 1J See Business Heek, ]J Week, y2/ Ibid., p. 77. '}I ~f See Roger Brinner, (June 9, 1973), pp. 76 - 77. — "Potential ‘Potential Grov1th Growth to 1980," 1980,” Otto Eckstein et. al., Economic Issues and Parameters of the flext Next 4 Years, Lexington, nassachusetts: Data Resources, Inc., Inc., Economic Study t~’xington, NassacThii~’E~:Data Series, 1977, pp, 17. pp. 99 - 17. — especi a 11y Denis S. s. Karnosky, "The 4/ See especially “The Link Between Money and Prices - 1971—76,” 1971-76," Federal Reserve Bank of St. Louis, Review June 1976, pp. 17 - 23. — — §j ~ References to our May and June papers throughout are: Robert H. Rasche and John A. /\, Tatom, "The Regir.ie “The Effects of the flew New Energy Regime on Economic Capacity, Production, and Prices," Prices,” and "Energy “Energy Resources and Potenti a 1 GNP,” GllP," Feder a 1 Reserve Bank of St. Louis Potential Federal Review ilay [lay and June, 1977, pp. 2 - 12, and pp. 10 - 23, respectively. respecti vely. — §/ y — Brookings Papers on Economic Activity, 1:1977, :1977, pp. 11 —47. - 47. Brg9ki! ?.~zem,~!LEconomis.Activit, ZJ This 2/ This contrasts contrasts §/ The derivation ~/ with his his opinion cited above. above. with opinion in in 1973 1973 as as cited of this specification is indicated in our June Review paper. Note in particular that it assumes that the aggregate demand for energy is on a de1nand demand curve 1-1ith with unitary elasticity with respect to both output and relative price. This is aa relatively comon common assumption assu~1ption when working with time series data generated over annual intervals. intervals, On the other hand, this condition is less likely to be satisfied over shorter time quarter. Under such circumstances, it is intervals such as a quarter, more conrion corrmon to specify partial adjustment models 1·1hich which have smaller impact elasticities. For a discussion of the biases in our estimates of the output elasticities which which result from impact elasticities which ,1hich are smaller than unity, see Appendix II of our June paper. In addition, partial adjustment mechanisms for factor demands, such as that specified by 11. it I. I. Nadiri and S. Rosen, “Interrelated ''Interrelated Factor Demand Functions,” Functions,'' 457-71, would American Economic Review, September 1969, pp. 457—71, suggest that the above equation may he be misspecified by the omission of lagged values of all factors. It is not clear that such a source of potential specification error error would systematically bias our regression coefficients in one direction. y9/ See John C. i·lusgrave, "Fixed Nonresidential Business and Musgrave, “Fixed Residential Capital in the United States, 1925-75," 1925—75,” ~Survey of Current Business April 1976, pp. 46 - 52. — 104 10/ The minor lQ/ minor differences from the results in Appendix III of our June article arise due to BLS data revisions and revisions in the GIIP GNP accounts. jjJ 111 The constrained estimate yields a measure rr.easure of the output elasticity elasticity of labor equal to 64.3 percent and quarterly trend growth rate of .41 ,41 percent. The Durbin-Hatson Durbin—Watson statistic for the equation is 1.91, the estimate of rho is .78, ,78, and the standard error of the regression is ,0077. .0077, lY j~/The standard error error of the equation below is identical to that using the unemployment rate to four decimal places. places, 13/ The error pattern without any adjustment indicated a smooth phasing in and out of the shift shift over a four quarter period, thus, the dummy variable was allowed to increase from zero to one in steps of one—fourth one-fourth and conversely to decrease at the end of the period in the same v1ay. way, Of course, this phasing in and out led to a reduction in the standard error of the hours per worker equation. with such an hours per 14/ The weakness associated ,lith especially with the adjustment adjustment for the unusual the 1960s, a 1so been noted by Perry (1977, ( 1977, l960s, has also a similar equation for hours per worker in the sector. sector, worker equation, developments in p. 31). 31), He used nonfarm business A Clark's method for deriving the full-employment A description of Clark’s full—employment unemployment rate and the potential labor force may be found in GNP," Council Council of Peter K. Clark, "A “A Nevi New Estimate of Potential GNP,” Economic Advisers, 1977; processed. Output," 16/ See Ronald J. Talley, Tailey, "Some “Some flew New Estimates of Potential Output,” ]!I forthcoming in American Statistical Association, 1977 Proceedings of the Business and Economic Statistics Statistics Section, · ia~ ]lJ See \·Jalter U. Heller, "Productivity fl/ Walter W. ‘Productivity and GNP Potential," Potential ,“ Wall Street Journal, June 29, 1977. l1. Heien, 18/ See Albert J. Eckstein and Dale ii. Fleien, "Estimating “Estimating Potential .1§./ in a Model Framework,” Framework," Achieving Output for the U.S. Economy in Achievin the Goals of the Employment Anniversary Review, Em lo ment Act of 1946-Thirtieth 1946—Thirtieth Anniversar eview, U.S. Congress, Joint Economic Committee, 94th 9 t Cong., 2ncfsess., 2n sess,, 25. December 3, 1976, pp. 11 - 25, — Y1J 12/ See the the comment Otto Eckstein, Eeks tei n, ~Brookings Papers on Economic See comment by by Otto on Economic Activity,(I:1977, p. 53. ~jvit,(I:1977, 20/ G. Cain & eds., "Income & H. Watts, eds,, “Income Maintenance and Labor Supply," Supply,” (Nevi (New York: Academic Press, 1973). 105 21/ The model by Eckstein and Heien (1976) suggests a slight positive increased energy cost. cost. effect on the labor supply due to the increased However, it is not clear clear whether they estimate the relevant net effect or one of the components. W ggj/ Capital embodying different technologies is not differentiated here. Presumably some substitution toward less energy intensive processes would stimulate demand for certain kinds of capital goods while reducing that of other capital and the total demand. Also, the analysis follows the usual convention in assuming the • real rate of return demanded by lenders and equity owners and that used by investors in discounting income streams is unchanged. unchanged, 23/ This result is demonstrated in our 11ay article. lay (1977) Review article, 24/ The graphical analysis is adopted from fror1 Leonardo Auernheimer, "Rentals, “Rentals, Prices, Stocks and Flows: AA Simple 11odel," Model,” Southern Economic Journal, July 1976, pp. 956-59. 956—59. — g~/If price of new capital goods is independent of the long-run long—run supply price price and the output rate, the result is unambiguous as both the price rental price of capital goods are higher in the new steady-state solution. 26/ AA discussion of the unusual behavior of non-residential non—residential fixed investment in tile the recent past may he found in Jal—Hoon investnient Jai-Hoon Yang’s, Yang's, "A Recovery," Federal “A G1lide Guide to Capital Outlays in the Current Recovery,” Reserve Bank of St. Louis, Review, February 1977, pp. 22 - 7. — evaluation of recent energy regulation regulation in the United 27/ AA review and evaluation States may he found in Paul W. \/, flacAvoy, rlacAvoy, ed., Federal American Energy Administration Regulation, Washington, llashington, D.C.: Arierican Enterprise Institute for Public Policy Research, 1977. 1977. Enterprise 28/ In the levels form, the equation has a RR2 of ,93 .98 and standard error ,012. The price of crude crude oil coefficient, .45 ,45 (t == 5.65), is of .012. Is in agreement agreement with that reported above. The rho statistic has a value of .97. • 97. Thus, the first difference form is cited in the below. text and used be lov1. assumrtion is also made by PP.rry 29/ This assumption Perry (T11ble (Table 14, p. 45) and may unusually r11pid contribute to his unusually rapid rate of growth conclusion. 106 WILL BOTTLENECKS SLOW THE EXPANSION? William D. Nordhaus The topic of this conference is indeed an ii1portant important one, one. Although currently the United States economy suffers from considerable excess caca— pacity, both in labor and in product markets, we hope that this condition forever. will not last forever, What II would like to discuss this afternoon is marthe state of utilization and the extent of imbalance imbalance in different markets, and possible strategics strategies for avoiding bottlenecks during this rec ov cry. recovery. of the Recession The Legacy L sLfthef~s~ssiofl Starting Starting roughly five years ago the world economy was struck by a cu1m 1nated in the worst inflation, and thereafter series of shocks which culm~nated the worst recession, of the post—war post-war era, era. A A simultaneous boom in all the industrial countries led le,1 to severe capacity shortages in major ininindustries, dustries, especially materials materials industries. Following the 1973 boom, and the food and oil inflation that succeeded succeeded it, virtually every major inindustrial country suffered aa severe recession, recession. 1972—73, together The imbalances that developed in the boom of 1972-73, with the devastating effects of the oil and grain shocks were enough to downturn, cause a downturn. But the fiscal and monetary authorities added their influences, own restrictive influences. Thus, examining examining the OECD area: area: DFN~flEiusiTTmeThBeroTf1irPresident’sCouncil Dr. Nordhaus is a member of the President's Council of Economic Advisers. 107 o The narrowly defined money supply (M-1) (N-i) decelerated frOm an annual growth rate of over 12 percent at from the beginning of 1973 to below 77 percent at the end of 1974. oo OECD were rising at approxiapproxiSince prices in the OECO mately 13 percent annually, this means that the real money supply was falling f~]jjn at almost 6 percent per anannum by the end of 1974. oo short-term As a result of the monetary stringency, short—term rates rose from around 66 percent at the beginning of 1973 to 11 percent in the second half of 1974 1974 and long-term rates rose about a point and aa half, half. o afFiscal policy turned sharply toward contraction after the 1973 boom. Real government expenditures from 1973 to 1974 rose only 11 percent in the United States and United Kingdom, 44 percent in Germany, and fell in Japan, Japan. The effect of these forces is by now well known. One particularly disturbing legacy of the recession has been its effect on investment and thereby the level of capacity in the United States and abroad. abroad, Although measurement of capacity is quite difficult, the estimates show a signisignificant decline in the growth of capacity in manufacturing industries industries over the last ten years. From 1948 to 1968, 1968, growth of capacity in manumanu- facturing averaged 4.5 percent per year, year. rate dropped to 4 percent per year. year. From 1968 to 1973, the growth But in the period from 1973 1973 to 1976, 1976, capacity grew at only 33 percent per annum. annum, This means that over the last three years, the growth of capacity has not matched what we would would need to keep up with potential output -- even with the most pessimistic -- of the estimates of potential that we have heard today. capacity Examining the latest data, we see that the growth in capacity leaves much to be desired, as is shown in Table 1. 1, 108 108 1. TABLE 1. Rate of Growth in Capacity 1976:2 to 1977:2 Manufacturing Manufccturing 2.8% Primary processing 3.2% Advanced processing 2.5% Materials Basic Basic metal Textile Paper Chemical Energy • 1.4% 2.1% 3,0% 3.0% 4.8% 2.3% For all of manufacturing the growth rate of the last year has been less less than 3 percent, with basic metal materials showing the smallest increase and chemicals showing the most rapid. It is clear that the recent slowdown in the growth of capacity must be reversed. noninflationary expanexpanWe cannot hope to sustain a noninflationary withsion over the next three or four years, reaching high employment, without a major acceleration in the growth of capacity. The Current Imbalance in Labor and Product Markets Given the slow growth of capacity during the current recession, it is inevitable that an imbalance between labor markets and capital or product markets arises. clear what the nature of this imTo make more clear im- balance is, let us consider capacity output as it is distinguished from potential output: o ~Potential output is conventionally defined as the would be produced at a reference level of output that would unemployment rate, or weighted unemployment 109 rate, conrate. In computing potential potential output, it has been conventional (up until recently) to assume that there are no capacity constraints. This can be rationalized by assuming that, that, through the accelerator principle on investment, the level of capacity adjusts with a disdisdemand, If this is the tributed lag to the level of demand. case, and if labor is inelastically ultinelastically supplied, the ultimate constraint on output is labor input rather than capital. oo ~ Capacity output can be viewed as the level of output which can be produced with the current capital stock. Clearly, the definition of capacity differs across inindustries, both because of flexibility of productive techniques in some industries and the possibility of others, Nevertheless, expecially in shift work in others. continuous continuous processing industries, capacity has a defdefinite meaning. It is useful to compare the state of utilization of labor and product markets by looking at what will be called the full employment ~ capacity utilization. To calculate the full employment capacity utiliutili- zation index, we need to know the relation between capacity utilization and the level of unemployment. capacity,” Then, using an “Okun’s "0kun's law for capacity," we can estimate what the level of capacity utilization would be if the 1’ We have taken the "full unemployment rate were at “full "full employment. employment.”,,y “full employment" weighted-average unemployment rate used employment’ definition to be the weighted—average in the CEA potential output series, this corresponding approximately to a 55 percent rate today. We have investigated the relationship between these two series over the last twenty years, using a number of alternative techniques. Figure 11 shows the result of one of these experiments. According to this graph the period from the mid-1950s mid-l950s until the mid-l960s mid-1960s showed aa gradual upward creep in the full employment capacity utilization index. Then starting in 1965 and lasting until approximately 1970, there was a 110 capacity Utilization Utilization (Percent) (Percent) Capacity iL 92 90 / 88 — 86 — I /I I/ I/ I/ / /\ 1\ \ ‘I \ Full-Employment Full—Employment \ \ '--, I/ /I - I — /I ~ /I I/ -J ~ 82 80 80 ,, — — 76 I I — — \ I/ I /I I\ I ' I \\ \ / 78 78 I/ I I \ I/ \‘I I \I I I II I I/ I I I I I \ \ /I 84 '\ N II I Actual Actual —> I I I \N '' I "I N, I I \ \ \ I I I \ \ I I I I /I \ \ I1~ It I/ \I! I, I — \ I \I I 74 \ ~ 1' — I I I I 72 '--...l...-1---l..-.L.-...L.-1--...L.-.L--1--'---1--'----'--'---'--'----'--'---' I I I 1 .1 I 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1964 1966 1968 1972 1974 1976 Figure 1. Figure Actual Capacity Utilization Actual and Full—Employment Full-Employment Capacity Utilization (Federal Index) (Federal Reserve Reserve Index) dramatic decrease in this, reflecting the potency of the investment boom of the 1960s. Starting in 1970, however, there was a definite and sharp upward rise in the full employment capacity utilization index. Over the last six years full employment capacity utilization has risen from approximately 81 percent to the current level of 91 percent. percent. indication The full employment capacity utilization index is an indication of how tight product markets in manufacturing would be, today, if we were at full employment (and potential output as defined above). can be compared with historical experience. These The capacity utilization rates of the Federal Reserve Board, which we are using here, averaged about 83 percent for manufacturing for the period 1955 through 1975, and about 86 percent for industrial materials from 1967 through 1975. 1975. The highest level of the capacity utilization index for manufacturing perwhich has been experienced for an entire year since 1948 was 91 per1966. The most recent period of high utilization was in 1973. 1973. cent in 1966. In that year capacity utilization in manufacturing averaged 88 percent, while primary processing industries had aa utilization rate of 92 percent. What are the implications of this apparent rise in the full employment capacity capacity utilization utilization index? index? The obvious point point is is that that we employment The obvious we cannot expect to have high levels of employment without one of the three following possibilities: 1) a significant investment boom; 2) change in in the from manufac2) aa major major change the composition composition of of our our output output away away from manufacturing and materials; or 3) operating rates in manufacturing and materials which are well above those which are normally experienced. materials 112 Or put differently, it is clear that if the capacity and utiliutilization data of the Federal Reserve Board are relatively accurate, then zation,data there is currently insufficient capacity to sustain an inmediate imediate gallop to full employment. Returning to our earlier discussion, we noted that one of the most significant bequests of the recent recession was insufficient investment. Before the recession got underway, our index index was considerconsider- ably lower: thus in 1973 the full employment capacity utilization index was 88 percent, 33 percent lower than it stands today. The inin- vestment slump and associated problems of the last three years have apparently raised our full employment capacity utilization index three full points. The Break-even Capacity Utilization Why has the growth of capacity not kept up with the growth of potential output? This is the other side of the question, "why “why has badly during the current recession and recovery." investment lagged so badly recovery.” CEA has studied the reasons behind the investment lag, using aa number of economic theories of investment. Although different models give different different answers, the basic reason -- and one that can hardly be -— surprising -- is that the demand for future capacity is performing —— today’s capacity has been been poorly because the level of utilization of today's so low. There are other factors as well -- environmental regulations -- certainly have raised the cost of additional capacity in many heavy industries (steel, utilities, and chemicals being among the most 113 heavily affected). affected). In addition, the depressed state of the stock market hardly is conducive to new ventures, although -- to be sure --- -- stock market prices (and in particular the ratio of market value to Q) have reflected quite closely levels of utilizautilizareplacement costs, Q) tion of capacity capacity over recent years. Finally, there has been aa clear shift in the composition of investment away from long-lived investinvestmen ts ments -- especially structures -- and toward equipment. -- Nonwithstanding the caveats, however, it is probably the case that the major reason for the depressed state of investment is the low levels of capacity utilization the economy has experienced over the last three years. We know that very low levels of utilization -—- operating through the accelerator mechanism -- lead to aa slowdown in -- investment and in the growth of capacity. If capacity is below below some "break-even" “break—even” point, and investment therefore insufficient to keep capacity growing rapidly enough, we may actually be in the situation where capacity is growing gro~iing less rapidly than potential output. The full employment capacity utilization rate would therefore rise and the imimbalance between labor and product market would widen. It is ironic that in pursuing an anti-inflation policy which keeps the level of slack in the econoroy economy very high, we have created a situation in which future bottlenecks become more likely. As aa way way of of illustrating illustrating the the relation the growth of As relation between between the growth of capacity and capacity utilization, we have run aa standard investment investment to the rental equation. The equation relates the level of investment cost of capital and non-linearly to the level of capacity utilization. 114 Then by taking into account the historical relation between investinvest- ment and capacity growth, we can ask whether the level of capacity h·igh to assure that capacity growth utilization has been sufficiently high is as rapid as potential output. relation between Figure 22 shows a graphic representation of the relation difference, capacity utilization and capacity-potential difference. In making this relationship, we have assumed that the real cost of capital (in terms of percent per annum) was at its post—war post-war average, so that there tenns was no extraordinary push or pull from monetary or fiscal incentives. difference between the growth The figure figure shows quite clearly that the difference of measured capacity and potential output is positively and nonnonlinearly related to the appropriately lagged rate of capacity utilizautilization. One can calculate from such aa relationship, assuming no change in the historical level of capacity utilization at which the capital break-even stock has grown at the same rate as potential output, the break—even utilization rate. This presumes, as has been the case recently, that -- if imbalances are not to appear -- potential output will grow -- -- approximately 1.2 percent per annum faster than the capital stock. According relationships, the the break-even break-even capacity capacity utilutilAccording to to these these relationships, ization point is around 84 percent. That is to say, when capacity utilization is 84 percent, and assuming the relationship is the same as in the historical period, capacity output will be growing as rapidly as potential output. output, On the other hand, if utilization is lower than this -- say the 74 percent in 1975 or 80 percent in 1976 115 —— • 08 Difference Difference between growth of capacity capacity and growth of potential outoutput (per annum) ·~-..,., ~--- I I • 0 I .06 I I • • I .04 . 04 S l I .a .02 .02 -1 -a • -S S .a " ------=-•..,_,-------=+1~.,----''0 • S C S 0~ • S -.02 —.02 •a .04 -.04 St • S ~a S — • eec a a.,, •1 • a •• S * • • • a a $ .•. S S S C— 40 0 C a0 . . • a C .. I -- • 5 I. S .ISI .. S .4.. . S. a 0 ...•• •• 5, I 0~’ • • • Break—even Break-even capacity capacity utilization utilization a a0 ‘a. ••• • S _.0670 .75 85 .90 .95 .95 Average capacity capacity utilization Figure 2. Relationship between lagged lagged average capacity utilization utilization and and the growth of capacity relative to potential output, 1955-76. The level of growth of capacity relative to potential output, 1955—76. The level of capacity at which capacity and potential output grow at the same rate capacity at which capacity and potential output grow at the same rate ("break-even utilization") is is estimated to be be 84 84 percent. percent. (“break—even capacity capacity utilization”) estimated to incentives to invest are insufficient to keep the capital stock growing as rapidly as potential output. It appears that even today we are below the break-even break-even point, for over the,last thelast few months capacity utilization in manufacturing has been averaging only 83 percent. To summarize the evidence up to date: doxical situation. we are faced with a parapara- Capacity output has been growing more slowly than potential output for some time now. Yet, we are constrained from having a rapid growth in actual output because of the fears in many quarters other hand, of getting too close to the inflationary shoals. On the other if we stay too far away from our objective objective of high levels of employment and utilization, we see that capacity will grow too slowly for us to reach our ultimate target. Thus, historical Thus, again according to historical relationships, if we were to stay at a utilization rate of 83 percent for an extended period of time, our full employment capacity utilizautilization index would continue to rise. Speed Limits to Growth? - We have seen that there is aa fundamental dilemma which the econeconomy faces over the next four years. A wfll A path of immediate recovery will clearly lead the economy onto the shoals of capacity bottlenecks. On the other hand, a path of very slow growth, with capacity utilization break—even point, will lead to an increasing below the break-even increasing secular diverdivergence between potential output and capacity output. optimum lies somewhere in between. 117 Clearly the The problem is what I will somewhat whimsically call the flyflypaper problem. AA hungry fly sees a delicious morsel of fly food across the room, but unfortunately the morsel is very close to aa sticky piece of flypaper. The fly wishes to get as much of his tasty dinner as he can, but in doing so he risks the danger of overshooting his dinner and getting stuck on the flypaper. Of course the flypaper problem is exactly the problem we have been discussing up to now. If we stay too far away from potential outout- put and capacity in the hopes of avoiding the inflationary shoals, we inflation, but we are risking future ininwill indeed not risk present inflation, flation by building insufficient capacity to prevent future bottlebottle- necks. On the other hand, if w,, immediate we pursue the strategy of irrniediate recovery, we risk encountering inflationary bottlenecks immediately. immediately, if recovery,-we we encounter exogenous disturbances which lead us to overshoot capacity and trigger inflation. Ther,.,fore, like our friendly fly, we must get Therefore, employment close enough to capacity t) to get investment, output, and employment high, but at the same time not overshoot our target. ies for The Recovery Recover Strategies Given our current economic situation -- high levels of unemployunemploy—- ment and the high level of the full employment capacity utilization rate -- this suggests aa strategy for the recovery must take into -— account both factors. oo We must assure producers that they will have adequate markets to sell their output. This imimplies that the levels of capacity utilization must plies that the levels of capacity utilization must be above the break-even point -and soon. be above the break-even point and soon. -- 118 The problem is what I will somewhat whimsically call the flyflypaper problem. A A hungry fly sees a delicious morsel of fly food across the room, but unfortunately the morsel is very close to a sticky piece much of his tasty dinner as he of flypaper. The fly wishes to get as nijch can, but In in doing so he risks the danger of overshooting his dinner and getting stuck on the flypaper. Of course the flypaper problem is exactly the problem we have been discussing up to now. If we stay too far away from potential out out- put and capacity in the hopes of avoiding the inflationary shoals, we will indeed not risk present inflation, but we are risking future inin- flation by building insufficient capacity to prevent future bottlebottlenecks. On the other hand, if we pursue the strategy of inrediate immediate necks.. immediately, recovery, we risk encountering inflationary bottlenecks inrediately, if we encounter exogenous disturbances which 1-1hich lead us to overshoot caca- pacity and trigger inflation. Therefore, like our friendly fly, we must get close enough to capacity to get investment, output, and enploy— employment high, but at the same time not overshoot our target. Strategies for The Recovery Given our current economic situation -unemploy-— high levels of unenploy— ment and the high level of the full en~loymentcapacity employment capacity utilization must take into rate -—— this suggests a strategy for the recovery iwst account both factors. o We must assure producers that they will have adequate marmarimplies that the levels of kets to sell their output. This iwlies -capacity utilization must be above the break-even point —— and soon. prooo At the same time we must recognize that our labor and pro— 119 means that o duct markets are badly out of balance. This means our recovery must proceed in an orderly fashion as investment accelerates and capacity output recovers its growth. imbalance between capacity and potential output oo The inthalance must be taken into account in our overall fiscal and monetary policy. It would be extremely untimely for the monetary authorities to slam on the brakes at that point investment. And our fiscal and when we so badly need Investment. tax policy oust must recognize the central importance of spespecial incentives to invest during the next few years. is the one on which I would like to close. This last consideration Is As I have indicated today, our capacity output does not dovetail with the social and economic needs of today. In designing the major fiscal pol icy actions over the next two years, we must taken into account the policy needs for capacity expansion. careThe Administration is considering care- fully the possibility of giving special incentives for investment in short-run to aid the growth of capacity. the short—run It should be emphasized that -- while it is always nice to have -— -- the needs over the next few years are particuparticuadditional capacity —- larly critical. If we are to succeed in reaching a noninflationary profull employment economy, we must assure that capacity expansion pro- ceeds at a sufficient pace. I expect that the Administration will propose tax measures especially designed to encourage the growth of ity over the next few years. capacity We hope that a climate of cooperacoopera- on from the monetary authorities and the business coni,unity community will tion make sure that, in fact, capacity bottlenecks do not slow the current recovery. Footnote l/ jJ "straight up” up" increase in outoutNote that the calculation is a “straight calput and utilization, like that customarily employed in cal''gap.'' culating the “gap.” 120 SHORT TERM PROJECTIONS OF MANUFACTURING CAPACITY UTILIZATION James F. F, Ragan, Jr. As the papers presented presented at this conference demonstrate, there there is a divergence of views as to whether current measures of capacity utiliutilization overstate or understate the amount of untapped capacity remainremaining in the economy. I want to sidestep sidestep this issue, concentrating concentrating instead on one widely used measure of capacity utilization: Fedthe Fed- eral Reserve Board’s Board's capacity utilization rate for manufacturing, manufacturing. rn In particular, II want to discuss aa simple model which can be used to project manufacturing capacity utilization, as published by the Board, over the next couple of years. Those, including certain members of the Board’s measure of Federal Reserve Bank of St. Louis, who feel the Board's probcapacity utilization is biased downward may argue that capacity problems will develop sooner than our model predicts; those who feel current measures of capacity utilization are biased upward may take the opposite view, arguing that capacity problems will not emerge until later. Nonetheless, examining when capacity is likely to become strained--at Board's capacity utilization strained——at least on the basis of the Board’s statistic--is statistic—-is an interesting experiment and provides a useful benchmark for discussions about prospective capacity problems. Indeed, to leak Or. professor of Economics Economics at Kansas State UniUniDr. Ragan is an assistant professor versity. The author thanks Robert Falconer and An-loh Lin for comments on an earlier version of this paper and Debbie Jamroz for research assistance. 121 one of our findings, it is not necessary to take the "St. Fed's" “St. Louis Fed’s” position that currently published measures of capacity utilization are artificially low to show their concern about about potential capacity probproblems within the next couple of years. The model presented in this paper is for the key sector of manumanufacturing, although the technique can be applied to other sectors of the economy as well, well. First, an equation equation is estimated linking growth in manufacturing output to growth in GNP. Next, additions to manumanu- facturing capacity are estimated, based on projections of investment. The forecast of output is then divided by the forecast of capacity to yield projections of capacity utilization. utilization, The model is first used to project capacity utilization from 1977 III - 1978 IV, based on aa — "consensus" forecast of GNP growth. “consensus” Next, implications are drawn concon- cerning the impact on capacity utilization of strong protracted econeconomic growth, the Administration’s Administration's assumption. Specifying the Model Manufacturing output and GNP tend to move together. In growth terms, the relationship between these two variables can be specified betweenthese as)/ as:1’ p I°P t = aa + a 1 GNP t + (1 1) t 0 Percentage change in the manufacturing index of indus— indusII°P Pt == trial production, i.e., 100 / trial production, i.e., 100 (IPt - IPt_ 1) / IPt-l. - 0 0 GNP GNPt t = = Percentage change in constant-dollar constant—dollar gross national product. product. 122 "t = Error term. = Error term. The change in capacity from one period to the next depqnds dep~nds positively on the volume of investment and negatively on the extent of depreciation, which in turn depends on the level of capacity last period. (See Appendix l1 for greater elaboration.) The change-inchange—in- depicted as follows: capacity equation can therefore be depicted C~ Ct1 = biCt1 - + b2It + (2) 6t Ct== Capacity index for manufacturing. It== Real Real manufacturing manufacturing investment net of of pollution pollution investment net control expenditures. "t = Error = Error term. As many economists have observed, investment accelerates as utilizathe volume of unused capacity shrinks, i.e., as the capacity utilization rate (CU) rises. Changes in investment investment are therefore specified to be a function of past changes in capacity utilization: 00 = c c C°U~ +e~ I =c 1 1 t 00 (3) 00 It== Percentage change in investment investment (I) from the previous calendar year. C°Ut1= Percentage change in capacity utilization (CU) over the previous year (fourth quarter to fourth quarter). Et= = Error Error term. Although this equation greatly abstracts from the underlying determindeterminants of investment, it performs well empirically. Another advantage of this specification is that that the capacity utilization rates generated by 123 our model can be used to project investment in subsequent years. That is, when combined with GNP projections, equations (1), (2), and (3) constitute a closed system capable of projecting capacity utilization rates indefinitely into the future. Empirical Results Equations were estimated estimated over over the the perio4 periorl 1954-1976. 1954-1976. Equations (1) (1) -- (3) (3) were apparResults are reported in Table 1. From the first equation, it is apparis more volatile than GNP; the large ent that manufacturing output Is 0 0 coefficient for GNP indicates that rapid GNP growth is on average accompanied by even more rapid growth in manufacturing output. The Ct-l in equation (2) indicates that, in the absence of coefficient of ~ investment, capacity declines 3.35 percent per year, the result of It indicates that depreciation and obsolescence. The coefficient of I~indicates each one billion dollars of non-pollution-control investment investment expendexpenditures, measured in 1972 dollars, increases the capacity index (1976 IV IV= 163.2) by 0.29 percentage point. Converting the investment coefficient to an elasticity, each 1.0 percent change in real investinvestment net of pollution control is on average associated with aa 1.0 per1.0 percent change in gross additions to capacity (as opposed to net additions, i.e., additions net of depreciation4~ depreciation.)Y Finally, as expected, equation (3) indicates that investment accelerates as capacity utilization rises. The fit fit of all three equations is quite good, as judged by the R2 from zero. R2 values, and all coefficients are statistically different from 124 Table 11 Regression Results* (t-statistics in parentheses) (t—statistics in parentheses) 0 I°Pt 1°pt S= —.756 -.756 + 2.203 GNP GNPt ( 4.22) (16.46) ( 4.22) (16.46) ~ 2 SE= RR2 = .751 SE = 1.37 DW == 1.85 Sample Period: 1954 I 1976 IV Sample Period: 1954 I 1976 IV (1) C—C Ct-Ct-l = —.0335C -.0335 Ct-l +.29231 + .2923 It (3.23) (7. 10) t t—l (3.23) t—l (7.10) ~ R2 == .935 SE= p = .578 SE = .576 Sample Period: 1954 - 1976 Sample Period: 1954 1976 (2) - DW == 1.70 - 0 It= C Ut l I = 4.196 + 1.754 c°u (2.97) (2.97) (6.55) ~0 (3) 2 R == .671 SE= SE = 6.74 Sample Period: 1954 - 1976 DW == 1.71 - r00 p = Note: I°P GNP= GNP CC = I = I= Quarterly growth of manufacturing output Quarterly growth of real gross national product Index of manufacturing capacity Real manufacturing investment net of pollution control expenditures 0 cC°U u == Growth Growth of of capacity capacity utilization utilization in in manufacturing manufacturing The Cochrane-Orcutt Cochrane-Orcutt iterative iterative technique technique was was used used to to adjust adjust * The first—order autocorrelation in equation (2). for first-order * 125 Even more important, simulation results (presented in Appendix 2) indicate that the model model does aa good job of tracking capacity utilizautilization during the current recovery, recovery. Having passed this test, the model model was then used to run two experiments, described in the following sections. sections. ~ec~nCaacitUtilization, Projecting Capacity Utilization, 1977 III - 1978 IV - Based on the increase in investment projected for 1977 0 3/ -- capacity is projected to increase by 3.2 = 12.9 12.9 percent~/ r177 percent77 percent between 1976 IV and 1977 IV, The increase is assumed to be 1977 IV. -- distributed equally throughout the year, implying a quarterly growth percent.11 Output growth is projected using the in capacity of 9.791 percent.~” median of eight prominent forecasts of real GNP growth, as published 1s Statistical in Board's In the September 1977 issue of the Conference Board Bulletin (see Table 2). The output growth and capacity expansion propro- jections are brought together in Table 3.Y 3,~.1 Based on the Conference Board median GNP forecasts, our model projects that capacity utilizautilization in manufacturing will increase steadily to 85.8 percent in 1978 IV. Capacity will expand at aa 3.2 percent annual rate in 1977 and at a 3.9 manupercent rate in 1978, compared to the 2.3 percent rate of 1976; manufacturing output will increase over the forecast period at an average annual rate of 6.3 percent. The 85.8 percent rate projected for 1978 IV is but 22 percentage points below the 1973 quarterly peak and 2.8 percentage points below the highest peacetime peak recorded, peak recorded. 126 Hence, based upon the median Table 22 Real GNP GNP Growth, Real Growth, Median Conference Board Forecast Compound Annual Growth Rate 1977 III 4.4 percent 4,6 4.6 IV 1978 1.082 l. 082 percent 1.131 1. 131 4,45 4.45 II II 1.094 1. 094 1.082 1.034 1.034 .839 .839 4.4 4.2 4,2 3.4 III IV Source: Quarterly Growth Rate The Conference Board, Statistical Bulletin, September 1977. Table 33 Capacity Utilization Projections Based on Median Conference Board Forecast 1977 0 0 CC GNP IIPp p CU cu ..791 791 ..791 791 1. 082 1.082 1.131 1 . 131 1.628 1.628 1.736 1. 736 1. 008 1.008 1.009 1. 009 82.6 82.6 83.3 84.0 •.961 961 .961 .961 .961 1.094 1.082 1.082 1.034 .839 1. 654 1.654 1.628 1.522 522 1. 1.092 1.092 l. 007 1.007 1.007 1.006 l. 006 1.001 1. 001 II III IV 1978 0 0 II II IIII II IV 84.6 85.2 85.7 85.8 0 Note: C == Percentage change in capacity capacity (from previous quarter) 0 GNP == Percentage change in real GNP 0 II°P P == Percentage Percentage change in manufacturing output pp = = (IP~ (C~/ Cti) (IPt / !Pt-l) / (Ct/ Ct-l) CU CU== Capacity utilization (percent) 127 forecast fQrecast of GNP growth, as published by the Conference Board, our model indicates that the manufacturing sector is likely to contain some modest amount of untapped capacity at the end of 1978, yet little enough so that concern over bottlenecks in 1979 seems warranted. Utilizati on as Im liedb the Administration’s Capacity Utilization as Implied by the Administration's Projections of GNP Growth The Administration recently set aa goal of reducing the aggreaggre4.6 percent by the end of 1981. gate unemployment rate to 4,6 To achieve this goal, they estimate that real GNP must grow from 1977 1977 through 1981 5,1 percent per year,~ by an average of 5.1 year.§! The implications for capacity utilization can be examined by plugging the 5,1 5.1 percent growth rate Into our model -- an experiment which indicates the Administration’s Administration's into .~— goal is apparently overly optimistic. Based on the Administration’s Administration's GNP figures, our model model projects that capacity utilization would reach its 1973 peak in 1978 IV, its peacetime peak in 1979 1979 I, and its all— all- time peak in 1980 II {see (see Table 4). Assuming 5.1 percent GNP growth could be sustained, capacity utilization would rise to 96 percent in 1981 IV. Historical experience, however, indicates that a utilization rate high is unattainable for manufacturing; widespread shortages rate this this high is unattainable for manufacturing; widespread shortages aod bottlenecks would emerge well before such a rate could be achieved. and Of course, investment is not actually predetermined through 1981. based on the historhistorThe investment values forecast by the model were based ical relationship between changes in capacity utilization and changes in investment growth. growth, But investment investment growth can can be influenced by other other But 128 factors as well; e.g., by changes in tax policy or in the degree of factors uncertainty facing businessmen. Therefore, if the Administration wants to foster prolonged economic growth it must attach increased importimportance to stimulating investment, thereby slowing the rise in capacity strained, utilization and postponing the time when capacity will become strained. Administration's goal may still Yet, even if investment is spurred the Administration’s prove elusive. Our model suggests that, on the basis of continued strong GNP growth, capacity problems are likely to appear within the next two years. 129 Tab 1e 4 Table Capacity Utilization Projections Based on the Administration's GNP Scenario Based on the Administration’s GNP Scenario 1977 1979 1980 1981 0 0 0 CC GNP I p IP pp cu Cu 1. 012 1.012 82.6 83.6 84.6 II IV • 791 .791 .791 • 791 1. 251 1.251 1.251 1 • 251 2.000 ~.00Q 2.000 1.012 1. 012 I II III IV .985 .985 .985 .985 1. 251 1.251 1. 251 1.251 1. 251 1.251 1. 251 1.251 2.000 2.000 2.000 2.000 1.010 1.010 1.010 l. 010 1.010 85.4 86.3 87.2 I II III IV 11.131 • 131 1. 131 1.131 1. 131 1.131 l. 131 1.131 1. 251 1.251 1. 251 1.251 l. 251 1.251 1. 251 1.251 2.000 2. 00') 2.000 2.00,J 2.000 2. oo,J 2.000 1.009 1.009 1.009 1.009 88.8 89.6 I II III III IV l. 251 1.251 1.251 1.251 l. 251 1.251 l. 251 1.251 i. 251 1.251 1. 251 1.251 2.000 2.000 2.000 2.000 1.007 1.007 1.007 1.007 91.8 92.5 93. 1 93.1 93.8 l. 251 1.251 1. 251 1.251 1. 251 1.251 2.000 2,000 2.000 2.000 2.000 1. 006 1.006 1. 006 1.006 1. 006 1.006 1. 006 1.006 94.4 94.9 95.5 96.1 IIII II 1978 0 0 I II III IV l. 251 1.251 l. 348 1.348 1. 348 1.348 1. 348 1.348 1.348 l. 251 1.251 130 130 88.0 90.4 91.2 Appendix 1: Projecti Capacity Growth Projecting Additions to manufacturing capacity are estimated from investment data. Investment is measured in real or constant-do11ar constant—dollar terms, since addicapacity is related to real rather than nominal investment. In addition, pollution control expenditures are netted out, since these expenditures do not augment productive capacity. Yet, even with these adjustments, translating investment data into capacity growth can be tricky tricky.. composition as well as volume of One difficulty difficulty is that the ~p2~jtion investment is important. Investment which eliminates aa production bottleneck may have a tremendous impact on capacity. On the other bottleneck hand, investment which expands plant size may, while providing addiadditional office space, leave plant ant capacity unchanged. A machine, A new riachine, if added to the existing existing sstock of equipment increases capacity, but but if if some existing equipment is retired red when the new machine is put in place capacity capacity need not be increased. Furthermore, expenditures on modernization generally provide for smaller capacity growth than outoutnally, the impact of an investment lays on new plant and equipment. Finally, dollar is likely to vary from industry to industry. industry, AA dollar spent in an industry approaching capacity will have a more pronounced impact on on aggregate capacity than aa dollar spent in an industry possessing abundabundant unused capacity. /i.lso Also complicating the investment-capacity relationship is the fact that investment frequently increases capacity with a lag. 1ag. ProProjects requiring years to finish are likely to add to capacity only when completed or nearly so. AA plant half-completed may not augment a firm’s firm's capacity at aall, 11. Moreover, the lags 1ags invo 1 ved may vary both over time involved and by type of investment. The severity of these problems is difficult difficult to assess a priori. While the composition of investment may vary substantially substantially over time tithe for a particular company or industry, in in the aggregate the composition of investment investment may remain relatively relatively stable. Therefore, how well aggreaggregate investment investment explains capacity growth is ultimately an empirical question. So is the question of whether capacity growth this period is significantly related to previous investment. Each period, invest— invest~nt dollars are spent which increase capacity only in the future. At rn~ntdollars capacity the same time, however, certain projects started in the past are finished, adding to capacity in the current period, period. If If these two lag effects wash out sufficiently, empirically capaci sufficiently~ then capacity growth may not be related to previous investment, but only to current investment. To investigate the relationship relationsraip between investment investment and capac~ty, capacity, the following model was developed. ven period (C ) is identically Capacity in a given identically equal to capacity last period minus the loss in capactty capaccty due to depreciation and 131 13l obsolescence (Dt) plus the gross additions to capacity (CADDt)t (CADDt)' i.e., i.e., Dt ct~ Ct_i ct-1 —- Dt + + CADDt. CADDt. (4) (4) It is assumed that capacity depreciates at a constant rate each period: Dt = a C~_1. (5) In addition, it is assumed that gross additions to capacity are related to current and and possib.ly previous investment: to current possibly previous investment: CADOt = Bilt + ~21t—l + ... (6) (6) where It refers to real investment net of pollution control expendiwhere It refers to real investment net of pollution control expenditures. Combining the above equations, capacity can be rewritten as: C,~—(l-.a)Ct_1 B1I~+ B2It..i + ... Ct. (7) where In cawheres represents the error term. Alternatively, the change in capacity can be expressed as: pacity can be expressed as: (7 Ct_1 = ..aC~_1 + ~l1t+ 821t—1 + ... e~. (7’) + I ) — cf a and ~ s are arc the same whether v:hcther obtained by estimating Estimates of equation (7) or equation (7'). (7’). or.equation caThe capacity variable of this study refers to manufacturing capacity as measured by the Board of Governors of the Federal Reserve paci System.y System.7/ Capacity values were obtained for the fourth quarter of each Fourth-quarter values were chosen because both year (see Table 5). Fourth—quarter the Board of Governors and McGraw-Hi McGraw-Hill 11 estimate capacity growth on an end-of-year basis. end—of—year The equipment spending. investment variable (I) is an estimate of real plant and expenditures over the calendar year net of pollution control The variable is defined as follows: POL)/P II== PE (100 - P01)/P — where PE p1ant and equipment, in == Expenditures for new manufacturing plant bi 11ions of (current) dollars; billions of (current) dollars; POL== Percent of plant and equipment expenditures for air and P01 water pollution control; water pollution control; Implicit ClIP GNP price deflator deflater for business fixed investment. PP;= Inplicit The PE data are published by the Bureau of of Economic Economic Analysis, Analysis, the the P01 POL The PE data are published by the Bureau 132 data by McGraw-Hi1l.~ McGraw-Hill.!!/ Equation (7’) (7') was estimated over the period l954_76.2/ 1954-76.Y Lagged investment terms did not contribute to the explanatory power of the equation, nor were their coefficients statistically significant. Only current investment proved to be statistically important. Therefore, the lagged investment terms were dropped. (Regression results are reported reported in Table 1, equation (2). Table 5 Values Values of of the the Investment Investment and and Capacity Variables Investment (I) 33. 1 33.! 28.9 28.7 33. 33.2:'. 36.2 33.0 62.7 65. 1 65.1 67.9 71.4 74.2 76.6 79.3 83.0 85.9 88.8 92. 92.11 96. 1 96.1 102.7 102.1 110. 2 110.2 117. 9 117.9 124.7 124.7 131 . 1 131.1 136. 1 136.1 140.0 140.0 144.7 150.3 155.7 159.5 34.2 163.2 18.9 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 J965 1966 1967 1968 1969 1970 1910 1971 1972 1973 1974 1975 17. 17.88 18.3 22.2 22.7 17. 2 17.2 17. 2 17.2 20.3 19.4 20.2 21.6 25.4 30.4 35.6 35.0 33.3 35.1 1916 1976 133 Capacity (C) ( C) Appendix 2: Simulating the Model To test its predictive ability, the model was simulated over the first nine nine quarters of the current recovery. The equations, estimated over the period 1954 I to 1974 IV, were used to generate forecasts for 1975 II - IV. After updating the equations through 1975 IV, forecasts for 1976 were made. Finally, after extending the sample period through 1976 IV, capacity utilization was forecast for the first two quarters ~uarters of 1977. - Althougn Simulation results uncovered no apparent bias. Al thougA capacity utilization was somewhat underpredicted during 1976, the t~e model did get back on track. Capacity utilization was recorded to be 82.6 percent in 1977 II, compared to a projected rate of 82.7 percent (see Table 6). Thus, capacity utilization rose 11.7 percentage points during the first 11. 8 percentage pe,·centag" points nine quarters of the recovery, compared to the 11.8 projected by our model. zation and the mode1. Moreover, Moreover, actual capacity uti ciizatio1 1. 3 percentage p':rcentage rate predicted by our model never di verged by more than 1.3 capacii;y utilization uti1 iwtion point. Our model even picked up the decline in capacity registered in 1976 IV. registered in 1916 IV. Table 66 Simulations of Capacity Utilization 1975 1976 1977 Note: 0 0 0 0 0 0 CC GNP IPP . p cu cu CU . CtMCU CU-CU 1I II II I III IV .668 .668 .668 1. 573 1.573 2. 735 2.735 .745 2.627 2.621 5.008 1.930 l. 930 1.019 1. 019 1.043 1. 043 1.003 72.2 75.4 7!i. 4 75.6 7S.6 70.9 71.33 71. 75.3 75.8 .9 •9 .1 .1 —1.2 - 1.2 I II II IIII IV .595 .595 .595 .595 2.130 1.234 .959 .288 3.946 1. 965 1.965 1. 357 1.357 1.033 1. 033 1.014 1. 014 1.008 .993 78.1 78. 1 79.2 79.8 79.3 79.0 80.2 8D.2 80.8 80.6 —.9 -.9 —1.0 -1.0 —1.0 -1.0 —1.3 -1. 3 II II • 791 .791 . 791 .791 l. 503 1.503 1. 025 1.025 1. 018 1.018 s:. 3 8..3 82.7 81.2 82.6 .l .1 .l .1 . 1. 833 1.833 —.126 - . 126 3. 282 3.282 2.555 cu = Capacity utilization uti ization as simulated by the riodel nodel CU CU cu== Actual capacity capac ty utilization 134 Footnotes jJ 11 reNonlinear versions of equation (1) were also tried, but their tee sults were empirically inferior. y ~ The te~ term Gt= C~. Ct —- (l—u)Ct_i (1-a)Ct-l == C~ Ct - .9665 C~_ Ct-l 1measures the gross change in capacity, i.e., the difference between actual capacity and the level which would have prevailed in the absence of any I, the elasticity of G res.e_ect investment. Since CG = .2923 I, C with resp~ct to I, evaluated at the mean, is n~== .2923 T/ T / ~whereTand G where T and GG refer to the mean values of I and CGover over the estimation period (1954-76). T = 26.92 and CG and 7.89. 7.89, Hence, t~ n = .2923 (26.92 / invest7.89) = 1.0. In other words, each 1l percent change in our investment variable is associated with withal percent change in gross a 1 capacity growth. Therefore, if in a certain year $25 billion in investment would increase capacity by 5 percent in gross terms, then raising investment to $30 billion (an increase of 20 percent) can be expected to raise gross capacity growth to 66 percent (also cap an increase of 20 percent). The finding that investment changes and changes in capacity growth are linked in such a manner is appealing on theoretical grounds, and suggests that our investinvestment variable does a good job of capturing gross additions to capacity. p 11 Interestingly, the 1977 projection of investment derived from equation (3) fa11s in between the investment plans reported by falls in the BEA and by McGraw—Hill McGraw-Hill in late spring. Based on the growth in capacity utilization between 1975 IV and 1976 IV, equation (3) projects that manufacturing investment in 1917, 1977, as measured by our investment variable (I), will exceed investment in 1976 by about 12.9 percent. The BEA investment survey figures transtranslate into a 9.3 percent increase in in!; the McGraw-Hill figures, I; McGr~w—HiIlfigures, inGrease in into a 14.3 percent increase (assuming aa 66 percent increase the price of investment goods (P) and using the 1977 estimate of pollution control expenditures reported by McGraw-Hill). (l .00791 )4 =s 1.032. That is, (l.0079l)~ ~/ !/ — Capacity utilization in time period ttis Is defined as the ratio of actual output to capacity output, i.e., Cut cut== IPt/Ct. it can easily be shown By lagging this relationship one period, it. that the utilization rates in successive periods are related as follows: 135 135 Cut cut wherep= where p = = P P = cut_ 1 !Pt/ IPt-1 ‘Pt / iPt_i —---- Ct / Ct_i This is Is the formula used to project capacity utilization. 6/ §j ZJ LI See Office of Management and Budget, Midsession Review of Fiscal 1978 Budget, Special Supplement, July 1, 1977. j9~judet, ~Q~al Su lement, Jul yT~fl77~~ A.'llong other other advantages, the Board Board of of Governors Governors series is readily readily Among advantages, the series is available to the general public, has a long track record, and lacks any apparent cyclical bias. For a discussion of the major series of capacity uti utilization, “Measuring 1i zat ion, see James Ragan, "Measuring Capacity Utilization Federal1 Reserve Bank of Uti 1i zati on in Manufacturing,” Manufacturing," Federa New New York 9~r~rlReview, Quarterly Review, Winter 1976, pp. pp. 13-20. 8/ '§I The pollution control data are actually available only since 1967; 1967; but, because pollution control expenditures did not begin their rapid ascent until untfl the late sixties, the fraction of investment probexpenditures devoted to pollution control prior to 1967 was probably close to the fraction spent in 1967. This was the assumpassurnption made. Thus, the pre-1967 values of POL were set equal to the 1967 value (2.8 percent). The BEA also publishes publishes a series on pollution control expenditures, but it does not begin until six years after the McGraw-Hill McGraw~Hili series. 9/ The estimation estimation period was annual, rather than quarterly, because truly independent capacity values were available only once per year. Both the Federal Reserve Board (whose series is used in this study) and McGraw-Hill year-end. McGraw—Hill obtain capacity values at year—end. these are simply Although quarterly estimates are available, these interpolations between annual annual observations. observations. interpolations between 1 36 136 COMMENTS ON RASCHE ANO AND TATOM, "THE EFFECTS OF THE NEW ENERGY REGIME .... " “THE REGIME., AND “ENERGY RESOURCES AND POTENTIAL GNP" GNP” "ENERGY Frank de Leeuw These two studies have performed aa timely service in reminding us that aa major rise i.n natural resource costs can have a sizable negative impact on potential GNP. The purpose of this note is not to question that central proposition. Rather, it is to argue that (1) the impact on potential GNP takes place only gradually as production techniques and consumption patterns change, not all at once as these studies imimply, and (a) (2) that the ultimate ultimate impact may not be as large as the 4 or 55 percent estimated in these studies. studies. At the present time, the note will conclude, production techtech- substantiniques and consumption patterns do not seem to have altered substantially in response to higher energy prices. Potential GNP has therefore not yet declined appreciably; rather, what has happened is that aa larlar- ger fraction of GNP (or claims against GNP reflected in balance-oftrade deficits) must be paid to the owners of energy resources. PoPo- tential ~consumption after subtracting out this fraction has been rereduced, but potential production has not. It is, however, important to Fiscal Analysis inte in the ~r. deLeeuw is the Assistant Director for Fisca ~euw Office. Views expressed in this note are those Congressional Budget Office, those of his employers. of the author and not necessarily those 137 watch for signs that production techniques are beginning to respond to high energy prices and to take any such response into account in formuformulating economic policy. The Rasche-Tatom Assumptions and Their Implications ~sche-TatomAssumtionsandTheirImlicatiOns The reasoning used in these studies to translate higher energy prices into reduced potential GNP can be explained in aa few sentences. potential output deSince output requires lator, labor, capital, and energy, potential depends on available supplies of these three inputs. For labor and for capital, it is possible to measure at least approximately maximum availavail- particiable inputs, determined by population and expected labor force participation rates in the case of labor and by the initial capital stock, its rate of depreciation, and the expected fraction of new output devoted to fixed investment investment in the case of capital. Since energy is traded internationally in huge amounts, it does not make sense to think of a fixed quantity of potential energy consumpconsumption by any one country, analogous to potential labor and potential capcap- ital. Rather, it makes sense to think of producers and consumers consumers chooschoos- ing aa ratio of energy to output on the basis of relative prices, technotechnoinfluences. The higher this logical developments, and perhaps other influences, ratio is —-the --the more energy-intensive production is-—the is--the more output can be produced with given amounts of labor and capital. A sizable increase A in the relative price of energy should lead producers to conserve energy and consumers to shift shift purchases away from energy-intensive goods and services--should, services—-should, in other words, reduce the ratio of energy to output. supplies of High energy prices should therefore mean that available supplies 138 labor and capital will not yield as much output as they would have if the cheap-energy years of the past past had continued. A major increase in energy prices should, in this view, cause cause A (1) energy conservation, or a fall in energy consumption per unit of outout(2} a rise in both labor consumption per unit of output {equivalent put; (2) (equivalent to aa fall in productivity as usually measured) and capital consumption per unit of output; and (3) a reduction in productive capacity or potential output, properly measured. The specific relationships used by Rasche and relaTatom, furthermore, imply imply that these reactions occur occur at once when relative energy prices go up. The Evidence Since 1973 Have these consequences actually taken place since the OPEC price price rise of 1973-1974? 1973-1974? The evidence is, at best, mixed. The first conseconse- quence, a fall in the energy-output ratio, is a central one. Only as this ratio falls do producers need to use more capital and/or labor per unit of output and hence reduce potential GNP. But there is no evidence of a drop below trend in the energy-output ratio since 1973. Table 11 shows two measures of energy per unit of output from 1970 through 1976. 1976. They both display a trend toward conservation over these years amounting to a reduction of 11 to 33 percent in energy per unit of output each year. But they show this trend before the dramatic increase in energy prices as well as afterwards and there is no sign of any acceleration after the energy price increase took place. Examination of a longer period than 1970-1976 suggests that some movement toward conservation may have taken 1970-1976 place in recent years, but nothing like the 33 percent drop in the ratio 139 TABLE ‘I1 ENERGY CONSUMPTION AND GNP (CONSTANT DOLLARS) Energy Consumption (thousand trillion BTIJs) tri 11 ion BTUs) Total Total Industrial GNP, 1972 Prices (billions of dollars) Total "Industri a1"l/ “Industrial”~-” Ratio of Energy to Output (indexes, 1973 1973 == 100) Total Industrial 1 1970 ' 68.3 23.3 1075. 1075.33 370.2 104.5 107 .8 107.8 1971 69.5 23.0 1107. 5 1107.5 374.9 374,9 103.2 103,2 105.1 105.1 1972 73.3 23.8 1171. 1171.1l 399.7 399,7 102.9 102,9 102.0 102.0 1973 75.1 75. 1 24.9 1235.0 426.4 100,0 100.0 100.0 1974 73.2 24.2 1217. 8 1217.8 402.6 98.8 102.9 1975 71.5 21.6 21,6 1202.1 379.9 97.8 97.4 1976 75.0 75.0 22,9 22.9 1274,7 1274.7 416.2 416.2 96.8 96.8 94.2 94.2 rce: Source: Energy Consumption--FEA, Consumption——FEA, by telephone, through 1975-1976, 1975—1976, CBO estimates based on data in FEA'ss ~ Monthly Energy Review FEA’ GNP--Commerce Department 140 roughly 50 percent IninRasche and Tatom would expect in response to the rpughly crease in the relative price of energy in 1973-1974. With respect to labor and capital, the evidence is not quite so negative. Output per unit of labor did fall in 1974 and has not yet caught up to its earlier trend, even after correction for cyclical inin- fluences. After cyclical correction, however, output per unit of capicapital does not appear to have fallen. Output per unit of combined labor— labor- capital did fall, and the Rasche-Tatom regression results reflect the fact that this shift in the relation of output to labor and capital inin- puts occurred at the same time as the rise in oil prices. But it is hard to interpret these labor and capital changes as responses to energy developments when there is no evidence of a shift in the energy-output ratio. Preferred Rates of Capacity Utilization The third implication of aa rise in relative energy prices is a decline in capacity and potential GNP. As Rasche and Tatom point out, manufacturing capacity as measured by the Federal Reserve Board did not fall when energy prices went up. Capacity, as Rasche and Tatom define output-it--namely, the cost-minimizing or profit-maximizing level of output-— should have fallen by 44 or 55 percent, according to their calculations. They interpret the failure of the actual indexes to fall as due to a difference in definition. Published capacity statistics, they believe, refer to maximum output feasible under customary operating conditions, not to the concept of cost-minimizing output which they prefer. 141 I believe they are probably probably right in their interpretation of pubpublished capacity statistics. statistics, In the short run at least, the failure of published capacity indexes to fall is not a serious argument against their view of the impact of energy prices. confirmation of their view. Neither, however, is it aa It is simply not relevant to evaluating their hypotheses. The Bureau of Economic Analysis does, however, collect another set set of statistics in its capacity survey which are more relevant to testing the Rasche-Tatom Rasche—Tatom views. These are manufacturers' manufacturers’ views of the percent of capacity at which they would prefer· ~f~r to operate. is: The exact question "At manufacturing capacity would would your company “At what percentage of manufacturing have preferred to operate in order to achieve maximum profits or other objectives?"Y objectives?”~’ Now if high energy prices do not change rated capacity but do have an immediate impact on production techniques and input propro- portions as Rasche and Tatom Tatom maintain, maintain, then the minimum—average—cost minimum-average-cost rate of operation operation should decline when energy becomes much more expen— expensive.~ sive.Y1 AA reduction reduction in preferred operating operating rates looks like a promising candidate for an empirical counterpart to this theoretical concept. ManMan- ufacturers might be expected to prefer not to operate equipment which was extremely energy-intensive, and to prefer to operate other equipment in ways which conserve energy and hence sacrifice some output. manufacturers’ view In fact, however, nothing much has happened to manufacturers' of their preferred rate of operation. For all manufacturers taken togetoge- ther, preferred utilization was 95 percent of rated capacity from 1970 through 1974 and 94 percent in 1975 and 1976. There is no sign of a 44 to 55 percent drop after the run-up of energy prices in 1973-1974. 142 Unpublished detail supplied by the Bureau of Economic Analysis, furtherfurthermore, does not suggest that the aggregate conceals any shifts fts at indusindustry levels that would bear out the Rasche-Tatom view. ew. For example, energy— there does not appear to have been aa drop in preferred rates in energy- intensive industries offset by a rise elsewhere. Thus, statistics on preferred operating rates, like statistics on energy per unit of output, suggest that so far there has not yet been a significant restructuring of production techniques in response to higher energy prices, flBeenHaenin? What Has Been Happenin2? There is no doubt that higher energy prices have created incenincentives to change production processes. So far, however, the evidence significant energy indicates that these incentives have not yet led to significant conservation and substitution of labor and/or capital for energy. ProbProb- ably one reason for the delay is that many of the possibilities possibilities for enenergy conservation require new plant and equipment. Frequently it will pay to continue to operate existing capital goods for a time even if they utilize uneconomic uneconomic processes because they have already been paid exfor and because conversion to a more energy-conserving process is ex- tremely costly. Another possibility is that large-scale energy conserconser- vation is awaiting more certainty about future technological change. and about government actions affecting affecting energy costs. While it is tempting to associate the recent productivity slowslowdown with the rise in energy prices, the facts about energy per unit of output do not bear out this connection. The most likely explanation for the productivity slowdown appears to lie elsewhere--lower e1sewhere--1ower rates of 143 capital investment in relation to GNP, shifts in the experience—mix experience-mix of the labor force and the industry-mix of output, lower grcv1th growth in research and development spending, and the severity of the 1974-1975 1974—1975 recession. With respect to potential GHP, GNP, the short run conclusion is that until production techniques begin to react significantly to the change in energy prices, it would be aa mistake to translate higher energy prices into reduced potential and lower output targets. So far, high energy prices have not altered production techniques but have caused this prothis country to pay sizable amounts to oil producers in order to produce GNP by old production techniques. In paying for oil, 1, we have incurred large balance—of—trade balance-of-trade deficits which represent growing foreign claims against domestic output. While potential output has not yet been reduced substantially, these foreign claims mean that potential conconsumption by U.S. citizens has been reduced by high energy prices. The conclusion about potential GNP in the long-run conlong—run is more con- jectural. conserEventually high energy prices should lead to energy conser- vation, substitution of other inputs for energy, and hence less potenpotential 61W GNP from given supplies of labor and capital. Qualitatively, the Rasche—Tatom results siem Rasche-Tatom seem quite plausible as a long—run long-run proposition. long-run effect may I suspect, however, that quantitatively the long—run not be as large as 4 or 55 percent. The 4 to 55 percent estimate assumes no response of labor or technology to changing productivity and real inincome. In actuality, low real wages due to high fuel costs could cause the supply of secondary workers or other dimensions of labor supply to increase. conRecent labor market statistics and analyses seem to be con- sistent with behavior of this kind, in which lower productivity is 144 partipartly offset (as it affects potential GNP) by higher labor force participation. Furthermore, future technological advances could on balance be energy-saving energy—saving rather than neutral (as Rasche and Tatom assume) with respect to input proportions. These offsets are no more than possibilities, possibilities, however. It It would be a mistake to ignore the danger of a substantial eventual impact of high energy prices on potential GNP. GNP, The Rasche and Tatom studies will have served a highly useful purpose if they remind us to monitor closely trends in energy conservation and productivity and be prepared estimates of potential GNP when U.S. production techto adjust our estimates techniques show signs of significant reaction to the new energy regime. Footnotes 1/ 11 "Industrial" sector covers manufacturing, mining, agriculture, and “Industrial” construction for comparability with energy consumption data. y2/ See Marie P. Hertzberg, Alfred I. Jacobs, and Jon E. Trerathan, “The 1965—73,” Survey of "The Utilization of Manufacturing Capacity, 1965-73," Current Business, July 1974, p. 49. 3/ Rasche and Tatom, “The "The Effects of the New Energy Regime,” Regime," pp. 3—4. 3-4. 145 ISSUES CONCERNING THE DEFINITION, MEASUREMENT AND FORECASTING OF PRODUCTIVE CAPACITY Laurence H. Meyer It would be difficult to identify a more important macroeconomic issue (and one about which there was more disagreement) than recent developments affecting the level and rate of growth in the nation’s nation's productive capacity. Knowledge of the gap between actual and potential immediate relevance to the design of short-run stabilizastabilizaoutput is of ininediate tion policy and recent projections of slower growth in productive capacity and the possible relation to slower growth in capital stock appear to have heightened the interest of both the Administration and Congress in tax reform keyed to expanding Incentives incentives for capital spending. The Perry, Clark, Sd and Rasche—Tatom Rasche-Tatom papers are attempts to provide the empirical evidence on the level and anticipated rate of growth of potential output that is essential to designing such policies. However, their approaches leave the question of the level of potential output quite unsettled and their projections of future growth rates are mostly conjectural. Dr. Meyer is Associate Professor of Economics at Washington University. 147 Three issues Issues Concerninq the t~e Definition, Definition, Measurement, and Forecasting of Potential Output Forecastinq_pjJ’otential It is useful to identify at the outset Putset three separate issues concernconcern- ing the definition, measurement, and forecasting fore,:asting of productive capacity. Each of the three papers deals at leas: leas·. to t,) some extent with each of the three issues: 1. The determinants of the _leve,. leve of potential output; output. 2. 2, The cyclical behavior of actual relative to potential output-- i.e. Le. the analysis of the cyclical pattern patt1ern of participation rates, hours per worker, and productivity which along with ,1Lh cyclical pattern in employemploy- ment explain the cyclical pattern in prod;ction. prodiction. The determination of the rate of growth of potential output In in 3. g ov1th over the next five years. the past and the projection of rates o · growth 0 thP level and rate of growth iss~Jes. i SS\Jes. The papers are most concerned wLh wi::h Lhe The Level of Potential ~ Output There are three problems relatinç relatinc to t,J the determination of the level of potential output th.at thLtt arise in conmctior connEct~orwith with the three papers. a. inputs. rates. identif~inj The first problem is idem:ifJ in•J ‘potential” 'potential" levels of factor Employment, hours, and partic~pa’:icn partic,pa·:irn rates vary vary with unemployment To define the potential potential level cf ~aLor ator input, we need to know know the “full employment" employment” level unemp1o~rne:~trate. "full leve1 of the unemplo}me,1t If capital is explicitly cefne treated in the analysis, we must also cef ne a potential level of capital input. b. The second problem i: is the r,os ;ibility of a once and for all shift ~os;ihi1ity 19’3-5 due, for example, to a once in the level of productive capacity in fo 19'3-5 and for all change in the relative price of qf energy. This must be picked up by dummy variables unless energy develcpments develcpm,!nts are explicitly integrated into the model. 148 148 c, The third problem is the possibility of incompatibility between i.e., potential levels of labor and capital inputs; i.e 4 the possibility that ful1 utilization of the capital stock may occur prior to full utilization full of the labor force and that the capital stock, not the labor force, force, is the real binding constraint constraint that determines potential output. This in much of the recent discussions d·iscussions of view seemed to be prevalent In capital shortage. It seems to me that such an incompatibility can arise only under the assumptfon putty-clay technology, but the assumption of a putty—clay Cl ark and Rasche-Tatom Rasche- Tatom papers identify separate and conceivably conflictconflictClark ing measures of potential pot2ntial labor and capital inputs in a putty—putty putty-putty model. of_.E_o_tenti a 1 Output Projecting Rates of Growth of Potential The second issue-—the issue--the cyclical cyclic,il behavior of actual relative to potenpotential output, while interesting in its ;ts own right and the issue that motivamotiva• ted early research in this area, Is i,; most important in the three three papers as part of the methodology for providing provid·ing evidence onthe on the third issue, the output. trend rate of growth in potential oytput. The basic approach used to isolate the secular trend in the rate of growth of potential output is to time trend purge the actual data of cyclical influences and then to fit a time dummies) to determine ex post rates of growth. When (with appropriate duninies) crLical role, projections of future growth growtti time trend dummies duimiies play such a critical rates become treacherous. Should we vie assume the time tim<, trend relevant to wi 11 be the sane sar1e as that over the most recent homogenhomogenthe next five years will 1il1 earlier period, or be different from st·· ous period, revert to that of a st either? Plausible stories can be told to rationalize each possible choice evi de nee in the th 1: papers often doesn’t dm,sn' t help in making but the econometric evidence the choice. 149 ~p~flc Specific Comments on the Three Papers Perry papers The Clark, Rasche-Tatom and Perry papers are closely interrelated. The Clark paper, for example, follows the approach Perry employed in excep: that Clark Clark introduces earlier papers to measure potential output, except capital explicitly and therefore therc,fore investigates investigatE,s weighted weightr,d factor input productivity rather than lator l ator procuctivity. proc:uctivity. Rasche—Tatom And the Rasche-Tatom capita: as 2,5 an explicit exp"!icit factor input paper introduces energy as vell yell as capital in the analysis. discu,,ses t~erole tre role o'' Finally Perry discusses o~capital and energy in the measurement of potential output and concludes his simpler method remains capable of a more reliable r1oliable estimate of po:ential potentia1 output. The Clark Cl ark and Perry Papers The Clark paper provides evidence thaI: tha1: the level of potential output, if appropriately set in relation to t(J a 4% agqregate aggregate unemployment rate in 1955, now should be defined in relation rel,tion to a,, 4~9%unemployment 4.91! unemployment rate, concon- siders some possible explanations of the unexpectedly unexpectediy poor performance of rang,, of estimates estimates of the growth productivity in 1973-4, and provides a raj~~~ the inability inabi",ity to isolate ·,c.olate the source souq::e of of potential output that reflects i:he behavior of procuctivity in 1974. ;97L_ the unusual behavior Perry simply excludes 1974 and reports small resicuals 1(175 and 1976. 1Cl76. resic.uals in 1973,. 1973, 1975 Clark tries alternate dummies which invclve dowm1ard :,hift in potential output in involve a downward ~hiftin 1974 and a more intense cyclical cyc1ical p,,ttern producti"itY in the 1974—5 1974-5 pattern of producti”ity recession. Clark is unable to choose betweun between the two explanations (although they have very different difforent ·'rnp·,icat mpl icat ons for ·;he the current level of potential output) and the evidence presented in the tile Clark and Perry papers does not permit a definitive evaluation evalua·,:inn of o-f i:lark''.i Clark’s dummies relative to Perry's Perry’s dummy: dummy The approach used 11 ii both the Pery Per·y ,ind Clark papers ~ndClark 1 50 suffers from inadequate attention attention to the selection of potential levels of labor input and provides little solid evidence on which to base projections of future growth rates. rates, I. 1. The full employment unemployment rate. When the CEA designated 1962, they rationalized the choice in 4% as the target unemployment rate in 1962, terms of aa simple {nonexpectational) (nonexpectational) Phillips Curve trade-off model. The trade-off model doesn’t doesn't yield a unique choice for the full employmentemploymentmodel is valid, it at least unemployment rate (Uf) (hf) but if we assume that model represents a well defined methodology for selecting hf. Uf. Since that time, there has been a great deal of additional theoretical and empirical research on the relation between inflation and unemployment, including the development of the natural rate model. This research raised the possibility natural, or of a vertical Phillips Curve and an (in principle) well defined natural, if not so natural natural,, at least unique noninflationary (or nonaccelerating) rate of unemployment. Yet studies like Clark, Rasche-Tatom and Perry’s Perry's continue to key the full employment levelof level of labor ,abor input to the 4% unemployment rate in 1955 and then to translate this via changing changing demographic composition of the labor force to a 4.9% rate in the 1970's. 1970’s. Clark and Perry tell us that this unemployment rate series represents aa constant degree of labor market tightness, but does not bear aa direct relation to the noninflationary rate rate of unemployment and is not the outcome of aa study of inflation dynamics. If their potential unemployment rate should not be confused with the noninflanoninfla- tionary rate of unemployment, should their measure of productive capacity be confused with the traditional concept of potential output? Until a more serious effort is made to introduce a meaningful concept of potential labor input, the level of potential output estimated by such studies must be 151 treated with skepticism. level of Having identified their potential level these labor input with aa constant degree of labor market tightness, these studies may still still be able to yield meaningful insights about cyclical behavior of output relative to employment (e.g. as in Okun's Okun’s Law) and the trend rate of growth in output, but the level of potential output remains arbitrary as long as there is no rationale raticnale (aside from historical continuity) for the specific constant degree cf labor market tightness employed in the analysis. 2. Explaining and forecasting growth rates of potential output. The methodology used to determine growth rates of potential output is ical component (by relating relating to purge the data on actual output of its eye' cyclical participation rates, hours per worker and productivity to cyclical variables} and then to extract the secular trr·nd variables) trend via estimation of aa time trend. Ex post, this method seems capable capab1e of reasonably determining determining historic trend rates of growth grov1th of output. Hovever, changes in the trend Hov·ever, rate of growth can only only be observed ex post and corrected for by time trend dummies. extrapoladun~iiies. And projections of future grov1th growth rates involve extrapolarates~~ con.:'ectures about whether tion of recent historic growth rates pjy~sconcectures recent rates are 1likely i ke ly to continue, the recent slowdown wil will1 worsen, or future rates will rebound. There is typically little direct support for such conjectures in the econometric analyses i:hemse l ves. ~:hemse1ves. 3. The explicit treatment of capital s':ock data in projections of ~ potential potential output. Clark’s paper from earlier studies What differentiates Clark'~ and from Perry’s Perry's paper is Clark’s Clark's explicit treatment of the capital capital stock stock as an input rather than capturing its influence via vi a time ti me trends in the labor productivity equation. My initial reaction reaccion was that this was a 152 reasonable extension, although the whole history of debate over the bafbaffling problem of developing aa meaningful measure of the aggregate capital stock made me aa bit wary. My initial favorable reaction was quickly included--by imposing cooled by the arbitrary way in which capital was included——by fixed weights on labor and capital to form an index of total factor inputs. Perry’s discussion of the difficulty in obtaining statistically Perry's significant coefficients on capital in aggregate production functions Clark’s procedure procedure and reinforces my disdisboth provides a rationale for Clark's trust of the imposition of the arbitrary weights and my concern that problems of measuring capital may indeed make Perry’s Perry's approach more reliable. Clark's Clark’s defense of his treatment of capital seems to be that it does not make a great deal of difference so that using labor input and labor productivity would not have altered his results. I don't don’t know whether this is a defense for usin9 using or not using capital! The Rasche—Tatom Rasche-Tatom Paper The Rasche-Tatom paper employs a methodology which allows for explicit treatment treatment of one of the alleged vi11ains villains in the mystery--the th~ relative price of energy since since 1973. influence of the sharp rise in thç 1973. By including the relative price of energy to proxy the input of energy developin production, Rasche and Tatom find that they can explain recent developments through 1975 without resort to special dummies for the 1973-5 period. Their results also suggest aa lower level of potential output in recent years even without any allowance for such factors as the increasing proportion of capital spending devoted to pollution abatement and the slowdown in the trend rate of growth in productivity after after 1967 or 1969. 1969. 153 substiAn increase in the relative price of energy should induce substitution out of energy into labor and capital thus Increasing increasing the labor and capital requirements per unit of output and decreasing the energy r~qui requirere— ments per unit of output. Both Perry and Clark accept this as a plausible response to a change in the relative price of energy but attempt to dummies for aa given capture it in their productivity equations with either dunnies year or time trend dummies. dinitfes. The issue then is the magnitude of this effect on the level of potential output and the duration of its effect on growth rates. Given the fixed levels of potential labor and capital inputs at the point of change in the relative price of energy, the increase in capital and labor requirements per unit of output translates into a decline in potential output. The greater the substitutability of capital and labor for energy, the greater the resulting decline in potential output. This procedure raises a number of questions: l. 1. The results can be heavily influenced by the form form of the production function and the constrai nts imposed in estimating the coef— constraints coeffiçients. How did the Cobb Douglas form and restrictions imposed on the ficients. parameters influence the sensitivity of potential output to the change in the relative price of energy? I share Perry’s Perry's view that the Rasche-Tatom methodology has led to a serious overestimate of the resulting decline ip in potential output. Yet II do not find their general approach, that of input, including the relative price of energy as a proxy for energy Input, unreasonable. Additional research with less restrictive assumptions about the production function would be useful. 2. No allowance is made for an influence of the relative price “54 154 of energy on either the "potential" 1eve 1 of capacity utilization or “potential” level technical conconthe rate of obsolescence. Both factors are treated as technical, stants rather than choice variables subject to influence by changes in relative prices. Yet one of the most important channels through. through which the sharp increase In in the relative price of energy might be expected to operate is by affecting the optimal rate of replacement investment. The theory or nontheory of replacement investment is an important gap in our conventional macro treatment of investment. Rasche-Tatom paper, a 1975 paper by Myers and In contrast to the Rasche—Tatoni 1/ 11 Nakamura on the effects of energy on productivity explicitly models the Nakamura- influence of a rise in energy priceson prices on the optimal rale ra1;e of replacement. "vintage" (putty clay) model, and their vintage production They use a “vintage” function includes energy along with labor and capital. An increase in the price of energy is shown to effect the condition which determines the shutdown point for aa given vintage. Increased energy prices result in an accelerated obsolescence of existing plant and equipment; the’ the increase in the rate at which old vintages are replaced with new (higher technology) vintages tends to increase labor productivity. On the other capihand, the increased energy prices induce a substitution to labor and capi- tal from from energy so that the amount of labor and investment required per unit of output tends to increase, reducing labor productivity. The net result on productivity is ambiguous and therefore must be determined empi ri ca lly. empirically. Thus the effect of energy on productivity may be more complicomp 1i- cated than the Rasche—Tatom Rasche-Tatom analysis suggests. 155 3. Another questionable feature of the Rasche-Tatom analysis is the failure to allow for aa gradual response pf of production decisions to changes in the relative price of energy. The discussion discussion of their projections of futurç futur~ growth rates, on the other hand, hand, makes referreference to a gradual response to energy developments. This seems to involve conjectures quite unrelated to the explicit econometric analysis of the paper. Summing Up outset——the level Of the three issues we identified at the outset--the level issue, the cyclical issue, and the rate of.growth of, growth issue--the issue——the papers do the best job on the cyclical issue. But the main focus of the papers is on the level and rate of growth issues and their approaches shortcomings, to these issues have serious shortcomings. The level issue is simply unanswered due to failure to identify an appropriate potential level of 'labor labor input. In addition, the effect of the increase in the relarela- tive price of energy on the level of potential output remains ununsettled. Rasche and Tatom have probably overstated the effect, but Clark's Clark’s results also suggest the possibility of a substantial once and for all decline in potential output. As for the projection of future growth rates in three papers, at least in this case, time will permit us to judge whose conjectures were most accurate. Footnotes Footnotes Jj jJ J. Po 11 uti on Effects on 3. G. Myers and L. Nakamura, "Energy “Energy and Pollution Productivity: AA Putty Clay Approach," Approach,” National Bureau of Economic Research (mimeo), 1976. 156 U.S. PRODUCT! PRODUCTIVE VE CAPACITY: A A COI44ENT COMMENT Pham Chi Thanh I find myself in a somewhat unenviable position. The three papers you heard this morning represent three different approaches to the measurement of U.S. productive capacity. The three conclusions conc1 us ions that another--almost to the emerged differ quite significantly from one another——almost point of being contradictory. It is virtually an impossible task as a discussant, therefore, to be nice to all of the authors since if one agrees with one, one will wi 11 probably haye have to disagree with another. The perhaps is to play the role of a Devil’s Devil's Advocate to all easiest way out perhaps three papers, and so I will try to catch the Devil by the ideological tail first. Although the three papers represent three different approaches, it is in the conclusions and their ideological implications that they differ sharply. George Perry is obviously a liberal Keynesian who shows his concern about the high level of unemployment. To say that there is a large gap between actual output a,,d a1,d potential output is to imply that there is room for the government to interfere. Bob Rashe and Jack Tatom are just about at the opposite pole, being conservative. When they show that actual output is only a little below potential and that it will reach its potential maybe within a year, they are trying to sell us the American Dr. Thanh is Chairman of the Department of Economics, The knerican University, Washington, D.C. University, Washington, D.C. 157 idea that no government interference is good interference. “Demand "Demand inflationary," as they management policies are both unnecessary and Inflationary,” put it. It seems quite strange to hear such an idea expressed so Keynes' famous lecture on “The "The End of seriously some fifty years after Keynes’ Laissez Faire.” Faire." Peter Clark, on the other hand, seems to have managed to put himself in the middle. I have the impression that he may be a little bit conservative at heart but, being on the staff of the PresPresident’s ident's Council of Economic Advisers, he cannot very well advocate nonnoninterference by the government. Having speculated about the the three three ideologies, it is is probably probably fair fair ideologies, it Having speculated about to reveal now with whom I side, if only to indicate the line of critique that II am going to deploy. Coming from from a radical center like American University, it would seem to be aa foregone conclusion. I certainly share George Perry's Perry’s concern about high unemployment and his belief that the government might be able to interfere to alleviate it, although II do not agree with his approach to the problem of measurement of potential output. From a pure theoretical point of view, Rashe and Tatom’s Tatom's deapproach is a better one, although there are a number of technical details that I found unsatisfactory. By way of substantiating these remarks, let me begin with some background first. To relate a given level of effective demand to aa unique volume of employment is one of Keynes’ Keynes' famous contributions to short run analysis. This is perfectly legitimate. If the period is sufficiently short, one can reasonably assume that other factors of production are fixed. After all, the structure of social capital, for example, can only change 158 gradually. gradually, Thus Keynes believes that there exists aa level of aggregate effective effective demand that would generate full utilization of the labor force. Keynes was aware, of course, that even in the short run full employment does not imply a zero rate of unemployment since there will always be frictional and/or voluntary unemployment. On the other hand, to relate a given rate of unemployment to aa unique level of aggregate supply is Arthur Okun's Okun’s contribution (now commonly Okun’s La'(I), Law), which commonly referred to as Okun's potential GNP debate. led to this potential Okun's Okun’s Law is about the long-run long—run rather than the short-run short—run -- at least that is how it has been used, with -- some confusion, in the last fifteen years or so. The general consensus seems to be that, even in the long-run, full utilization of the labor force implies aa certain rate of unemployment. stenining from recent revelation. stem~ing This is not a new belief Marx, for example, wrote intensively “reserve army of the unemployed" unamployed’ and, in modern times, one often on the "reserve often hears the term the "natural" “natural’ rate of unemployment from a conservative 1like i ke Milton Friedman as well we 11 as from a liberal 1 i bera 1 like Edmund Phelps. Indeed, no rate of unemployment in the short run can be regarded as "natural" “natural” by anyone in any sense. Almost everyone seems to subscribe to the view that in the long run there is a minimum rate of unemployment that can not be reduced permanently by fiscal or monetary measures. ''benchmark" unemployment rate. ~ This is the so-called so~called The obvious initial confusion was that it was measured by the simple head count, which is appropriate only in the short run. run, This is because unemployment, like sickness, takes its toll in all age groups and in both sexes, sexes. 159 For some reason, certain groups like female and young males, particularly black, are more susceptible to the disease than others. It follows, therefore, if the composition of the labor force changes, the benchmark rate will change accordingly. George Perry’s accordingly. Perry's ingenious device of unemployment weights is well—known well-known and addresses itself directly to the long—run long-run measure of the benchmark rate. In this respect, Peter Clark’s Clark's work on labor force and participation rates also deserves praise. The only thing that I am unhappy about is Is the easy and convenient reference to “cyclical "cyclical variavariations” and "trend “trend break.” tions" break." Of course, they are necessary to suit their econometrics but these do not provide an explanation. To give an example, Perry’s Perry's 1967 trend break happens to coincide with a remarkable phenomenon. Beginning in about 1966, the hourly wage of unskilled and semi-skilled workers falls relatively to the hourly wages of skilled workers. This was also the beginning of a period of expansion by U.S. multinational firms. They started to move their labor-intensive propro- cesses of production abroad where unskilled and semi-skilled labor was cheaper than in the U.S. One can imagine that a great proportion of females and young males, who were new in the ‘labor labor market, belonged to this unskilled and semi-skilled category. To emphasize the point more strongly, one could say that the largest 500 firms in the U.S. can propro- vide the same level of output with significantly less manpower by moving some of their production away from the U.S. and then importing these goods back to the home market. Clark's exclusion of That is why Peter Clark’s "output generated from the rest of the world,” world," which seems perfectly “output reasonable and innocent on the surface, might have an important bearing 160 on domestic output and the actual volume of employment. It would be nice to know how much of the “output "output generated from the rest of the world” world" was actually produced by U.S. multinational corporations. Now, since all ail authors in their revision of the Council of economic Economic Advisers' early estimates of potential GNP have looked upon the benchbenchAdvisers’ mark unemployment rate as a long-run concept, they all naturally extend their analyses to take account of other factors of production production such as capital and energy resource. capEveryone knows that the presence of cap- ital input will pose great difficulty. Even in pure theory, the capital problem has never been quite settled. When one wants to deal headache and I can sympathize with with it econometrically it is a real hea4ache Perry's complaint. Unlike non-durable non-durah1e goods which can be handled George Perry’s with the help of an index, the evaluation of capital involves not only a1so one that involves invo 1ves time itself. itse 1f. AA new a cross-section index but also o'ld airplane as it is from aa new motor airplane is as different from an old spea.king, however, the evaluation of capital is nomore no more car. Strictly speaking, complicated than the evaluation of labor. vionders how are we to Perry wonders since, withwithmeasure the flow of the productive services from a machine sinèe, out incurring any new investment, we can expand its productivity by extending the number of hours it is used. Well, the same applies to aa worker. f1 ow of productive services from a worker can always a1ways be The flow 1anger hours rather than adding an expanded by making him or her work longer extra person to the labor force. Although Peter Clark and Rashe and Tatom recognize the difficulty, strange tthing they brave a try at the capital stock. The s~~ange hing is that they 161 pay so much attention to obtaining a measure of the labor force and unemployment that is better than the simple head count, but did not do the same with respect to capital stock. Just as the simple head count will give a wrong picture of labor input and unemployment, counting machines will give a wrong picture of capital input. Indeed, as far as capital stock is concerned, it is very important to know the composition of social capital, and the average life of plants and capital equipment as well as the rate of accumulation. Any change in these will affect the volume of output produced, and therefore the relationship between output and labor utilization. Knowledge of the rate of accumulation is of particular importance when energy resource is recognized as another factor of production. Even in the theoretical literature the introduction of an exhaustible energy, into the analysis of growth is aa new thing. resource, such as eneryy~ Tatom's effort in incorporating this into their II applaud Rashe and Tatom’s estimate of potential output. Their use of a Cobb-Douglas production function is however unfortunate. While there are good reasons in the theoretical literature for the use of a Cobb-Douglas production function, particular there is no compelling reason for using it here. This partjcular production function is used in the theoretical literature because it is the best weapon to defend growth theory. Two of the special properties of the Cobb-Douglas production function are that (i) output falls to zero whenever one input falls to zero so that every input is essential and (ii) the Average Product and Marginal Product of every input goes infinity as the Input input falls to zero. to Infinity 162 Unless we make such assumptions, there cannot be any economic growth when exhaustible resource is a factor of production If if the stock of resource is finite and If if its average product is unbound unbound only a finite quantity of output can ever be produced. Therefore if growth is to be possible, this must not happen. But that is the theoretical literature, where one is at liberty one's line of defense. to make any assumption to suit one’s For the problem at hand, the use of a Cobb—Douglas Cobb-Douglas functon, besides its usual distorted involves also the implication of Infinite infinite subsubpicture of production, Involves stitution possibility between energy and other input. If the price of one Input input rises relative to another, we substitute. Well, the concept of substitution at the macro level in economics is a strange one. It misunderis like a beautiful woman, always being loved and always being misunder- stood. Substitution in the macroeconomic sense does not mean input can always be ininedlately,transformed. immediately transformed. AA drastic fall in wage rate today does not mean that each Greyhound bus from Chicago to St. Louis will now be driven by ten drivers. A drastic rise in the price of fuel does not mean that the bus will be pushed liy 1w drivers to St. Louis all the way from Chicago. Substitution in the th" macro sense must be understood as scrapping old equipment and replacing it with new, accompanied by aa 1abor intensity and/or different fuel consumption. different labor Very high fuel wiil lead to the production of new cars which consume less gas. fuel cost will In other words, it takes time and needs the help of technical progress. Tatom make the In using a Cobb--Douglas Cobb—Douglas production functon, Rashe and Tatoni same specification error as the followers of Solow made more thap than a decade ago. The elasticities a, 8Band and yyin in their paper are the 163 "surrogate" elasticities. They do not tell us about the so-called “surrogate” information about the production relative shares nor do they give any Information process. Therefore, the calculation of potential output based on such is quite futile. Th.eir estimate of demand for energy is a function function is~quite futile. Their based on the assumption that energy will always be used, up to the point where the value of its marginal product is equated to its price. Beside the implication of Instant instant substitution that I mentioned earlier, this assumption also carries another extraordinary implication. Their estimated value of the output elasticity for energy is 94% 9.4% which implies that the share of energy factor in the U.S. national income is almost 10%! 1O%t Earlier, I said that the knowledge of the rate of saving is of particular importance when energy resource is recognized as another factor of production. If income is to grow steadily, then the rate of saving affects not only the rate of growth but also the level of output produced. Tatom's calculation of growth rate of 3% is Rasche and Tatom’s therefore more of a wild guess than aa meaningful estimate. In general, I think it is possible to estimate potential output, which corresponds to given levels of input utilization. But to do so, one needs to be able to estimate capital stock and energy resource utilization correctly. The three papers do not give satisfactory calcalculations in this respect. From a more general viewpoint, I also think the debate misses aa fundamental point. Everyone obviously likes to see a lower rate of unemployment and it may be possible to reduce it to aa long run minimum rate by expanding output. But this is a shortsighted 164 view. Faster growth rate or higher production also means using up exhaustit)le resources resourcr::, at a faster rate. rate., exhaustible i:Jrong choice may mean mr~an drying dr~ring Af\ wrong up resources at a point in -in time time that we cannot afford to tn do so. so~ The prob1 em is not a simple s imp·i e trade-off like the inflation-unemployment problem trade-off of the good old days of the Phillips’ Phillips' curve. It is more like 1 ike a trade-off between betwi!en 1low ow unemployment unenip I oyment now and high unemployment later 1ater or relative abundance abundo.nce now and poverty later. later" The choice involves future generations gcneratioris yet to be born, and crnd therefore, tht:reforeo\ no r.o social socia"I contract contra.ct is possible. I think that unless we can address ourselves to this prob— len, the estimation of potential GNP or benchmark unemployment rate does not have much meaning beyond an intellectual exercise. 165