The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
SUMMER 1982 Opinions expressed in the Economic Review do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco, or of the Board of Governors of the Federal Reserve System. The Federal Reserve Bank of San Francisco’s Economic Review is published quarterly by the Bank’s Research and Public Information Department under the supervision of Michael W. Keran, Senior Vice President. The publication is edited by Gregory J. Tong, with the assistance of Karen Rusk (editorial) and William Rosenthal (graphics). For free copies of this and other Federal Reserve publications, write or phone the Public Information Department, Federal Reserve Bank of San Francisco, P.O. Box 7702, San Francisco, California 94120. Phone (415)974-3234. 2 Time and Economics I. II. Introduction and Summary 4 On Federal Deficits and their Economic Impact 6 Rose McElhattan ... We have found that deficits have significant real output effects when we measure changes in their components, high employment expenditures and receipts, relative to their past average rates of change. I ll. Consumption and the Term Structure of Interest Rates and Inflation: An International Comparison 18 Joseph Bisignano ... The aim of this paper was to determine whether real consumption was influenced by the term structure of interest rates in a model which captured the influence of permanent income and the impact of price level movements. The answer was found generally to be in the affirmative. Editorial Committee: Charles Pigott, Brian Motley, Roger Craine 3 In economics, timing is often all-important. The articles in this Review demonstrate that economic decisions made today, and their outcomes, can depend crucially upon what happens, and can be expected to happen, in the future. In the first article, Rose McElhattan discusses the duration of the impacts of fiscal· budget deficits on prices, inflation, and real output. She distinguishes between the long-run and the short-run effects of deficits caused by spending and tax policies. McElhattan notes that, by conventional reasoning, federal deficits are likely to have short-run stimulative effects on aggregate demand but that, in the long-run, deficits occurring without an accommodating change in the money supply can have one of three impacts: (I) they may crowd out an equal amount of private sector spending and leave the nation's productive capacity unchanged; (2) they may crowd out a greater amount of private spending because the government expenditures replacing private investment are less productive; (3) they may increase the nation's productive capacity if they have positive supply-side effects. McElhattan explains the concept of crowding out as the result of "additional demand related to the deficit which places upward pressures on market prices and interest rates which then reduce some private sector interest-sensitive spending." To test the impact of federal spending and tax policies upon real GNP and inflation, McElhattan used reduced form equations in which changes in real GNP and inflation are functions mainly of changes in money growth and changes in real federal high employment expenditures and tax revenues. The sample period used to estimate the equations range from the second quarter of 1966 to the fourth quarter of 1979. McElhattan found significant real output effects when deficits were measured by their components -high employment expenditures and tax receipts -relative to the components' past average rates of change. She suggests that these measures may serve "as crude estimates of unanticipated changes in the deficit." Anticipated changes, she notes, can be immediately incorporated by the public into higher market prices and interest rates and have but little effect on real output. The results of McElhattan's study point to the existence of temporary output and inflation effects and lasting price level changes. An increase in tax revenues or real expenditures steadily raises the level and the rate of growth of real GNP. After I to 3 years, the effects die out leaving no traces after another 2 to 3 years. McElhattan cites the lack of a long-run output effect as evidence of complete crowding out. A sustained tax cut, or an increase in real government spending, also temporarily raises the inflation rate. The rate change peaks in 2 to 3 years and disappears at the end of 5 years. Price levels, however, remain at a higher level. A I percent tax cut leaves the price level 1.2 percentage points higher than its initial level, while a I percent increase in real expenditures leaves the price level .4 percentage points higher. In summary, McElhattan writes: "The real GNP response to changes in federal deficits appears to be transitory, and in the long-run, changes in the deficit appear to crowd out about an equal amount of some private sector spending ... Changes in deficits also appear to change the rate of inflation in the short-run and the price level in the longer run ... " The second article, by Joseph Bisignano, extends the simplest consumption model in which an individual decides between present and future consumption on the basis of the "price" at which a dollar of consumption today can be transformed into a dollar of consumption in some future period. The price is the real interest rate relevant to the two periods but the exact effect on present consumption is ambiguous because of opposing income and substitution effects. Bisignano's article extends this 4 simple model beyond two periods by using the entire term structure of real interest rates-because consumption depends "on the entire structure of interest rates over the horizon of potential future consumption.' , Bisignano's analysis uses the framework of the Life Cycle-Permanent Income model in which "individuals make consumption decisions about what individual goods to purchase in a temporal framework" subject to wealth or permanent income constraints. He points out that while economists have recognized that consumption decisions depend on the term structure of real interest rates, they have seldom used the term structure in their studies. His analysis therefore attempts to isolate the effects on consumption of permanent income, term structures of interest rates and inflation, and price levels in four countries: the United States, the United Kingdom, Canada, and the West German Federal Republic. Changes in price levels, Bisignano notes, can create a "money illusion" of short-run changes in relative prices. Inflation affects consumption in three ways: by changing real taxes if the tax system is not indexed to inflation, by changing real liabilities and therefore real net wealth, and, when future inflation is uncertain, by reducing current consumption in risk averse individuals. The results of his analysis show that, in general, in the context of a permanent income consumption model, the term structure of real interest rates plays a significant role in consumption behavior. He found that a rising term structure of real interest rates decreases current consumption in the cases of the United States, the United Kingdom, and Canada and increases current consumption in the case of Germany. Holding real interest rates constant, he found that a rise in inflation "appears to increase real consumption in the U. S. and Canada and to reduce consumption in the U.K. and W. Germany." These findings, Bisignano notes, have implications for the national economy. According to Bisignano, recovery from the recession of 1982 will likely be led by improvements in the consumption sector, with the response of that sector to future expectations embodied in the term structure of real rates determining the strength and duration of the recovery. Based on his findings, he concludes that the quantitative effect of the major decline in real interest rates from June to August 1982 should outweigh the effect of the decrease in inflation and "should provide a major boost to consumption and to the general economic recovery." 5 Rose McElhattan* There is considerable concern these days about the prospect of high and perhaps rising federal deficits and their effects on intlation, private investment and other aspects of our economic lives. In this paper, we will review some considerations regarding the economic impact of deficits and provide some empirical results derived from reduced form equations for real GNP and inflation which we have estimated. The best known reduced fonn estimates of fiscal policy effects are those from the St. Louis equation. In that equation, the impact of fiscal actions upon nominal GNP, when unaccompanied by changes in the money supply, essentially disappear within four to five quarters. A recent update by the Federal Reserve Bank of St. Louis reaffirms these results and also finds that fiscal variables exert no statistically significant effect in any quarter. I Thus, fiscal policy has no significant short or long-run effect upon nominal GNP according to these results. Our question is whether the seeming ineffective- ness offiscal policy will hold up in an analysis of the impact of fiscal policy upon the components of nominal GNP, that is, real GNP and inflation considered separately. Eve!} should nominal GNP be unaffected by changes in deficits, there may be considerable movements in real output and inflation. Changes in these measures concern policy makers and form the focus of theoretical discussion. The results of our analysis suggest that deficits do have significant effects upon real output and inflation. Deficits, when measured relative to their past average values, appear temporarily to stimulate real output, although they have no permanent real impact. This finding suggests that federal deficits ultimately crowd out some private spending. Increases in deficits also seem to raise inflation in the short run, and the price level in the long run, as convention theory suggests. In our analysis, we measure deficits by their high employment estimates, and consider separately the leverage of expenditures and revenues. I. Theoretical Considerations Economists are divided as to whether fiscal actions matter in the determination of aggregate output. Some argue along convention lines that federal deficits, which are the result of changes in spending or tax policies, or some combination of these, will lead to changes in aggregate demand at least in the near future. Many believe, however, that deficits will have little or no impact upon aggregate output in the long-run. Still others contend that deficits which are matched with increases in the national debt do not increase the nation's net wealth and therefore will not lead to any sizable changes in aggregate output in the short or longer run. *Senior Economist, Federal Reserve Bank of San Francisco. Eileen Dixon provided research assistance. 6 II. Conventional View The model from which the conventional argument stems is the aggregate demand-aggregate supply paradigm discussed in standard economic textbooks and the foundation of several forecasting models of the U.S. economy.2 Along the supply schedule, greater quantities of output are supplied only at higher prices. This is because the greater quantities of labor and other resources needed to increase output places upward pressure on unit production costs. The aggregate supply schedule is derived from the assumption of constant money wages and other factor prices and a fixed stock of business plant and equipment. According to conventional models, money wages will increase with tighter labor market conditions and with expectations of future increases in inflation. Any increase in money wages will shift the entire supply schedule upward and to the left so that higher prices are associated with the original supplies of output. Alternatively, an increase in the capital stock will shift the supply curve in the opposite direction, indicating lower prices for the original quantities. The aggregate demand schedule indicates that total demand for goods and services will be greater at lower price levels. This occurs because lower prices tend to ease monetary conditions and hence to lower interest rates which in tum stimulate interest-sensitive spending. Similarly, higher prices tend to tighten monetary conditions, raise interest rates and depress aggregate demand. The demand schedule is derived from the assumption of constant monetary and fiscal policy. A change in either wili shift the entire schedule. To clarify terms, a constant monetary policy means the maintenance of a given level (or rates of change) in the supply of money as measured in this paper by MI (currency and checkable deposts). A constant fiscal policy means no change in federal government expenditure programs or the federal tax structure. An increase in federal spending or a reduction in taxes, or any combination of these that increases the federal deficit shifts the entire demand schedule upward and to the right. This means that greater quantities of output will be demanded at the original prices. A decrease in the deficit is associated with a downward shift in demand and consequently with smaller quantities of output demanded at the original prices. The same can be said for changes in monetary policy: an increase (decrease) in the money supply is associated with more (less) output demanded at the original prices. III. Deficit Impact Second, the increased demand for output increases the need for greater quantities of money balances to cover the increased volume of transactions. With a constant monetary policy, this increased demand for money balances will tend to tighten monetary conditions and raise (real) interest rates. These higher interest costs, in tum, will limit interestsensitive spending and the expansionary impact of the original deficit. Real GNP, then, will tend to increase after an increase in the federal deficit, but the stimulative effect will be partly mitigated by crowding out of some private spending through higher prices and interest rates. This behavior is illustrated in part in Figure 1. The increase in the deficit raises aggregate demand from DI to D2. At the original equilibrium level of prices, PI, more output (Yl) will be demanded. Given the aggregate demand and supply schedules, we may now analyze the effect of an increase in the federal deficit, assuming no accommodating change in the money supply, i.e., holding monetary policy constant. This assumption is equivalent to financing the increase in the deficit through the sale of bonds. The increased deficit (due to increased expenditures, reduced taxes or some combination) shifts the entire demand schedule upward and to the right. Although demand rises, two important effects limit the stimulative power of the deficit. First, the increased demand tends to drive prices up, since greater quantities of output will be supplied only at higher prices according to the supply schedule. Consequently, part of the expansionary effect of fiscal spending will be dissipated in higher prices. 7 and demand. An equal increase in both would leave the price level the same as the original equilibrium level while greater (smaller) increases in supply relative to demand would lead to lower (higher) prices than initially. However, positive supply-side effects are not a certain outcome. For instance, according to the theory of labor supply, while a lower tax rate may encourage some people to enter the work force (an income effect), others may be encouraged by the higher take-home pay to spend less time at work and more at leisure (substitution effect). As a second case, we consider that the fiscal programs underlying the deficit may have no net effect on aggregate supply in the long-run. In this second case, complete crowding out of some private ,spending by federal deficits occurs. The increased level of output, (Y2) in the previous chart, leads to increased demands for labor, tighter labor market conditions, higher money wages and higher prices. Furthermore, the higher prices tend to increase expectations of future prices. These increased wage and price expectations produce an upward shift in the supply schedule and a further increase in final product prices. Higher prices, in tum, reduce real money balances, tightening money market conditions, and leading to further increases in interest rates. As a result, real GNP is reduced further as additional private sector spending is crowded out of markets by the combined effects of continuing price and interest rate increases. Economic adjustments in prices, interest rates, wages and price expectations will continue until the initial increase in the deficit displaces an equal amount of some private sector spending. Ultimately, real output equals the level that would have existed without the change in fiscal policy because, as long as output is greater than it otherwise would have been, employment will be higher and so will money wages and prices. These higher prices, given an unchanged monetary policy, will lead to higher real interest rates which reduce interestsensitive spending. Only when aggregate output equals the level it would have been had the deficit not occurred will we observe no further adjustments in employment, wages and prices. Although complete crowding out occurs in real GNP, it will not occur in nominal GNP. Some permanent effect will be left on the level of nominal Figure 1 Transitory Effect of a Deficit Increase ... more output but at higher prices Prices P, P, Y, Y, Output But, with an upward sloping supply curve, a new short-run equilibrium occurs at E2. As a result, the increase in the deficit leads to higher prices (P2) and more output (Y2) than before the change in the deficit. However, output rises less than would have occurred if no change in prices had taken place (Yl). This new equilibrium is likely to be short-lived because the changed prices (P2) and output (Y2) will set off a chain of reactions in factor costs and investment that lead to further shifts in aggregate demand and supply and which will culminate in any one of three final cases. In the first case, deficits have positive supply-side effects increasing the economy's productive capacity. Public expenditures, for instance, may finance projects such as road construction, rural electrification, research and development which, on balance, may be more productive than equal sums spent by the private sector. Alternatively, deficits may result from a reduction in income taxes which may increase incentives to enter the work force and, on balance, increase the supply of labor and again, the nation's productive resources. In these situations, the supply curve is shifted downward and to the right. Ultimately, more output will be produced, and the change in the price level will be determined by the size of the relative shifts in long-run supply 8 deficits will raise the price level, but have an ambiguous impact on nominal GNP. From these alternatives-positive supply side effects, crowding out and super-crowding out, we observe that the ultimate effect of a deficit will depend materially upon the particular types of spending and/or tax policies associated with the deficit. Deficits due to corporate tax reductions, for instance, are likely to have a more stimulative impact on aggregate supply than an equal-size deficit that results from aid to state and local govemments who then incur surpluses in their own budgets. In partial recognition of this, we separate the effects of spending and tax policies in our estimations. GNP because the crowding out process is achieved largely through a rise in prices and interest rates which increases velocity and money income. 3 In the final case, deficits may ultimately lead to a decline in aggregate supply and an increase in prices-the case of super-crowing out. This occurs when the reduction in private sector spending which the deficit crowds out of markets is replaced with government spending that is less productive. The increased interest costs associated with the initial stimulative impact of deficits may .reduce private spending on productive capital and lead to a decline in the nation's productive resources. In this case, IV. Some Further Considerations In formulations of the conventional model, individuals are assumed to change their expectations of future prices on the basis of past price behavior. Peo'ple, however, may be wiser and use all relevant information to form their expectations, not just the past behavior of inflation. Their rational expectations open further possibilities with regard to fiscal policy effectiveness in the short run. For example, anticipated deficits may quickly crowd out private spending, and have little if any impact upon real GNP in the short run. This occurs because individuals, applying their knowledge of the conventional model, anticipate the initial stimulative effects of an increased deficit. These anticipations are quickly incorporated in market prices and interest rates. The existence of rational expectations suggests that we consider the different leverages of anticipated and unanticipated fiscal deficits. 4 Another case in which federal deficits may have little real stimulative impact in the short run is one in which the public does not regard the increase in the national debt associated with the deficit as an increase in its net wealth. According to this view people will anticipate the compensating future taxes implied by their holdings of government debt. The important implication of this view is that federal deficits which are bond-financed will have the saIne impact upon aggregate demand as a balanced budget. Consider a tax cut and the corresponding deficit which is financed by selling bonds to the public. Each dollar of tax reduction is matched with an extra dollar of federal debt that will be repaid and serviced with future taxes with the same present value. Since there is no change in the present value of net tax liabilities, there will be no change in household permanent income, net wealth or spending. The same type of reasoning applies to federal deficits which are the result of increased government spending which is bond-financed. To the extent that bonds are not net wealth, fiscal policy will have little if any significant impact upon either real or nominal GNP in the short run. 5 V. Summary According to conventional reasoning, federal deficits are likely to have short-run stimulative effects on aggregate demand, although the size of these effects is an empirical issue. Ultimately, in the long run, deficits may (a) crowd out an equal amount of private sector spending, (b) crowd out a greater amount of private spending because deficits lower the nation's capital stock, or (c) increase the nation's productive capacity as in the case of federal investment in the economy's infrastructure. Fiscal actions may have little, if any, leverage in the near or more distant future when individuals equate federal deficits with future tax liabilities, or anticipate future price and interest rate changes caused by deficits and change market prices quickly to conform to their expectations. 9 VI. Estimation Model and Data (dHR). In equation (lB) changes in the rate of growth of real GNP occur when monetary or fiscal policies deviate from their anticipated courses. The dHEBAR and dHRBAR represent changes in real high employment expenditures and receipts relative to their respective anticipated rates of change. Equation (2) states that changes in money growth determine changes in inflation, reflecting earlier considerations that both anticipated and unanticipated monetary changes will have an impact upon inflation. This equation also allows changes in the growth of real high employment expenditures (dHE) and receipts (dHR) to determine changes in the inflation rate. It incorporates the assumption that both anticipated and unanticipated fiscal policies may affect the rate of inflation. The monetary variable is M I, currency in the hands of the public plus checkable deposits. To approximate unanticipated monetary changes, we estimated dMBAR as the current rate of change in Ml less its average rate over the past two years, measured in logs. The high employment budget estimates are provided by the U. S. Commerce Department. 7 The measures we use here are adjusted to exclude the automatic effects of inflation on revenues and expenditures. There are several reasons why revenues and expenditures respond automatically to price changes. For example, federal interest payments depend on interest rates which tend to change with What has been the impact of federal spending and tax policies upon real GNP and inflation? To answer this question, we tum to reduced form equations in which changes in real GNP and inflation are determined mainly by changes in money growth and changes in real federal high employment expenditures and tax revenues. These reduced form equations may be derived from a model which was detailed in a previous paper. 6 Underlying this model is an aggregate demand-supply framework similar to that discussed earlier, with the additional assumption that anticipations of future monetary and fiscal policies are determined by the past history of these policies. The real economy is assumed to be stable in the sense of growing at a steady pace; it will depart from this pace temporarily when either monetary and/or fiscal policies deviate from their anticipated patterns. The economy is also assumed to have the classic long-run neutrality property associated with money growth. This means that a permanent change in the rate of growth of money ultimately results in an equal change in the rate of inflation, but in no change in the rate of growth or level of real GNP. Also, in the long-run, the leve of prices will change with the stock of money so that the public's holdings of real money balances are at desired levels, given interest rates and the level of income. The reduced-form equations (lB) and (2) below follow directly from this model. Equation (lA) is a more general depiction of how fiscal policy might affect real growth for it makes no distinction between anticipated and unanticipated fiscal values. dRGNP = F(c, dMBAR, dHE, dHR) (lA) dRGNP = F(c, dMBAR, dHEBAR, dHRBAR)(lB) dP = F(c, dM, dHE, dHR) (2) Equation (IA) states that the rate of change in real GNP (dRGNP) deviates from a constant rate, c, when money growth differs from its anticipated rate, (dMBAR). In addition, the rate of growth in real GNP will change with changes in real high employment federal expenditures (dHE) and changes in real high employment federal receipts 10 changes in the rate of inflation. If not corrected for such inflation-induced components, the high employment measures would provide biased estimates of the economic effects of fiscal policy. To approxi- mate changes in unanticipated real high employment expenditures (dHEBAR) and receipts (dHRBAR), we measured the current rate of change in each variable less its average rate of change over the past two years, measured in logs. VII. Empirical Results The estimation of equation lA, in which changes in real GNP are determined by changes in money from its past average (dMBAR) and by changes in real high employment expenditures (dHE) and tax revenues (dHR), indicated that neither of the fiscal variables contributed significantly to the determination of real GNP. The results are summarized n Table I. This should be interpreted with some caution, however, since there is reason to believe that the monetary and fiscal policy measures were closely related during the sample period. For instance, we estimated changes in real GNP with the monetary measures alone. These were statistically significant, and adding fiscal policy measures did not matter in the determination of real GNP. Conversely, when estimating changes in real GNP with the fiscal measures alone, they were also significant and adding monetary measures did not matter. Each policy variable did about as well as the other in II forecasting changes with real GNP both in and outside the sample. These results, along with the estimation in Table I, suggest that one type of policy served as a fairly good proxy for the other. The close relationship among the variables implied by these results presents an estimation problem in that their covariance prevents us from obtaining a very precise estimate of any independent effects. In the case of monetary measures, this was not as important and the results were still statistically significant. But the case of fiscal policy is, at best, inconclusive. We may approach this estimation problem by respecifying the fiscal policy measures. Fiscal variables may have significant real output effects when fiscal policy deviate~ from anticipated patterns, as suggested earlier. This appears to be the case in the results shown in Table 2. Both fiscal measures are statistically significant (revenues at the 5 percent level and expenditures at the 10 percent level) and, together with money, account for 39 percent of the variation in real GNP growth rates. According to these results, fiscal policy has an independent and temporary influence upon real GNP. 12 To illustrate this, Chart 1A shows the response of the real GNP growth rate to a I percentage point reduction in tax revenues in the initial quarter only. This is equivalent to a permanent cut beginning in 1980 in constant dollar tax revenues of approximately $3.2 billion. This tax cut steadily increases the rate of growth of real GNP from the initial quarter to a maximum response of .3 percentage points (annual rates) by the end of the first year. Thereafter, the effect steadily dies out, following a cyclical pattern, and ceases to have any impact within 5 to 6 years. The tax multipliers derived from these estimates are illustrated in Chart 2A. The level of real GNP steadily increases for about two and a half years. By that time, each billion dollars of the tax cut has increased the level of real GNP by $2.4 billion. Thereafter the impact upon the level of GNP dies out. Within 5 to 6 years, the additional deficit resulting from the tax cut has completely crowded out some types of private expenditures not favored by the tax reduction. Chart 1 Response of Annual Rates of Growth in Real GNP and Prices to a One Percentage Point Reduction in Real Tax Revenues in the Initial Period Only Percent Chart 2 (A) Response of the Level of Real GNP and Prices to a One Percent;:ige Point Reduction in Real Tax Revenues in the Initial Period Only .40 .30 (A) .20 Billions of Constant Dollars .10 2.5 0.0 I----~r-------~ 2.0 -.10 1.5 -.20 1.0 -.30&...1..; o ~ 5 10 15 ........J 20 Real GNP .. Tax Multipliers 0.5 25 Periods following Initial Change 0.0 &...I..; o ~...... """- 5 10 15 20 25 Periods following Initial Change Percent (8) .40 (8) Percent .30 1.4 .20 1.0 .10 0.5 0.0 o ......" 5 10 O.OIllll;" .... 15 20 o 25 Periods following Initial Change ........J ...o..l. 5 10 15 20 25 Periods following Initial Change 13 out an equivalent amount of some private spending within 5 to 6 years. Table. 3 provides the inflation equation estimates. Both. fiscal. measures are statistically significant in these estimates. Chart 1B illustrates the impact upon the rate of inflation ofa sustained decrease in real tax revenues of $3.2 billion. The inflation rate steadily increases and reaches its maximum of .34 percentage points (annual rate) about 2 years after the initial tax cut. Thereafter, the inflationary response dies out, with no further impact after Charts 3A and 4A illustrate the effects on real GNP of a sustained $3.2 billion increase in federal high employment real expenditures. (Again, this is the equivalent of a 1 percentage point increase in these expenditures in the initial quarter only.) The real output response is somewhat less inthis case than. in the case of tax cuts.. The real expenditure multiplier reaches a maximum value of $1. 7 billion about 3 years after the initial change, relative to a maximum of $2.4 billion in the former case. The increase in federal expenditures eventually crowds Chart 3 Chart 4 Response of Annual Rates of Growth in Real GNP and Prices to a One Percentage Point Increase in Real Expenditures in the Initial Period Only Percent Response ofthe Level in Real GNP and Prices to a One Percentage Point Increase in Real Federal Expenditures in the Initial Period Only (A) Billions of Constant Dollars .20 (A) 2.0 .10 1.5 0.0 r------~-----, 1.0 -.10 0.5 -.20 O.OIl..l-............................_ _......._ ..............i:IIiIooI 5 10 15 o 20 25 5 10 15 20 Periods following Initial Change 25 Periods following Initial Change Percent Percent (8) .30 .30 .20 Price Level ~ .20 .10 .10 0.0 0.01----':1------- -.10 -.10 -.20 -.30 (8) .40 -.20 ....._ o 0 _...&.lo_....._ ................ ...... 5 10 15 20 25 Periods following Initial Change 10 Periods following Initial Change 14 Chart 5 approximately 5 years. However, the sustained tax cut does have a permanent effect upon the price level, as shown in Chart 2B. The tax cut increases the price level by 1.2 percentage points from its initial level. Together, Charts 2A and 2B illustrate the croWding out features of the initial federal deficit. The tax cut first increases aggregate output and prices. These price increases, in tum, act as a major vehicle by which other types of spending not favored by the tax cut are crowded out of markets. Eventually, the initial deficit crowds out an equal ameunt of some private sector spending and has no long-run effect on the level of real GNP. It does, however, have a permanent effect on the price level, and consequently, on nominal GNP. The results for nominal GNP are shown in Chart 5. The inflationary impact of a $3.2 billion increase in real expenditures is shown in Chart 3B. At first, the inflationary response is negative, but it steadily increases to a maximum of .2 percentage points within approximately 3 years of the initial spending increase. Eventually, the inflation rate response dies out, leaving the price level .4 percentage points higher, as shown in Chart 4B. Together, Charts 4A and 4B illustrate that the increased deficit crowds out some private spending, but will permanently raise the price level. Response of Nominal GNP to a One Percentage Point Reduction in Real Tax Revenues in the Initial Period Only Percent (A) Annual! Rates of Change .40 .30 .20 .10 0.0 I------~r__---_ -.10 -.20 o .... 5 10 15 20 25 Periods following Initial Change Percent (B) 1.5 1.0 0.5 0.0 0 5 10 15 20 25 Periods following Initial Change 15 VIII. Summary andiConclusions Economists are divided regarding . whether changes in the federal deficit willproduceany.significant effects upon the total level of real GNP and prices. According to conventional reasoning,. the impact of deficits in the long-run dependsuponthe particular expenditure and tax. programs which make up the deficit. Certain govemmenttaxand expenditure. policies may induce positive supply side effects through incentives which incr~ase the supply of labor, for example. Other polici s ay l11 transfer the nation's resources away from productive private spending towards less efficient government expenditures and reduce the nation's productive capacity. This would be a case ofsl.lpercrowding out. Another case is that of complete crowding out, when deficits replace private sector spending. In any case, an increase in federal deficits initially is likely to stimulate aggregate demand. Thereafter, if some degree of crowding out takes place, it occurs as the additional demand related to the deficit places upward pressures on market prices and interest rates which, then, reduce some private sector interest-sensitive spending. We must make a further distinction, between anticipated and unanticipated deficits. An anticipated deficit may quickly crowd out private spending as the public incorporates the stimulatory effects into higher market prices and interest rates. Anticipated deficit changes, then, may have little effect on real output even though they raise prices. Unanticipated deficits, on the other hand, may have dynamic effects on real output and inflation as explained in the conventional model. \Vehave found that deficits have significant real output effects when wcITleasure changes in its components, high employment expenditures and receipt~, r~lative totheir past averag~ratesofshange. Thes~measures.111ayserve .as ciUde estimates of unanticipated changes ~in. the deficit. The real GNP response, then, to changes in federal deficits appears. to be. transitory, \ and in the long-run, changes in the deficit appear to crowd out about an equal amount ofs?me private sector spending. Changes in deficits also appear to change the rate of inflation in the shortrunand the. price levelin the lon~er run,according to our estimates.These results are consistent with the conventional view that federal crowding out of private sector spending achieved in part through higher prices. A few final statements are probably in order. It should be clear that the deficits we are addressing in both theoretical discussions and empirical w?rk are those best measured at high employment because these are adjusted for automatic changes in the deficit due to changes in business conditions. Consequently, the high employment measures, more accurately than their corresponding actual deficit measures, depict fiscal policy initiatives which can alter market conditions. Lastly, we regard our results as tentative, at best a depiction of experience during the sample period from 1966.2-1979.4. The exact timing and size of fiscal economic effects remains uncertain subject to a fairly wide range of estimation error, probably in part because historical movements in fiscal measures have been closely related to monetary growth. This correlation may reflect responses ofthe respective policy makers to unfolding economics developments of the past. r 16 FOOTNOTES 1. SeeHafer, R.W, "The Role of Fiscal Policy in the St. Louis Equation," FRB of St. Louis Review, vol. 64, no. 1, Jan. 1982. 5. For a discussion olthe theoretical and empirical results, see Barro, Robert J., "Are Government Bonds Net Wealth?" Journal of Political Economy, Dec. 1974; Arak, Marcelie, Are Tax Cuts Stimulatory?, The Review of Economics and Statistics, Feb. 1982, and Feldstein, Martin, GovElrnment Deficits and Aggregate Demand, Journal of Monetary Economics (9),1982. 2. As examples of the standard model, see Rudiger Dornbusch and Stanley Fischer, Macro-Economics, McGraw Hill Book CompanY,1978. 3. For further discussion of how fiscal actions completely crowd out real spending but have a permanent effect on the level of nominal spending see Franco Modigliani and Albert Ando, Impact of Fiscal Actions on Aggegate Income and the Monetarist Controvery: Theory and Evidence, Monetarism, editor Jerome Stein. 6. McElhattan, Rose, The Response of Rea! Output and Inflation to Monetary Policy, FRB San Francisco Economic Review, Summer 1981. 7. Frank de Leeuw and Thomas M. Holioway, "The HighEmployment Budget: Revised Estimate and Automatic Inflation of Effects, Survey of Current Business, UnitEld States Department of Commerce, Bureau of Economic Analysis, April 1982, volume 62, NO.4. For empirical results from a large econometric model see, Probyn, Christopher, Properties of the 1981C Edition of the DRI Macro Model, Data Resources U.S. Review, September 1981,especially pages 1.34-1.39. 8. For details of this method see Appendix A in McElhattan, Rose, op.cit. 4. Hall presents a model assuming rational expectations and price stickiness. Hall, Robert E., The Macroeconomic Impact of Changes in Income Taxes in the Short and Medium Runs, Journal of Political Economy, vol. 86, no. 2, April 1978. 17 Joseph Bisignano* and West German Federal Republic. The plan of this paper is, first, to discuss the life cycle-permanent income model of consumption and the role of interest rates, irl particular the spectrum of interest rates over all maturities, in influencing real consumption behavior. Second, several related theoretical issues are discussed, such as the concept of "money illusion" and the alternative ways a change in interest rates may influence consumption. Third, the form of the consumption equation to be empirically estimated is considered, followed by an analysis of related data measurement problems. Fourth, we discuss the estimated consumption equations for the U.S., U.K., Canada and Germany and, fifth, we conclude with the implications of the empirical findings, given the recent behavior of interest rates and inflation in the U. S. (Chart I). This paper seeks to provide an answer to the question of how interest rates affect the level of real consumption. The framework used to solve this problem is one in which consumers make consumption plans using information about the entire term structure of interest rates. In theory, real consumption decisions are determined by relative prices and real interest rates, the latter defined as the nominal rate less the anticipated rate of inflation. Both determinants will be tested in the following analysis, the former, by testing whether the level of prices influences consumption, and the latter, by attempting to separate the effects on consumption of the term structure of nominal interest rates from the effects on consumption of the term structure of inflation. The same empirical analysis is undertaken for four countries: United States, United Kingdom, Canada I. The life Cycle-Permanent Income Model The life cycle-permanent income model of consumption characterizes the consumption behavior of individuals as maximizing the utility or satisfaction derived from consumption over their entire lifetimes. The "constraint" on this implicit maximization we all perform in planning current consumption is not simply current income but current and all expected future income. So-called "lifecycle" models emphasize that aggregate wealth, defined as current wealth plus discounted expected income from labor and non-labor sources, is the variable which determines the "scale" of current consumption. Since an estimate of aggregate wealth involves an estimate of discounted future income, the models implicitly require some consideration of current and future interest rates. Thus, the term structure of interest rates is important for its influence on both the determination of the consumption allocation decision and the calculation of the wealth/income constraint. As total wealth may be considered the "stock" constraint on the total lifetime consumption decision, the "flow constraint," that is, period-byperiod, may be considered to be "permanent income. " The concept of permanent income rests on the argument that consumers "smooth" the estimates of their income and base their consumption decisions on the non-transitory components of income. The basic idea to keep in mind, nonetheless, is the stock/flow distinction between wealth and permanent income. Both are constructed concepts to determine the lifetime and period-by-period constraints on the consumption decision. *Yice President and Associate Director of Research, Federal Reserve Bank of San Francisco. Margaret Saunders provided research assistance. My thanks to Professors Roger Craine, Milton Friedman, Kevin Hoover, Thomas Mayer, Jurg Niehans and John Seater for comments on earlier drafts of this paper. 18 In short, the consumer is pictured as maximizing his utility of consumption over his lifetime. The constraint on this exercise is his total wealth. Permanent income is then some fraction of total wealth which the consumer uses as the constraint on period-by-period consumption. Empirically one can use, with equal justification, either total wealth or permanent income in the consumption equation. For practical reasons one often finds in estimated consumption equations some estimate of permanent income, for example, some weighted average of recent income levels, rather than wealth because of the difficulty in estimating the latter with readily available data. In theory, at least, either permanent income or wealth can be used in consumption functions since they are part and parcel of the same theoretical paradigm. II. A Simple Model of Consumer Choice consumption decisions depend. Rather, there is a spectrum of interest rates depending on the maturities of financial assets. I (We abstract here from default risk and its effect on interest rates and assume default-free securities.) The interest rate for each maturity corresponds to the price at which today's consumption can be' 'traded" for consumption at some specific date in the future. Decisions to consume today or to postpone consumption thus depend on the entire structure of interest rates over the horizon of potential future consumption. To achieve an optimal allocation of consumption expenditures in each time period, a consumer must consider relative prices by equating the desired rate of substitution between consumption today and consumption in each future period to the ratio of the Individuals make consumption decisions about what individual goods to purchase in a temporal framework, now and in the future. Consumption decisions at anyone point in time are determined by the prevailing relative prices. The intertemporal decision to consumer a good today or at some time in the future must necessarily involve interest rates, as well as the current and expected future prices of goods. The reason is simple enough: interest rates are th~ unit of exchange, the' 'prices," determining the rate at which a dollar of consumption today can be transformed into a dollar of consumption in some future period. One problem offormulating and empirically testing a model of consumption decisions over time is that there is no unique interest rate upon which Chart 1 Percent Three-Month Treasury Bill Rate and Inflation 3-Month Treasury Bill Rate .. 14 12 Consumer Inflation:, 10 8 6 4 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 'Consumer inflation is measured by the twelve-month percent change in the implicit price deflator for personal consumption expenditures. 19 prices. In the simple two-period case this is given by equation (I). The marginal rate of substitution is the slope of the indifference curve at a given point. When the rate of consumption substitution is made equal to the ratio of discounted prices the consumer achieves the highest level of satisfaction for a given level of expenditure. In the case where the decision is between consumption today, period I, and consumption in some future period beyond period 2, say, perioqk, the relevant relative price is given by the product of one-period interest rates, as seen in the expression given in Chart 2. This describes the role of the entire sequence of future interest rates in determining today's consumption decision vis-a-vis consumption in all future periods. Since consumption deci~ions involve more than the choice of consumption today versus consumption in the immediately adjacent period, equation (1) must be generalized to display the decision rule which permits an optimal allocation of consumption between the current period (say, period t) and any and all future periods over the consumers' consumption horizon. This equilibrium relationship can be expressed as simply discounted prices for each period. Consider the simplest two-period case. Here the consumer achieves an optimal allocation of his expenditures between the two periods by allocating his consumption according to the rule: MRS u PI P 2/(1+l r 2) 1+ h = 1+ IZJ 1+, 7T 2 - (I) where MRS 1.2 is the marginal rate of substitution between consumption in period I and consumption in period 2; Pi the price level in period i, Ir2 the short-term nominal interest rate between periods I and 2, and 17T2 the rate of change in prices between periods I and 2. The marginal rate of substitution, MRS 1.2' is the ratio of the incremental satisfaction derived from consuming an additional unit of consumption in period 2 to that derived from consuming the good in period I. Equation (I) reveals that the consumption allocation decision between periods I and 2 depends on the "real interest rate," IZ2. Hence both nominal interest rates and the rate of inflation must be considered in the consumption decision. As equation (1) reveals, I plus the nominal interest rate divided by 1 plus the rate of change in prices is defined as I plus the real interest rate. Equation (1) may be described graphically, as seen in Chart 2. Alternative combinations of consumption between any two periods leading to the same level of consumer utility describe the' 'indifference curve," U O , along which the consumer is equally satisfied. Where optimal consumption takes place along this curve is determined by relative MRS U +N = P'+N/[(l +l,+) (I +'+l r'+2)··· (I +'+N+lr'+N)] _ [(l+lrt') (l+'tl r ,+?)···(1+'+N_l r'+N)] (1 +,7T N) ==(I+,Z'+N) (2) Equation (2) states that the optimal allocation of consumption between period t and period t + N is given by the ratio of the discounted prices between the two period, where, for example, '+lr'+2 is simply the one-period interest rate between period t + I and t + 2. Rewritten we see this to be equal to the "real interest rate" prevailing over the period t to t+N. Equation (2) provides a link between optimal consumption allocation and the term structure of interest rates. Consumers do not know with certainty what short-term interest rate will prevail at some future time. However, because individuals arbitrage between yields available on short-term and long-term financial assets, there must necessarily be an equilibrium relationship between currently known long-term yields and current and expected short-term interest rates. If there are no market Chart 2 Representation of Consumer Equilibrium HN-1 r(1+,r I s H ) t c, 20 the future in order to maintain an optimal allocation. It has long been recognized that in theory, intertemporal consumption-saving decisions involve the entire maturity structure of interest rates. Empirically, however, investigators have, for the most part, swept the issue under the rug, often opting for the use of a particular interest rate for a given maturity, or attempting to model interest rate expectations by some ad hoc weighting of selected interest rates. These are but two of the reasons that the empirical I iterature on interest rate effects on consumption finds little agreement on the direction of the effect, let alone any agreement on the magnitude of such effects. As seen in equation (2) above, real (pricedeflated) consumption decisions are dependent on real interest rate. Thus, in principle, it is the term structure of real, and not nominal, interest rates whose influence on consumption we wish to determine. To uncover such a structure, however, presents several theoretical and empirical difficulties, since it requires the approximation of not only an anticipated rate of inflation over a given time period, but the approximation of the term structure of inflation for all maturities on the available financial assets. In principle, such an undertaking is required if we are to identify the effect on real consumption of changes in the term structure of real interest rates. As will be seen shortly, the procedure we chose will allow us to measure indirectly the effect of changes in the term structure of real interest rates on real (price deflated) consumption. impediments to traders of financial assets of different maturities, and investors have similar risk preferences, the long-term interest rate will be approximately equal to the geometric average of the yield on current and expected future one-period securities, given by equation (3). (l +,rN) = [(I +trt+l) (11+lrt+2)"· -(I +t+k_lrt+N)]ilN (3) A bar over a variable in equation (3) denotes a currently known value. Equation (3) is often referred to as the Hicksian "expectational model" of the term structure of interest rates because it defines an equilibium relationship between a known long-term interest rate and current and expected future shortterm (one-period) interest rates. The r's in equation (3) without bars over them are implied "forward" (or expected) one-period interest rates. Equation (3) can be used to obtain any forward one-period rate by simply taking the ratio on known adjacent longterm bonds, as follows: 1+ r. = (l +,r,+j+l)j+1 '+1 '+1+1 (l-L, ,r'+jY " . (4) Equations (2), (3), and (4) show the natural interrelationship between intertemporal consumption theory and term structure theory. A change in any long-term rate, or any expected future short-term rate, with given current and expected future prices, will change the right-hand side of equation (2). Hence the left-hand side must change, implying the consumer must marginally adjust his consumption between current consumption and consumption in III. Other Theoretical Issues Regarding Consumption It has been traditional in empirical studies to assume that the demand for real consumption is unaffected by the level of all prices. This proposition, called the "absence of money illusion," is held to apply in both the short-run and the long-run. In theory, this proposition is quite logical: a scaling of all nominal quantities, prices, income and wealth, should leave real demands unchanged. In recent years, however, changes in price levels have been argued to have short-run impacts on perceptions of changes in relative prices, and thereby to give rise to a real demand or supply response. Little empirical work has been attempted to isolate these so-called money illusion effects on consumption. 2 Our empirical estimation, however, will attempt to capture both the effects of the term structure of interest rates and the effects of the price level, or money illusion, on aggregate real consumption. Above, we noted that inflation played a role in the consumption decision by determining the real interest rate. (See equation 2.) But aside from this role, is there any independent role inflation can play in determining real consumption? Such a role is possible in that, because they lack information on all prices, consumers may interpret a rise in a particular good's price with a change in its relative price. This confusion may lead to systematic changes in their consumption patterns. Thus, inflation has re21 consumption. The higher interest rate means that "forward prices," or, equivalently, discounted expected future prices, have fallen, creating the incentive to postpone current consumption in favor of purchasing goods in the future at lower effective prices. It is an empirical matter whether this latter negative substitution effect will dominate the posi- cently been given an additional separate role in consumption studies, that of confusing consumers' perceptions of relative prices. Inflation can also have an independent role in consumption via three other avenues: real taxes, wealth and uncertainty. First, income taxes are not indexed in many countries, so inflation often results in a rise in rea! tax rates. This rise reduces the rea! returns on financial assets and may possibly induce increased expenditures on goods where the tax is either absent or sheltered, such as real estate. The deductibility of interest payments from taxable income is hence a greater haven from taxes when the real tax rate is raised by inflation. Second, while inflation reduces the real value of financial assets, it also decreases the real value of financial liabilities. It is thus possible for changes in consumption to result from changes in the real net wealth position of households. Such effects would not likely be captured with available income or wealth data series. Third, uncertain knowledge of future inflation makes future real income uncertain, and, it has been argued, reduces current consumption in some cases. That is" with an equal probability that future real income will be greater or less by the same amount, risk-averse individuals will choose to reduce current consumption. 3 Is it possible, a priori, to establish what the sign of the effect of a rise in rea! interest rates will be on consumption? The answer in general is no. 4 The reason is well known and common to all empirical demand equations. A rise in an interest rate has two opposing effects on consumption: a positive income effect and a negative substitution effect. With a higher interest rate, individuals can increase current consumption without decreasing future consumption because a higher interest rate will generate greater future cash flow; this is the positive income effect. In contrast, a rise in the interest rate also induces a negative substitution effect on current tive income effect. The ambiguity of the effect of interest rates on consumption is readily apparent in the empirical literature on consumption as well. To mention two recent studies. Michael Boskin and Warren Weber have estimated aggregate consumption functions with interest rates as explanatory variables and have obtained opposite results, Boskin finding that a rise in interest rates decreases consumption, and Weber finding that consumption is increased with a rise in interest rates. 5 Weber's results are interesting because they hint at the importance of future interest rates. Weber finds that a rise in the weighted average of current and past nominal interest rates increases consumer expenditures on nondurables and services. He argues that the increase in the weighted average of current and past nominal interest rates suggests that nominal interest rates will be lower in the future. Hence the expected decline in interest rates would tend to increase the present value of future income. This rise in the present value of future income, Weber argues, causes consumers to increase current consumption. Weber's study suggests that understanding how changes in current interest rates may affect current consumption requires that one posit a relationship between currently observed interest rates and expectedfuture interest rates. Such a relationship can resemble the expectational theory of the tenn structure, as was seen above. We now tum to the specification of our estimable consumption function. IV. Consumption Function Specification The arguments above have concentrated on why the tenn structure of real interest rates should iNfluence consumption decisions. This section will briefly outline the functional specification of the consumption equation. The basic life-cycle consumption argument is that consumption depends on total wealth. This relationship may be specified as simply: (5) where c is real consumption, W real wealth, and k a parameter. Equation (5) represents current consumption as dependent on the stock of current real wealth. Assume that the consumer's planning hori- 22 zon is infinite so that his stock of real wealth generates a flow of income each period proportional to his stock of wealth. This permanent income flow from wealth can be written as yf iW( inflation rate over the past m years. Equation (8) also includes the effects of changes in the price levels on disaggregate price indices. The reason for the inclusion of the disaggregate price indices is primarily statistical. Statistically, we can capture the presence or absence of money illusion with one aggregate price level variable only if all subaggregate indices move proportionately. Since this requirement is unlikely to be satisfied by the observed components of the aggregate index, we disaggregate the price level variable. Recall that by definition the nominal interest rate, r, includes both a real interest rate and an anticipated inflation component. Equation (8) includes as independent variables both the level of nominal interest rates and the rate of inflation. This implies that in a regression equation the effect ofthe real interest rate on consumption is captured by the coefficient on the nominal interest rate while the independent effect of inflation on real consumption is captured by the sum of the coefficients on the nominal rate and the inflation variables. Since the nominal interest rate includes an anticipated inflation component its effect must be added to the separately measured observed inflation variable. Three measurement problems remain before subjecting equation (8) to statistical estimation. The first is to devise some approximation for the time series data on permanent income. The other two measurement problems are related. Statistically, it is difficult to obtain reliable estimates of the effects of interest rates and inflation on consumption if more than a couple of these variables, for different maturities, are included in the equation. Hence, we devised proxy variables for the term structure of nominal interest rates and inflation. (6) where yP is permanent income and i a real interest rate. Solving for W in (6) and substituting into (5) yields ct = k yPt == K(i) y\ (7) The proportionality factor K is now written as a function of the real interest rate. If we define the real interest rate i as the nominal interest rate, r, less the rate of inflation, and argue, as above, that real consumption depends on the tenn structure of real interest rates, the function K( ) would have to include more than one real interest rate. In addition, if individuals' real consumption decisions are altered by movements in the price level and the rate of inflation, these variables should also be incorporated in the K function. These considerations lead to the general specification of our real consumption equation. C{ == K(r 1t ,f 2t" .. ,rmt,1Tit,1T2t"" ,7T rn t' (8) Equation (8) captures the effects of the term structure of real interest rates but includes both nominal interest rates and inflation variables. The first subscript on the interest rate and inflation variables represents the maturity of the financial assets. Thus, r!TIt is the nominal interest rate at time t on an asset with m years to maturity and 1T !TIt the observed V. Measurement Problems In theory, what we wish to explain is consumption, not expenditures on consumer goods. The latter would include expenditures on durable goods, only a part of which can economically be considered current consumption. We therefore define consumption as expenditure on consumer nondurable goods and services plus a portion (the current service flow or depreciation) of the stock of consumer durables. This definition of current consumption was implemented where data provided a means of approximating the service flow of consumer dur- able goods, namely, in the U.S. and the U.K. (see Appendix). For Canada, our consumption variable is simply expenditure on consumer nondurable goods and services. In the case of Germany, disaggregated consumption data were not available and hence what we will seek to explain will be total consumer expenditures. As described above, the hypothesis that consumption is dependent on permanent, and not currently observed income, derives basically from the argument that consumption is, in principle, dependent 23 on wealth, where wealth is defined as discounted future income from both labor and non-labor sources. Wealth is thus a stock and permanent income, the flow; permanent income being some proportion of total wealth. Real consumption is thus dependenrohtheflow ofcurrent permanent income} Permanent income is assumed to adjust by some amount each> period· in proportion to the flo\v of current income. Hence, the estimate of permanent income is updated by the flow of observed income. To estimate permanent income according to this assumption, we use> the following well-known approximation. 7 y; = Ay, (I-A) (l+o)Y~i Equation (9) is nonlinear in the adjustment parameter A. Hence alternative consumption functions are estimated for different meaSures of permanent income, iterating on values of A. between 0 and I. That value of A. is selected with minimizes the standard error of the estimated consumption equation. 0 is obtained by first regressing the log of income on a quadratic in time. Economic theory provides little guidance as to which interest rates are appropriate to. include. in consumer demand equations. Intertemporal consumer choice implies that the entire term structure of interest rates is relevant, yet, since many interest rates, particularly those along the short-end of the maturity structure, move together, using mOre than a couple of interestrates in demand equations often leads to statistical problems of multicollinearity. Multicollinearity, in tum, results in the inability to distinguish statistically among the effects of different interest rates on aggregate demand. Instead of including several interest rates in an aggregate demand equation, we adopt a parameterization suggested by Heller and Khan and used with some success in their study of money demand. 8 The procedure, for each time period, is to regress the vector of short-to-long-term interest rates, the term structure, against a quadratic in the maturities of the (9) In equation (9), y represents observed real per capita personal disposable income, yP permanent income, 0 the estimated growth rate (trend) of real per capita income and A the permanent income adjustment coefficient, revealing what proportion of current observed personal disposable income is used to update the estimate of permanent income. The notion behind equation (9) is that future income is uncertain and variable, so the consumer obtains a "smoothed" estimate of his permanent income by updating it by some portion of recently observed income. Chart 3 Representation of the Term Structure of Observed and Expected Inflation Rate of Inflation Observed Inflation t-N t-j Expected Inflation HI t-I 24 Hj HN Period ture. The intercept term does not by itself determine the average level of interest rates because a change in the other two terms can change the average level even with the intercept term held constant. (The same applies to the B coefficients for the term structure of inflation.) Note that the slope of the term structure is a function of its maturity. The signs and will determine whether the term structure is positively or negatively sloped. 'rheestimatedcoefficients •AO'.r\I,ir\2 andB. o ' BI , B 2will all be entered as explanatpryvariables to explain.the. behaviorofreal·consumption.lfreal interest rates were important in determining consUlIlption, the A coefficients would have to. be of opposite signs from the B coefficients.Thereason is .as follows: Assume that arise in real interest rates decreases current consumption. This implies that with inflation held constant (the B coefficients), a rise in the nominal structure of interest rates (the A coefficients) would cause real consumptiontofall. Thus the estimated consumption function coefficients on the A variables should be negative. That is, a rise in the nominal term structure ofinterest rates, holding the term structure of inflation constant, increases the term structure of real interest rates, reducing consumption. Alternatively, assume that the term structure of nominal interest rates remains fixed and the term structure of inflation rises. Real rates thusfall and, by assllmption, should increase real consumption. This implies that the consumption functioncoefficients on the B variables should be positive. Moreover,the effect of the inflation rate on real consumption is .given by the sum of the respective coefficients on the Ai and B i variables. Since nominal .interest rates inclllde the anticipated rate of inflation, this effect, captured in the coefficient on theA variables, must be added to the impact of inflation captured by the coefficients of the B variables. Because the forward-looking. term structure. of inflation is being proxied in the estimationofequation (ll) by the. observed past term structure of inflation, •the •ass~mption is that a reasonableestimate of future inflation, for allmaturities,<isthe observed inflation rate. The term. structure ofpast inflation .is thus assllmedtobe symmetrical to the tennstructure of future inflation expectations. (See Chart3. ) Given this empirical assumption, the esti- corresponding financial assets. The estimated coefficients are then saved and used as explanatory variables in the estimation of the consumption function. Formally, assuming that there are financial assets of N different maturities and T observations on the market yields, we estimated the following regression for each sample period: (10) 1, •... , t= "',T where r mt is the market interest rate ·at timeton the asset with a maturity of myears. We thusapproximate the maturity structure of rates with a quadratic in the maturities (v t is the stochastic error for the term structure vector, (r" ... ,rN)p for which coefficients are to be estimated). The estimation of (10) will yield three term structure coefficients for each time period, t = I, ... ,T. These coefficients are saved and used as explanatory variables in our consumption function. 9 Since real interest rates determine real consumption decisions, we require either an estimate of the term structure of real rates or a direct estimate of the term structure of inflation. Given the effects of the latter on consumption, the former can be derived. We, therefore, follow a procedure similar to the above for estimating inflation. In this case, we estimate coefficients of the term structure of inflation over the immediate past and assume that this is the best directly available. estimate of the future term structure of inflation. This assumption is represented graphically in Chart 3. The following equation is estimated for each sample period. 2 (rrrr mt = Bat + B!tm + B 2tm + u t (II) m = I, ... ,N; t = I, ... , T where 1T mt is the annual rate of inflation which has occurred over the immedediate past m years, and where the maturities of the measured past inflation rates correspond to the maturities of the financial assets. The coefficientsB o , HI and B2thuscapture the prevailing (past) tenn structure of inflation. Equations(1Q)and (II) may be viewed as an exponential approximation to the term structure of interest rates· and inflation respectively. Thecoefficient Ao is the. shift or intercept parameter for the interest rate term structure, A 1+ 2A 2m the slope of the term structure at maturity m and 2A 2 its curva25 year interest rate term structure, requires that the term structure of inflation estimation use ten years of past inflation observations. Thus the availability of price data constrains the period for the estimation of the consumption function. The above consid~rations lead us to the specifica- mation of equation (II) required to obtain the three B coefficients, for each period over which the consumption function is to be estimated, limits the sample period for consumption function estimation. The reason is obvious enough: to estimate a consumption function which uses, for example, a ten- 26 tion of our consumption function with tenn structure effects, specified in logarithmic fonn: tYJ c t = 4> + AtYJY; + YJ/YJP lt + ... +YJNt'l')P N t + lYoA Ot + lYlA it + O!zA Zt + .BoB at + .B I B it + .BzB Zt + e t (12) Recall in equation (12) that AOt ' Alt' and A Zt are the parameters derived from the regression of the log of the vector of interest rates on a quadratic in the maturity structure. Similarly, Bot' B It and B Zt are the parameters derived from the regression of the log of inflation on a quadratic in the maturity of inflation. The A's and B's in equation (12) are treated as explanatory variables representing the tenn structure of nominal interest rates and inflation, respectively. To summarize, we argue that if real interest rates, more specifically the tenn structure of real interest rates, affect real consumption, then the coefficients on theinterest rate and inflation proxy variables in (12) ought to be of opposite sign. If a rise in the term structure of nominal interest rates, represented by the Ai s (representing a rise in real interest rates with inflation held constant) decreases current consumption, a rise in the tenn structure of inflation, which implies a decline in real interest rates, should increase current consumption. We now tum to estimation of equation (12) for the U.S., Canada, the U.K. and Germany. VI. The Empirical Consumption Function addition of rent, food, consumer industrial products and services price components. The next thing to note in the empirical estimates is that the coefficients of the tenn structure of nominal interest rates and of inflation are of opposite sign, as suggested above, for three of the fourcountries considered, the exception being the U.K. A Empirical estimates of equation (12) appear in Table I. Note that the price level explanatory variables are the disaggregate components of consumer price indices. For the U.S. and the U.K., these components are the price indices for durables, nondurables and services; for Canada, these three plus a series for semi-durables; and for W. Gennany, the Chart 4 U.S. Real Consumption Equation Dollars Per Capita 4800 4600 4400 Without Term Structure Effects ... 4200 4000 3800 3600 3400 _-a.._.. . ._. . . . . . _. . . . 3200 ...........,j"-....... 1970 _ . . L . _ . . . L - _.............IL......... 1974 1976 27 1978 1980 1981 rise in the term structure of nominal interest rates, holding the term structure of inflation constant, is seen to decrease real consumption for the U.S., Canada and the U.K. The t-statistics reported below the coefficients in Table I suggest that the shape of the term structure, and not simply the general level of interest rates, helps determine current real consumption. As suggested earlier, the coefficients on the Ao' AI and A 2 variables can be interpreted as representing the impact of the term structure of real interest rates on consumption. All three coefficients on these variables are negative for the consumption equations estimated for the U. S., Canada and the U. K. These results suggest that a rise in the term structure of real interest rates will decrease real consumption. The estimated coefficients are significant at the 5-percent significance level for a twotail test except for the Ao coefficients for Canada and the U. K. This last result suggests that it may be the shape of the term structure, as well as of the general level of interest rates, which is affecting real consumption. In the case of Germany the coeffi~ cients on the interest rate variables are all positive, but none are significant. We conclude from this that neither the level nor the shape of the term structure has any impact on real consumption in Germany. The estimated equations capture expected inflation in two ways. First, the Bo' B I and B2 variables act as proxies for inflation. Second, since the nominal interest rate is assumed to contain an expected inflation premium, the expected inflation rate is also captured in the coefficients of the A variables. Hence, summing the respective coefficients on the A and B variables yields the total impact of expected inflation on real consumption. Carrying out this exercise, we find that a rise in expected inflation increases real consumption in the U.S. and Canada and reduces real consumption in the U.K. and W. Germany. The coefficients on the directly estimated expected inflation proxy variables for the W. German equation are all negative and very significant at conventional significance levels. The assumption that real consumption can be represented with a permanent income hypothesis is supported by the estimated equations. The permanent income variable coefficients are all significant. The elasticities of real consumption with respect to permanent income are very similar for Canada and the U.S. Except for the German case, the appearance of widespread "money illusion," displayed by significant coefficients in the price level variables, is not confirmed by the equations for the U. S., Canada and the U.K. While nne or more individual price level coefficients are found to be significant, we do not establish the degree of money illusion in consumption which other investigators have uncovered. 1O Only in the case of Germany is significant money illusion evident, and here it may be due to a data. measurement problem, since the dependent variable includes expenditures on durable goods which cannot accurately be called consumption. The general impression one obtains from viewing the results in Table I is that, within the context of a permanent income consumption model, the term structure of real interest rates plays a significant role. This result has not previously been noted in other empirical consumption functions studies but it is implicit in much of the theoretical literature on intertemporal consumption. Furthermore, in three of the four countries considered, a rise in the term structure of nominal interest rates, holding constant the term structure of inflation, was found to decrease real consumption. In the U.S. and Canada, inflation appears to increase real consumption, but it appears to decrease real consumption in the U.K. and W. Germany. As we have stated above, it is not possible to determine a priori whether a rise in real interest rates will increase or decrease current consumption. The sign of the effect depends on whether the negative substitution effect is greater or less than the positive income effect. Our empirical results appear to indicate that for the U.S., Canada and the U.K. the substitution effect dominates the income effectReal interest rates have a negative impact on real consumption. The quantitative impact of the term structure of interest rates and inflation can be summarized for the U.S. by viewing Chart 4. Here, the estimated equation in Table I is fitted over the estimation period with and without the effects of the interest rate and inflation term structure variables, along with the actual values of real per capita consumption. As can readily be seen, real per capita consumption is considerably overstated when the term structure effects are dropped from the equation. I! 28 VII. Conclusion cients on permanent income or price levels, that is, the term structure coefficients changed much less when alternative functional specifications or sample periods were chosen for ~stimation. Consider now the policy significance of the empirical results. From June to August 1982, interest rates in the U.S. declined dramatically. The threemonth Treasury bill rate fell from an average .12.47 percent in June to 7.92 percent in September. From the results .of this analysis, this rapid declille in interest rates in the U.S. can be expected to contribute to growth in real consumption spending in the U. S. Since we found that inflation generally had a positive impact on real consumption in the U.S., we can expect the declining inflation rate .to weaken real consumption. Quantitatively, however, the major decline in real interest rates, particularly at the short-end of the maturity spectrum, should provide a major boost to consumption and to the general economic recovery. The aim of this paper \Vas to determine whether real consumption was influenced by the term structure of interest rates in a model which captured the influence. ofpeflllanentincome and the impact of price level movements. The answerwasfotindgenerally to be in the affirmative for the four countries considered. A rising term structure of real interest rates was found to decreasecllrr~nt consumption in the case of the U.S., U.K., and Canada, and to increase current consumption in the case of Germany. By holding constant the effect of inflation over varying maturities, it was found that the impact of the term structure comes from the impact of changes in real interest rates. Holding constant the term structllre of real interest rates, a rise in inflation appears to increase real consumption in the U. S. and Canada and to reduce consumption in the U.K. and Germany. Although not reported here, it was also found that the coefficients on the term structure variables were often more robust than the coeffi- Data Appendix This appendix briefly describes the source of the data and the construction of variables used in the test. The constructed variables include the real per capita permanent personal disposable income for the U.S., U.K., Canada and Germany, the Ao' AI and A 2 and Bo' B I and B1 variables used to approximate the term structure of interest rates and infla- tion, respectively, for each country, the generation of the service flow from durable goods forthe U.S., and the quarterly interpolation from annual data on capital consumption at current replacement cost for the U. K. personal sector. Population series for the U.K. and Gef!l1anyareobtained by interpolating annual to quarterly observations. I. Data Sources (A) Germany Statistical Office's National Income and Expenditures, 1981 Edition (the Blue Book), an annual series, and interpolated to quarterly figures. Series for personal disposable income, consumer expenditures and prices are taken from the Central Statistical Office's. Monthly. Digest of Statistics, various issues. An annual population series was obtained from the GSO'sNationallncome and Expenditure, 1981 Edition, interpolated to get quarterly series. Interest rates used to construct the term structure variables inclUded. the threemonth interbank rate, the gross redemption yield on local authority one-year bonds, and the gross redemption yield on 5-, 10-, and 20-year gilts. These interest rate data were kindly provided by Mr. Paul Temperton, Economics Division, Bank of England. Disposable income, private consumption and the cost of living index for all households are taken from Statistical Supplement, Series 4, to the Monthly Reports of the Deutsche Bundesbank. Annual population, series •were obtained from the annualO.E.C.D. country reports for Germany. Disaggregate prices for Germany were obtained from the Data Resources, Inc., databank. Interest rates are Eurocurrency deposit rates for 1-, 3-, 6... and 12-rnonth, and 2-, 3-, 4-. and 5-year maturities, taken from the Data Resources, Inc., databank. (B) United Kil19c1om Capital consumption at current replacement cost for the personal sector is taken from the Central 29 (D) Canada Data for personal disposable income, personal expenditures on consumer and services goods (services and nondurables), and price indices for personal expenditures on consumer goods and services were .• obtained from the. National Income and Expenditure Accounts, Volume 2 (the quarterly estimates 1947-1974), and quarterly issues of the same publication. The population series came from the. Canadian Statistical Review, various issues. Canadian term structure of interest rate data on Canadian Government securities: 3- and 6month Treasury bills, bonds of 1-3,3-5,5-10 and IO-years-and-over maturities were kindly provided by the Bank of Canada. (C) United States Personal disposable income, consumer expenditure series for durable and nondurable goods and services, the personal consumption expenditure price deflator, and its components, are taken from the National Income and Product Accounts. Interest rate data used to approximate the term structure included the 3- and 6- month Treasurj bill yields and the yields on constant maturity 1-,3-,5-, 10- and 20-year Treasury bonds, available from the Federal Reserve Bulletin. The benchmark stock of consumer durables used to derive the service flow from durable goods held by the household sector was obtained from the Flow of Funds Accounts. II. Construction of the Service Flow from Durable Goods for the U.S. DDt = .02761(ED t) + .21784(SDt-!) SD t = .25(ED t - DDt) + SDt-! where . EDt = expenditure on durable goods in period t SD t = the stock of durables in period t DDt = the service flow from durables in period t This methodology is that used in the Board of Goven:wrs staff's quarterly econometric model. Personal consumption is then defined as expenditures on nondurables and services plus the derived DD variable. To accurately reflect total quarterly consumption of durable goods, it is necessary to separate the service flow derived from these goods in each period from total expenditures on durable goods in each period. To do so requires a benchmark estimate of the stock of durable goods held by households. This stock was taken as the end of the year 1953 value from the Flow of Funds Accounts of October 1981. For the period 1954.1 to 1981.II we then solved the following two equations: III. Construction Quarterly Data from Annual Observations Where interpolation is used to derive quarterly series from annual observations, the method of cubic splines interpolation was employed. FOOTNOTES Bank of England Quarterly Bulletin, (March 1976). The role of. the misperception of the price level on aggregate demand is considered in a theoretical model by R. J. Barro, "Rational Expectations and the Role of Monetary Policy," JOurnal of Monetary Economics, (1976). 1. There are several ways of formulating the simple intertemporal consumption problem. Two of the clearest presentations can be seen in J. M. Henderson and R. E. Quandt, Microeconomic Theory, 2nd edition, pp. 293309, and E. F. Fama and M. H. Miller, The Theory of Finance, Holt, Rinehart and Winston. The role of discounted prices in intertemporal consumption analysis is emphasized in Henderson and Quandt and in A. Deaton and J. Muelbauer, Economics and Consumer Behavior, Cambridge University Press, (1980). 3. See especially Deaton (1977), cited above, with respect to the role of inflation in causing confusion over true relative price changes. With respect to the effect of uncertain real income on consumption, see J. Dreze and F. Modigliani, "Consumption Decisions Under Uncertainty," Journal of Economic Theory, (December 1972). 2. An early attempt to estimate money illusion effects on real consumption is W H. Branson and A. K. Klevorick, "Money Illusion and the Aggregate Consumption Function," American Economic Review, (December 1969). See also A. Deaton, "InVOluntary Saving Through Unanticipated Inflation," American Economic Review, (December 1977), and J. C. Townsend, "The Personal Saving Ratio," 4. An illustration of the theoretical ambiguity in determining alternative interest rate effects on consumption can be seen in J. I. Bernstein and D. Fisher, "Consumption, the Term Structure of Interest Rates and the Demand for Money," Working Paper #1976-8, Concordia University. See also Henderson and Quandt, cited above. 30 5. See M. J. Boskin, "Taxation, Saving and the Rate of Interest," Journal of Political Economy, (April 1978, Part 2), andW E. Weber, "Interest Rates, Inflation and Consumer Expenditures," American Economic Review, (December 1975). 6-month and 1-, 2-, 3-, 4-, and 5-year Eurocurrency deposit rates from 1973.IVto 1981.111. Data for the U.K., supplied by the Bank of England, included the three-month interbank rate, the gross redemption yield on local authority one-year bonds, and the gross redemption yields on 5-, 10-, and 20-year gilts. While the U.K. data are not all yields on government securities, they were suggested to be representative market yields for the given maturities. For Canada yield data were for 3- and 6-month Treasury bills and government bond yields for 1-3-year, 3-5-year, 5-1 O-year and 10-year-and-over maturities, provided by the Bank of Canada. 6. See Milton Friedman, A Theory of the Consumption Function, Princeton University Press, (1956). A recent test of the permanent income hypothesis which utilizes the concept of rational expectations is given in M. A. Flavin, "The Adjustment of Consumption to Changing Expectations About Future Income," Journai of Poiitical Economy, (1981). 10. The arguments for introducing the disaggregated components of the general index of consumer prices instead of the aggregate index were offered by A. Cukierman, "Money Illusion and the Aggregate Consumption Function: Comment," American Economic Review, (March 1972). Cukierman argues that the "plim" of the money illusion coefficient is dependent on the individual components of the aggregate price index. J. van Daal argues that the appearance of money illusion in macro consumption functions may likely be due to aggregation bias. See J. van Daal, "Money Illusion and Aggregation Bias," De Economist, (1980). 7. See M. R. Darby, "The Permanent Income Theory of Consumption-A Restatement," Quarterly Journal of Economics, (May 1974), and J. J. Seater, "On the Estimation of Permanent Income," Journal of Money, Credit and Banking, (February 1982). 8. H. R. Heller and M. S. Khan, "The Demand for Money and the Term Structure of Interest Rate," Journal of Political Economy, (February 1979). 9. Quarterly term structure of interest rate data were used to estimate equation (10) for each quarterly sample period over which the consumption function was to be estimated. Data for the U.S. included yields on the three-month and six-month Treasury bills and the constant maturity 1-, 3-, 5-, 10-, and 20-year government bonds. In the case of Germany the term structure data were yields on 1-, 3-, and 11. Both fitted series plotted in Chart 3 utilized the second order autocorrelated error structure estimated for the U.S. in Table I. 31