Tax Incentives, Material Inputs, and the Supply Curve for Capital Equipment

Tax Incentives, Material Inputs, and the Supply Curve for Capital Equipment Karl Whelan Division of Research and Statistics Federal Reserve Board (cid:3) May 4, 1999 Abstract Theslope ofthe supply curve for capitalequipment has importantimplications for the macroeconomics of investment and the e(cid:11)ects of tax reform on capital accumulation. Goolsbee (1998) has used changes in investment tax incentives to identify whether this supply curve is signi(cid:12)cantly upward-sloping and has concluded that it is. This paper shows that investment tax incentives are a poor instrument for identifying this supply curve because they are spuriously correlated with supply shocks for equipment producers. Once input costs for equipment producers are controlled for, there is no evidence of a relationship between tax incentives and equipment prices. In fact, the evidence favors a flat supply curve interpretation. (cid:3)MailStop80,20thandCStreetsNW,WashingtonDC20551. Email: kwhelan@frb.gov. Iwishtothank, without implicating, Darrel Cohen, Spencer Krane, Stacey Tevlin, and, in particular, Austan Goolsbee for comments on a previous draft. The views expressed in this paper are those of the author and do not necessarily reflect theviews of theBoard of Governors or the sta(cid:11) of theFederal Reserve System.

Despitetheoreticalpredictionsthatthecostofcapitalshouldhaveanimportante(cid:11)ecton investment, traditionally empiricalresearch hasfoundthis e(cid:11)ect toberelatively small. One potential explanation for this pattern has been that these estimates are biased downward due to the endogeneity of interest rates: The monetary authorities tend to lower interest rates in response to negative shocks to investment. Thus, a popular alternative approach to identifying the e(cid:11)ect of user cost on investment has been to instead focus on variations in the tax component of the cost of capital. However, the evidence on the e(cid:11)ect of tax incentives has been mixed. While some studies, such as Cummins, Hassett, and Hubbard (1994)havearguedthat,aroundmajortaxreforms,theelasticity ofinvestmentwithrespect to the tax portion of the cost of capital is about -2/3 or larger, estimates based on time series regressions, such as those of Peter K. Clark (1993), have been far smaller, on the order of no more than -0.4. In an important contribution, Austan Goolsbee (1998) has provided a potential explanation for why tax incentives may impart only a limited stimulus to investment: If the supply curve for capital equipment is su(cid:14)ciently upward-sloping, then the outward shift in the demand curve for equipment induced by investment tax incentives could mainly result in higher equipment prices rather than higher quantities. In testing this hypothesis, Goolsbee’s empirical analysis focused on the e(cid:11)ect on equipment prices of tax incentives and revealed a robust relationship consistent with a strongly upward-sloping supply curve for capital equipment. This result has important consequences not only for the macroeconomics of investment but also for public (cid:12)nance since the proponents of fundamental tax reform often stress the bene(cid:12)cial e(cid:11)ects on capital formation of improved tax incentives for investment. Interestingly, however, another detailed empirical study of supply curves by John Shea (1993) reports downward-sloping supply curves for the two capital good industries in its sample (construction machinery and aircraft). An important di(cid:11)erence between these two studies is their choice of \identifying" demand shock used to trace out the supply curve. Goolsbee uses measures of investment tax incentives while Shea’s demand instruments arechosen fromadetailed search for variables thatsatisfy two critera, oneindicating they are an important component of an industry’s demand, the other suggesting they are likely to have a low correlation with the industry’s supply shocks. This paper re-examines the supply curve for capital equipment and concludes that tax incentives are a poor instrument for identifying this curve because they substantially fail Shea’s second criterion of low correlation with supply shocks. Speci(cid:12)cally, I show that 1

starting in 1974-75 and continuing until the early 1980s, relative prices for almost all types ofequipmentroseatafastpace(orfasterthantheirtrendrate)andthenforsomeyearsafter this period, this pattern was reversed. Since the investment tax credit was strengthened in the mid-1970s and eliminated in 1986 this resulted in a correlation between equipment pricesandmeasuresofinvestmenttaxincentives,implyingastronglyupward-slopingsupply curve. However, I show that that these gyrations in equipment prices were far more highly correlated with movements in pricesof intermediate inputs(energy andmaterials) andthat once these supply shocks are controlled for, there is no evidence of a relationship between equipment prices and tax incentives. In fact, I argue that the response of equipment prices to these supply shocks is instead broadly consistent with a flat supply curve. Section 1 gives a brief theoretical discussion. Section 2 presents the evidence on the behavior over time of equipment prices, the investment tax credit, and intermediate input prices for equipment producers. Section 3 contains the basic econometric results, which extend Goolsbee’s analysis to account for the e(cid:11)ect of intermediate input prices. Section 4 examines whether the observed fluctuations in intermediate input prices for equipment producerscould berelated tochanges over timeinthetaxtreatment of investment. Section 5 concludes. 1 Supply, Demand, and Equipment Prices As I will focus below on the e(cid:11)ect of materials prices on the price of equipment, consider the case of (cid:12)rms producing capital equipment with materials being the only variable input (Q = M(cid:11)) implying a marginal cost curve of the form cQ(cid:11) 1−1, where c is the price of the (cid:11) materials. Suppose there is a large number of (cid:12)rms, n, producing equipment, each taking the price, p, as given and determining their supply by setting marginal cost equal to this price. This implies an equipment supply curve of the form (cid:18) (cid:19) (cid:11) QS = n (cid:11)p 1−(cid:11) (1) c Suppose now the demand curve for equipment is QD =(p(1−s)) −(cid:12) (2) wheressummarizesinvestmenttaxincentives.1 Settingsupplyequaltodemandandsolving 1Amorecomplete modelofequipmentdemandwould of coursealso includeinterest rates, depreciation, 2

for the equilibrium price gives " (cid:18) (cid:19) #! 1−(cid:11) 1 1 c 1− (cid:11) (cid:11) (cid:11)+(cid:12)(1−(cid:11)) p = (3) n(1−s)(cid:12) (cid:11) Re-written in terms of logs we get: (cid:18) (cid:19) (1−(cid:11)) (cid:12)(1−(cid:11)) (cid:11) c log(p)=− log(n)− log(1−s)+ log (4) (cid:11)+(cid:12)(1−(cid:11)) (cid:11)+(cid:12)(1−(cid:11)) (cid:11)+(cid:12)(1−(cid:11)) (cid:11) Considernowthetwoextremecasesof(cid:12) = 1(unitelasticinvestmentdemand)and(cid:12) = 0 (price inelastic investment demand). When (cid:12) = 1 the elasticities of the price of equipment with respect to the tax term and the price of materials are 1−(cid:11) and (cid:11) respectively. When (cid:12) = 0 the elasticity with respect to the tax term is zero while the elasticity with respect to the materials price is 1 or, more accurately, a coe(cid:14)cient equal to materials’ share in total variable cost, here assumed to be 1. Of course, if the equipment industry is better approximated by free entry and so the number of (cid:12)rms is not (cid:12)xed, then zero pro(cid:12)ts implies a price of equipment that is independent of tax incentives. In this case, we have F1−(cid:11) log(p) = log( )+(cid:11)logc (5) (cid:11)(cid:11) −(cid:11) where F measures (cid:12)xed costs. The regressions below suggest that this competitive price equation appears to (cid:12)t the data well. Note that, when (cid:12)rms are price takers, then (cid:11) can be observed as the ratio of total materials costs to the value of output since cM = (cid:11) is a pQ (cid:12)rst-order condition. The estimated elasticities with respect to material prices reported in Section 3 reveal coe(cid:14)cients similar to this observed ratio. 2 The Data 2.1 The Relative Price of Equipment Figures 1A and 1B show the behavior of equipment prices relative to the GDP deflator over the period 1959 to 1997. Through 1994, these data can be obtained from Table 7.8 of Department of Commerce (1998a); data from 1995-97 can be found in Department of and so on. However, since we are only focusing on prices and taxes, this is a reasonable simpli(cid:12)cation for our purposes. 3

Commerce (1998b). This sample is longer than the 1959-1988 sample used by Goolsbee, which was taken from Department of Commerce (1993) and since many of these series are based on hedonic adjustment methodologies that have changed over time, one cannot exactly replicate Goolsbee’s results with this data set. However, as I show below, the qualitative features of his results can still be obtained from these data. Two patterns emerge strongly from Figures 1A and 1B. First, many of the relative prices have substantial trends over time. While some types of equipment appear to have upward trends, more noticeable are the downward trends for computing equipment and other \high-tech" categories such as communications equipment and instruments. The declineintherelative priceofcapital equipmentis, ofcourse, duetotheradicalimprovements in productivity in high-tech industries, a well-known fact that has featured prominently in recent attempts to explain the process of aggregate productivity growth.2 Thus, it is necessary to control for these long-term trends and, in the regressions below, we follow Goolsbee in including a time trend on the right-hand-side of the relative price regressions. The second noticeable pattern is that, for a wide range of equipment types, there was a sharpriseinrelativepricesbeginningin1974-75andcontinuinguntiltheearlytomid-1980s. Even those types of equipment that did not see large increases in relative prices appeared to rise relative to their trend level: After detrending using a simple regression of the log of the relative price on a time trend, 20 of the 22 relative equipment prices rose substantially beginning with the 1974-75 period with most falling back again at varying speeds during the 1980s. However, this simple characterization masks a signi(cid:12)cant diversity in terms of the magnitude of these swings. For example, while the timing of the swings in the relative prices of construction machinery and mining machinery are similar, the detrended relative price of the former increased 19 percent between 1974 and 1980 while the increase for the latter was 37 percent. 2.2 Equipment Prices and the ITC Figures 2A and 2B compare the detrended relative equipment prices with the good-speci(cid:12)c ratesfortheinvestmenttaxcredit(ITC),takenfromGravelle (1994); thedetrendedrelative equipment prices are the solid lines and the ITCs are the dashed lines. The ITC was (cid:12)rst introduced in 1962 and set at varying rates for di(cid:11)erent types of equipment. Thecredit was 2Seefor instance, Greenwood, Hercowitz, and Krussell (1997). 4

Figure 1A Equipment Prices Relative to the GDP Deflator Indexes (1992=1) Furniture Fabricated Metals Engines and Turbines 1.4 1.4 1.1 1.2 1.2 1.0 1.0 1.0 0.9 0.8 0.8 0.8 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Tractors Agricultural Machinery Construction Machinery 1.2 1.2 1.1 1.0 1.0 1.0 0.8 0.9 0.6 0.8 0.8 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Mining Machinery Metalworking Machinery Special Industrial 1.4 1.1 1.2 1.2 1.0 1.0 1.0 0.9 0.8 0.8 0.8 0.6 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 General Industrial Office and Computing Service Industry Machinery 1.2 150 1.4 1.1 100 1.2 1.0 50 1.0 0.9 0 0.8 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 5

Figure 1B Equipment Prices Relative to the GDP Deflator Indexes (1992=1) Electrical Transmission Communications Electrical Machinery 2.0 2.0 2.0 1.5 1.5 1.5 1.0 1.0 1.0 0.5 0.5 0.5 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Trucks Autos Aircraft 1.4 3 1.1 1.2 2 1.0 1.0 1 0.9 0.8 0 0.8 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Ships and Boats Railroad Equipment Instruments 1.1 1.4 1.4 1.2 1.2 1.0 1.0 1.0 0.9 0.8 0.8 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Other Equipment 2.0 1.5 1.0 0.5 1960 1970 1980 1990 6

Figure 2A Detrended Relative Equipment Prices and the ITC Dashed Line is the ITC (Scale on Left Axis) Furniture Fabricated Metals Engines and Turbines 0.10 0.2 0.2 0.10 0.10 0.10 0.05 0.0 0.0 0.05 0.05 0.05 0.00 0.00 -0.05 0.00 -0.2 0.00 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Tractors Agricultural Machinery Construction Machinery 0.1 0.1 0.1 0.10 0.10 0.10 0.0 0.0 0.0 0.05 0.05 0.05 -0.1 0.00 -0.2 0.00 -0.1 0.00 -0.1 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Mining Machinery Metalworking Machinery Special Industrial 0.4 0.1 0.1 0.10 0.10 0.10 0.2 0.0 0.0 0.05 0.05 0.05 0.0 -0.1 0.00 -0.2 0.00 -0.1 0.00 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 General Industrial Office and Computing Service Industry Machinery 0.1 0.2 0.1 0.10 0.10 0.10 0.0 0.0 0.0 0.05 0.05 0.05 -0.2 0.00 -0.1 0.00 -0.4 0.00 -0.1 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 7

Figure 2B Detrended Relative Equipment Prices and the ITC Dashed Line is the ITC (Scale on Left Axis) Electrical Transmission Communications Electrical Machinery 0.1 0.2 0.2 0.10 0.10 0.10 0.0 0.1 0.0 0.05 0.05 0.05 -0.1 0.0 0.00 -0.2 0.00 -0.1 0.00 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Trucks Autos Aircraft 0.2 0.2 0.1 0.10 0.10 0.10 0.0 0.0 0.0 0.05 0.05 0.05 0.00 -0.2 0.00 -0.2 0.00 -0.1 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Ships and Boats Railroad Equipment Instruments 0.10 0.2 0.05 0.10 0.10 0.10 0.05 0.00 0.0 0.05 0.05 0.05 0.00 -0.05 0.00 -0.05 0.00 -0.2 0.00 -0.10 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Other Equipment 0.1 0.10 0.0 0.05 0.00 -0.1 1960 1970 1980 1990 8

briefly discontinued duringthe late 1960s, re-introduced in 1971, strengthened in 1974, and then strengthened again for a small number of equipment types in the early 1980s. The ITC was abolished in 1986 and has not been re-introduced since, although it did form part of the failed 1993 Clinton stimulus plan. Thatthereisapositivecorrelation between theITCandthedetrendedrelative pricefor many types of equipment is apparent from Figures 2A and 2B; a number of the equipment prices have correlations higher than 0.4 although the average correlation with the ITC across the 22 equipment types is only 0.09. It is also clear that, for most equipment types, the positive correlation is driven by the hump-shaped pattern for relative prices since the mid-1970s. Relative prices for equipment were high during the mid-1970s and early 1980s when the credit was at its most generous; they were low during the late 1980s and 1990s after the repeal of the ITC. However, the year-to-year movements in equipment prices and the ITC are usually not closely related. In particular, detrended relative equipment prices fell throughout the 1960s despite the introduction of the ITC and the timing of the 1980s decline in relative prices does not line up well with the 1986 repeal of the ITC. 2.3 Equipment Prices and Intermediate Input Prices The relationship between the ITC and equipment prices suggests the possibility that the swings in these prices have been due to shifts in the demand for equipment caused by changes in tax incentives. An alternative possibility is that these swings weredueto supply shocks. Indeed, the 1974-75 surge in equipment prices lines up exactly with the initial OPEC energy price increases. The 1975 Economic Report of the President (pg. 39) noted the rapid growth in equipment prices and explained it as being due to rising costs. As can be seen in Figure 3, during this period, (cid:12)rms had to deal with more than just a surge in energy prices. The abolition of price controls in 1974 and a worldwide jump in commodity prices also contributed to rising materials costs. The PPI for intermediate materials rose an average of 11.2 percent per year over the period 1974-81, compared with an average rise of 8.1 percent for the GDP deflator.3 This likely contributed to equipment prices rising faster than GDP prices, since the bundle of goods and services making up GDP contains a numberoflargecategories, mostnotablypersonalconsumptionexpendituresonnon-energy 3See Bruno (1984) for a discussion of the rise in materials prices during this period and its potential relationship with theproductivity slowdown. 9

Figure 3 PPIs for Energy and Materials Relative to the GDP Deflator, Indexes (1992 =1) Energy 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 1960 1965 1970 1975 1980 1985 1990 1995 Materials 1.35 1.35 1.30 1.30 1.25 1.25 1.20 1.20 1.15 1.15 1.10 1.10 1.05 1.05 1.00 1.00 0.95 0.95 1960 1965 1970 1975 1980 1985 1990 1995 10

Figure 4A Detrended Relative Prices for Equipment and Intermediate Inputs Dashed Line is the Detrended Relative Price for Intermediate Inputs (Scale on Left Axis) Furniture Fabricated Metals Engines and Turbines 0.2 0.10 0.2 0.2 0.2 0.2 0.1 0.05 0.0 0.0 0.0 0.0 0.0 0.00 -0.1 -0.05 -0.2 -0.2 -0.2 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Tractors Agricultural Machinery Construction Machinery 0.4 0.1 0.2 0.2 0.2 0.1 0.2 0.0 0.1 0.1 0.0 0.0 0.0 -0.1 0.0 0.0 -0.2 -0.2 -0.1 -0.1 -0.2 -0.1 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Mining Machinery Metalworking Machinery Special Industrial 0.4 0.4 0.2 0.1 0.4 0.1 0.2 0.2 0.2 0.0 0.0 0.0 0.0 0.0 0.0 -0.1 -0.2 -0.2 -0.2 -0.1 -0.2 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 General Industrial Office and Computing Service Industry Machinery 0.2 0.1 0.2 0.2 0.2 0.05 0.0 0.0 0.1 0.00 0.0 0.0 -0.2 -0.2 0.0 -0.05 -0.2 -0.1 -0.4 -0.4 -0.1 -0.10 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 11

Figure 4B Detrended Relative Prices for Equipment and Intermediate Inputs Dashed Line is the Detrended Relative Price for Intermediate Inputs (Scale on Left Axis) Electrical Transmission Communications Electrical Machinery 0.2 0.1 0.05 0.2 0.1 0.2 0.1 0.0 0.00 0.1 0.0 0.0 0.0 -0.1 -0.05 0.0 -0.1 -0.2 -0.10 -0.1 -0.1 -0.2 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Trucks Autos Aircraft 0.2 0.2 0.2 0.2 0.1 0.1 0.0 0.0 0.0 0.0 0.0 0.0 -0.2 -0.2 -0.2 -0.2 -0.1 -0.1 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Ships and Boats Railroad Equipment Instruments 0.2 0.10 0.2 0.2 0.4 0.05 0.1 0.05 0.2 0.00 0.0 0.0 0.0 0.00 0.0 -0.05 -0.1 -0.05 -0.2 -0.2 -0.2 -0.10 1960 1970 1980 1990 1960 1970 1980 1990 1960 1970 1980 1990 Other Equipment 0.2 0.2 0.1 0.1 0.0 0.0 -0.1 -0.1 1960 1970 1980 1990 12

services, which require very little raw materials.4 To assess the importance of movements in input costs, I used price deflators for intermediate inputs from the NBER manufacturing productivity database (Bartelsman and Gray, 1996). These price deflators, available for all 4-digit manufacturing industries from 1958 to 1994, were calculated using Input-Output tables and price deflators for 529 types of material inputs and 6 types of energy inputs. By matching each type of equipment with the two, three, or four-digit industry that produces it, I derived price deflators for the intermediate inputs required to produce each type of equipment.5 Figures 4A and 4B show the detrended relative equipment prices as the solid lines and the detrended relative price of intermediate inputs as the dashed lines.6 It is fairly clear from these charts that the input price deflators match up closer with the behavior of equipment prices than the ITCs. The average correlation of 0.42 is far higher and 12 of the 22 equipment prices have correlations higher than 0.54, which is the highest correlation between any price and the ITC. However, while certainly suggestive, these simple summary statistics do not rule out the hypothesis that investment tax incentives also a(cid:11)ect equipment prices, but rather suggest the need for inclusion of intermediate input costs as explanatory variables in equipment price regressions. 3 Regressions The basic regression speci(cid:12)cation is log(PE ) =(cid:11) +(cid:12) (TAX )+γ log(PM )+(cid:18) t+γ GROW +(cid:14) NIXON +(cid:15) (6) it i i it i it i i t i t it where PE is the relative price of equipment of type i (where relative means in relation it to the GDP deflator), PM is the relative price of intermediate inputs for equipment of it type i, t is a time trend, GROW is GDP growth, NIXON is a variable accounting for the Nixon price controls from 1971-74, and TAX is a variable measuring investment tax it incentives. This di(cid:11)ers from Goolsbee’s speci(cid:12)cation only in including PM and omitting exchangeratevariables,whichIdidnot(cid:12)ndhadasigni(cid:12)cante(cid:11)ectorinfluencedtheresults. 4In 1974 non-energy services accounted for 25 percent of GDP, and the price deflator for this category fell substantially relative tothe GDPdeflator. 5Thedetails of thismatching exercise are provided in AppendixA. 6It was necessary to detrend the relative intermediate input prices since for some industries, such as computing, material prices havefallen rapidly over time. 13

All equations were estimated using the Seemingly Unrelated Regression technique and the estimation sample was 1961-1994. 3.1 Static Regression Results Table1reportstheresultsfromestimationofequation6bothwithandwithoutintermediate input prices, using the ITC as the TAX variable. While Goolsbee’s reported results used it anAR(2)correction,theresultsreportedherewithoutanycorrectionforautocorrelationare very similar, both in terms of which categories appear to respondstrongly to tax incentives and in terms of the magnitudeof the coe(cid:14)cients: Restricting the e(cid:11)ect of the ITC to bethe same across all equipment types gives a coe(cid:14)cient of 0.30 (standard error, 0.02) compared with Goolsbee’s estimate of 0.39 (standard error, 0.03). However, once PM is added as an explanatory variable, this apparently strong relationship does not hold up. Most of the signi(cid:12)cant coe(cid:14)cients on the ITC disappear, 14 of the 22 coe(cid:14)cients on the ITC are negative, and the pooled coe(cid:14)cient is signi(cid:12)cantly negative. Conversely, there is consistent evidence for a signi(cid:12)cant positive e(cid:11)ect for PM. Moreover, the addition of PM improves the(cid:12)tof mostof theregressions, mostnotablythosethat hadpreviouslysuggested astrong relationship between equipment prices and the ITC. Of course, the ITC summarizes only one aspect of the e(cid:11)ects of the tax code on the incentive to purchase capital equipment. A broader measure also used by Goolsbee is the well-known Hall-Jorgenson tax term that features in the user cost of capital formula. This term is de(cid:12)ned as 1−ITC−(cid:28)z, where (cid:28) is the marginal corporate income tax rate and z is 1−(cid:28) the present discounted value of depreciation allowances per dollar invested. Table 2 repeats the regressions using this full tax term in place of the ITC and shows similar results.7 A signi(cid:12)cant relationship, although negative this time, between the tax term and the price of equipment is evident in the simple regressions, with an estimated pooled elasticity of -0.11 (standarderror,0.006), whichcompareswithGoolsbee’sestimateof-0.1774 (standarderror 0.02). However, again, once we include PM as an explanatory variable, these results are overturned. 7ThedetailsbehindtheconstructionofthistaxtermareinAppendixB.Myempiricalcalculationdi(cid:11)ered slightlyfromtheformulainthetextintakingintoaccountthefactthat,forsomeyears,(cid:12)rmswererequired to reduce theirdepreciation base bysome proportion of theirinvestment tax credit. 14

3.2 Dynamic Regression Results Tables1and2showthatthesebasicregressionshavehighlyautocorrelated errors(although those including PM tend to have higher Durbin-Watson statistics). This suggests that the simple speci(cid:12)cation is missing some important dynamics. Goolsbee’s solution to this problem,anAR(2)correction, impliesaveryspeci(cid:12)cformofdynamicspeci(cid:12)cationinwhich both lagged dependent and explanatory variables a(cid:11)ect the current period’s value of the dependent variable. However, simple tests suggested that including lags of the explanatory variables did not improve the residual autocorrelation problem but that including two lags of the dependent variable did. Thus, Tables 3 and 4 repeat the regressions from Tables 1 and 2, this time with two lags of the dependent variable. The tables report the long-run e(cid:11)ects for the tax and input price variables; in other words, they report the estimated coe(cid:14)cients divided by one minus the sum of the coe(cid:14)cients on the lagged dependentvariables. Standarderrorsfortheselong-rune(cid:11)ects werecalculated bytheDelta method. The message from the results in Tables 3 and 4 is very similar to the earlier results. Without PM, there appears to be a strong relationship between the prices for a large numberofequipmenttypesandtheITCortaxterm; oncePM isincludedthesecoe(cid:14)cients usually become insigni(cid:12)cant. One exception is the \pooled estimate" for the tax term elasticity which is -0.09 (standard error 0.03). However, this estimate is something of an anomaly since none of the individual regression coe(cid:14)cients for this term are signi(cid:12)cantly negative. One question concerning these results relates to the econometric speci(cid:12)cation. Kevin Hassett and Glenn Hubbard (1998) have critiqued Goolsbee’s results as being the result of a spurious regression, arguing that the relative equipment prices and tax term variables are both I(1) variables but are not cointegrated. In this case, the regression should be run in di(cid:11)erences and they show that when this is done, they do not obtain signi(cid:12)cant coe(cid:14)cients onthetaxvariables. Thatthetaxvariablesareinsigni(cid:12)cantoncetheregressionisestimated in di(cid:11)erences should not be too surprising since Figures 2A and 2B show that while some of the relative equipment prices are correlated with the level of the ITC, the year-to-year movements tend not to be closely related. Conversely, I (cid:12)nd that if one estimates the SUR equation system in (cid:12)rst-di(cid:11)erences including PM then one still obtains highly signi(cid:12)cant coe(cid:14)cients on PM, similar in size to those obtained from the levels estimation. Thus whether the regression should be run in levels or di(cid:11)erences does not a(cid:11)ect our conclusion. 15

Given that econometric tests are notoriouslypoor atdistinguishingunitrootbehavior from trend stationarity (which may describe relative equipment prices well) or stationarity with regime shifts (which may (cid:12)t the corporate tax code well) it is also very possible that the levels estimates reported here are the appropriate ones. 3.3 Interpretation Theseresultssuggestthatregressionsrelatingequipmentpricestoinvestmenttaxincentives but excluding input prices are mis-speci(cid:12)ed. The correlation between the tax variables and input prices produces spurious estimates of a large e(cid:11)ect of tax incentives on equipment prices, estimates that disappear once one controls for input prices. It is important to note that despite the coincidence of thecommon pattern displayed by most of the equipment prices with shifts in the price of oil, these results are not obtained because of an omitted aggregate variable that could be captured using year dummies. Indeed, Goolsbee reports regressions including year dummies designed to pick up aggregate e(cid:11)ects and (cid:12)nds that the inclusion of these dummies does not a(cid:11)ect his estimates. This result is still qualitatively true with this updated data set. While I found that the inclusion of year dummies produces a weaker estimated relationship between tax variables and equipment prices, many of the tax coe(cid:14)cients are still signi(cid:12)cant. However, again, once PM is included these results disappear and the (cid:12)t of the equations is noticeably improved. That equipment-speci(cid:12)c input prices explain the behavior of equipment prices better than aggregate year dummies should not be surprising given the facts documented in Section 2. The magnitudes of the swings in relative equipment prices and in input costs di(cid:11)ered markedly across di(cid:11)erent types of equipment. Thus, one would not expect that aggregate year-dummies would capture these e(cid:11)ects as well as the inclusion of the appropriate equipment-speci(cid:12)c input price variable. Do these results imply that investment demand is una(cid:11)ected by tax incentives? Not necessarily. As discussed above, there were two cases in which tax incentives had no e(cid:11)ect on equipment prices. The (cid:12)rst was one in which the equipment supply curve was upwardsloping and (cid:12)rms had price-insensitive investment demand ((cid:12) = 0). The other was the case in which free entry led to a flat supply curve, no matter what value (cid:12) took. Note, though, that the equipment price elasticity with respect to PM di(cid:11)ered in these two cases: With an upward-sloping supply curve, the elasticity should equal the ratio of intermediate input 16

Table 1: Equipment Prices, Material Prices, and the ITC No Input Prices Including Input Prices Asset Class ITC DW R(cid:22)2 ITC PM DW R(cid:22)2 1. Furniture 0.11 (.06) 0.39 .51 -0.25 (.05) 0.44 (.03) 1.07 .83 2. Fabricated Metals 1.02 (.12) 0.61 .57 0.16 (.09) 0.76 (.04) 0.72 .78 3. Engines 0.50 (.13) 0.69 .73 -0.62 (.18) 1.05 (.11) 0.88 .83 4. Tractors 0.49 (.08) 0.71 .93 0.13 (.09) 0.37 (.04) 0.61 .95 5. Agric. Machinery 0.65 (.09) 0.79 .87 0.29 (.09) 0.37 (.05) 0.62 .91 6. Constr. Machinery 0.56 (.08) 0.71 .86 -0.43 (.03) 0.78 (.03) 1.38 .96 7. Mining Machinery 1.43 (.16) 0.73 .79 0.07 (.11) 1.18 (.07) 0.77 .92 8. Metalworking Mach. 0.21 (.06) 0.75 .83 -0.30 (.05) 0.45 (.04) 1.16 .91 9. Special Ind. Mach. 0.17 (.05) 0.71 .95 -0.27 (.05) 0.45 (.03) 1.08 .98 10. General Ind. Mach. 0.27 (.09) 0.66 .51 -0.35 (.08) 0.63 (.05) 0.91 .72 11. O(cid:14)ce & Computers -0.72 (.32) 0.39 .99 -1.61 (.28) 0.81 (.05) 0.75 .99 12. Service Ind. Mach. -0.15 (.05) 1.21 .94 -0.05 (.06) -0.09 (.04) 1.21 .94 13. Electrical Distrib. 0.07 (.06) 1.10 .90 -0.09 (.07) 0.27 (.05) 1.32 .91 14. Communications 0.20 (.10) 0.51 .96 0.12 (.11) 0.29 (.16) 0.55 .96 15. Oth. Electr. Equip. 0.03 (.11) 0.34 .76 -0.06 (.12) 0.16 (.07) 0.35 .77 16. Trucks and Buses -0.03 (.11) 0.46 .64 0.10 (.12) -0.11 (.08) 0.46 .64 17. Autos -2.33 (.17) 0.95 .97 -1.42 (.15) -0.51 (.08) 0.82 .97 18. Aircraft 0.12 (.09) 0.70 .68 0.17 (.09) -0.07 (.08) 0.71 .68 19. Ships 0.39 (.08) 0.59 .88 0.11 (.08) 0.30 (.04) 0.64 .92 20. Railroad Equipment 1.38 (.15) 0.77 .68 -0.16 (.09) 1.37 (.04) 1.92 .96 21. Instruments -0.29 (.04) 1.52 .94 -0.18 (.05) -0.10 (.03) 1.51 .94 22. Other Equipment -0.20 (.12) 0.49 .87 -0.11 (.13) -0.09 (.06) 0.49 .87 POOLED 0.30 (.02) -0.25 (.02) 0.55 (.01) Sample is 1961-1994. Standard errors in parentheses. The dependent variable is the log of the equipment price minus the log of the GDP deflator. PM is the log of the relative equipment-speci(cid:12)c price of energy and material inputs. Each equation also includes a time trend, GDP growth, and a Nixon price controls variable and the 22 equations were estimated jointly using SUR. The pooled coe(cid:14)cients were restricted to be the same across all equations. 17

Table 2: Equipment Prices, Material Prices, and the Hall-Jorgenson Tax Term No Input Prices Including Input Prices Asset Class Tax Term DW R(cid:22)2 Tax Term PM DW R(cid:22)2 1. Furniture -0.13 (.02) 0.58 .56 0.01 (.02) 0.33 (.03) 0.54 .77 2. Fabricated Metals -0.40 (.05) 0.42 .42 0.08 (.04) 0.84 (.04) 0.75 .79 3. Engines -0.25 (.07) 0.64 .69 0.31 (.07) 0.95 (.09) 0.81 .83 4. Tractors -0.03 (.03) 0.38 .90 0.21 (.03) 0.57 (.03) 1.00 .97 5. Agric. Machinery -0.08 (.03) 0.32 .78 0.15 (.03) 0.64 (.04) 0.77 .93 6. Constr. Machinery -0.20 (.02) 0.58 .83 0.06 (.02) 0.66 (.03) 1.16 .96 7. Mining Machinery -0.28 (.06) 0.30 .63 0.31 (.03) 1.37 (.05) 0.95 .94 8. Metalworking Mach. -0.11 (.02) 0.72 .81 0.09 (.02) 0.39 (.04) 1.03 .90 9. Special Ind. Mach. -0.10 (.02) 0.69 .94 0.10 (.02) 0.40 (.03) 1.03 .98 10. General Ind. Mach. -0.23 (.03) 0.78 .52 -0.02 (.03) 0.47 (.05) 0.85 .69 11. O(cid:14)ce & Computers -0.15 (.10) 0.37 .99 0.23 (.15) 0.55 (.08) 0.32 .99 12. Service Ind. Mach. 0.10 (.02) 1.34 .95 0.07 (.02) -0.06 (.04) 1.34 .95 13. Electrical Distrib. 0.00 (.03) 1.04 .90 0.08 (.04) 0.29 (.05) 1.38 .92 14. Communications -0.15 (.07) 0.52 .96 -0.11 (.08) 0.34 (.18) 0.58 .96 15. Oth. Electr. Equip. 0.11 (.03) 0.34 .78 0.26 (.04) 0.39 (.06) 0.50 .81 16. Trucks and Buses 0.34 (.03) 0.70 .76 0.36 (.03) -0.09 (.06) 0.74 .77 17. Autos -0.20 (.05) 0.25 .92 -0.04 (.03) -0.90 (.08) 0.63 .96 18. Aircraft 0.06 (.04) 0.74 .67 0.10 (.04) 0.09 (.07) 0.80 .67 19. Ships -0.15 (.03) 0.42 .85 0.01 (.03) 0.33 (.04) 0.64 .92 20. Railroad Equipment -0.65 (.06) 0.57 .56 0.08 (.04) 1.37 (.04) 1.90 .96 21. Instruments 0.11 (.02) 1.26 .92 0.06 (.02) -0.15 (.03) 1.46 .94 22. Other Equipment 0.18 (.04) 0.58 .89 0.24 (.05) 0.04 (.04) 0.64 .89 POOLED -0.11 (.006) 0.09 (.005) 0.48 (.01) Sample is 1961-1994. Standard errors in parentheses. The dependent variable is the log of the equipment price minus the log of the GDP deflator. PM is the log of the relative equipment-speci(cid:12)c price of energy and material inputs. Each equation also includes a time trend, GDP growth, and a Nixon price controls variable and the 22 equations were estimated jointly using SUR. The pooled coe(cid:14)cients were restricted to be the same across all equations. 18

Table 3: Estimated Long-Run E(cid:11)ects of the ITC No Input Prices Including Input Prices Asset Class ITC (cid:26) ITC PM (cid:26) 1. Furniture 0.20 (0.12) .62 (.05) -0.25 (0.07) 0.42 (0.05) .40 (.05) 2. Fabricated Metals 1.46 (0.34) .74 (.04) -0.04 (0.20) 1.09 (0.12) .62 (.03) 3. Engines 0.64 (0.29) .58 (.06) -0.49 (0.37) 0.97 (0.25) .54 (.04) 4. Tractors 0.93 (0.19) .73 (.03) 0.11 (0.21) 0.69 (0.14) .68 (.03) 5. Agric. Machinery 1.18 (0.24) .75 (.04) 0.23 (0.19) 0.72 (0.12) .65 (.04) 6. Constr. Machinery 0.66 (0.20) .63 (.04) -0.33 (0.13) 0.72 (0.07) .44 (.04) 7. Mining Machinery 1.97 (0.39) .75 (.03) 0.05 (0.27) 1.35 (0.15) .60 (.03) 8. Metalworking Mach. 0.39 (0.16) .62 (.04) -0.11 (0.16) 0.38 (0.09) .52 (.05) 9. Special Ind. Mach. 0.46 (0.13) .64 (.03) -0.05 (0.12) 0.39 (0.08) .54 (.03) 10. General Ind. Mach. 0.40 (0.16) .58 (.04) 0.14 (0.17) 0.28 (0.10) .46 (.03) 11. O(cid:14)ce & Computers -2.89 (1.16) .79 (.05) -3.07 (1.17) 0.39 (0.36) .77 (.08) 12. Service Ind. Mach. -0.05 (0.09) .56 (.06) 0.01 (0.12) -0.03 (0.09) .56 (.06) 13. Electrical Distrib. 0.19 (0.10) .45 (.05) 0.02 (0.13) 0.22 (0.10) .45 (.04) 14. Communications 0.54 (0.61) .89 (.06) 1.19 (0.89) -2.00 (1.69) .89 (.06) 15. Oth. Electr. Equip. 0.92 (0.40) .85 (.03) 0.49 (0.41) 0.60 (0.37) .84 (.03) 16. Trucks and Buses 0.52 (0.26) .73 (.05) 1.05 (0.44) -0.36 (0.28) .75 (.04) 17. Autos -2.17 (0.48) .61 (.06) -0.10 (0.48) -0.85 (0.21) .56 (.04) 18. Aircraft 0.37 (0.20) .63 (.07) 0.31 (0.33) 0.15 (0.30) .69 (.08) 19. Ships 1.14 (0.29) .80 (.04) 0.22 (0.17) 0.47 (0.12) .65 (.06) 20. Railroad Equipment 1.48 (0.47) .77 (.04) -0.21 (0.14) 1.35 (0.07) .33 (.04) 21. Instruments -0.27 (0.07) .42 (.07) -0.16 (0.08) -0.09 (0.06) .41 (.06) 22. Other Equipment 0.35 (0.56) .85 (.06) -1.51 (3.02) 4.31 (7.26) .96 (.06) POOLED 0.49 (0.09) .76 (.06) -0.02 (0.05) 0.49 (0.08) .71 (.04) Sample is 1961-1994. Standard errors in parentheses. The dependent variable is the log of the equipment price minus the log of the GDP deflator. PM is the log of the relative equipment-speci(cid:12)c price of energy and material inputs. Each equation also includes two lags of the relative equipment price, a time trend, GDP growth, and a Nixon price controls variable and the 22 equations were estimated jointly using SUR. The pooled coe(cid:14)cients were restricted to be the same across all equations. Note: (cid:26) is the sum of the coe(cid:14)cients on lagged relative equipment prices. 19

Table 4: Estimated Long-Run E(cid:11)ects of the Hall-Jorgenson Tax Term No Input Prices Including Input Prices Asset Class Tax Term (cid:26) Tax Term PM (cid:26) 1. Furniture -0.17 (0.07) .65 (.05) 0.02 (0.04) 0.32 (0.05) .45 (.05) 2. Fabricated Metals -0.97 (0.24) .79 (.04) -0.09 (0.09) 1.03 (0.10) .62 (.03) 3. Engines -0.21 (0.21) .63 (.06) 0.54 (0.17) 1.06 (0.18) .55 (.04) 4. Tractors -1.13 (0.54) .90 (.03) -0.02 (0.09) 0.80 (0.11) .70 (.04) 5. Agric. Machinery -0.75 (0.31) .88 (.04) 0.01 (0.07) 0.87 (0.10) .65 (.04) 6. Constr. Machinery -0.49 (0.16) .79 (.03) -0.05 (0.05) 0.56 (0.06) .49 (.04) 7. Mining Machinery -1.44 (0.45) .86 (.03) 0.06 (0.11) 1.37 (0.10) .58 (.03) 8. Metalworking Mach. -0.23 (0.09) .67 (.04) -0.06 (0.07) 0.31 (0.07) .54 (.05) 9. Special Ind. Mach. -0.38 (0.10) .72 (.03) -0.11 (0.07) 0.33 (0.07) .59 (.03) 10. General Ind. Mach. -0.26 (0.08) .60 (.03) 0.23 (0.17) 0.22 (0.09) .53 (.04) 11. O(cid:14)ce & Computers 2.50 (2.45) .90 (.07) 2.31 (2.85) -0.53 (1.49) .91 (.09) 12. Service Ind. Mach. -0.01 (0.06) .62 (.07) -0.01 (0.06) -0.02 (0.08) .58 (.07) 13. Electrical Distrib. -0.09 (0.07) .48 (.05) 0.03 (0.07) 0.29 (0.09) .43 (.05) 14. Communications 0.14 (0.53) .90 (.03) 0.47 (0.84) -0.87 (1.55) .92 (.05) 15. Oth. Electr. Equip. -0.42 (0.25) .87 (.03) -0.20 (0.20) 0.77 (0.33) .84 (.03) 16. Trucks and Buses 0.16 (0.12) .67 (.05) 0.18 (0.13) 0.11 (0.17) .68 (.05) 17. Autos -0.57 (0.22) .77 (.05) -0.12 (0.07) -0.87 (0.14) .56 (.04) 18. Aircraft -0.20 (0.19) .72 (.08) 0.05 (0.17) 0.36 (0.24) .71 (.08) 19. Ships -0.79 (0.32) .87 (.05) -0.09 (0.08) 0.55 (0.11) .67 (.06) 20. Railroad Equipment -0.71 (0.27) .80 (.04) 0.08 (0.06) 1.31 (0.07) .31 (.04) 21. Instruments 0.11 (0.04) .53 (.06) 0.05 (0.04) -0.13 (0.06) .43 (.06) 22. Other Equipment 0.15 (0.22) .83 (.07) 0.66 (0.43) 1.41 (1.04) .89 (.06) POOLED -0.29 (0.03) .78 (.03) -0.09 (0.03) 0.44 (0.07) .72 (.04) Sample is 1961-1994. Standard errors in parentheses. The dependent variable is the log of the equipment price minus the log of the GDP deflator. PM is the log of the relative equipment-speci(cid:12)c price of energy and material inputs. Each equation also includes two lags of the relative equipment price, a time trend, GDP growth, and a Nixon price controls variable and the 22 equations were estimated jointly using SUR. The pooled coe(cid:14)cients were restricted to be the same across all equations. Note: (cid:26) is the sum of the coe(cid:14)cients on lagged relative equipment prices. 20

costs to total variable cost, while with the flat supply curve, the elasticity should equal the ratio of intermediate input costs to the total value of shipments. The estimated coe(cid:14)cients are closer to that implied by the flat supply curve. For the 22 types of equipment, the average value for the ratio of intermediate input costs to total variable cost (de(cid:12)ned as the sum of energy, material, and labor costs) is 0.677 (standard error 0.083) while the average value for the ratio of intermediate input costs to the total value of shipments is 0.489 (standard error 0.084). This latter average value is close to the pooled estimates of the equipment price elasticity with respect to PM reported in Tables 3 and 4. Thus, the evidence favors the flat supply curve interpretation, implying a highly competitive market structure with free entry keeping economic pro(cid:12)ts low. Is this a credible conclusion? While the extreme assumptions of competition and free entry may not match the reality, the market for equipment in the U.S. is extremely open to international trade compared to other markets, and this probably helps to keep prices near the level consistent with zero economic pro(cid:12)ts. 4 Equipment Tax Incentives and Materials Costs The results so far have used variations in input costs to show that the equipment producing industry is well approximated by the assumption of a flat supply curve. However, I have implicitly assumed that the shifts in the prices of the intermediate inputs used to produce equipment are true \supply shocks" which are independent of the tax treatment of equipment purchases. It is possible that this assumption is false. If equipment demand is price sensitive then an increase in investment tax incentives for good i will raise demand for the materials used to produce good i. Thus, if the producers of this good represented a su(cid:14)ciently large proportion of the demand for their material inputs and the supply curve for theseinputsisupward-sloping, thensuch anincrease indemandcould signi(cid:12)cantly raise the price of their inputs. Indeed, it may still be that through this mechanism there is a signi(cid:12)cant \crowding out"of thepositive demande(cid:11)ect of tax incentives, even if the supply curves for producers of capital equipment were flat. Not surprisingly, given the correlations evident in the charts shown earlier, simple regressions of the same form as equation 6 but instead using log(PM ) as the dependent it variable produce signi(cid:12)cant coe(cid:14)cients on tax incentives, which could be construed as evidence in favor of this interpretation. However, for a number of reasons, it seems far more 21

likely that this relationship is spurious. Firstly, I found that if one augments these simple input price regressions with related variables likely tobeexogenous todomesticequipmentdemand,suchas theaggregate PPIs forenergyandsteel(bothofwhicharelargelydeterminedbyworldwidesupplyanddemand conditions) then tax incentives are no longer a signi(cid:12)cant explanatory variable. Secondly, informationonquantitymovementspointsagainstthedemand-shockinterpretationofinput price movements. In particular, the correlation between tax incentives and input prices for equipment producers is largely driven by the common U-shaped pattern starting with the 1974-75 surge in materials prices. However, the behavior of investment quantities during the 1974-75 period suggests the exact opposite of a demand-driven boom: After growing 18 percent in 1973, real equipment investment grew only 2 percent in 1974 and fell 10.5 percentin 1975, thelargest decline intheperiod1959-98 (andamuch larger declinethanin the deeper recession of the early 1980s). These quantity movements are far more consistent with a negative supply shock. Finally, there is the question of whether equipment producing industries are, in fact, large enough to have a signi(cid:12)cant influence on the prices of their intermediate inputs. This question could be answered very easily if each equipment industry used only one intermediate input. In this case, we could compare intermediate input usage for each equipment industry with the total production of that speci(cid:12)c input. Since, in reality, each type of equipment requires a number of intermediate inputs, we instead need to calculate a weighted average estimateofhowlargeeach equipmentproducingindustryisrelative tothe supply of these inputs. To construct such a \size" measure for each equipment-producing industry,Iusedinformationfromthe1992two-digit Input-Outputtables. Thesizemeasure for equipment-producing industry i is de(cid:12)ned to be XN M ik S = ! (7) i ik M k k=1 where N is the number of intermediate inputs, M is total purchases by industry i of ik intermediate input k, M is total production (for both intermediate and (cid:12)nal use) of input k 22

k, and ! is the share of input k in industry i’s materials costs:8 ik M ! = P ik (8) ik N M k=1 ik Thismeasureisbestunderstoodusingasimplenumericalexample. Supposeanindustryhas two inputs with outlays on each being the same (! = ! = 0:5) and the industry demands 1 2 10 percentof thetotal productionof input1 and70 percentof the total productionof input 2. In this case, our measureof size equals 0:5(cid:3)0:1+0:5(cid:3)0:7 = 0:4, which implies that shifts in this industry’s demandare likely to have a sizeable impact on the weighted-average price of its intermediate inputs. Table 5 presents the estimates of this size measure for each of our 22 equipment industries. These estimates show that each of the equipment industries, on average, demands a very small proportion of the production of the industries that supply their inputs. The largestestimates, forAutosandO(cid:14)ceandComputingMachineryarestillbelow6percent.9 Thus, if we were looking for good demand instruments for the set of \synthetic" weightedaverage industries that supply their inputs to each equipment producing industry, in each case, the equipment producers would fail John Shea’s (cid:12)rst criterion, which requires that they demand a high proportion of the supplying industry’s output.10 As such, it would strain credibility to suggest that more generous tax incentives for these equipment industries could signi(cid:12)cantly raise the weighted average price of their intermediate inputs. One can note, though, that the construction method behind the NBER input deflators implicitly assumes that the price of each individual input is the same for all equipment producers (since they are constructed by weighting aggregate deflators for input prices according to each input’s share in costs). It is possible, though, that for some equipment producers an increase in demand could result in higher prices for a speci(cid:12)c input for those producers, even if the the price of that input is unchanged for all other (cid:12)rms. For example, 8These calculations were derived in three steps. First, I constructed \industries" based on each of the NIPA equipment categories used in this paper using Table E of Lawson (1997a), which de(cid:12)nes each of the NIPA categories in terms of input-output commodities. Second, I used the commodity-by-commodity \use" table in Lawson (1997b) to estimate the total intermediate input usage of each commodity by each equipmenttype(theM s). Third,totalcommodityoutput(theM ’s)wastakenfromTable2.1ofLawson ik k (1997a). 9Ifthese(cid:12)guresseems too small, one should notethat totalnominal privateexpenditureson equipment averaged only 6.6 percent of GDP over oursample. 10Shea’s cuto(cid:11) rule required his instrumentsto represent at least 15 percent of the industry’sdemand. 23

suppose that all tractors were constructed using steel supplied by one speci(cid:12)c steel mill. In that case, an increase in the demand for tractors could strain production at this steel mill and could possibly raise the price of steel for producers of tractors; this increase, though, would not show up signi(cid:12)cantly in the NBER input price deflator for tractors. However, if such speci(cid:12)city were important then one would expect to estimate a signi(cid:12)cant e(cid:11)ect of tax incentives on equipment prices, even if one includes the NBER input deflators, and this is not what we observe. 1. Furniture .008 12. Service Ind. Mach. .005 2. Fabricated Metals .005 13. Electrical Distrib. .005 3. Engines .003 14. Communications .035 4. Tractors .004 15. Oth. Electr. Equip .003 5. Agric. Machinery .004 16. Trucks and Buses .033 6. Constr. Machinery .005 17. Autos .055 7. Mining Machinery .001 18. Aircraft .006 8. Metalworking Mach. .008 19. Ships .001 9. Special Ind. Mach. .011 20. Railroad Equipment .001 10. General Ind. Mach. .010 21. Instruments .009 11. O(cid:14)ce & Computers .058 22. Other Equipment .004 Table 5: The Size of Equipment Industries Relative to Their Input Suppliers 24

5 Conclusion Thee(cid:11)ect of tax incentives on capital investment is avery importanteconomic policyissue. Beyond speci(cid:12)c policies such as the investment tax credit, understanding the response of investment to tax incentives is crucial for assessing the likely long-run e(cid:11)ects of tax reform proposals, many of which stress their bene(cid:12)cial e(cid:11)ects on capital formation. Thus, the hypothesis of a steep upward-sloping supply curve for capital equipment, as proposed by Goolsbee, has profound implications for a number of policy debates. This paper has re-examined the evidence on the link between equipment prices and tax incentives and concluded that Goolsbee’s result that investment tax incentives drive up equipment prices appears to be spurious. Once one controls for variations in prices of energy and material inputs, there is no evidence that tax incentives a(cid:11)ect equipment prices. In fact, the evidence is broadly consistent with a flat supply curve for capital equipment. An important implication of a flat supply curve is that one can only identify the price elasticity of investment demand by examining quantity movements. Thus, the challenge for macroeconomists is to reconcile the macroeconomic evidence of a weak e(cid:11)ect of the user cost of capital on investment quantities with microeconomic evidence such as that of Cummins, Hassett, and Hubbard (1994) which suggests a large e(cid:11)ect. References [1] Bartelsman Eric J. and Wayne Gray (1996). The NBER Manufacturing Productivity Database, NBER Technical Working Paper No. 205. [2] Bruno, Michael (1984). \Raw Materials, Prices, and the Productivity Slowdown", Quarterly Journal of Economics, 1-30. [3] Clark, Peter K. (1992). \Tax Incentives and Equipment Incentives", Brookings Papers on Economic Activity, 1, 317-339. [4] Cummins Jason, Kevin Hassett, and R. Glenn Hubbard(1994). \A Reconsideration of Investment Behavior Using Tax Reforms as Natural Experiments", Brookings Papers on Economic Activity, 2, 1-59. [5] Goolsbee, Austan (1998). \Investment Tax Incentives, Prices, and the Supply of Capital Goods", Quarterly Journal of Economics, 121-148. 25

[6] Gravelle, Jane (1994). The Economic E(cid:11)ects of Taxing Capital Income, Cambridge: MIT Press. [7] Greenwood, Jeremy, Zvi Hercowitz, and Per Krussell (1997). \Long-Run Implications of Investment-Speci(cid:12)c Technological Change", American Economic Review, 342-62. [8] Hassett, Kevin and R. Glenn Hubbard (1998). Are Investment Tax Incentives Blunted By Changes in Prices of Capital Goods?, NBER Working Paper No. 6676. [9] Lawson, Ann M. (1997a). \Benchmark Input-Output Accounts for the U.S. Economy, 1992: Make, Use, andSupplementaryTables", Survey of Current Business, November, 36-82. [10] Lawson, Ann M. (1997b). \Benchmark Input-Output Accounts for the U.S. Economy, 1992: Requirements Tables", Survey of Current Business, December, 22-47. [11] Shea, John (1993). \Do Supply Curves Slope Up?", Quarterly Journal of Economics, 1-32. [12] U.S. Department of Commerce, Bureau of Economic Analysis (1993). Fixed Reproducible Tangible Wealth in the United States, 1925-1989, Washington DC: U.S. Government Printing O(cid:14)ce. [13] U.S.DepartmentofCommerce,BureauofEconomicAnalysis(1998a).NationalIncome and Product Accounts of the United States, 1929-94: Volume 2, Washington DC: U.S. Government Printing O(cid:14)ce. [14] U.S. Department of Commerce, Bureau of Economic Analysis (1998b). \National Income and Product Accounts Tables", Survey of Current Business, August, 36-118. [15] U.S. Department of Treasury, Internal Revenue Service (1998a). Publication 534: Depreciating Property Placed in Service Before 1987. [16] U.S. Department of Treasury, Internal Revenue Service (1998b). Publication 946: How to Depreciate Property. 26

A SIC Codes for Equipment Producers Equipment Class SIC Codes 1. Furniture 25 2. Fabricated Metals 34 3. Engines 351 4. Tractors 3537 5. Agric. Machinery 352 6. Constr. Machinery 353 ex. 3537, 3532-3 7. Mining Machinery 3532-3 8. Metalworking Mach. 354 9. Special Ind. Mach. 355 10. General Ind. Mach. 356 11. O(cid:14)ce & Computers 357 12. Service Ind. Mach. 358 13. Electrical Dist. 361 14. Communications 366 15. Oth. Electr. Equip. 36 ex. 361, 366 16. Trucks and Buses 3711 17. Autos 3711 18. Aircraft 372 19. Ships 373 20. Railroad Equipment 374 21. Instruments 38 22. Other Equipment 359 NotalloftheequipmentclassescouldbeexactlymatchedupwithanSICcodeandsoacoupleofthesematchesarebasedoninformedguesses. Afulldescriptionofwhattypesofequipment are covered by each category is contained in AppendixE of Benchmark Input-Output Accounts of the United States, 1992, available at http://www.bea.doc.gov/bea/an1.htm. 27

B Construction of the Full Tax Term The full tax term was de(cid:12)ned to be (cid:20) (cid:21) 1−ITC −(1−(cid:18)(cid:3)ITC)(cid:28)z 1−(cid:28) where(cid:28) is the marginal corporatetax rate, z is thepresent discounted value of depreciation allowances, ITC is the investment tax credit, and (cid:18) is the proportion of the investment tax credit that needs to be deducted from the depreciation base. Table 6 displays the investment tax credit for each type of equipment, taken from Gravelle (1994). The parameter (cid:18) was set equal to zero for all years apart from 1962 (for which it was set equal to 1) and the period 1982-86 (for which it was set to 0.5). The present discounted value of depreciation allowances was calculated based on the service life assumptions shown in Table 7, again largely taken from Gravelle. Prior to 1981, the income stream of depreciation allowances for each type of equipment was calculated based ontheassumptionthat(cid:12)rmsclaimed allowances usingthedoubledecliningbalancemethod switching to the so-called \Sum of the Year’s Digits" method. For 1981-86, the stream of allowances for each type of equipment was taken directly from IRS Publication 534, while thecalculations for1987-1994 weretaken fromIRSPublication 946. Presentvalues of these depreciation allowances for each year’s tax code were calculated using that year’s value for the mean BAA corporate bond rate. 28

Equipment Class 59-61 62-68 69-70 71-73 74-80 81-86 87-97 1. Furniture 0 7 0 7 10 10 0 2. Fabricated Metals 0 7 0 7 10 10 0 3. Engines 0 5.1 0 5.6 10 10 0 4. Tractors 0 6 0 6 9 10 0 5. Agric. Machinery 0 7 0 7 10 10 0 6. Constr. Machinery 0 4.6 0 4.6 6.6 10 0 7. Mining Machinery 0 7 0 7 10 10 0 8. Metalworking Mach. 0 6 0 6 8.6 9.4 0 9. Special Ind. Mach. 0 7 0 7 10 10 0 10. General Ind. Mach. 0 6.4 0 6.4 9.1 9.6 0 11. O(cid:14)ce & Computers 0 7 0 7 10 10 0 12. Service Ind. Mach. 0 7 0 7 10 10 0 13. Electrical Dist. 0 4.8 0 5.7 10 10 0 14. Communications 0 4.6 0 5.2 10 10 0 15. Oth. Electr. Equip. 0 7 0 7 10 10 0 16. Trucks and Buses 0 4.6 0 4.6 6.6 10 0 17. Autos 0 2.3 0 2.3 3.3 6 0 18. Aircraft 0 7 0 7 10 10 0 19. Ships 0 7 0 7 10 10 0 20. Railroad Equipment 0 7 0 7 10 10 0 21. Instruments 0 7 0 7 10 10 0 22. Other Equipment 0 7 0 7 10 10 0 Table 6: Investment Tax Credit Rates 29

Equipment Class 59-61 62-70 71-80 81-86 87-97 1. Furniture 14 10 8 5 7 2. Fabricated Metals 25 18 14 5 7 3. Engines 29 22 18 5 7 4. Tractors 12 9 7 5 5 5. Agric. Machinery 14 10 8 5 7 6. Constr. Machinery 10 7 5 5 5 7. Mining Machinery 16 11 9 5 5 8. Metalworking Mach. 14 9 8 5 7 9. Special Ind. Mach. 16 11 9 5 7 10. General Ind. Mach. 14 12 10 5 7 11. O(cid:14)ce & Computers 10 7 7 5 7 12. Service Ind. Mach. 17 12 10 5 7 13. Electrical Dist. 22 17 14 5 7 14. Communications 19 14 12 5 5 15. Oth. Electr. Equip. 16 11 9 5 7 16. Trucks and Buses 10 7 5 5 5 17. Autos 4 3 3 3 5 18. Aircraft 16 12 9 5 5 19. Ships 28 20 16 5 10 20. Railroad Equipment 26 19 15 5 7 21. Instruments 18 13 10 5 7 22. Other Equipment 15 11 9 5 7 Table 7: Tax Service Lives 30