onom IIo'w S. I 1 -. () 0r/ FILE COPY TK1001 .M416 ..... .E56 Archives TECINOLOGICAL CHN:CE, TAX POLICY, AND THE N S T. TEc8 . AUG 161983 DERIVED DE.l-AND FOR ENERGY LIBRARmW by Ernst R-- Berndt David 0. Wood* Energy Laboratory Report No. MIT-EL-75-019 First Draft .Comments Welcome Not for Citation *University of British Columbia and Massachusetts Institute of Technology. The theoretical development in this paper has drawn heavily on E.R. Berndt In addition, we gratefully acknowledge the and D.W. Jorgenson [1975]. Byron... Neil of research assistance AUG 16 1983 L' / AR MITLibraries Document Services Room 14-0551 77 Massachusetts Avenue Cambridge, MA 02139 Ph: 617.253.5668 Fax: 617.253.1690 Email: docs@mit.edu http://libraries.mit.edu/docs DISCLAIMER OF QUALITY Due to the condition of the original material, there are unavoidable flaws in this reproduction. We have made every effort possible to provide you with the best copy available. If you are dissatisfied with this product and find it unusable, please contact Document Services as soon as possible. Thank you. Due to the poor quality of the original document, there is some spotting or background shading in this document. __~~__ _ __ __ _~___ ~--- ~.~------.. 1 ~b _____ IntrAteuct ion The aszuription of Hicks-neutral factor aucmnenting technical change has prcved to be extremely useful to empirical researchers interested in applying production theory. technical change is Hicks-neutral, ~z: en the analytical. task is con- siderably simplified; for example, substitution possibilitieamong inputs are independent of the level of technology. There have been a number of studies which have attempted to test whether the Hicks-neutral technical change (HTC) asJumption is valid. Almost all of these studies have dealt with traditional two input CES duction models. or Cobb-Douglas capital-labour pro- In recent years more flexible functional foraTms with multiple (more than two) inputs have been estiated, although almost all such studies have assumed Ricks-neutral technical change. The single published exception is P. Binswanger [1974a3, that of HanZ who estimated a multiple-input t anslog "cost function using U.S. agricultural data. Binswanger did not derive his functional form and estimating equations from an explicit statement on the forn of technical change; rather, he simply inserted 'time' into his cost share equations and, baSed on the sign of the estimated coefficient, observed whether technical change was input i share-saving, share-neutral, or input i share-using. In the first part of this paper we develop an explicit theory of factor augmenting (input price diminishing) technical change ard inc.orate it into the trianzlc func;-ical We test whether technical change is -actor augmenzing; _ d-=- -2that technical change is of the constant factor augmenting form, we assume constant returns to scale and simultaneouslf identify and estimate the techn:ological substitution parmameters and the rates of factor augmentation for each input. Finally, we test trather technical change is Hicks-neutral. The data we use in this study is that on capital (1-), !a- bour (L), energy (E), and other interediate materials (M) for total U.S. manufacturing 1948-71. This data was eimployed i; our earlier study which assumed H -TC. The data is of consid- erable interest: for a salient result we obtained in our earlier study was that energy and capital tended to be cmplewentary rather than substitutable inputs. In the present study- we test for INTC and assess the reliability of our earlier re-entarity. sult on capital-energy co.pr We find that over the 1948-71 time period in U.S. raanufacturing, technical change has been energy-using and labour-saving; further, the energrcapital complementary relation persists when non-noutral ten:Ch- nical change is allowed. We cannot, however, reject the n'ull hypothesis that thichnical chang. has been Hicks-neutral. Thus D= results provide some rather weak support for the hypoth.-esis that the post-war increase in demand for energy is d:.e Partly to the energy-neing bias of technical change. : si :e we cannot reject HNTC, applied researchers may feel justified in continuing to simplify their work by assuming ENTC. An additional reason we provide for the increase in zostwar encrgy demand is that it .icy. Specifically, is due partly to federal tax pl- since energy and capital are complement.ary, federal investment incentives have tended to lower the price 074674 of capital and thecrby increas both the capital and energy in- tensiveness of U.S. manufacturing. In the final portion of the paper we produce simulation results which attempt to depict what the pattern of energy demand would have been like had the postm war inves- t .ent incentives not been implemented. - 111M, -3m II. The Translo. Function with Factor Auamentina Technical Change Assume there exists a twice differentiable aggregate pro- duction function characterized by constant returns to scale, (1) Y = F(K*,L*,E*,M*) where Y = gross output K* = capital services measured in augmented units L* = labour services measured in augmented units = energy services measured in augmented units E M* = services of all other intermediate goods measured in augmented units. Corresponding to (1) there exists a cost functicn which reflects the production technology: (2) C G(YrP*P,,P'P where C is total cost and P*, Pf, P* and P the augmented inputs K*; L*, E* and M*, are prices of __ _n_____ I ~ ______~Z=~;~~_______q_____S__ 4We now assume that the cost function (2) 2 approximated by the translog function: (3) In C = In a + + hBKK (in I n Y + aKln P* + aLln P* + aEln P* 0K P L In + TPIn P!S + i xf . + hLL~nhL 2 .- -q k Liri4 - i ,- LE KE 4 in P* "AM P1i -in P* -i e 1 *M E + 8IEln P-ln P* + + ~iS Linear homgen: can be closely BE (ln P*) E 2 in In A. T Kt P*~n P* + B n P*In P* + 4e (In P.) EM 4M E . es imposes the following restrictions .ri on (3): (4) L + aK + C + E + CM =1 OM + OL + BLL + BLE + KE + BE+ BKM + 8M+ = + ' E + -=0 l = 0 M = 0 0 EY + 0B. I 4 = 0. From the point of view of the production function, it has been traditional to specify certain types of technical change as factor augmenting. Viewed from the vantage of the cost func- tion, such factor augmenting technical change corresponds with S . .- I (-, *CC% C.5 .. '- , g4,,. -5- input price diminishing technical change. Let us assume that factor augmenting technical change occurs at a constant exponential rate: (5) ... Xit = where X =- services of input j at time t measured in augmsnted units, Xjt = services of input j at time t measured in n atural units, S constant e-ponential rate of augmentation f cr in- put j T ,= t-to, where to is an initial point in time. The corresponding specification for input price dimlinishing technical change is (6) P3t =Pjte where Pjt and Pjt are the prices associated with Xjt and ::, respectively, and Xj is the constant rate of price dinminution for input j. .. -6"It is of course desirable that Pt X = P X , i.e., jt jit t jt that input value be invariant to units of augmentation mea3ure- ment. 1A For this to hold, it is necessary and sufficient that =-Pj, i.e., that the rate of factor augmentation is the negative of the rate - f -df Inserting (6) 5-rZL-n into (3) and utilizing the restrictions in (4), we rewrite the cost function taking account of technical change: (7) In C = In o + In Y + aKln P + %$,(in 2 + + in PIl aLln PL + an P an + iT PL + SBln PK n PE + 11 I " + 14L(In P )2 BLE n n + Ln PE + h4B (In P,) LM n PL 2 (9) *T = KK LL + OE'E + = S~kK +K LL - SKK0K1) M +KL + MM OKE E (AL M PM + BITT In P. + 4 where (8) n + BEM+ln P 1n PM P gE( + n PKI n PM + RMxEE KE E M + 0 + LT E 1"2 T In PK in PL In PE I . KLE (XK-M) SLT + 7 - s(LL(A-M) ) LE (ILLT + !BT + PT and 0 ) 4 0E(XE- ET)' - (10) + BLE(E-XM) U = KTA1K TTT + BLT L + $ET E + BIT M BKT(AK-M) + BLT(L-XM) +O ET0(E-AM) - We now derive.estima'ting equations. If input prices and output level are fixed, cost minimizing input demand functions are obtained by logarithmically difsrentiating (7), 8an C -81- P I aJ + jk S In Pk + 0 jT = KL,E,M. UsingShephard's Lemma, we obtain (.1) QM C M ax + .= BK1n PK + RL 1n SPL 5r'---= ,,,. + BELn PK $. + a. Eaa aE + PEE - - - E + = PL + n PL KE I n PE .+ B0KMln 0 LE1n PE + 8LIn 2 M + KTT M + BLTT XKEln PK + O:T.ln PL + KE1 n PR +OLEIn PL + BEEln PE + 8EL4In PM + KM PL + PD + in BEMI n PE + Oin P.+8111 ETT BMM1n PM + 'T M 4-' PMT -8- where the total cost C = P + PL + PM. +PE The M are of course the cost shares of the inputs in the total cost of producing Y. We now differentiate (7) with respect to T: (12) nC aT T + P + XT n K Tn PL + 8 LT L PIT I n PE + 8b In P + 6T Equation (12) can be rewritten by taking account of the relations (), (I), and (10): a1n C S ~e erate of total cost dimisnution (output fixed) is e..al to a weighted average of .the rates of input price diminution, the weights _einfictor shares. In practice, we shall mea- sure a1n C/at as the negative of the rate of total factor productivIty.3 If rates of input price diminution are equal for all factors, i.e., KT 0LT if - 8ET j = Ak, j,k = K,L,E,M, MT = TT 0. then by ( and (10) a) We term this form of tech- U% nical change Ilicks-neutral. From (11) it is obvious that when v." technical change is Hicks-neutral, cost shares are independent of time; further, the rate of total cost diminution will then - d" bic~ ~ a TT- 111,11111111911 -9simply be the common rate of input price diminution, i.e., aT = j = Ak , J,k = K,L,E,M. Thus Hicks-neutral technical change involves a set of testable parametric restrictions. If one interprets the cost function %7) as a function in its own right, it is possible to test for a weaker form of technical change, namely that technical change is simpl tor augmentiLg. This is so. because factor augmentation in- volves repararmeterizations restriction in fac- (8) and (9), but imposes the single (10). The above framework allows us to test for other types of technical change as well. 4 For example, Harrod- neutral change implies Ag = AE = )N = O, IL 4 0; Solowneutral change implies AL = AE = I M = O, Kr 0. Other forms can easily bu developed. To our knowledge, the first published empirical study of technical change biases with many factors of production and 'flexible' functional forms is that of Hans Binswanger [1974a1; he also uses a translog function. Binswanger esti- sates only the share equations (11) and defines technical change as input j-saving if and input j-using if 8T > 0. T < O neutral if jT = 0, Thus his definition of biases is in terms of factor shares, not in terms of factor augmentation (input price diminution). Because Binswanger esti- iates only the share equations (11), he can only obtain - I0 - estimates of differences in rates of augmentation; from (s), for example, it is clear that XK , AL' AE , and XM cannot be separately identified, only the differences XK-XI, and I ').1 are estimable. Identification of- M L-M' and thus of the remaining Ai can be attained, however, by estimating in the rate of total coSt diminution ecr - (see eluation (8) tion (12) T AZ , . a t E3---- returns to scale in input.quantities ) constant ct o simultaneousiv estimate and identify technological parameters and rates of augmentation for each input. 5 Estimates of the Allen partial elasticities of substitution (AES) bet 1 ;.een inputs j and k are obtained from the translog parameters and the data: . + M -M (13) . jj (14) °jk ,' , 6A9 + M M. , jJ K,L,E,M, Jrk D - ,LE,M. The corresponding price elasticities, E jk are computed as Mj al Xj/ amn P , k. A useful interpretation of the translog parameters 8jk is obtained by defining a share elasticity. The first partial derivative of In C with respect to in P equation (see (11)). is the share The croas partial derivative, 22 1n C/8ln P jln Pk is a constant share elasticity equal to Bjk. Thus the translog parm.eters 0 jk sunm-marize the response of cost share Ij. to a chang3 in In Pk" We now turn to a discussion of econometric concerns. We propose to estimate parazmeters in the cost share aquatic-s (11) and the total cost diminution equation (12) subject tc the restrictions (4). To each of the equations in (11) and (12) we append an additive distu rbanc ts-. in. e of the four equations in (11) are 2±depenen drop the M share equation. A o t're we arbitrar_:y We specify that the distuizbarnces from the remaining three share egiations and (12) form a 421l disturbance vector c(t) swhich is teciporally independently and identically normally distribu=ed with muean vector zero and non-singular covariance imatrica 1l, t=l,... ,T. The final issue on stochastic specification concerns the issue of simlntaneity, Two considerations suggest that the aggregate prices PK' PL' PE and PM may be ccrrelated with components of the disturbance vector r(t). First, although input prices may be fisd at the level of an individual firm, the extent of aggregation in this study (total U.S. manufacturing) is considerable; at this industry level, it is less likely that the supply of inputs is perfectly elastic. Second, even if, for example, the supply of er. i -It , coal, - 12 natural gas, crude and refined oil were perfectly elastic at the industry level, the price index of aqqregate energy would be endogenous, for it would vary with changes in the mix of energy inputs demrnded. Similar arguments could be made for the endogeneity of PK' PL and PM. veloped by Fuss [19753. This argument has been de- If our estimation procedure is to provide consistent estimates of the parmeters in (12), (11) and account rust be tJaken of this possible simultaneity.v As in Berndt-l.ocd E19753, we again use an instrumental vaziable estimator. Our estimator is the iterative nonlinear minimum distance estinator for simultaneous equations; when the model is linear in the parameters, theqminimu'm distance estimator is simply the iterative th-ree stage least squares estimator.7 Statistical inference is based on the Wald test procedure. Before presenting epirzical results, we refer back again to the estimated Allieu partial elasticities of substitution (AES); see (13) and (14). Since the etimated S are a function of the estimated parzn.eters and data, it possible to a-; should be derive their estimated standard errors. The problem, however, is that the fitted Mj depend on the parameters a , jk and the data. The foimula for the stan- dard error therefore is rather complicated. 9 Estimates of these standard errors could be quite useful, however, for 11116 -I - 13 - they would provide insight into the precision of the estimated AES.10 Thus it appears useful to e:pend some e eet in obtaining standard errors for the estimated AES. = PL Suppose we scale our data so that P T = 0, in som,-year (say, 1947). PE P = 1.0, The fitted or predicted M in that year would then simply be the estimated a j, for all the 12 Pj and T terms would be zero (see (11)). Now substitute cj for M. in (13) and (14): =(13a) ~jj ("3" jk " - k k + ajo K,L,E,M = KoLEI . + Gja, S(14a) Thus for the year 1947, the estimated AES would only be a function of the estimated parameters. Denote.the column vector of first partial derivatives of the AES in (13a, 14a) with re- spect to the parameters c~, jk as SB --and the estimated variancecovariance matri .. of the parameter estimates as V(B). Follow- ing: Ementa [1971, p. 4443, a first order Taylor series approximation to the variance of the estimated AES is then computed as (15) SV (B)S . - 14 - Since fitted M. and estimated AES are invariant to data scal.ing, the data could be appropriately rescaled at any an-nual observation in order to obtain estimates -* the standard errors for the AES at that point. bo .L. • - -. - Il nllh - 15 - III. Data and mri-rical Results on Technical Chance The data used in this study is the same data used by us in our earlier analysis.11 The additional data required here, however, is a time series on total factor productivity. Based 12 we separately compute Divisia indexes on the Faucett data, of total output (sales of total manufacturing to all sectors of the U.S. economy, including final demand) and total factor input (purchases of total manufacturing from all sectors of the U.S. economy, plus labour and capital services). Follow- ing Jorgenson-Griliches [1967, 19723, we obtain a measure of total factor productivity as a Divisia index of inputs. * Since the Divisia index is a chained -index,we lose one observation. therefore " Divisia index of output minus The data for this study is the period 1948-71. In Table 1 below we pre- sent the index of the rate of total cost diminution for U.S. manufacturing A is computed simply as the negwhich ;, ative of the rate of tctal factor productivity. If technical "progress" has taken place, the expected sign of the rate of total %ost diminution is negative. Table 1 Rate of Total Cost Dimrinution in U.S. Manufacturing, 1948-71' (Denoted as C/C) Year 1948 1949 1950 1951 1952 1953 Year YC -.01525 -.00500 -.01124 -.01700 -.00029 -.00759 We first 1951 1955 1956 1957 1958 1959 Year -~Q.,555 -.02376 .0115 .00782 .00667 -.00893 1960 1961 1962 1963 1964 1965 6c/ -.01639 -. 00939 -.02443 .00077 -.03181 -.00815 Year ;/C 1966 1967 1958 1969 1970 1971 -.0095 .00577 -.00376 -.01406 .01241 .01256 directly estimate parameters in three of the four share equations (11) (12). /C plus the total cost diminution ecuation Although there are twenty-four parameters in these four equations, numerous restrictions must be imposed. For examrGe, *KL in the K-share equation must equal KL in the L-share equation; similarly, 0B in the X-share equation must ecual 8KT in the total cost dimLinution equation. There are ten such symmetry restrictions; this reduces to fourteen'the nnmber of free parametersto be estimated. 13 In the first 2 we present parameter estimates and t-ratios with these ten sjymetry resqAbctions imposed. col-n of Toble for the model It should be noted that this model does not impose the restriction that technical change is of the factor augmenting (price diminuting) form. ~ I ~I^ _~ _____ ^ -- ~gill Table 2 Parameter Estimates fcr Translog Model with Technical Ch.ange U.S. Manufacturing, 1948-71 (t-ratios in parentheses) Factor Au ymentation and Svrett;r7 Parm n~t r .0593 (34.002) .2587 (65.100) 14* K BKL OKE OKM SLL OEE EMI Sp '*~' - .0433 (28.555) .6387 (114.624) .0381 (4.708) .0163 (1.151) -. 0073 (1.612) -. 0471 (1.979) .1449 (2.021) -. 0267 (1.329) -. 1345 .(1.574) .0429 (3.378) -.0090 (.432) .6 .1906 (1.765) -. 0005 (1.518) .0594 (35.348) .2589 (65.742) .0434 (28.850) .6382 (116.724) .0381 (4.735) .0162 (1.148) -.0073 (1.622) -. 0470 (1.997) .1435 (2.014) -.026 (1.334) -. 1331 (1.576) .0429 (3.394) -.0090 (.437) .1891 (1.772) -. 0005 (1.453) Hicks Teutral Techiical Change and S,-nt:i .0589 (37.654) .2541 (99.950) .0443 (33.375) .6427 (155.139) ..0329 (4.781) -. 0052 (1.257) -. 0083 (1.982) -. 0193 (1.679) S .0721 (9.143) -.0045 (1.370) -.0624 (4.983) .0265 (3.096) -. 0136 (1.015) .0953 (3.342) .0 Paraee'ter S-~ etrv -. 0018 (1.175) .0005 (1.160) .0018 tLT BMT (.958) ON -.0018 .0005 (1.068) .0018 (.975) .oooo -. 0059 (1.891) (2.071) I UI UI R2 -m Share .5953 -L Share S-E Share .8007 .7347 ;-14 Share -t/C .6177 .0383 fqicks Neutral Tecynical Change and S retr_.0 (1.093) .0004 (.976) .0101 Ul itote: Table 2-continued Factor Augmentation S znd Sremetr .0 .0 t.911) -. 0105 (1.909) -. 0137 (2.840) .0018 (.299) -. 0028 -. 0060 (2.321) -. 0060 (2.321) -. 004o (2.32.) -. 0060 (2.321) -.0060 (.888) (2.321) .5950 .8005 .7344 .6175 .0005 .6057 .7824 .6980 .0000 UI denotes unidentified. '' - S-19 To impose and test for factor augmenting technical change, we reparwneterize the Sjt in terms of the Sjk and j, T in terms of the aj and Aj, and then impose the additional nonlin-ear restriction (.C, TT in terms of the The chi-square test statistic for the-single Zjk and the 1j. restriction in which reparaeterizes (10) is is 3.841 (6.635). 4, while the .05 (.01) critical value Thus we cannot reject the null hypothesis of factor augmenting technical change. It is of considerable interest to examine the estimated rates of factor augmentation (input price diminution). From the second column of Table 2 we obtain estimated annual rates of factor augmentation (the negative of the estimated XA) as 1.05% for K, 1.37% for L, -.18% for E, and .28% for M. The estimated is significantly different from zero, and AK is marginally significant. The high estimate for labour is somewhat surprising, because the hours worked data series has already been adjusted for changes in educational attainment over time.1 6 An intriguing result we obtain is that the rate of factor augmentation is larg*-st for labour, the input whcse •price has risen the most, and is lowest fnr energy, the input whose price has risen the least. Indeed, the estimated AE suggests that factor augmentation for energy over the 1948-71 period has actually been negative (-.18%), 2 timateis insignificantly different from zero. this esThus our re.- sults suggest that part of the post-war increase in industrial demand for energy is due to the fact that technical change augmented capital, labour, and other intermediate material - 20 materials more rapidly than it augmented energy. In this sense ; the historical technical change can he characterized sonmew;hat . loosely as labour saving and energy using. An alternative mizthod for analyzing is c ical change bias to measure the effsct of technical change on equilibrium factor cost shares. This method is more involved than the simple measurement of rates of augmentation, because the net effect of technical change depends on the technological substitution parcmeters and differences in rates of factor augmentation. For example, the parameter L (=SX 0- M) sunmarizeS the effect of technical XLL(AL-M) + 8LE(X'E-M)) change on the labour share, taking into account differences in rates of augmentation and technological possibilities for factor substitution. Fronma Table 2, Coltun 2, we note that the estimate of 0 LT is -. 0018 (labour share saving) which is .insignificantly 1 is statistically different from zero, sigrnificant. (and t-values) fcr ST, SET' (1.068), and .001(.975). even though the estimat d The corresponding esti.ate3 and a are-.0005(1.453), r .0003 Thus the pattern of technical cnange appears to have been labour share saving and senergy share using. It is tempting to suggest that our results lend supportc to the induced innovaticn hypothesis, which states that factor prices significantly affect the rate and bias of technical change. We tend to resist this temptation, simply because we ii - 21 - believe that the process of research, innovation, and technical change is extremely complex. Hans P. Einswanger [1974b2, for example, has considered invention possibilities on the basis of potential payoffs to alternative research lines. In his type of model, the biases and rates of technical change are determined jointly by (i) the relative productivity of alternative research lines (for example, If labour saving research results'are easier to achieve than e._ y saving ones, then technical change will tend to be labour saving), the relative (ii) costs of alternative research lines, (iii) the size of the research budget (which may be affected by scale of output), and (iv) the relative present values of expected factor costs. Although our results suggest that technical change in U.S. manufacturing 1948-71 has been labour saving and energy using, it does not seem reasonable to attribute this bias to a simple induced innovationIypothesis. Indeed, since our model assmes that factor augmenting t2chnical change has constant rates and biases, we cannot really address ourselves satisfactorily tc the empirical issues concerning varying endogenous rates and biases of technical change.7 We now examine in further detail the strength of our result that technical change appears to have been labour saving and energy using. To do this, we test the null hypothesis that the rates of factor augmentation are equal, i.e., that K = L = E = Hicks neutral. SLY LT ET = , which implies that technical change is Note also that when AK = AL XE = XM' OKT 0 i.e., technical change is neutral with , = 0, MAT r 'V - 22 respect to factor shares. the three restriction The chi-square test statistic for \.2_2 3 = A =L is Ih, hie the .05 (.01) critical value is 7.815(11.345). Thus, although point estimates suggest that technical change has been labour saving and anergy using, we cannot reject the null hypothesis that techn-ical change has been neutral. This result is of some comfort, however, for most e.pirical an.alyses (including our Berndt-Wood [319753 study) have assumed that technical change is of the Hicks-neutral factor augmenting form. Before leaving the topic of technical change, we note briefly that when technical change is assoumed to be .icks neutral, the estimate of the common rate of factor aumenk-tion (see column 3 of Table 2) .a is 2.321, is .60%; since the t-valuh this rate of overall technical change is icantly different from zero. on si-ni - The figure of .60% as an annual . rate of technical change may at first glance appear a bit lo. Spencer Star 19074], however, has shown that previous stud.i s measuring technical change have tended to croduce somewhat larger figures because they -- "m .t v.. aded 'capita .and labour), and ignorelintermediate materials.18 Specifi- cally, in his study based on 1950 and 1960 census data for seventeen U.S. manufacturing industries at the two digit level, Star found that the overall annual rate of technical change was reduced from 1.51% to .59% when h/.ong other thin.,s, .intermediate inputs were property taken into accunt. __ IIIINIMIili fiYil - 23 S. . Star's figure of .59% is remarkably close to (= our estimated ) = of .60%.. We now turn to a discussion of Allen partial elasticities of substitutinn (.}S). For the sake of brevity, we report esti- * uates o. the AS for ohly the most restricted model accepc'E bby our data, i.e., the mrodel with Hicks-nautral technical imposed. h . . ochange mated a We note in passing, however, that the estt- basel on the syr.metry and symety pls fatrctor- augmenting technical change parameter estimates always in - ated that capital and enery were complementary inputs. ,; The energy-capital complentarity was of course a salient .-.; result of our earlier paper, a study which assumed FHicks". neutral technical chanre. - ". In the first three column-s of Ta'1e 3 we present est-20 mated AES for selected years (1948, 1960, 971) based on he the parameter estimates from column 3 of Table 2 -- model with Hicks-neutral technical change imposed. Using the linearized appro::ia-tion as discussed in Section II above, we also corpute and present i. Table 3 estimated standard errors and t-ratios for the AES. The estimates of the own AES are all negative; aM 21 the and aLL estimates are significantly negative in all three years, while UKK and a. estimates tend to become less sig- nificant statistically tow;ards the end of our sample. The _ , estimates (around .64) are significantly different from zero, but not significantly different from unity. %A EstLma-tes of aKE becor.e increasingly negative and statistically significant towards the end of our samzple; a 953 . confidene interval, h-oweer, wsculd inzlude zero and slightly nificartly of aLE and aI are sig- tU.es & . cr positive val'jUs! ositive, while tites St fica n .y dlfrent .tcii itive but and a, .22 " zero.22 " from o~ finajlly, we copare our przent r. u~s reported in c- are pos- -with ho:e .~_- n the fout h 2olumn of ea.-ier effor'.: Table 3-we rep~acca estimat4d 1 S for 1971 frcm Table 4 o Berndt-ood [19753. esults are sub-stant'.ally the T1he atimUates of o 7 OLL and 0o. in the pre3snt study are sli...- ly smaller (in absol'ute value) than thori the same t3:sad is obs 2ved fo ard L~ same. of our first stu. the estimates of cuL, tKE' 9..i' Th-q're-nt eutimates of and ..... ., - are slightly la:ge_- than those of the earlier analys;.'. DL.-* ferences in the -wo sz::; of r-e.lts can Ib attributed to t ;.: variationz: '(1) In the present study we add a total cost bninuti ,n if dutonr disturba.nces in ~ltis equation are *correlatid with .i3sturbances 2 ge~neralWT "t e s harc eqvatios, ""rical estimates will diff r when the total c.. dimjiution equauion is .added. in presient.and if toto- efficienc .f such residuia e sarple s:.~za were the se, corre..ti.' the asp:. of results in this paper would be greater stady. 2 3 than those of our -arlir The simpe correlation between re-iduals in the total cost diminution amd th equations is , 2.162 , -. 0S(L), *216(E), the residual correlatioas are largest (in in the ir end E equation-. shar : and -. 047 (). Tht.: absolute value) (2) Because tho present study - .25 - employs a total cost .iminwutionr measlure based 3n log-diff-er- enced Divisia indeig procedures, drop one obseri- atin. it h-a been necassary to Che reLults in this pap3r aze based on 1948-71 data, iwhile those in our earlier study were base-d on the years 1947-71. This small change in data base coul produce a slight var-ation in the parameter and. 1E estimates "' "T-.ble 3 Esim.ted A .. .ate Translog Kodzi -: (,. E.".st'c.5tiS3 of c" --..- -:n ms the C ,-? itU- posed .tanda-d %.zor) 197. 1948 * 8E. t-ratio 960 .2.77.7 -8.33 (2.3 =) 2.657 (3.45) -. 55. -1 73 C 8 .E. * (.11.) -1.53 '.110) . t-rciti 15.570 15.2 i3 AE -8.403 -0.03 " (4.63) I T: 7 8.E. t-'ratio 8.E. (3.E62) 2.3~.I -. 338 (.070) a.i t-ratio S.E. c ,5 33 .658 2.142 I.Co c 7 2.2S4 -2.040 (1.731) -10 "-.355 -. S2 (.073) (.C74) 4,835 t-rad& (..07) 1.955 -2.420 -3.53 .447 .438 .327 t-ratio (,2323) 1.32 (.333) 1.317 .626 S.E. S.649 (.272) 2.375 (.2L8) 2.178 (.20) 2.452 .622 .637 .655 (.073; (.0059) (.075) -. 39 .530 (:.973) t-ratio . 5.152 (1.CI6) 1,311 Ciz S.E. s;tim ts4 1971 -6. 3:!3 (2.3 2) 1.433 .49 .68 .61 S.762 .511 (.4-2' 1. :.. -Ion 1.C -$ 3..3 ) .510 .S t -. .. 73 i - ----lr-- ~ __- -~--~.p-------;; ~-~C=-iSi-i~--- ---- -- ~--il~L__ .5 - 27 - • • • IV. Simulated Effects of Variations in Tax Policy on Enercy _ --- Demand In the previous section we presented empirical evidence rWhich provided some weak support for the h: _thesis that the . posTar increase in industrial demand for energy was due in . This result part to biased (energy-using) technical change. S s not conclusive, for we are unable to reject the hypothesis that technical change was Hicks-neutral. . We now assune that technical change is • 'nd-i :: examine the effects of certain tax policies on the deand for energy in U.S. ".. . Hicks-neutral, manufacturing. Specifically, we investigate the effects on energy demand of: --- (i) .- removal .* f the investment tax credit (w-hich took place in 1962) and ..a return to the pre-1954 accelerated depreciation allowances S.on producers' . ..- tures; and (ii) -- * " -Our durable equipment and non-residential struc- '' imposition of a 5% tax on energy. simulation procedure is similar to that of Hall- S orgenson L1967). Specifically, using the formulae for cap- :ital service prices as presented in Berndt, Kesselman and -Williamson 119753, we estimate the capital service price S(denoted as PK) had the investment tax credits and post-1954 5 -changes c. ' in depreciation allowances not taken place. .Table 4 below, we compare the two price series. * fN' In ''i There it r is ~, seen that removal of these investment incentives would have implied a sizable increase in the price of capital services. r ~ .. ; 55 ~ ;. i~r 75 ~ ~ 6'. 5; ~-. 2, r r . Table 4 P ) and Tax-Adjusted (PI) Capital Service Prices amnufacturing, 1948-71 ,U.S. Actual Px Year . 1948 S.-1949 1950 1.00270 1951 1.04877 .99744 1.00654 1.08737 -1953 1954 1.10315 1956 .99607 1.06321 1.15619 1.30758 9-57 2959 . "195 . Tear 1.00270 " " .74371 .92497 . 1.04877 .99744 1.C0H54 1.18321 1.20993 1.10043 o1.1S42 1.30087 1.50035 .74371 .92497 .-1952 p... i . - _i 1960 1.25413 1961 1962 1963 1964 1965 1966 1967 19681969 1970 1971 1.26329 1.26525 1.32294 1.32798 1.40659 1.45100 1. 33618 1.49901 1.44957 1.32465 .'1.20178 •3. . .. "**. . - °.. 1. 43750 1. 43 401 1. '5724 53838 . * 1.1 1. 71959 1. 75797 1. 65735- ", 1. 1. 7-625 1. 53204 1. 43467 • ** - - 3 :: :0k. . .J • T.o sIbalate the effec-ts of this input price change, we se parameter estimates from Column 3, Table. 2 and insert them sIt ... . t n .with the new input price data into (7). . This provides s with A new price of output 'since with zero profits and consist - returns to scale, marginal cost = average cost = price of ct-.. put) z ries. We then solve the share equations (11) for K. L,. ' , and M in terms of the new P, historical output V, the ero:. "genous input prices, and the parame-ers. -. - eat result S..... years. centives -. .mandj ~~ In Table 5 we pre- from this simulation (Simulation I) In 1971, for examp for selectd removal of the investment in- Awou haBM resulted in a 2.4% decrease in a use of the energy-capital compleentaitt energy dci capital . - -- '~-Yull --- - 29 - K,)~ demand would have decrease 5.0%, while the quantities of L and M demanded would have increased .6 and .3 of 1%, respectively. Thus removal of the investment incentives would have resulted in a less energy and capital intensive produc-tion process. In our second siauilation, we obtain derived demands un-'er the ass,~mticn that in addticn to removal of the invest--ent incentives, a 5% tax was imposed on energy. tive scenario which might hfave ~ to An alterna- increase in P is the renmoval of price ceilins s on cert in energy types (e.g., na&ural gas). The joint effect of th--- :o tax policies in 1971 is to decrease the quantity of E demanded by 4.1%, de. crease K by 5.6%, increase L by .7% and increase M by .4%. In our final simulation (Simulation III), we take acScount of output demand conditions. Specifically, the effect: of the removal of investment incentives and the 5% energy Stax is to increase the cost of producing output. Given zea.ro profits, this implies an increase in the price of output ('. -(denoted aP). If output cqantity d-mandd is price sensi- tive, the effect of the two tax policies is to reduce the quantity of output demanded. For convenience, we assume an output price elasticity of -1, which implies that revenue would be unaffected. Taking account of these output price effects, we find that in 1971 the effect of the two tax r;- S 30 policies is , to - . -- ..--.--. . d--- - . . . . - -~ ..... ., e -i=Pase '." price of output by? 1. % and reduce output demand by 1.0%. The de- rived demands for E, K, L, and M fall by 5.1%, .3% 6.5%, and .6% respectively. We conclude that because of the energy-capital ccmplementarity, investmrent incentives have contributed sig- nificamtly to the post-war increase in industrial dmand. for energy. .1 Table 5 Effects of Changes in Tax. Policy on Price of Cutput and Derived Demands for Inputs in U.S. Manufacturin-. muirlation I: Year 1954 1960 1966 1971 AP AV .005 .0 .003 .012 .0 .0 .009 .0 AK A1 -. 029 -. 049 .003 -. 011 .C02 .005 .004 -. 074 .008 -. 018 -. 027 .-. 050 .006 -. 024 .A03 .003 .GC6 8iul2tion I i . 1954 1960 1966 1971 .007 .010 .014 .011 .0 .0 .0 .0 -. -. -. -. 034 054 078 056 .004 .007 .009 .007 -. 029 -. 008 -. 010 -. -. -. -. 041 063 090 065 -. 003 -. 003 -. 004 -. 035 -. 044 -.- -. 05 1 -. 051 -,.'06 .76 -. 035 -. 042 -. 041 .C4 Simzulation III: 1954 1960 1966 1971 .007 .014 -. 012 .011 -. 010 -. 003 C4 SmSimlation7I: Removal of Post-1954 investment Incentives using Eistorical Cutput Smlation II: Rea.oval of Post-1954 In-estment Incnti -s and IZiposition of 5 Energy Ta, Eis .ozfcal Output S&ulation III: ,.emcval of Post-154 -Invetment Incent:.ves and I positTion of 5% EneiegyU Tax, COttpu Quntity Demanded Subject to Unitary Price Elasticity. Footnotes i1. See, for example, P.A. David and Th. van de Klundert [1965], and the references cited in Hans Binswanger [1974a3. S ~\ t * 2. The translog function can of course in interpreted as a second o-rder approximation in logarithms to an arbitrary function. For further discusicn of this, see Berndt-Jozgenson [19753 and W.E. Diewert 119743. 3. For further discussion. see W.E. Diewert (1975a,b), Jorgenson-Lau [1975, Jorgenson-Griliches [19(7, 19723 ota ftor pr-AoducTh e i4,6x and Nadiri [1970. .neex of outtivity is typically co-mputd as a Livisa puts (in constant dnllars' minus a Divisia index of inputs. 4. For further discussion, see Jorgenson-Lau S5. This reveals the error in Binswanger (p. 964): "The substitution parameters of the production process have to be estimated befoze any biases can be measured." Binswanger's statement is true, however, if returns to scale are unknown. 6. For further discussion of share elasticities, see Jorgenon-.Lau [1975]. 7. For f-uther discussion, see Berndt-Hall-Hall-Hausman .1974). 19753. .. 8. See A. Zellner and H. Theil [19623 and E. Malinvaud [1970], ch. 9. It is worth noting that the Wald test statistisLc. is asymptotically equivalent to the likellhood ratio cziterion. j The method used by Hunphrey-Moroney [1975] is incorrect, for it fails to recognize that the M :are functions of --tt.- stochastic parameters. 10. **"11. 12. The standard error -information would be particularly useful in assessing, for example, if estimated own AES which happen to be positive are "significantly nonnegative. See Berndt-Wood [1975), Section III and Tables 1 and 2. The prices are scaled to unity in 1947. See Faucett [19733. I 6. . *m 00 .0m * * '4: O 5 Lr4a O l odrtD(0 * Dois (M (D (f t (D r afo * 14 (Drnu"4 k 0 ot 0rt2 .3 (A 0 r! ft *t On t NI 0 r J0'1 O r; In P (A 4* !$ I, Dj40 (D P. a AoWM isp ' as ra 0V( ) ft 11D (0db' tr . J p vN toa 1 0 n 2 C- tti.4!:1 4D Di totC~ 10 C* 0 0 rk 4 ( L-41 t 1im 0 0 n rt v0 -1 c, t4 t0 ! I mIS ::r.JW (D 19'dCV0** s 1 -- 1-p) 0ur*r mg PMt'3 bi 0 Q -j D t W 0 rt srp *w M I'M 1- 1h m 0i 10to o:0-6 I-A% C4 Mi 4- t- 6J ato01 rt a. 1 * tp. Al l) 0 (1) * P Clt@1 ftI-I0C) ('C cf-t-I mC 0 C0)0ll P.L%.QO. r*I, Ht r-(D p. t N 0 t Po M' 0 i $bHs 0, (D in fl 1* ( -r rO0DALID P (D 0 IM M 110: ' 0 * rIO 9-t I? (1) Cj oAjlp (+119o 6 (D:to 1D n Q Sd %D Dt-tf ID M oM arri* P. t rdW(4 0C1 b (D' q 0 WiCf rt 4oIJ( :I 0 T ri inDM P* VI -V Ut1 (0 0 O Li 0 0 ID IC) 0 "4 Of r-a til PM 1 rto (i1 sU .4 ,* L-J ID CI) HQ.e C O L-j Pa1*H rn 01. (D W H ILI 0 'Ohl P 0 r1%to 0 to(' 0D Pi %0DH ri.-ad -i a 5 '4a *0 It 0 1( c Di to'.91 (0N H fZ Na p4 .. !* 1 -*0P. -(Li*m L. pi D4 Co 0 M- Di dra0 t- 0 P* :j. ONI- Sr~i t) 0 0 :3 - r.J WML.,,* H\ 'I .4 -4 I4 tw "o ) 44 (1 % NJ 0 *..r 0 9+" f ID~'I (HC Pa $- 00 mulW in * 0 M USO0 pP)1 P ri- ridt 0 P-PDi 0, Wj rb 411 0 mO0 Ut *th (9 Pi IV CD ri- All t *1 (p WD * 0 :P1O !P-- 1-0 Oa <) to~ I m ft ) t P) ft X4 o* WOCs 4 0 0 i. C)i H 0 He t* LRDI (0t t O t- to0 0! m 1 0 * a*0 He w to r0H (1) 601 S(0 7 0 H t ( E. C) P1.(0 (0 DI 1J t 0..0 Pa )0 61 00 O r 0a Ad (4 Pi 00( D3 4 * H0 0 11CDLA( t~J Dit.S (8(D M He* V) :s o im % Sto3 W 0 I1 V Pt C Prt ( De Utto a oW) N chf t i Ol IP (0 2 , rt: P' '4: N-'t(D DI ' ,a 0SP .ri -D Pa OM(I (D 0N' CI t::JC a 'd U) .*i' N* * tr [> r-i Di 'd CD') N* 0t SI-tj Di ::J' HM P. D to (OS)to tT (D I * tb 0 Caler P t Hm:rU </4 O N t N Di '. (I(D ( 10. r+IIn (a rI to :3 0J SN ft j trat i't 0(D Hi t) (D f b* Mi ~ac ~r, O 0C+ w j ft D 1*. 60 lrt 1-j ai j a) (9 i-nt rt ct H 4 4 00 11o*irrt* Ii3O 1-01 m * : 3 -* o** rS (1 tC 4* rt r' tf ft0* 0 N' C)O 0rer _oP rtrtO Q .. N' :' *0 lU '43 r:D ( *P -. O L/ 0 u r't If0. (1t IN. ft U <.4 <aNAi-r1C ,a rI rr rt fL it( rU) As it.i 0* p. r to 6 (a A 0V ;Oa rt 1. at-) e-44t ft ft t to- aftW M* 0 'H 13 ST'e fi) t - r O ri (i3 :H . o0rs P.tt 0 t rtl 111 I-(0 *rn( P O Eq 0 (3 It P- ri * alCt 0Y U)k o 0*0 :3 t C 1 0 I1 . 0 : :1 r* s 0t+ W 09 .0 0 ..ii - 4 - Footnot-es (corntinu. d) 21. As noted in footnote 19, monotonicity and convexity conditions'are satisfied at all annual observations. 22. To conserve space, we do not report estimates of the price elasticities. The interested reader can obtain approxinrte estinates by multiplying our estimated AZS by the cost shares, Eij = ij The cst share data is found in Table 2 of 1erndt-Wood [1975]. 23. See Zellner-Theil [1962]. * ? ^~ -l^-~'~I^- '-- "" -"~ ''~~~ ---~ - - i C-)--r -- -- -r.^Mcl r-r.-~c~-.* ~..~.",_,_~,,__~~'_1~F.--r-rrp.X- -. l X1-- -^---~- Referenccs Syed Ahrad [19663, "On the Theory 'of Induced Innovation, Ecorcric Journl, Vol. 76, No. 302, pp. 344-357. Ernst R. Berndt and L.R. Christensen 11974], "Testing for the Existence of a Consistent Aqgregate Index of Labor Inputs," fA-rican Econo:..c Revie' , June, pp. 391-404. Ernst R. Pcerndt, B.H. Hall, R.E. Hall, and J.A. Hausman (19743, "Estimation and Inference in Nonlinear Structural Iodels", Annals of Sccial and -econi 4d.eaur ement, 3/4, October, pp. 53-665. Ernst R. EBerndt and Dale W. Jorgenson, (1975, "Characterizing the Structure of Technology", ch. 3 in Dale W. Jorgenson, E.R. Berndt and E. Hutson, Enery r-esources and U.S. Economic Growth, unpublished manuscript, 19.T 5 Ernst R. Berndt, Jonathan R. Xesselman, and Samuel H. Williamson [19753, "Ta.: Credits for Employment Rather Than Investment", unpublished manuscript, University of British Columbia, June. Ernst R. Berndt and David O. Wood E19753, "Technology, Prices and the Derived D..and for Energy", Pe- iew of Eonomics and Statistics, August. Hans P. Binswan-ter "1974a]. "The Measurement of Technical Change Biases with Many Factors of Production", Imer-can Economic Reiview, December, pp. 964-916 Bans P. Binswan-er L1974b3, "A Microeconomic Approach to Induced Innovacion", pp. 940-958. Economic Journal, December, John Conlisk [19693, "A Neoclassical Growth Model with Endogenously Pcsitioned Technical Change Frontier ', Eccnmi Journal, 'Vol. 79, No. 314, pp. 348-362. P.A. David and Th. van de Klundert E1965], "Biased Effici-ency Grcwt- and Capital-Labor Subsrtitution in the U.S. 1899-1960", A.erican Economic Revie, June, pp. 357-239 w. Erwin ie-ert 1974], ".Application of Duality Theory", in Ii. Intriligator and D. 'endrick, eds., Frontiers of Veantitative Economics, "olland.rh- rr~r~~ ..-: r..,~.,~.~,~,,,~.~~ ~,,, Vol. 2, Amsterdam: x-~'~~C ~'~~~1-~ -"-'"~~~~'-~~-~ ~- ~~" " ~~~Tlx~nr r e r -~-- ^--r,..,,,,.~,, North- ,~.,,,~,.,~.,,.,~,~ .,,,,.~, ~..~._., References (continued) W. Erwin Diewert [1975a], "Exact and Superlative Index %umn,0rs", forthcoming, Journal of Econometrics. W. Erwin Diewert [19755], "Ideal Log Change Index Numbers and Consisterncy in Agqre-ation", Discussion Paper 75-12, University of Eritish Columbia, Department of Economics, August. Jack Faucett Aszcciates [1973], 'Data Deveiopment for the 1-0 Energy Model: Final Report", Chevy Chase, Md.: Jack Faucett Asscciates, Inc., May. William FelLner [ 151], "To.Propositions in the Theory of Induced Innovation", Econcaic Journal, Vol. 71, No. 232, .pp. 305-30a. William Fellner 19563, "Profit Maximization and the Rate a.d Direction of Technical Change", American Economic reviw, Vol. 56, No. 2, pp. 24-32. .Melvyn A. Fuss [1975), "The Demand for Energy in Canadian Manufacturing: An Example of the Estimaton of P ;odu%tion Functions with Many Zrn;uts", inpubiished mimeo, University of Toronto, August. A.R. Gallant [1975. "Seemingly Unrelated Nonlinear Regras- Ssons", Journal of Ecno-metrics, Vol. 3, No. pp. 35-50. 1, March, .Robert E. Hall and Dale W. Jorgenson [19673, "Tax Policy and Investment Behavior", American conomic Reviey¢, June, pp. 394-414. D.B. Humphrey and John M. Moroney [19751, "Substitution Among Capital, Labor and Natural Resource Products in Americ"a Manufacturing", Journal of Political Sconomv, Vol. 83, No. 1, February, pp. 57-2. Dale W. Jorcgenson and Zvi Griliches [19673, "The Explanation of Productivit Change", Review of Economic StudCies, July, pp. 249-283. Dale W. Jorgenson and Zvi Griliches [1972], "Issues in Grcwth Accounting: A Reply to Edward F. Deniscn", Survey of Current Business, Vol. 52, Number 5, Part II, iiay, pp. 65-94. Dale W. Jorgenson and L.J. Lau r19753, Dualitty forthcoming, ;Ansterdam: North-iHolla. and Technrol.o, "" - -- Reforerces YIYII iLIII IYIY I^YIIIYIIYIIII IIIIYIYYIIIIIIY I. IYYIYIYIYIYIY (contin-ed) Charles Kennedy [19641, "Induced Bias in Innovation and the Theory of Distribution", Econozic Journal, Vol. 74, No. 295, pp. 541-7. n~nedy [12"T3 5, "A Generalization of the Theory of Charles Induced rLiase in Technical Progress", Economic Journ.--i, Vol. 83, No. 329, pp. 48-57. Jan menta r1971), Ele-ents of conomtrics, New York: *MaclMillan Co. L.J. Lau [1975), "The Econometrics of Mcnotonicity, Convexity and Quasi-Concavity", forthcoming, Econometricap. Edmond Malinvauzd (19703, Statistical Methods of Economeitri:s, Chicago: Rand McNally and Company, ch. 9. M.I. Nadiri (1970, "Some Approaches to the Theory and Measurement of Total Factor Productivity: A Survey', Journal or Economic Literature, December, pp. 1137-77. Willicm D. Nordhaus [1967J, "The Optimal Rate and Direction of Technical Change", in Karl Shell, ed., Essays on t s Theory of Optimal Economic Growth, M.I.T. Press, Cabrid2ge, Mass. William D. Nordhaus (1973], "Some Skeptical Thoughts on the Theory of Induced Innovation', Quarteriv Journal of Economics, Vol. 87, No. 2, pp. 208-219. "A Theory of Induced Innovation Paul A. Samuelson (19653, Along Kennedy-Weizsacker Lines", Review of Economics cand Statistics, Vol. 47, No. 4, pp. 3T3-56. R. Sato [1970), "The Ettimation of Biased Technical Progress" i-bernational Economic Review, June, pp. 179-207. R.M. Solow [1957, "Technical Change and the Aggregate Production Function", Review of EconCmics and Statistics, August, pp. 312-20. Spencer Star [1974], "Accounting for the Growth of Output" American Economic Review, March pp. 123-135. A. Zellner and I. Theil 119623, "Three Stage Least Squares: Simultaneous Estimation of Simultaneous Equations", Econometric , January, pp. 54-78. i