Macro Dynamics and Labor-Saving Innovation: US vs. Japan

Transcription

1 CIRJE-F-528 Macro Dynamics an Labor-Saving Innovation: US vs. Japan Ryuzo Sato New York University an University of Tokyo Tamaki Morita National Grauate Institute for Policy Stuies (GRIPS) November 2007 CIRJE Discussion Papers can be ownloae without charge from: Discussion Papers are a series of manuscripts in their raft form. They are not intene for circulation or istribution except as inicate by the author. For that reason Discussion Papers may not be reprouce or istribute without the written consent of the author.

2 Do Not Quote without the Authors' Permission MACRO DYNAMICS AND LABOR-SAVING INNOVATION: US VS. JAPAN By RYUZO SATO an TAMAKI MORITA * New York University an the University of Tokyo National Grauate Institute for Policy Stuies (GRIPS) JEL Classification: O33, O47, O57 Keywors: Technical Change, Labor-Saving Innovation, Prouctivity Abstract This article eals with the empirical analyses of the growth for the Unite States an Japan from 1970 to Following our analysis in Quantity or Quality: The Impact of Labor-saving Innovation on US an Japanese Growth Rates, (March 2007), we applie the same metho to a ifferent ata series in orer to confirm our previous finings. As with the previous paper, the results shown in this paper support our view that Japan s eclining population can be compensate for by aitional quality improvement of the existing labor force. * Corresponing author. Tamaki Morita, National Grauate Institute for Policy Stuies Roppongi, Minato-ku, Tokyo , Japan. aress: morita@grips.ac.jp Phone: Fax: Preliminary, October,

3 Contents 1. Introuction The Data Neutrality Tests Estimates of Prouction Functions Metho Simulation Results Biase Technical Change of Japan an the Unite States Stability uner Biase Technical Change Summary an Conclusion REFERENCES Appenix Preliminary, October,

4 1. Introuction 1. The contrast between the population changes in the Unite States (US) an Japan has recently become more istinct: Japanese population growth has slowe an is expecte to continuously ecline, while the population in the US reache 3 billion in 2006 an is expecte to continue to expan ue to subsequent immigration. 2. A recent stuy by Sato an Morita (2007), (hereinafter previous paper ) attempte to explain whether population changes severely affect economic growth. Extracting the labor-saving innovation from the economic growth le to the conclusion that it is not the quantity, but quality of growth that matters. By iviing conventional total factor prouctivity (TFP) into labor efficiency an capital efficiency, we pointe out that improve labor efficiency woul compensate for the shrinking of the labor force in the near future. 3. We begin by reviewing the previous paper s key points through several excerpts. (1) The novel contribution of [the] paper, accruing from those of Sato an Ramachanran (1987) an Sato (1970), is that we analyze not only how the TFP has increase or ecrease, but also analyze separately the efficiency of capital an the efficiency of labor. The result of this analysis will allow us to make a policy proposal that in orer to raise TFP growth, we have to consier how an how much the efficiency of either or both of capital an labor must be increase. Merely knowing TFP is generally consiere sufficient for economic analysis. However, our comparison of the two countries will show that because each country s composition of TFP is funamentally ifferent, knowing only total efficiency oes not suffice. (p. 5) (2) In orer to analyze the efficiency of capital an labor, we nee to know the prouction function or the elasticity of (factor) substitution, which is the summary inex of prouction function. In general terms, elasticity of substitution is a technology inex. As Sato Preliminary, October,

5 an Beckmann (1968) an Rose (1968) iscovere, elasticity of substitution plays a critical role in the analysis of the efficiency of each input factor. Our growth analysis uses the concept of elasticity of substitution an applies the concept to the ata of the two countries. (pp. 5 6) (3) We contrast the ifference in the economic structures of Japan an the Unite States (US) by comparing the rate of factor-augmenting technical progress. Our investigation reveals that whether or not the capital an labor are efficiently use has a strong impact on economic growth. (p. 6) (4) After the theoretical explanations, we conucte the estimation using both countries macro ata. The ata were taken from 1960 to 2004 an then ivie into two perios: Perio I ( ) an Perio II ( ). Perio II for Japan inclues the lost ecae, while that for the Unite States is often escribe as the new economy. The analysis on Perio II was particularly effective in highlighting the characteristics of each economy. (p. 39) (5) Concisely state, the source of Japan s economic growth was quality improvement rather than quantity increase of population an labor force. In contrast, for the US, what supporte its economic growth was quantity increase rather than quality improvement of population an labor force. (p. 39) (6) Overall, we iscovere that Japan s high growth in Perio I was not so much ue to the increase of the population, but to improve labor efficiency. Japan s stagnation, too, was not explaine by the population ecrease or shortage of effective emans, but by the slowown of the improvement of labor efficiency. Broaly efine innovation has been an will be the engine of the evelopment an growth of the Japanese economy. Thus, Japan oes not have to be pessimistic about the eclining birth rate. (p. 40) (7) We test the equilibrium conition to etermine each economy s performance. In Japan, [the actual value of output per Preliminary, October,

6 effective labor] is much higher than [the equilibrium value of output per effective labor]. Japanese capital stock has grown very fast, but it was not utilize to increase the economy s total income (GDP). In Perio I, especially before the first oil crisis in 1973, an extremely high rate of investment accumulate Japanese physical stock at a very rapi pace, which supporte the country s miraculous economic growth. As for the US, [the actual value of output per effective labor] is very close to [the equilibrium value of output per effective labor], an the actual output in [each perio] was just below the optimal output. The growth rate of effective capital an effective labor, an the economic growth rate were balance in both Perio I an Perio II. (p. 33) 4. In this paper, we apply the same methos to more specific ata on the US an Japan in orer to confirm the characteristics of the countries macro-ynamics escribe in the previous paper. This time, the prouction series Y is represente by net omestic prouct (NDP) instea of gross omestic prouct (GDP), which was use in the previous paper. This change allows us to more irectly fin the income shares of capital an labor, an enables us to omit the effect of consumption of fixe capital, which amounts to as much as about 20 percent of Japan s GDP an about 10 percent of US GDP. We selecte strictly private-nonagricultural sectors for all the series, incluing Japan s net stock. Because of ata constraints, we use ata from a slightly shorter range of years ( ). 2. The Data 5. The major ifferences between the ata use here an that use in the previous paper are: (1) as a series of an economy s output, Y, the real net omestic prouct, is taken instea of the real gross omestic prouct; Preliminary, October,

7 (2) the public sector an the agricultural sector are exclue from all the series; an (3) instea of OECD ata, each country s official ata are carefully selecte an ajuste to improve comparability between the US an Japan. We implemente the first change because we neee a more income-oriente value of the shares of capital an labor the estimation of biase technical progress epens largely upon them. In aition, omitting consumption of fixe capital helpe us to observe how new value is ae to each economy. Due to these changes in the ata, mostly influence by the first change, the series Y of each country is generally lower than that in the previous paper, as is epicte in Figure 1a (Japan) an Figure 1b (US). Preliminary, October,

8 Billion Yen (Base Year =2000) Figure 1a. Comparison of the Series Y: Japan Y in the previous paper (base on GDP) new Y (base on NDP) million Dollars (Base Year =2000) Figure 1b. Comparison of the Series Y: US Y in the previous paper (base on GDP) new Y (base on NDP) The average relative shares of capital an labor iffere consierably as well, an the labor share is smaller here. For example, take average labor share α after In Japan, it was formerly 75.17%, but in the new series, it is calculate as 72.07%. In the US, it also ecline from 67.61% to 61.97%. The new series escribe more accurately how the newly ae income each year is istribute to capital an labor. Precise comparison is shown in Table 1a (Japan) an Table 1b (US). Preliminary, October,

9 Table 1. Average Relative Share of Input Factors a. Japan New Data Previous Data Average Relative Share of Capital α 31.54% 34.42% 27.93% 29.34% 31.60% 24.82% Average Relative Share of Labor β 68.46% 65.58% 72.07% 70.66% 68.40% 75.17% Total 100% 100% 100% 100% 100% 100% b. US New Data Previous Data Average Relative Share of Capital α 37.62% 37.30% 38.03% 31.33% 30.80% 32.39% Average Relative Share of Labor β 62.38% 62.70% 61.97% 68.67% 69.20% 67.61% Total 100% 100% 100% 100% 100% 100% Notes: α an β are calculate as perio averages of α () t = r() t K() t Y() t an β () t = w() t L() t Y() t. 3. Neutrality Tests 6. Before estimating the prouction function using the new ata, we have to confirm whether the selecte series in both countries are appropriate for the analysis of the biase technical change. The prouction function with biase technical change can be expresse as equation (1), in an aggregative economy uner the neo-classical constant returns to scale technology, where at each year t, one output (Y(t)) is prouce by two factor inputs, capital (K(t)) an labor (L(t)). [ ] Yt () = F AtKt () (), BtLt () (), (1) where A(t) an B(t) are efficiencies of capital an labor, respectively. We followe the metho use in the previous paper an conucte two tests: one on whether the elasticity of factor substitution is unity, an another on whether the prouction function is Hicks-neutral. The results are shown in Tables 2 an 3. Preliminary, October,

10 Table 2. Average Elasticity of Substitution Metho Results R ( z/ w) R ( y/ r) Japan Unite States ( ) ( ) ( ) ( ) R( x/ ω ) ( ) ( ) Notes: Stanar Deviation is in parentheses. Extreme 1 3 ata are exclue in each series. Table 3. The Hicks-Neutrality Test Results Japan Unite States Average TT 0.74% 1.07% α β Equation (2) w k w k = = w k w k Regression ( 0.21) (11.18) (6.24) ( 1.00) Results Aj R 2 = Aj R 2 = Estimate TT -0.08% Estimate TT 1.66% Estimate σ Estimate σ Equation (3) r k r k = = r k r k Regression (1.22) ( 4.97) (3.87) ( 5.02) Results Aj R 2 = Aj R 2 = Estimate TT 1.60% Estimate TT 2.45% Estimate σ Estimate σ In Table 2, if the three variables converge to one, i.e., R( z w) = R( y r) = R( x ω) = 1, where r = return to capital, w = wage rate of labor, x = L/K, y = Y/K, z = Y/L, r/ w, ω =, an R(y/r) =( y y) ( r r) the elasticity of factor substitution σ in equation (1) is on average equal to unity 1. The results show that the three variables are not equal, an not even close to unity. Hence, the prouction functions cannot be Cobb-Douglas type. Table 3 presents the results of the following regression: w T α k = + w T σ k (2) 1 The reason is explaine in Sato an Morita (2007). Preliminary, October,

11 r T β k =. (3) r T σ k T is the inex of technical change in the prouction function with Hicks-neutral technical change ( Yt () TtF () [ Kt (), Lt ()] = ), which is equivalent to TFP. If the relationships of equations (2) an (3) simultaneously hol, an T Tin the two equations are equal in each country, then the country s prouction function is Hicks-neutral. The results suggest, however, that the relationships harly hol because the estimate variables are not significant, an the estimate two T Tiffer greatly in each country. This leas to our jugment that the technical change in each country is not Hicks-neutral, an it shoul be estimate with biase technical change. 4. Estimates of Prouction Functions 4.1 Metho 7. In accorance with the etaile explanation in Sato an Morita (2007), we take four steps in the estimation of the prouction functions with biase technical change for Japan an the US. Step 1: Estimation of Hicks-neutral technical progress First, we calculate T Tin each year t, using Tt () zt () () () kt = α t, (4) Tt () zt () kt () where z = Y / L, k = K / L. Step 2: Deriving average elasticity of substitution σ N Next, we estimate the average elasticity of substitution. In orer to analyze the efficiency of capital an labor, we nee to know the elasticity of (factor) substitution, which is the summary inex of the prouction function. As Sato an Beckmann (1968) an Rose (1968) iscovere, the elasticity of substitution plays a critical role in the analysis of Preliminary, October,

12 the efficiency of each input factor. Here, we estimate it uner the assumption of Hicks-neutral technical progress, σ N, by the following efinition: Kt () Kt () Kt () Lt () N Lt () Lt () Kt () Lt () σ () t = =. wt () wt () wt () rt () rt () rt () wt () rt () Step 3: Estimation of CES functions with Hicks-neutral technical change Our ata contains no trens of σ correlating with the values k or time t. Thus, we assume the constant elasticity of substitution (CES) prouction function an etermine how σ N fits the actual ata. Before we irectly estimate the prouction function with biase technical change, we estimate the function with Hicks-neutral technical change so that we can compare the fitteness of the two kins of estimate prouction functions. With Hicks-neutral technical change, the CES function shoul take the form of N N ρ Y () t = T() t αk() t + βl() t ρ N N 1/ ρ, (5) where σ N =1/(1+ρ N ). T(t) is assume to grow at a constant rate uring a perio, which is given as the average of each Tt () Tt () estimate in equation (4). We also assume the income shares of factors α an β are constant throughout the perio, an we apply perio averages of observe α an β. Step 4: Estimation of CES functions with biase technical change Finally, we estimate the CES function with biase technical change. Sato (1970) argue that theoretically, the elasticity of substitution of biase technical change has to be state as equation (6) B σ AK AK BL BL =, (6) F BL F BL F AK F AK because when technical progress is nonneutral, the value of the elasticity itself is influence by the efficiencies of capital an labor. However, because we cannot observe σ B Preliminary, October,

13 irectly, we use σ N as a proxy of σ B, an we substitute the estimates of elasticity σ N into σ in equations (7) an (8) to erive AA an B B 2, rt () Yt () Kt () σ () t At () rt () Yt () Kt () = (7) At () σ () t 1 wt () Y () t Lt () σ () t Bt () wt () Y() t Lt () =. (8) Bt () σ () t 1 Then, the prouction function becomes B Y () t = α A() t K() t + B() t L() t N ρ ( ) β( ) N ρ N 1/ ρ. (9) Estimate Y B (t) summarizes our moel. It represents both the form of the prouction function an the biaseness of technical change. 4.2 Simulation Results 8. In this subsection, we present the estimate values of actual Y, Y N an Y B. In Figure 2 (Japan) an Figure 3 (US), CES prouction functions with Hicks-neutral technical change Y N are plotte with thick gray lines. CES prouction functions with biase technical change Y B are shown by thin lines with markers. Estimate ata from 1970 to 2005 are shown in Panel 1 of each Figure, an in Panel 2, the ata are ivie into two at the year 1990, so that Y N an Y B in 1990 start from the actual Y in the year. Generally, Y N eviates from actual Y, an Y B fit better than o Y N. This supports our view that the economies of Japan an the US both experience biase technical growth. It suggests that estimation of TFP oes not suffice to iagnose economic performance or to prescribe any policy for either of these countries. We shoul note one point here. In Figure 2, Panel 2, the Y N an Y B in perio seem to be very close. One may conjecture that this similarity occurs because 2 Equations (7) an (8) are erive in Sato (1970). Preliminary, October,

14 AA B B TT. Actually, though, as shown in the next subsection, estimate capital efficiency growth AA was negative, an labor efficiency growth B B was positive, an was much larger than T T. In this case Y N an Y B coincientally appeare to be close. Billion Yen (Base Year =2000) Figure 2. Estimate Output of Japan Panel Panel σ = 0.46 actual Y B Y N Y Billion Yen (Base Year =2000) Panel an σ = 0.56 σ = B Y actual Y N Y Preliminary, October,

15 Figure 3. Estimate Output of the Unite States 10 million Dollars (Base Year =2000) Panel σ = 0.39 B Y N Y actual Y Panel an million Dollars (Base Year =2000) σ = 0.34 B Y σ = 0.46 actual Y N Y Preliminary, October,

16 4.3 Biase Technical Change of Japan an the Unite States 9. During the estimation process, in Step 1, we get some important values that explain economic ynamics of both countries. Table 4 lists the perio average values for some of the variables. Table 4. Growth Rate of Hicks-Neutral Technical Change an Other Factors a. Japan New Data Previous Data Growth Rate of Output YY 2.72% 4.20% 0.97% 4.65% 6.35% 1.03% Growth Rate of Hicks Neutral Technical Change TT 0.74% 0.99% 0.44% 2.13% 2.91% 0.45% Growth Rate of Capital K K 5.15% 7.45% 2.41% 7.32% 9.25% 3.18% Growth Rate of Labor L L 0.35% 0.86% -0.27% 0.28% 0.56% -0.31% Growth Rate of Output Per Labor zz 2.36% 3.31% 1.23% 4.35% 5.76% 1.34% b. US Ne w Data Previous Data Growth Rate of Output YY 3.25% 3.29% 3.20% 3.08% 3.05% 3.13% Growth Rate of Hicks Neutral Technical Change TT 1.07% 0.76% 1.44% 1.07% 0.90% 1.43% Growth Rate of Capital K K 3.11% 3.55% 2.59% 3.31% 3.58% 2.72% Growth Rate of Labor L L 1.58% 1.88% 1.23% 1.39% 1.49% 1.19% Growth Rate of Output Per Labor zz 1.64% 1.38% 1.95% 1.66% 1.54% 1.92% The table compares the new ata with those in the previous paper, so we can observe not only the perioic change but also the effect of excluing consumption of fixe capital an of strictly selecting private-nonfarm sector. In Japan s case, when we compare the new an previous ata, the exclusion of the high growth ecae (1960s) ue to ata constraints strongly affecte the growth rate of output an other variables. Nevertheless, in comparisons from the 1990s on, the ifferences clearly show the effect of the change of ata source. From the 1990s on, capital growth in the new ata is 2.4% an 0.77 percentage point lower than that in the previous ata. This explains how the consumption of fixe capital affecte the Japanese economy. Preliminary, October,

17 Among those variables, there is a well-known an important relationship Yt () Tt () () () () t Kt = + α + β() t Lt, Yt () Tt () Kt () Lt () (10) which is equivalent to equation (4). Table 5 shows the perio average values of each term in equation (10). Table 5. Relative Contributions to Economic Growth by Technical Change an Factor Inputs a. Japan Ne w Data Previous Data T Y T Y 27.11% 23.55% 45.35% 45.76% 45.87% 44.30% K Y α K Y 59.64% 61.13% 69.32% 46.14% 46.05% 76.90% L Y β L Y 8.74% 13.49% % 4.30% 6.03% % Statistical Ajustment 4.52% 1.84% 4.99% 3.80% 2.05% 1.41% Total 100% 100% 100% 100% 100% 100% T T K α K L β L Y Y Y Y Y Y b. US Ne w Data Previous Data % 23.09% 44.82% 34.80% 29.58% 45.69% 36.02% 40.20% 30.81% 33.65% 36.11% 28.15% 30.34% 35.82% 23.71% 31.09% 33.73% 25.71% Statistical Ajustment 0.77% 0.90% 0.65% 0.46% 0.58% 0.45% Total 100% 100% 100% 100% 100% 100% Notes: To apply actual ata to the theory, we have to approximate ifferentiation by ifference. Thus the weighte sum of the increase of each factor is not equal to the growth rate. We show such iscrepancy as "Statistical Ajustment" ( K K) In Japan, capital ( α ) contribute most to the economic growth. Capital investment, though excluing the consumption of fixe capital, still supporte about 70 percent of economic growth uring In this perio, capital an technical change compensate for the ecline of labor ( β ( LL ) ). In the US, the three terms contribute almost evenly to the growth on average. In the Preliminary, October,

18 perio , the TFP growth rate (1.44% per annum) explains about 45% of the growth, which suggests prouctivity improvement uring the perio. 10. Once σ is etermine in Step 3, the growth rates of capital an labor efficiencies can be estimate in Step 4. The results are shown in Table 6. In both countries σ < 1 an AA < B B, so both countries are experiencing what Hicks (1932) originally efin e as labor-saving innovation 3. We shoul also mention here that T, A, an B are relate as follows: Tt () At () Bt () α + β. Tt () At () Bt () (11) This is a simplifie version of equation (13) in Sato an Morita (2007). The values of each item in equation (11) are presente in Table 7. Table 6. Growth Rate of Biase Technical Change a. Japan Ne w Data Previous Data Growth Rate of Hicks Neutral Technical Change TT 0.74% 0.99% 0.44% 2.13% 2.91% 0.45% Estimate Elastisity of Substitution N σ AVG Growth Rate of Capital Efficiency A A -1.24% -1.04% -0.93% -1.61% -1.63% -1.36% Growth Rate of Labor Efficiency BB 1.65% 1.77% 1.00% 3.86% 5.11% 1.01% b. US New Data Previous Data Growth Rate of Hicks Neutral Technical Change TT 1.07% 0.76% 1.44% 1.07% 0.90% 1.43% Estimate Elastisity of Substitution N σ AVG Growth Rate of Capital Efficiency A A -0.04% -0.30% 0.20% -0.41% -0.59% 0.08% Growth Rate of Labor Efficiency BB 1.74% 1.40% 2.18% 1.74% 1.56% 1.97% 3 Definition of the labor-saving innovation also appears in Sato an Morita (2007). Preliminary, October,

19 Table 7. Relative Contributions to Hicks-Neutral Technical Change by Biase Technical Change a. Japan New Data Previous Data A T α A T % % % % % % B T β B T % % % % % % Statistical Ajustment Total -0.32% 100% 19.04% 100% -5.08% 100% -5.95% 100% -2.30% 100% 7.00% 100% A α A B β B T T T T b. US New Data Previous Data % % 5.22% % % 1.86% % % 94.14% % % 92.95% Statistical Ajustment 0.01% -1.00% 0.64% 0.52% 0.38% 5.19% Total 100% 100% 100% 100% 100% 100% Notes: See the notes of Table 5. In Japan, the results in the previous paper showe the AA to be on average negative throughout the perio. Consiering the results pr esente the previous paper, we presume one of the reasons for this negative value coul be a relatively large amount of consumption of fixe capital (approximately one fifth of the GDP). However, even when we exclue the factor, the results i not change significan tly. The AA in is a scant 0.43 percentage point higher than the previous result, but still negative. From 1990 on, espite the ecline of labor (-0.27%, Table 4), the labor efficiency grew at 1% per annum (Table 6), which contribute % to TFP. In other wors, the labor efficiency growth compensate for the ecline of labor. We emphasize this point in the previous paper, an it is still persuasive with the new ata. In the US, both experiments showe the AA to be slightly positive from the 1990s on. This change from that in the perio , together with the s increase in B B, is a Preliminary, October,

20 key to the rapi TFP growth in We presume the possibility that it coul be a structural change relating to the New Economy. The contribution rate of capital efficiency growth α ( AA ) turne positive but remains low. It may be interesting to point out that if this contribution rate oes not grow much in the future, we will be able to juge the US technical change to be more like Harros-neutral. 4.4 Stability uner Biase Technical Change 11. In the economy with biase technical change efine as equation (1), the stability conition shoul be as in equation (12). * * * Y = t ( AK) t ( BL) Y = AK BL, (12) where Y = output after technical change AK = effective capital BL = effective labor. The precise explanation of equation (12) is presente in Sato an Morita (2007). Here we woul like to etermine whether the economies of Japan an the US satisfy this conition. We show the growth rate of AK, BL, an the economic growth rate in Table 8. Preliminary, October,

21 Table 8. Growth Rate of AK, BL an Y a. Japan New Data Previous Data t ( AK) Growth Rate of Effective Capital AK 3.90% 6.41% 1.48% 5.71% 7.62% 1.82% t ( BL) Growth Rate of Effective Labor BL 2.00% 2.63% 0.74% 4.14% 5.66% 0.70% Growth Rate of Output YY 2.72% 4.20% 0.97% 4.65% 6.35% 1.03% Y t ( AK) Y AK Y t ( BL) Y BL t ( AK) t ( BL) AK BL b. US New Data Previous Data t ( AK) Growth Rate of Effective Capital AK 3.08% 3.25% 2.79% 2.90% 2.99% 2.80% t ( BL) Growth Rate of Effective Labor BL 3.32% 3.29% 3.41% 3.13% 3.05% 3.16% Growth Rate of Output YY 3.25% 3.29% 3.20% 3.08% 3.05% 3.13% Y t ( AK) Y AK Y t ( BL) Y BL t ( AK) t ( BL) AK BL We can see that in Japan, compare to the economic growth rate, the growth rate of effective capital was higher an that of effective labor was lower. Consequently, the effective capital growth is twice as high as the effective labor growth. Thus, the Japanese economy s growth path is far from balance. In contrast, in the US, the growth rate of each factor is almost even, especially uring the perio By estimating the capital an labor efficiency, we can clearly observe how much Japan nees to raise its labor efficiency in the era of the shrinking labor force. Preliminary, October,

22 5. Summary an Conclusion 12. We have confirme the finings in Sato an Morita (2007) by using more specific ata the nonfarm-private sector s net omestic prouct (NDP) instea of gross omestic prouct (GDP). As our neutrality tests confirme, the new series in both countries were also appropriate for the analysis of biase technical change. We illustrate that the estimations of prouction functions with biase technical change fit better to actual output than those with Hicks-neutral technical change. 13. To compare the new results with the previous results for Japan, the ifferences are as follows: The economic growth rate an some other new series uring an are consierably ifferent from the previous series. Most of the ifferences are explaine by the exclusion of the ata for the 1960s owing to ata constraints. The strong influence of the ata (or lack thereof) for the 1960s inicates that the extremely high growth in that ecae too strongly affecte the analysis of or in the previous paper. In , capital growth was lower in the new series, which suggests that the increase of consumption of fixe capital affecte the growth rate. Excluing it, in the new ata, growth rate of capital efficiency in was still negative, but its egree is smaller. It follows that the contribution of effective capital to the economic growth went up with the better capital efficiency. The new ata more clearly shows the role of newly accumulate capital. Estimate elasticity of substitution in was lower in the new ata. The results common to the two stuies on Japan are: Japanese growth rate was backe by both capital growth an growth of labor efficiency throughout the observation perios. The Japanese economy remains far from the steay-state. It is highly possible that the economy will autonomously ajust itself eventually. Preliminary, October,

23 For the US, there is just one major ifference: the estimate elasticity of substitution was lower in , but higher in The results common to the two stuies on the US are: Contrastive to Japan, no effects of exclusion of the 1960s ata were foun except the value of elasticity of substitution. The two stuies in general showe closely corresponing results espite the ifference in the perio. The reason why only the elasticity iffers shoul be investigate in future stuies. Capital efficiency turne positive in , which hints that there may have been a technical change that may relate to the so-calle New Economy. The US economy has been in the steay-state since the 1960s. 14. In conclusion, what has our application of the theory of biase technical change reveale? The US economy has been in the steay-state, but the Japanese economy is far from it. Since the ratio of the growth rate of effective capital to the growth rate of effective labor is too high in Japan, there will be two possible processes to reach the steay-state. One is negative ajustment ecline of effective capital, which may lower the level of steay-state growth rate. The other is positive ajustment increase of effective labor, which may realize higher steay-state growth. Accoring to our stuies, Japan has continuously improve labor efficiency until it coul compensate for the ecline of labor. As its labor force continues to shrink, Japan shoul make a perpetual effort to raise labor efficiency with a view to avoiing the negative type of ajustment. Preliminary, October,

24 REFERENCES Hicks, John R. (1932, Secon eition in 1963) The Theory of Wages, Macmillan. Rose, Hugh (1968) The Conition for Factor-Augmenting Technical Change, The Economic Journal, Vol. 78, No. 312, pp Sato, Ryuzo (1970) The Estimation of Biase Technical Progress an the Prouction Function, International Economic Review, 11, pp Reprinte in R. Sato, Growth Theory an Technical Change, The Selecte Essays of Ryuzo Sato, Vol. 1, Economists of the Twentieth Century Series, Ewar Elgar Publishing, 1996, Ch. 12., an M. J. Beckmann (1968) Neutral Inventions an Prouction Functions, The Review of Economic Stuies, Vol. 35, No. 1, pp Reprinte in R. Sato, Growth Theory an Technical Change, The Selecte Essays of Ryuzo Sato, Vol. 1, Economists of the Twentieth Century Series, Ewar Elgar Publishing, 1996, Ch. 11., an T. Morita (2007) Quantity or Quality: The Impact of Labor-Saving Innovation on US an Japanese Growth Rates, , CIRJE Discussion Papers, CIRJE-F-483, The University of Tokyo; COE Discussion Paper Series, COE-DP F-168, Research Center for the Relationship between Market Economy an Non-Market Institutions, The University of Tokyo; Working Paper Series, N-07, The Center for Japan-U.S. Business an Economic Stuies, Leonar N. Stern School of Business, New York University., an R. Ramachanran (1987), "Factor Price Variation an The Hicksian Hypothesis: A Microeconomic Moel," Oxfor Economic Papers, 39, Reprinte in R. Sato, Growth Theory an Technical Change, The Selecte Essays of Ryuzo Sato, Vol.1, Economists of the Twentieth Century Series, Ewar Elgar Publishing, 1996, Chapter 15. Preliminary, October,

25 Appenix The ata sources are as follows. Japan Y: Net omestic prouct excluing agriculture, forestry an fishing, plus proucers of government services, eflate by GDP Deflator (Department of National Accounts, Economic an Social Research Institute, Cabinet Office, Government of Japan, Annual Report on National Accounts) K: Tangible fixe assets excluing wellings (net stock, real) multiplie by the ratio of private sector in proucing assets (gross stock, nominal) (Department of National Accounts, Economic an Social Research Institute, Cabinet Office, Government of Japan, Annual Report on National Accounts) multiplie by the ratio of gross capital stock by inustry, total minus agriculture, forestry an fishing. (gross stock, real) (Department of National Accounts, Economic an Social Research Institute, Cabinet Office, Government of Japan, Annual Report on Gross Capital Stock of Private Enterprises) L: Hours worke per employee multiplie by number of private employees excluing agriculture (The Japan Institute for Labour Policy an Training, Rouou Toukei Deeta Kensaku Sisutemu (Search System for Labour Statistics), an Statistics an Information Department, Ministry of Health, Labour an Welfare, Monthly Labour Survey) w: Compensation of employees excluing agriculture, forestry an fishing, an proucers of government services eflate by GDP eflator (Department of National Accounts, Economic an Social Research Institute, Cabinet Office, Government of Japan, Annual Report on National Accounts) ivie by labor force (L) Preliminary, October,

26 Unite States Y: National income without capital consumption ajustment by inustry, private excluing agriculture (National Income an Prouct Accounts Tables, Bureau of Economic Analysis, U.S. Department of Commerce, K: Current-cost net stock of private nonresiential fixe assets by inustry group an legal form of organization, private excluing farms eflate by chain-type quantity inexes for net stock of private nonresiential fixe assets by inustry group an legal form of organization (Fixe Asset Tables, Bureau of Economic Analysis, U.S. Department of Commerce, L: Average weekly hours of prouction an nonsupervisory workers on private nonfarm payrolls multiplie by employees on private nonfarm payrolls multiplie by 52 weeks (Employment, Hours, an Earnings from the Current Employment Statistics survey, Bureau of Labor Statistics, U.S. Department of Labor, w: Compensation of employees by inustry, private inustries excluing farms eflate by GDP eflator ivie by labor force (L) Preliminary, October,