- 1 These arguments attribute the causes of recession to either the demand side or the supply side. The (...)
1In this study, we investigate the macroeconomic performance of the Japanese economy, using the vector autoregression (VAR) model. Numerous standard economists have debated the causes of Japan’s low economic growth after the bubble burst (Iwata and Miyagawa, 2003; Hamada and Horiuchi, 2004; Hayashi, 2007; Wakita, 2008)1. In this study, we examine Japan’s macroeconomic performance in terms of growth and distribution because such an analysis is lacking in the standard economic analyses cited above.
2VAR is a dynamic econometric model that captures the relationship among economic variables over time without relying on a particular economic model. The theoretical viewpoint of the analyst is reflected, to a certain extent, in the variables included in the VAR model and the evaluation of the results. This study employs the post-Keynesian model as a theoretical reference. This model emphasizes income distribution between wages and profits in determining macroeconomic performance such as aggregate demand. The VAR analysis based on the theoretical implications derived from the post-Keynesian model will significantly contribute to research on the growth regime of the Japanese economy.
3Theoretical examinations by post-Keynesian economists have revealed two patterns of aggregate demand and economic growth (or demand regime and economic growth regime): a profit-led pattern and a wage-led pattern. The theme of growth and income distribution has also been important for Régulationists. Therefore, this theme is the most interactive one among post-Keynesians and Régulationists (Ponsot, 2007). Moreover, they also indicate that both profit- and wage-led demand and growth regimes are theoretically possible. In other words, determining which pattern is more dominant in the dynamic relationship between growth and income distribution is an empirical issue; it is only through empirical analysis that we can distinguish which regime is dominant in a period or country. Our empirical study focuses on the Japanese economy and utilizes post-Keynesian economics to contribute to the literature on this issue.
- 2 In this study, an individual demand formation pattern refers to the relationship between change in (...)
4Post-Keynesian economics emphasizes the relationship between income distribution and the aggregate demand formation pattern. In addition, it focuses on the relationship between individual demand (particularly investment demand) and aggregate demand formation pattern2. This viewpoint is critical for the understanding of Japanese macroeconomic performance since, as will become evident in the next section, both investment and export demand are important influences in growth and business cycles. Post-Keynesians consider that aggregate demand is important not only for short-run but also long-run dynamics, which distinguishes it from other economics. Such a view is empirically important, as the Japanese economy has in fact suffered from low economic growth, especially after the 1990s, largely due to lack of effective demand. Therefore, we employ their viewpoint in empirical analysis.
5Additionally, as Bowles and Boyer (1990) indicated, the dynamics of income distribution play a role in both the source of domestic demand and the cost of production. Several studies have examined these issues empirically. For example, Hashimoto (2002) explained that a decrease in profitability due to a large increase in wage share led to the stagnation of private investment activity, which was the major cause of Japan’s economic stagnation during the 1990s. Furthermore, Miyagawa (2005) highlighted the deterioration in the price competitiveness of Japanese firms against NIEs and China. Thus, the dynamic relationship of income distribution and individual demand also determines macroeconomic performance at an empirical level. Therefore, it is necessary to consider these integral relationships in order to understand the characteristics of the Japanese economic growth regime. Since the VAR model is a time series model that captures multiple economic variables, it is useful for the purpose of our analysis.
6Recently, Azetsu et al. (2010) described the growth regime of the Japanese economy using a Kaleckian framework. They estimated a structural VAR model that included profit share, capacity utilization rate, capital accumulation rate, exports, and imports. Using the shape of the impulse response function from their estimate, they identified the Japanese economic growth regime as being profit-led. However, a demand and growth regime depends on more than just the impacts of income distribution on investment, exports, and imports. Intrinsically, consumption demand is also important in the determination of growth regimes; however, this is ignored in Azetsu et al.’s analysis. In this sense, their analysis is inadequate. In order to detect the nature of a growth regime, the individual demand formation pattern—such as the relationship between income distribution and consumption, investment, and exports—must be considered. Our empirical analysis using the VAR model attempts to enrich recent post-Keynesian empirical researches. We will reconsider the demand and macroeconomic growth regime of Japan and the dynamic relationships among demand variables in greater detail by using Granger’s causality test, the accumulated impulse response function, and variance decomposition analysis.
7The remainder of this paper is organised in the following manner. In Section 2, we reconsider the macroeconomic performance in Japan by surveying macroeconomic data on income distribution and demand formation from the beginning of the 1980s to 2008. In Section 3, we briefly present the literature related to post-Keynesian empirical analysis and the VAR analysis of the Japanese economy. In Section 4, we explain the data and a few preliminary tests for VAR estimation. In Section 5, we estimate VAR models that include wage share, consumption growth rate, exports, GDP, capital accumulation, and exogenous government expenditure; moreover, we provide a plausible economic interpretation of the results. Our basic conclusion is as follows: the Japanese economy was characterised by a profit-led growth regime between 1985 and 2008. This finding is the same as that of Azetsu et al. (2010). In comparison with their study, our empirical analysis is clearer because it uses both impulse analysis and variance decomposition (which comprise ‘innovation accounting’). The analysis in this study indicates that the Japanese profit-led growth regime was accompanied by the profit effect for consumption, capital accumulation, and increase in exports. In addition, the fluctuation in GDP is explained by the order of consumption, capital accumulation, and export demand. In Section 6, we present the conclusion which summarizes the results of the analysis and mention a few remaining issues for consideration in future research.
8In this section, we provide an overview of Japan’s macroeconomic performance by referring to data on income distribution and effective demand components. Figure 1 describes the relationship between wage share and the growth rate of investment, consumption, exports, and GDP after 1981. Wage share and GDP growth vary when affected by various factors. It may be stated that growth is wage-led if the relationships have an upward-sloping curve and profit-led if they have a downward-sloping curve.
9Figure 1 illustrates the short-run cyclical pattern of wage-led and profit-led growth for all demand components. By applying a linear regression equation to the demand components, we obtain a downward sloping curve. Therefore, in general, each demand component has had a profit-led demand formation pattern since the 1980s. The last point of the series represents data from 2008. In 2008, the rate of demand growth was low because it was affected by the international financial crisis. Figure 1 also illustrates that the post-1990 growth rates for domestic consumption and investment were lower than growth rates in earlier years. Only exports recorded a higher growth rate after 2000. Wage share was relatively low during the 1980s; however, it began to rise after the bubble burst and remained particularly high in the mid-1990s. After the late 1990s, wage share began to decline. On average, wage share reached its lowest level during the recovery process from 2002 onwards, as indicated inTable 1 below.
Figure 1: Income distribution and cyclical pattern of effective demand components
Source: Economic and Social Research Institute, Cabinet Office, National Accounts for2008(93SNA, benchmark year = 2000).
Notes: The vertical axis is the annual growth rate of a variable. Both axes are measured in percentages. The broken line displays the linear regression for the data plot.
10As mentioned in the introduction, income distribution has a role in both the source and cost of demand components such as consumption, investment, and exports. Let us consider the dynamic relationship between income distribution and each demand component in detail. Table 1 presents the average growth rate of the demand component over a business cycle.
Table 1: Growth rate of demand component in a business cycle
|
Period
|
Phase of business cycle
|
|
Average growth rate
|
|
GDP
|
Private final consumption
|
Gross private fixed capital formation
|
Government expenditure
|
Exports
|
|
Wage share
|
|
83:1–85:2
|
Expansion
|
1.18
|
0.75
|
1.82
|
-0.52
|
2.70
|
73.62
|
|
85:2–86:4
|
Contraction
|
0.90
|
0.98
|
3.88
|
1.05
|
-0.63
|
73.88
|
|
86:4–91:1
|
Expansion
|
1.37
|
1.08
|
2.79
|
0.85
|
1.46
|
78.57
|
|
91:1–93:4
|
Contraction
|
0.27
|
0.56
|
-1.55
|
1.34
|
0.40
|
79.53
|
|
93:4–97:1
|
Expansion
|
0.62
|
0.76
|
0.80
|
0.47
|
1.54
|
78.27
|
|
97:1–99:2
|
Contraction
|
-0.22
|
0.01
|
-1.01
|
0.37
|
0.30
|
76.47
|
|
99:2–00:4
|
Expansion
|
0.56
|
0.25
|
5.85
|
-0.05
|
2.41
|
74.53
|
|
00:4–02:1
|
Contraction
|
-0.09
|
0.38
|
-2.47
|
0.21
|
-1.20
|
72.64
|
|
02:1–07:4
|
Expansion
|
0.50
|
0.31
|
1.14
|
-0.10
|
2.56
|
73.28
|
|
83:1–07:4
|
all
|
0.60
|
0.54
|
1.15
|
0.52
|
1.40
|
76.22
|
Source: Economic and Social Research Institute, Cabinet Office, National Accounts for 2008(93SNA, benchmark year = 2000).
Notes: Basic samples are from 1983:1–3 to 2007:10–12. Gross private fixed capital formation is measured by private fixed capital formation plus private inventory. The government expenditure is the total of government final consumption expenditure, public fixed capital formation, and public inventory. All series are seasonally adjusted. The units are in percentages.
11These variables will be included in our VAR model in Section 4. During expansion periods, the driving force behind growth changes in each period is evident. In the bubble period, gross private fixed capital formation was principally responsible for expansion. On the other hand, since 2002, economic recovery and expansion was mainly driven by exports. The growth in the bubble period (for example, in 1988) was basically led by domestic demand; the contribution of domestic demand (total of private consumption, investment, and government expenditure) to real GDP was 7.1 points, while that of net exports was -0.5 points. On the other hand, economic growth in recent years has been typically led by exports: in 2007, the contribution of domestic demand was 0.7 points and that of external demand was 1.3 points. The recovery process that began in 2002 reached its peak at the end of 2007, and the Japanese economy stagnated again due to the shrinking of world trade caused by the financial crisis.
12Now we focus on the periods of contraction after 1990. Two recessionary periods are evident in the 1990s. In both periods, a decline in the gross private fixed capital formation was primarily responsible for stagnation. However, it must be noted that the causes of investment stagnation were different. The investment stagnation that occurred immediately after the bubble was due to stock adjustment, whereas the decrease in investment during 1997–98 was due to a credit crunch (Yoshikawa, 1998). From 1993 to 1997, there was an expansion of private consumption, followed by stagnation after 1997. This stagnation was due to political factors, such as a rise in the consumption tax, and social factors, such as increased uncertainty with regard to employment security and social security (Nakagawa, 1999; Horioka, 2007).
13Numerous scholars have debated the effectiveness of Japan’s fiscal policy during the recession of the 1990s. Since the economic policy package (Sogo Keizai Taisaku) of 1992, over 100 trillion yen has been injected into the economy, a large share of which went towards funding public works. As a result, despite the fiscal structural reform by the Hashimoto government in 1997, the fiscal deficit expanded and the government’s debt grew during the 1990s. While the ratio of fiscal budget to GDP was 1.9 points in 1990, it reduced to -6.3 points in 2000. Moreover, the ratio of the balance of the accumulated government debt to GDP doubled from 64.6 points to 122.9 points in the same period (Nakazato, 2003). Nakazato (2003) surveyed studies of VAR analysis and examined the relationship between fiscal policy and GDP growth. According to his survey, government expenditure had a positive effect on GDP growth, although the effect was temporary in most cases. Therefore, it may be stated that the increase in government expenditure sustained Japan’s economic activity. For example, the recovery of private investment between 1994 and 1997 may be attributed to fiscal expansion from 1991 to 1993. On the other hand, the growth rate of government expenditure has been restrained since the structural reforms initiated by the Koizumi government. However, the accumulated debt of the Japanese government remains a big issue, while tax revenue is stagnant.
14This section surveys literature related to post-Keynesian empirical analysis and the VAR analysis of the Japanese economy. Bowles and Boyer (1995), Hein and Ochsen (2003), Naastepad and Storm (2007), and Hein and Vogel (2008) represent recent empirical analyses. These studies perform empirical analyses that are principally based on the post-Keynesian theoretical modeldeveloped by Bhaduri and Marglin (1990). In addition, Uemura (2000), Uni (2009), and Fujita (2009) are studies on the Japanese growth regime. Here, we survey the results obtained in studies conducted by Japanese economists and related literature, and our empirical results are reviewed from an international perspective in Section 5.3.
15Similar to Bowles and Boyer (1995), Uemura (2000) considers the growth regime and structural change in the Japanese economy after WWII. According to his study, the Japanese economic growth regime was characterised by a profit-investment-led pattern between 1963 and 1971. However, the growth pattern transformed to wage-led growth that was undermined by export-led growth between 1976 and 1995. This growth regime indicates that aggregate demand is profit-led due to the strong profit effect on exports, whereas domestic demand formation is wage-led. Uni (2009) compared the growth regimes of the Japanese and US economies. Evaluating both demand and productivity regimes, he illustrated the high growth rate of the US economy and low growth rate of the Japanese economy. Moreover, he explained Japan’s stagnation in the 1990s by focusing on the leftward shift of the demand regime curve due to a fall in investment and export demand. Further, Fujita (2009) also compared the growth regimes of the US and Japanese economies from the 1990s to the beginning of the 2000s by employing a cumulative causation model. He established a two-sector model comprising consumption and investment sectors; the model emphasized the path from the change in each sector’s productivity growth to the relative price by which the growth rate of demand is determined. Through estimations, his study found that in the US economy, high productivity growth in the investment goods sector yielded a large increase in the effective demand of this sector. However, in Japan, when productivity in the investment sector was high, it did not induce growth in this sector because of over-accumulation in the bubble period and bad-loan problems after the bubble burst.
16Japanese economists have also attempted to identify the causes of Japan’s economic stagnation using time series analysis. Harada and Iida (2004) and Harada and Masushima (2009) indicated monetary factors as the cause of the lasting recession. Through VAR estimations, they found that an expansion of the money base had a positive effect on economic activity. On the basis of this result, they insisted on the effectiveness of a quantitative easing policy to combat the economic stagnation. Bayoumi (2001) also explained Japan’s economic stagnation using monetary factors. His results indicated that the major explanation for the extended slump in Japan’s economic activity during the 1990s was the disruption in financial intermediation, which was largely the result of changes in domestic asset prices on bank lending. Taking demand and supply side factors into consideration, Miyao (2006) used VAR analysis to indicate that the demand shortage in the 1990s was due to a negative productivity shock after 1993. Further, the VAR analysis by Sadahiro (2005) illustrated that both the stock adjustment mechanism and a large increase in wage share contributed negatively to investment activity after 1990. Braun and Shioji (2007) revealed that investment shock, which decreases the relative price of investment goods, played an important role in the fluctuation of Japanese GDP.
17The preceding studies focus on variables affecting GDP; while highly sophisticated VAR analyses of the Japanese economy exist, in numerous cases, they fail to consider the role of income distribution. Post-Keynesians have emphasized the role of income distribution in the determination of aggregate demand. However, most post-Keynesian VAR studies focus on European or North American economies. Azetsu et al.(2010) applied the first post-Keynesian (Kaleckian) VAR analysis to the distribution and accumulation of the Japanese economy, and they concluded that profit-led growth was dominant. However, their analysis is inadequate since they did not consider individual or aggregate demand formation patterns. Furthermore, their study remains mechanical, since it only presents the impulse response function and does not compare their results with Japan’s actual economic performance. Therefore, in the current study, we employ VAR to estimate Japan’s macroeconomic performance in terms of post-Keynesian economics and examine this performance in detail.
18We introduce the variables, data sources, and processing below.
-
- 3 In the SNA of Japan, the entrepreneurial income is calculated by adding the net received property i (...)
WS (wage share): This is defined as employee compensation divided by the total of entrepreneurial income (after deducting receivables and payables of corporations’ distributed income)3 and employee compensation. The data were obtained from National Accounts for 2008,Economic and Social Research Institute, Cabinet Office. All series are quarterly and evaluated at current prices (benchmark year = 2000).
-
- 4 Private investment normally refers to gross private fixed capital formation in National Accounts. H (...)
ACCUM (capital accumulation rate): The capital accumulation rate is calculated by theratio of gross new investment to gross capital stock of all industries4. Both tangible and intangible assets are included in the definition of capital. The data were obtained from Annual Report on Private Enterprise Capital Stock: 1980 to 2007, Economic and Social Research Institute, Cabinet Office. All series data exclude on-going construction; furthermore, the series are quarterly and are evaluated at constant prices (at market prices for the calendar year 2000).
-
EXPO (growth rate of exports): This is defined as the percentage change in goods and services from the previous quarter. The data for this are obtained from National Accounts for 2008,Economic and Social Research Institute, Cabinet Office. All series are quarterly and are evaluated by the chain-linked method based on the values for the year 2000. This is also the case for consumption, GDP, and government expenditure, which are presented below.
-
CONS (growth rate of consumption): This is defined as the percentage change in private final consumption expenditure from the previous quarter.
-
GDP (growth rate of gross domestic product): This is defined as the percentage change in gross domestic product from the previous quarter.
-
GOV (growth rate of government expenditure): This is defined as the percentage change in government expenditure from the previous quarter. Government expenditure is the total of government final consumption expenditure, gross fixed capital formation, and changes in public sector inventories.
19Since we used quarterly data, we conducted seasonal adjustments using the Census X-12 method. After these preliminary adjustments, we did the above processing. The samples are from 1985:1–3 to 2008:1–3. As was evident in the previous section, an expansionary fiscal policy was implemented in the 1990s in Japan. Taking this into consideration, we introduce the growth rate of government expenditure as an exogenous variable into our VAR model.
- 5 When non-stationary series are included in the VAR estimation, we can use the vector error correcti (...)
20For these variables, we examined whether each time series variable has a unit root by using the Augmented Dickey-Fuller (ADF) test. Table 2 presents the results of this test. The null hypothesis (the time series has a unit root) can be rejected at the 1% significance level for the growth rate of exports, consumption, and GDP. However, wage share and the capital accumulation rate follow the integrated order of 1, i.e. . In addition, we also applied the Phillips-Perron test in order to address the possibility of a structural break in data. Table 3 presents the results for the existence of unit root in each variable in the Phillips-Perron test. Although non-stationary series are included, we will estimate the VAR model by first including them5.
Table 2: Augmented Dickey-Fuller (ADF) test for unit root
|
WS
|
d(WS)
|
EXPO
|
ACCUM
|
d(ACCUM)
|
CONS
|
GDP
|
|
t-stat.
|
-0.130
|
-10.050
|
-6.698
|
-0.7450
|
-3.560
|
-4.92
|
-9.046
|
|
p-value
|
0.6363
|
0
|
0
|
0.389
|
0.0005
|
0.001
|
0
|
|
exogenous term
|
N
|
N
|
C
|
N
|
N
|
CL
|
CL
|
Notes: Basic samples are from 1985:1–3 to 2008:1–3. The choice of the exogenous term is made in the following manner. First, we conduct the test by introducing both constant and trend terms. If they are both significant, we leave them in the test and test the null hypothesis. The significance level is set at five per cent. Second, if one of the terms is not significant, we omit the trend term and conduct the test again. CL indicates that the test includes both constant and trend terms and C implies that the test includes only constant terms. Further, the term N indicates that there is no exogenous term in the test. In order to take the first-order difference of the series, d is used as an operator. The lag in the ADF test is selected according to the Schwarz information criteria.
Table 3: Phillips-Perron test for unit root
|
WS
|
d(WS)
|
EXPO
|
ACCUM
|
d(ACCUM)
|
CONS
|
GDP
|
|
Phillips-Perron test stat.
|
-1.938
|
-13.964
|
-6.787
|
-0.936
|
-11.183
|
-13.085
|
-9.118
|
|
p-value
|
0.314
|
0.000
|
0.000
|
0.309
|
0.000
|
0.000
|
0.000
|
|
exogenous term
|
C
|
N
|
C
|
N
|
N
|
CL
|
CL
|
Notes: Basic samples are from 1985:1–3 to 2008:1–3. The choice of exogenous term is made using the same method as the ADF test. In order to take the first order difference of the series, d is used as an operator. The bandwidth in the Phillips-Perron test is automatically chosen according to the Newey-West estimator using the Bartlett kernel.
21The VAR model that we will estimate in this study can be formalized in vector and matrix form in the following manner:
22where yt = (WSt, EXPOt, ACCUMt, CONSt, GDPt)’, which is a column vector (5 × 1). The dash mark represents transposition of the vector. The order is determined by the degree of exogeneity in the post-Keynesian model and the Block exogeneity test given below.GOVt represents the growth rate of government expenditure at timet. α = (α1, α2, α3, α4, α5)’ is a constant vector, and β = (β1, β2, β3, β4, β5)’ is a coefficient vector that measures the impact of the growth rate of government expenditure on other endogenous variables.Фj is the coefficient matrix for lagged explanation variables; et is an error vector; moreover, we suppose that the mean is zero, the variance-covariance matrix is constant, and there is no serial correlation.
23Now we describe the method used to decide the order of the variables. According to Cholesky decomposition (or the so-called recursive system, i.e. decomposition of residuals in triangular fashion in VAR), variables must be ordered by degree of exogeneity. However, the Block exogeneity test described below posits that wage share precedes capital accumulation, and capital accumulation precedes wage share. Thus, it is not possible to determine the order uniquely using the exogeneity test. Therefore, we employ a standard post-Keynesian growth and distribution model for determining the order. In the standard model, income distribution is exogenously given, and its impact on effective demand is investigated (see Blecker, 2002). At the equilibrium, exports, capital accumulation, consumption, and capacity utilization rate are simultaneously determined and this yields the economic growth rate. In reality, exports are determined by foreign economic conditions and the exchange rate, which are not considered in this study. Therefore, exports are treated as more exogenous than consumption and investment. In addition, since the Block exogeneity test indicates that investment precedes consumption but consumption does not precede investment, we treated investment as a more exogenous variable than consumption. Finally, GDP growth is considered the most endogenous variable because it is determined by both distribution and effective demand.
24Since we estimated VAR with consumption and exports in the growth term, the estimation does not completely correspond to the basic post-Keynesian (Kaleckian) model. According to thestandardpost-Keynesian model, the levels of all variables are normalized by capital stock (K). Thus, there is a gap between basic theory and our method. Nevertheless, we employ the combination of level of wage share and demand components in growth rate. This is because VARs with level variables cannot generate economically interpretable results and would provide a less-robust estimation (for example, Granger causality, the shape of impulse-response, too-long lag length, etc.). However, although we do not include the level variables normalized by capital stock,we do considerthe estimation with all stationary variables, thereby avoiding the issue of spurious regression.
- 6 We checked the serial correlation in the estimated VAR (3) for lag order 8 by LM test. The null hyp (...)
25Using these settings, we estimate VAR (2) in advance, and then choose an optimal lag order based on Akaike information criteria (AIC) and Schwarz information criteria (SBIC). The results of the test are presented in Table 4. A lag of three is chosen using AIC, and a lag of one is chosen using SBIC as the best lag order. Therefore, we estimate VAR (3) and consider the results first6; thereafter, we compare them with VAR (1) and the VAR model in which all variables are stationary adjusted.
Table 4: Information Criteria
|
lag
|
0
|
1
|
2
|
3
|
4
|
|
AIC
|
-31.08
|
-36.18
|
-36.26
|
-36.46*
|
-36.38
|
|
SBIC
|
-30.80
|
-35.20*
|
-34.58
|
-34.08
|
-33.30
|
Notes:The asterisk indicates the minimum information value. The full form of AIC is Akaike information criteria and that of SBIC is Schwarz information criteria. A maximum lag length of four was used in the test based on the quarterly data.
26We conducted the block exogeneity test in order to confirm whether a variable has causality with any other variables included in VAR (3) in Granger’s sense. The results are presented in Table 5. If the p-value presented in the table is higher than the significance level (set at 10 per cent), the variable in a column is exogenous for the variables in a row. Table 5 indicates that each variable included in VAR (3) has causality in Granger’s sense with at least one other variable, and it also has Granger causality with other variables. Taking these results into consideration, all variables must be included in the VAR analysis.
Table 5: Block exogeneity test for VAR (3)
Notes:Block exogeneity implies that a variable does not have causality with any other variable in the Granger sense in VAR. In this test, the null hypothesis is that ‘a variable is block exogenous’. Therefore, if the null hypothesis is not rejected, it may be concluded that the variable is not necessary in the VAR model.
27Finally, the coefficient of the growth rate of government expenditure
β = (β1, β2, β3, β4, β5)’ is exogenous. Table 6 presents the results. According to this table, an increase in the growth rate of government expenditure raises wage share and has a positive effect on the domestic demand terms; however, it has a negative effect on export growth. Although they are not significant except the impact on GDP growth, we include these terms in the analysis since their inclusion generates results that are more economically plausible than the estimation without government expenditure variables.
Table 6: Estimation of coefficient on the growth rate of government expenditure
|
WS
|
EXPO
|
ACCUM
|
CONS
|
GDP
|
|
GOV
|
0.0403
|
-0.0890
|
0.0027
|
0.1117
|
0.2502
|
|
[0.3588]
|
[-0.4752]
|
[ 0.3784]
|
[ 1.5028]
|
[ 3.3539]
|
Notes: For VAR (3), the value in parenthesis represents the t-value.
28It should be noted that our estimation does not introduce explicitly financial variables in VAR. Post-Keynesians, especially Minskian, have emphasized the role of finance and debt deflation in economic growth or business cycles. In addition, nominal wage and price formation and the role of finance are not explicitly considered in this paper. It is important for Régulationists to investigate them, as these variables evolve under the so-called wage-labor nexus, forms of competition, and monetary regime, respectively. Therefore, one may wonder why our estimation omits these variables.
29Of course, the determination of these variables and its macro-economic impact should be also examined carefully. However, let us give three reasons for omitting these variables. The first reason is that it is not preferable to increase the number of variable in VAR because it reduces the degree of freedom. Each equation VAR model includes its own lag variable and any other variables with lag. Let us take an example of VAR (3) composed of 8 variables (e.g., wages, prices, finance variable, consumption, exports, capital accumulation and GDP). In this case, an equation needs at least 24 explanatory variables excluding constant terms. If so many explanatory variables are included in the analysis, the degree of freedom in estimation largely decreases. The lack of degree of freedom results in a statistically not confident estimation. Therefore, it is preferable that we set the VAR model as simply as possible by restricting the number of endogenous variables.
- 7 Moreover, we believe that the dynamics of wage and price determination should be examined more prec (...)
30Second, even if we include the impact of debt, price deflators, and equity prices in VAR, it is not easy to understand the dynamics among demand variables and financial variables. We may interpret the impact of debt on the capital accumulation in terms of a debt-led or debt-burdened regime. However, economic interpretation on the impact of exports on debt, of GDP growth on price deflators, and of consumption on equity prices, for example, is not so easy. The same reason applies to why we do not include nominal wages and prices in VAR. The economic interpretation of impulse response is not consolidated, for instance, for exports to nominal wages and exports to prices. Focusing on the income distribution share is the most appropriate variable for post-Keynesian oriented empirical analysis. Thus, we consider that the dynamics of prices and wages are reduced to the wage share7.
31Last but not least, we do not consider that the financial aspects are neutral for the macroeconomic performance of the Japanese economy. Therefore, we have integrally investigated the dynamics of finance and distribution on capital accumulation in another paper (Nishi, 2011). The results of structural VAR indicated that although profit share sustains capital accumulation, a high debt ratio restrains it. In other words, the income distribution-capital accumulation pattern of the Japanese economy over the past 20 years has been a profit-led one, and the debt-capital accumulation pattern over the same period has been a debt-burdened one.
32This section investigates Japan’s macroeconomic performance using VAR (3). Using innovation accounting, we consider the relationship between income distribution and each demand component; thereafter, we examine the impact of changes in individual demand components on aggregate demand formation.
33Figures 2 and 3 illustrate impulse response functions and accumulated impulse response functions derived from VAR (3), respectively. Table 6 presents the results of variance decomposition for VAR (3). In addition, Figure 4 indicates the accumulated impulse response functions derived from VAR (3) with all stationary variables. Estimation with all stationary variables is the preferred method since it prevents spurious regression. The following is a brief explanation of innovation accounting. Impulse response functions trace the time path of the various shocks on the variables included in the VAR model. The variance decomposition reveals the proportion of movements in a sequence due to its own shocks versus shocks to other variables (see Enders, 2010). The impulse is given by the Cholesky decomposition. Since income distribution is assumed to be predetermined, the first column of the second, third, and fourth rows (cells 2.1, 3.1, and 4.1) illustrates the relationship between income distribution and each demand component. The fifth row represents the aggregate demand formation pattern.
Figure 2: Impulse response functions
Note: The rows represent response functions in the order of wage share, growth rate of exports, capital, consumption, and GDP. The dotted lines represent a standard error of ±2 in the impulse response function; however, this is not appropriate since VAR includes non-stationary series. The existence of non-stationary variables might yield impulse response functions with rather large standard errors. Examples are provided in Enders (2010) and Hamilton (1994).
Figure 3: Accumulated impulse response functions
Note: The rows represent accumulated response functions in the order of wage share, growth rate of exports, capital, consumption, and GDP. The dotted lines represent a standard error of ±2 in the accumulated impulse response function; however, this is not appropriate since VAR includes non-stationary series.
Figure 4: Accumulated impulse response functions of VAR (3) with all stationary variables
Note: The rows represent accumulated response functions in the order of wage share, growth rate of exports, capital, consumption, and GDP. The dotted lines represent a standard error of ±2 in the accumulated impulse response function.
34Cell 2.1 indicates the relationship between wage share and growth rate of exports. From a theoretical viewpoint, there are two patterns for the dynamics of these two variables. First, Blecker (2002), using a post-Keynesian model, explains that an increase in wage deteriorates the price competitiveness of domestic firms, which lowers export performance. In other words, his model explains that a wage cut (fall in wage share) improves cost competitiveness and increases (net) exports. On the other hand, Nakatani (2009), using a similar post-Keynesian model, indicates that an increase in profit share also causes a reduction in firms’ price competitiveness, which decreases export volume. Our results support those of Blecker’s model; however, net exports are defined by the ratio of capital stock in his model. A positive change in wage share decreases export demand through promoting a negative response.Variance decomposition indicates that a change in income distribution explains three per cent of the change in exports in the long run, and this impact is the third largest among all the variables.
- 8 As indicated in recent post-Keynesian studies (see Taylor, 2004), both firms’ profit income and ext (...)
35Cell 3.1 indicates that an increase in wage share continuously restrains capital accumulation. Since the other side of wage share is profit share, this response implies a profit-led accumulation pattern. This result is probably because the internal funds of these Japanese firms are used to stimulate their investment activity, as discussed by Naastepad and Storm (2007) and Azetsu et al. (2010)8. Furthermore, we can also support the argument proposed by Hashimoto (2002), which states that a large increase in wage share during the middle of the 1990s decreased firms’ profitability and caused their investments to be restrained. According to the variance decomposition, income distribution can explain over 50 per cent of the fluctuation in capital accumulation, which has the largest impact of the included variables. In summary, wage share determines the cost of investment activities, whereas profit share provides the financial resources for such activities.
Table 7: Variance decomposition for VAR (3)
|
Growth rate of exports
|
|
Growth rate of consumption
|
|
Period
|
WS
|
EXPO
|
ACCUM
|
CONS
|
GDP
|
Period
|
WS
|
EXPO
|
ACCUM
|
CONS
|
GDP
|
|
1
|
0.11
|
99.89
|
0.00
|
0.00
|
0.00
|
1
|
1.25
|
0.83
|
7.68
|
90.24
|
0.00
|
|
5
|
3.12
|
90.80
|
1.83
|
1.09
|
3.16
|
5
|
5.67
|
7.66
|
10.09
|
68.84
|
7.73
|
|
10
|
3.11
|
89.35
|
2.14
|
1.91
|
3.49
|
10
|
7.59
|
8.13
|
10.72
|
66.01
|
7.56
|
|
15
|
3.17
|
89.13
|
2.25
|
1.95
|
3.50
|
15
|
8.42
|
8.21
|
10.73
|
65.07
|
7.57
|
|
20
|
3.17
|
89.02
|
2.35
|
1.96
|
3.50
|
20
|
8.87
|
8.20
|
10.71
|
64.62
|
7.60
|
|
25
|
3.23
|
88.88
|
2.41
|
1.96
|
3.52
|
25
|
9.06
|
8.24
|
10.75
|
64.37
|
7.58
|
|
30
|
3.34
|
88.73
|
2.44
|
1.96
|
3.54
|
30
|
9.07
|
8.28
|
10.85
|
64.24
|
7.57
|
|
Capital accumulation rate
|
|
Growth rate of GDP
|
|
Period
|
WS
|
EXPO
|
ACCUM
|
CONS
|
GDP
|
Period
|
WS
|
EXPO
|
ACCUM
|
CONS
|
GDP
|
|
1
|
0.00
|
5.94
|
94.06
|
0.00
|
0.00
|
1
|
0.77
|
4.56
|
31.84
|
26.80
|
36.02
|
|
5
|
6.04
|
17.67
|
67.60
|
2.92
|
5.77
|
5
|
5.06
|
14.86
|
23.42
|
30.29
|
26.37
|
|
10
|
23.12
|
13.55
|
51.30
|
2.72
|
9.30
|
10
|
8.21
|
15.67
|
22.69
|
28.86
|
24.57
|
|
15
|
37.60
|
9.43
|
39.28
|
2.00
|
11.69
|
15
|
9.26
|
15.54
|
22.46
|
28.53
|
24.21
|
|
20
|
47.45
|
7.28
|
31.08
|
1.55
|
12.65
|
20
|
9.59
|
15.57
|
22.51
|
28.33
|
24.00
|
|
25
|
52.66
|
6.57
|
26.66
|
1.36
|
12.75
|
25
|
9.59
|
15.63
|
22.69
|
28.21
|
23.88
|
|
30
|
54.32
|
6.73
|
25.15
|
1.37
|
12.43
|
30
|
9.60
|
15.64
|
22.87
|
28.10
|
23.80
|
Notes: WS represents the shock forwage share, ACCUMfor capital accumulation, CONS for consumption, EXPO for Exports, and GDP for GDP.
36It is evident from cell 4.1 that while an increase in wage share stimulates consumption slightly in the short run, it does not contribute to increasing consumption demand in the accumulated response term. This pattern implies a Kalecki-Kaldor-type macroeconomic consumption function does not apply to the Japanese economy. The variance decomposition indicates that a change in income distribution explains approximately 9 per cent of change in consumption for 30 periods. The fourth row of the impulse response function indicates that domestic demands such as capital accumulation, consumption, and GDP growth have a positive impact on consumption. Further, capital accumulation explains the 10.8% change in consumption at maximum, which has the largest impact of the variables. In innovation accounting, it may be assumed that both an increase in wage share and forward-looking factors, such as a favourable economic perspective for the domestic economy based on the recovery of private investment, are required for the stimulation of consumption.
37Further, cell 5.1 indicates that the aggregate demand regime is profit-led. This result is in agreement with Azetsu et al. (2010), although the sample periods and variables are different. The fifth row of the impulse response function is important because it indicates the relationship between individual demand and aggregate demand. The second, third, and fourth columns of the fifth row (cells 5.2, 5.3, and 5.4) indicate that increases in exports, capital accumulation, and consumption expanded the accumulated growth rate of GDP. Consumption is the most important factor for explaining fluctuations in GDP from a long-term perspective. According to Figure 6, consumption accounts for 28.1% of the fluctuations in GDP; investment, 22.8%; and exports, 15.6%. Income distribution has the least impact on the fluctuation of aggregate demand and accounts for 9.6%.
38Cell 1.5 of Figures 3 and 4 displays how a positive impulse to GDP growth has a cumulative effect on the behavior of wage share. The accumulated response of wage share (in both level and first difference term) falls with an expansion of GDP growth. This is because an increase (a decrease) in GDP growth restrains wage share especially in the short-run (cell 1.5 of Figure 2). That is, the evolution of wage share in the Japanese economy has generally been counter-cyclical, and that of the profit share has been pro-cyclical.
39In general, wage share falls in economic boom and increases in recession. Our results, obtained by innovation accounting, are consistent with this stylized fact. Uemura (2011) explains that one of the reasons for this dynamic is hoarding of labor during recession, and it was effective at least during the 1990s. The trajectory of response functions implies that this labor hoarding is effective especially in the short-run, as wage share falls for this period. This result is also true even if we relax the assumption that wage share is the most exogenous variable in a post-Keynesian framework (see also Figure A1 and A2 in Appendix).
40The Keynesian viewpoint is important because investment demand has driven growth and cycles in the Japanese economy. The third row illustrates accumulated responses concerning capital accumulation: capital accumulation is supported by an increase in exports, consumption, and GDP. Of these three determinants, the growth rate of export demand best explains the fluctuation in capital accumulation from the short-run perspective—it accounts for approximately 18% for five periods. Thus, in the short run, an expansion in exports has a strong impact on capital accumulation. This effect decreases gradually in the long-term, and, finally, capital accumulation is largely explained by income distribution, as mentioned above.
- 9 The result that consumption would be profit-led should not be surprising. In a basic post-Keynesian (...)
41From the results obtained above, the following dynamics of growth and distribution are evident. The profit-led demand formation pattern indicated in cell 5.1 is accompanied by a mechanism in which an increase in profit share sustains consumption, capital accumulation, and export growth. Taking the results of variance decomposition into consideration, this pattern is created not because an increase in profit share has a direct, positive effect on GDP growth, but because it has a strong impact, particularly on capital accumulation that contributed significantly to raising GDP growth. In addition, the autonomous growth of consumption, capital accumulation, and exports also contributed to GDP growth. Moreover, while the impacts from income distribution on exports and consumption demand are relatively small, they are both profit-led9. Considering these dynamics, both autonomous growth and growth affected by profit share may contribute to the profit-led accumulated response functions.
42We have ordered variables on the basis of Block exogeneity test and basic post-Keynesian modeling. The decomposition of residuals is calculated in triangular fashion by Cholesky decomposition, therefore variables are ordered by degree of exogeneity.
43The wage share has been ordered first in our analysis so far. However, the dynamics of exports may be exogenous in character both theoretically and empirically. As mentioned above, theoretically exports are affected by the exchange rate, which also depends on the condition of financial factors (Nishi, 2010). Since these factors are not included in our model, it is necessary to explain VAR dynamics based on the assumption that exports are exogenous in character. Empirically, the expansion phase of the Japanese economy is led by exports, as we have seen in Table 1. This fact implies that exports are also determined by the economic growth of trade partners. Especially after 2002 until the subprime shock, exports increased along with economic growth in the US and Asia (China above all) with which the Japanese economy has a large trade share (Uemura, 2011). Empirical facts also require explanation of VAR dynamics with exogenous exports.
- 10 To be more precise, under Cholesky decomposition, exports are determined last while they are indepe (...)
44Taking these possibilities into consideration, we conducted innovation accounting based on VAR (3) in which the order is EXPO, WS, ACCUM, CONS and GDP. We also investigated the accumulated impulse response function of VAR (3) with all stationary variables in which the order is EXPO, d(WS), d(ACCUM), CONS, and GDP. In this setting, exports are the most exogenous in character10. The accumulated response function is shown in Figures A1 and A2, and the variance decomposition is summarized in Table A1 in the Appendix.
45From the accumulated response function and variance decomposition, a profit-led demand formation pattern is still obtained, and increase in profit share sustains export growth, capital accumulation and consumption. The variance decomposition indicates that profit share had an important impact particularly on capital accumulation which contributed to the dynamics of GDP growth. Thus, the autonomous growth of consumption, capital accumulation, and exports also contributed to GDP growth in the long run, as in the results obtained from canonical VAR (3) in which the order is WS, EXPO, ACCUM, CONS and GDP.
46In order to confirm the robustness of the results obtained by VAR (3), we conducted a few additional tests and compared them. We compared VAR (1) with VAR (3). According to SBIC, the optimum lag length in VAR is one. In VAR (1), an increase in wage share stimulates the growth rate of consumption in the long run. However, the accumulated response function of consumption indicates that other variables have a larger magnitude in relation to consumption than wage share does. As to other variables, the accumulated response function shows little difference.
47Finally, we consider a VAR model in which all variables are stationary to prevent spurious results. Both ADF and Phillips-Perron tests indicate that the integrated order of capital accumulation rate and wage share is. Therefore, we stationalize them and include them in the VAR model. Thus, we can estimate the VAR model without fallacious regression. The accumulated impulse response function with all stationary variables is presented in Figure 4. In this case, while a positive shock to wage share increases consumption demand in the very short run, the effect disappears in the long run. With regard to other variables, we could not find any particular difference in the form of the accumulated response function.
48The results obtained by comparison indicate there is no particular difference as to the shape of the accumulated response functions of VAR (3), VAR (1), and the stationalized VAR. Therefore, we made our interpretation of Japan’s macroeconomic performance by employing VAR (3).
49Innovation accounting is, in effect, a mechanical analysis. Therefore, by comparing the evolution of Japan’s actual macroeconomic performance with the results obtained, we provide an assessment of that performance.
50With regard to the aggregate demand formation pattern, both external and domestic demands have a positive relationship with GDP growth rate. It must be noted that the accumulated impulse response function displays a positive relation among GDP growth, consumption, and investment demands (in block from cells 3.3–5.5). This implies that domestic demands positively stimulate each other. Therefore, an increase in one of these variables induces increases in other variables; however if one of them decreases, it causes stagnation in the others.
51Thus, it is evident that each demand component could ‘potentially’ induce GDP growth. Let us compare these mechanisms with historical events in the Japanese economy. In the bubble period, with the rise in asset prices, this potential was realized. On the other hand, after the bubble burst, growth rates of both consumption and investment stagnated. In the 1990s, consumption demand was not encouraged as much despite the rise in wage share. This is probably due to the increasing uncertainty regarding job security and social security during the long-run economic recession (Nakagawa, 1999; Horioka, 2007). As a result, the expenditure schedule of households deteriorated. With regard to the stagnation of investment demand, factors differ depending on the period. In the beginning of the 1990s, stagnation was basically due to the stock adjustment mechanism, as indicated by Yoshikawa (1998). During the middle of the 1990s, profit squeeze was the main factor, as explained by Hashimoto (2002). Motonishi and Yoshikawa (1999) indicated through an econometric analysis that the so-called credit crunch, particularly for SMEs, and the deterioration of the profit perspective were the main determinants of the investment stagnation at the end of this decade. Our results emphasize income distributional factors. We can interpret that combined with these factors, the Japanese economy failed to achieve a domestic demand-led economy. On average, there was a relatively good export growth rate, as indicated in Table 1. It is evident from the demand structure of the Japanese economy since 1990 that while domestic private demand stagnated, external demand and government expenditure sustained the aggregate demand. After 2000 (particularly since 2002), although the GDP growth remained low, exports encouraged investment demand and were the driving force behind economic growth.
52Taking these findings on the Japanese macroeconomic performance into consideration, we summarize our interpretation. In the 1980s, while wage share was relatively low, the domestic demand led to high economic growth. This is partially due to the bubble economy. However, after the bubble burst, wage share rose rapidly but did not stimulate the growth of individual demand. Each demand component stagnated while fluctuating. As a result, the macroeconomic performance was not favourable. However, in the 2000s, the expansion of exports and investment demand, stimulated by the recovery of profit share, has led to a favourable macroeconomic performance.
53This subsection reconsiders the macroeconomic characteristics of the Japanese economy from an international perspective, which is an important task for the Regulationist approach.
54As briefly surveyed in the related literature presented in Section 3, post-Keynesian studies have presented a variety of macroeconomic regimes in major economies.One of the theoretical references used as a post-Keynesian empirical study is Bhaduri and Marglin (1990), which explained the mechanism of both wage- and profit-led growth regimes. Inspired by their research, Bowles and Boyer (1995) tested investment, profit share, savings, and net exports by employing a single equation approach. They find that France, Germany, and Italy are weakly profit-led, whereas UK and USA are wage-led. The domestic sectors of the economies of all the studied countries are wage-led.
55Further, by introducing interest rates into Marglin and Bhaduri’s (1990) model, Hein and Ochsen (2003) re-examined the ‘profit squeeze’ explanation as a reason for the growth slowdown of thelate 1960s. In addition, the analysis in Hein and Vogel (2008) is also based on a demand-driven growth and distribution model for an open economy.They found that the growth patterns in France, Germany, the UK, and the US were wage-led,whereas those in Austria and the Netherlands were profit-led. Similar to Bowles and Boyer, theyconfirmed that opening up an economy might transform a wage-led growth regime (at the domestic level) to aprofit-led regime due to the effect of distribution on external trade.
56Naastepad (2006) integrated the demand regime and theproductive regimepresented by régulationists. Using econometric analysis, she explained that the growthtrajectory of the Netherlands was outright stagnationist. Naastepad and Storm (2007) extended the analysis to economies of the major OECD countries. They found that demandis wage-led in France, Germany, Italy, the Netherlands, Spain, and the UK and profit-led inJapan and the US. Moreover, they argued that the decline in world trade growth is a majorcause of the sluggish growth in all economies.
57These studies distinguish patterns of growth and income distribution based on theoretical models using single or simultaneous equations approaches. On the other hand, Barbosa-Filho and Taylor (2006) and Stockhammer and Onaran (2004) have performed post-Keynesian time series analyses. Barbosa-Filho and Taylor presented a VAR model comprising capacity utilization and income distribution. Using this framework, they indicated that the growth regime of the US economy has been profit-led. Stockhammer and Onaran developed a structural VAR model that includes five variables: capital accumulation, the capacity utilization rate, profit share, the unemployment rate, and productivity growth. In their empirical study of the US, the UK, and France, they indicated the validity of the Keynesian argument that labour demand is principally determined by effective demand.
58Thus, each of the previous studies attempted to reveal the growth regime types that exist in various countries, although the examined countries andanalytical approaches differ. An international comparison using a post-Keynesian oriented model reveals that growth performance varies according toeconomy and time; therefore, a ‘best’ growth regime may not exist. Moreover, regimes may include hybrid demand formation patterns comprising both wage- and profit-led regimes.
59According to our results, the Japanese profit-led demand formation pattern is characterised by a mechanism that causes increases in profit share to have a strong impact, particularly on capital accumulation that significantly contributes to increasing GDP growth. In addition, the autonomous growth of consumption, capital accumulation, and exports also contributed to GDP growth. Taking weak profit-led exports and consumption into account, both autonomous growth and individual demand simulated by profit share may contribute to profit-led accumulated response functions.
60Thus, the implication of our empirical results is that although Japanese capitalism is generally characterised by profit-led growth, it involvespeculiar demand formation patterns. Compared with other countries, Japanese capitalism has its own growth trajectory that differs from the growth patterns of other countries. Oursurvey of existing international research and results obtained by VAR support the idea that the capitalism has diverse growth regimes, as the Régulationistshave progressively indicated.
61This study estimated VAR models comprising income distribution, the growth rate of consumption, capital accumulation, exports, and GDP; it developed an empirical analysis based on the impulse response function and variance decomposition. Our results reveal that the aggregate demand regime is profit-led, as Azetsuet al. (2010) concluded. Moving beyond their analysis and using innovation accounting, we examined the properties of this profit-led pattern in greater detail.
62The impulse response function indicates that the profit-led regime was dominant between 1985 and 2008 in the Japanese economy. In this regime, both domestic and external demand components had a positive relationship with GDP growth. In other words, they had a potentially expanding effect on GDP. In the bubble period, characterised by a favourable economic perspective and rise in asset prices (which encouraged consumption and investment), this potential effect was realized. However, in the 1990s after the bubble burst, the increase in wage share did not generally contribute to consumption growth. In addition, the stock adjustment mechanism, deterioration of firms’ profitability, and credit crunch led to investment demand stagnation. As a result, the Japanese economy failed to realize a domestic demand-led economy. After 2002, exports demand expanded while profit share recovered, which induced an increase in investment demand and resulted in favourable macroeconomic performance.
63The results obtained from our empirical analysis have important policy implications. The shape of the impulse response function indicates that an increase in wage share does little to stimulate consumption. However, other demand components have a positive impact on consumption growth. Therefore, if the economic perspective in relation to demand components improves, it may stimulate consumption growth. Moreover, the accumulated impulse response function for the Japanese economy demonstrates that increases in consumption potentially stimulate the GDP together with other domestic demand components. In this sense, growth driven by domestic demand may be possible. However, if it is not realized, the economy can only grow from external demand. In order to realize a domestic demand-led economy, policies or institutional designs that eliminate factors contributing to instability for households’ consumption and deteriorating profitability for firms’ investment are required.
64Lastly, a few issues remain for consideration in future research. We provided our interpretation of Japan’s economic performance since the late 1980s using VAR models that had ‘constant’ parameters. Interpretations using VAR models must be made carefully since the Japanese economy has experienced numerous structural changes, including industrial changes, international production linkages, and surplus spillovers (Uemura, 2009). In order to consider the impacts of these changes, both VAR analysis at the macroeconomic level and panel data analysis that introduces individual heterogeneity or time effects would be useful. Our analysis essentially focuses on the realside of the economy; however, as we have occasionally mentioned in this study, an empirical analysis introducing the financial side is also required. This is important because numerous studies have indicated that financial panic, such as the credit crunch that occurred in 1997, contributed to the stagnation of investment and GDP growth. Finally, the samples in our analysis do not include the period after the so-called world financial crisis. Japan’s growth rate of real GDP in the fourth quarter of 2008 was -3.0%; the decrease in the growth rate of exports of goods and services was particularly severe, falling to -13.5%. The rapid fall in growth rate due to the negative export demand shock implies, according to our findings, that the increasing dependence on external demand is one of the weak points of Japan’s recent economic growth regime. The impacts of the world financial crisis on the Japanese economy must be investigated, as post-Keynesians have traditionally emphasized the impacts of the financial side on the real economy.