The Failure of Fiscal Policy and Japan s Lost Decades: The Role of Household Expectations

Transcription

1 The Failure of Fiscal Policy and Japan s Lost Decades: The Role of Household Expectations Geeta Garg April 2, 218 Abstract This paper examines the contribution of anchoring of fiscal expectations to the economic slowdown in Japan post While the monetary policy is constrained by the effective lower bound, the fiscal policy is characterized by large gaps between planned and implemented policies in this period. Continuous failure to achieve announced stimulus and consolidation targets affected household expectations. Using a Bayesian Structural Vector Autoregression Model, this paper finds that while expansionary fiscal policy failed to stimulate the Japanese economy, a contractionary fiscal policy in turn improved the macroeconomic performance by stabilizing household expectations. Given the high existing debt burden of Japan and large holding of debt by the private sector, the fiscal consolidation actions benefitted the private sector by reducing the riskiness of their debt holding. The coordination between monetary policy and fiscal policy influenced these outcomes. Keywords: key1, key2, key3 JEL Codes: key1, key2, key3

2 1 Introduction Post the asset price bubble collapse in the early 199s, Japanese economy experienced an uninterrupted economic decline that spanned over more than two decades along with brief periods of recovery (such as between and 22-8). Table 1 provides a summary of the macroeconomic situation in Japan starting 197s. While the economy was growing steadily in the decade of 7s and 8s, the performance deteriorated sharply in the decade of 9s and 2s on all fronts. The growth of capital formation and inflation in fact turned negative in 2s. Owing to the changes in the fiscal and monetary policy starting 213 onwards, the economy experienced a mild recovery between 21 to 17. Table 1: Macroeconomic Scenario from Period GDP Consumption Per Capita Gross Capital Consumer Price Growth Growth GDP Growth Formation Growth Inflation - All Items(%) Note: The numbers are average annual growth rates One of the factors that contributed to the slowdown was that the Japanese fiscal policy was characterized by sudden changes in the fiscal direction and large gaps between planned and implemented policies starting 199s. Posen (1998) argues that while claims are made of variously 65 to 75 trillion Yen spent in total stimulus efforts since 1991, the actual amount injected into the economy by the Japanese government - through either public spending or tax reductions - was only 23 trillion Yen, about a third of the total amount announced. In addition, on multiple occasions since 199s, the positive response to fiscal stimulus was was undercut by sudden fiscal contraction. Figure 1 (left) depicts these inconsistencies in the fiscal policy. The bars in Figure 1 represent the yearly change in the magnitude of government expenditure, tax revenues and bond issues in the General Account of the central 1

3 government. In the year 1991 and 92, higher government expenditures were financed by higher tax revenues and no bonds were issued implying the reversal of expansionary fiscal policy. Then in the year 1993 and 1994, fiscal policy turned expansionary. Although fiscal policy was expansionary in the year 1994, it was during this year the government raised the consumption tax from 3% to 5% which in turn stalled recovery and pushed the economy to a slump despite the overall expansionary stance of the fiscal policy. In the year 1996, fiscal policy was highly expansionary as higher government expenditure were financed by higher bond issues whereas tax revenues declined by a substantial amount during this year. The stance of fiscal policy shifted continuously between contraction and expansion until 28 despite the stagnant economy. After 21, the focus of fiscal policy appears to be on stabilizing debt as the bond issues declined every year and higher government expenditures were followed by even higher tax revenues. Despite the introduction of highly publicized three arrows of Abenomics : (1) Expansionary monetary policy, (ii) expansionary fiscal policy and (iii) structural reforms in December 212, we do not observe much expansion in fiscal policy. The government expenditure barely increased after 212 whereas tax revenues increased consistently. In fact, consumption tax was raised from 5% to 8% in April, 214 once again contributing to the slowdown of the economy. The Figure 1 (right) shows the difference between the magnitude of actually implemented and the announced fiscal policy stimulus. At times the difference between the announced and implemented fiscal policy was as high as % of GDP again suggesting inconsistencies in Japanese fiscal policy. 2

4 1 4 Government Expenditure Tax Revenue Bond Issues 2 Government Expenditure % GDP Tax Revenue % GDP Bond Issues % GDP Billion Yen.5 % GDP Year Year Figure 1: Change in Fiscal Stance (Left) and Difference between Announced and Actual Policy Stimulus (Right) While the above figure indicates contractionary fiscal policy for most part of the two decades, it is in contrast with the expansionary fiscal policy scenario as indicated by the primary surplus to GDP ratio. It stayed negative over the last two decades (Figure 2) suggesting an expansionary fiscal policy. One of the factors responsible for negative primary surpluses is the slow growth of the GDP that in turn lowers the tax revenue collection because fiscal policy during this period was not expansionary enough to have contributed to such high negative primary surplus. The decline in the growth of real GDP is evident from Table 1. The Japanese economy grew at about 4% per annum in the decade of 198s and declined to barely 1% per annum in the decade 199s and further to barely about half a percentage point per annum in the decade of 2s. However, it did increased to 1.2% per annum in indicating a slight improvement in the economic performance. Thus, negative primary surplus which is often suggested as an indicator of expansionary fiscal policy in Japan may not present the true picture of fiscal policy. 3

5 Figure 2: Primary Surplus as Percentage of GDP These sudden contractions and expansions in fiscal policies as discussed above failed to keep household expectations anchored rendering the fiscal policy unsuccessful. The result is over two decades of macroeconomic slowdown as is evident in Table 1. At the same time, the role of conventional monetary policy was muted during this time period. Figure 3 below presents the behavior of monetary policy in Japan. Figure 3: Policy Rate and Inflation (Left) and Quantitative Easing (Right) The fiscal policy cannot be viewed in isolation while estimating its impact on the real economy. The coordination from the monetary policy is important in reviewing the efficacy of fiscal policy for any country. However, as we will see the role of monetary policy was limited in case of Japan especially for the first decade after In Figure 3 (left), the uncollateralized call rate (policy rate) declined from about 8% in early 199s to about.5% in

6 after which it continued to stay at the effective lower bound of % with brief periods when it was raised above % such as the August, 2 (When Zero Interest Rate Policy (ZIRP) was lifted) and March, 26 (when Quantitative Easing (QE) ended). These short increases in interest rates have been viewed as some of the mistakes of monetary policy when interest rates were raised in false expectation of continued economic expansion (See Ito and Mishkin (26) and Ito (214) for more detailed review of Japanese monetary policy). However we observe huge fluctuations in the core inflation which cannot be attributed to the conventional monetary policy. As far as unconventional monetary policy is concerned, the Bank of Japan (BoJ) experimented with different forms of Quantitative Easing starting March 21. The BoJ adopted new instrument - Current Account Balances (reserve balances) in March 21 by targeting the amount of excess reserves which in turn forced the interest rates to go to zero while increasing the purchase of long term Japanese government bonds on the asset side. This can be seen as the blip in both the reserve balances and BoJ holding of government debt between 21 and 26 in Figure 3 (Right) above. However, the magnitude of this QE was not huge. But this QE did lead to a mild recovery. Inflation increased, although not persistently, during this period. Then in October 21, a version of QE - Comprehensive Easing (CE) was introduced in which BOJ introduced a temporary Asset Purchase Program (APP) as a part of which it purchases assets and provides liquidity with fixed-rate, funds-supplying operation. However, due to small size of this program, it didn t have much impact on the economic outcome. The asset purchase program (APP) is on the balance sheet but deemed temporary. However, this was the first time when the signs of coordination between fiscal and monetary policy were visible. The government and the BOJ signed a joint statement that they both would cooperate toward overcoming deflation when CE was introduced. In January 213, the BoJ officially introduced an inflation target of 2% in order to improve the credibility of the BoJ. It was only in April, 213 that the BoJ adopted Qualitative and Quantitative 5

7 Easing (QQE) in which it pursued an aggressive balance sheet expansion that doubled the size of its balance sheet and lengthened the maturity of Japanese government bonds (JGBs) BoJ hold on its balance sheet (Ito, 214). Figure 3 (left) above shows this expansion as the steep increase in reserve balances and BoJ holding of the government debt. The size of this expansion was substantial enough that inflation started increasing in expectation much before the QQE was implemented. The appreciation pressures on the Japanese Yen were also lowered in response to this QE. In fact Between mid-november 212 and mid-may 213, the yen depreciated by more than 2% (Table 2 below shows that the Yen/Dollar exchange rate increased from 87.8 Yen per Dollar in 21 to Yen per Dollar in 217). This was also the period when the objectives of the BoJ and the government were aligned with each other 1 QQE is considered to be much more successful than the earlier episode of QE by BOJ 2 One of the other factors that might have further prolonged the stagnation in Japan in addition to inconsistencies in fiscal policy could also be lack of coordination between fiscal and monetary policies. It is hard to find evidence especially before 21 of the two policies trying to coordinate to stimulate the economy. The rhetoric of fiscal policy had been contractionary throughout the two decades 3. Bernanke (217) argues that a key reason that a government might not approve an expansionary fiscal program at a time when it is warranted by macroeconomic conditions is concerns about the resulting buildup of the national debt. If legislators believed that monetary policies would be used to offset that buildup, they might 1 This statement is arguable as the cooperation between the government and the BoJ can also be viewed as lack of independence of the BoJ despite the legal independence achieved by the BoJ in There was a strong pressure on the BoJ by the Abe government to introduce QE and expand the size of balance sheet. 2 See Ito (214) for more details. 3 The effort to consolidate fiscal expansion began primarily in The government almost achieved primary balance in FY1997 (The consumption tax was increased during this year that further stalled the economic recovery). Then the Fiscal Structural Reform Act was enacted in November, 1997 to push efforts towards fiscal consolidation by bringing down fiscal deficit by FY23 and by reducing bond issues. It was however suspended in December, 1998 owing to weak economic recovery and the fiscal consolidation targets were shifted to FY25. Again in 21, Japan launched the Fiscal Management Strategy that targeted reduction in new government bond issues, reducing primary deficit by half by 215 and achieving a primary surplus by 21. Source: Economic and Fiscal Policy Management by Cabinet Office, Japan ( 6

8 be more willing to act. Moreover, they would understand that monetary policy would not lean against the expansionary fiscal actions, increasing the multiplier and providing more bang for the buck. To the extent that the central bank honors its commitments to allow a temporary overshoot of its inflation target, the impact of the combined program would ultimately be all the larger. While the data shows that the interest rates were stuck at zero lower bound, however, the monetary policy statements by the BoJ shows constant possibility of raising interest rates amidst fears of higher inflation 4 The lack of commitment by the monetary policy could possible be the source of contractionary fiscal policy. In addition to the failures of monetary and fiscal policy, the contribution of various structural factors to the Japanese economic slowdown cannot be denied. The slow response to deal with the problem of non performing loans post the 199s crisis eventually resulted in Banking crisis of and lasted until early 2s when serious steps were taken to rectify it. This led to the weakening of the financial system. In addition, the appreciation pressures on the exchange rate despite a weak economy also contributed to deflation (Ito and Mishkin, 26). The Asian currency crisis in 1997, the collapse of IT stock prices in Japan as a result of the collapse of IT bubble in USA in 2, the financial crisis of 28, the Great East Japan Earthquake in 211 added to the existing economic problems of Japan. Another oft-quoted concerns about the Japanese economy is the rising Debt to GDP ratio which is a little over 23% (Gross liabilities to GDP ratio) and is the highest among the 4 Statement by the BoJ on July 17, 2 : The majority of the Policy Board views that Japan s economy is coming to a stage where deflationary concerns are dispelled, which the Board have clearly stated as the condition for lifting the zero interest rate policy (Figure 3 (Left) shows that inflation was still less than zero around this time and the increase was implemented against protests from the government). Statement by the BoJ on July 14, 26 : The Bank has maintained zero interest rates for an extended period, and the stimulus from monetary policy has been gradually amplified against the backdrop of steady improvements in economic activity and prices. In this environment, maintaining the previous level of the policy interest rate may result in large swings in economic activity and prices in the future. Taking account of the current assessment of economic activity and prices from the two perspectives outlined in the New Framework for the Conduct of Monetary Policy (March 26), the Bank judged it appropriate to adjust the level of the policy interest rate at this juncture so that a desirable course of economic activity and prices was to be maintained. 7

9 OECD economies. This has often been presented as argument in favor of fiscal contraction. If the Japanese debt were risky, we would witness a decline in the debt holding of the private sector in expectation of a fiscal crisis and a rise in long term bond yield. If this risk is high, then the yield on long-term government bonds should be higher today in order to compensate bond holders for future expected losses due to default or inflation. In other words, the long-term bond market is designed to price inflation and default risk (Broda and Weinstein, 24). However, as the Figure 3 (Left) above shows, the 1 year government bond yield has been falling in Japan since 26 when QE was introduced and eventually turned negative in 216. Given these high levels of debt and such low interest rates, it seems the debt is considered safe and is expected to be paid off. 2 Government Debt There is a constant debate in the literature about the true level of debt in Japan. One of the reasons for this apparently very high level of debt is because of the difference in the way public accounting is done in Japan as pointed by Broda and Weinstein (24). They present several arguments against focusing just on the liabilities side of the government balance sheet. First of all, government borrowing used to finance government lending is not required to be paid with tax revenues because the revenue from the repayment of those loans can be used to pay off those borrowings. As of 217 (Quarter 4) about 16% of all lending in Japan is made by public financial institutions In the Japanese public accounting system the surpluses from one government account may be used to make deposits in other government account thus creating a pessimistic view of government s financial health by inflating the liabilities. Secondly, in the Japanese public accounting system, different government accounts are kept separate and therefore surpluses in one account (e.g. the social security funds that records a corresponding asset on its balance sheet) are lent to another government account (e.g. the Fiscal Investment and Loan Program that records a liability). This would again provide an overly 8

10 inflated view of the government debt if debt is viewed as gross liabilities of the government 5. Thus to get a proper idea of the true burden of the government debt, it is important to make some adjustments in calculating the level of debt that takes into account the double counting due to intra-governmental lending. The literature in this area debates over different methods of calculating the net debt of the Japanese government. The Figure 4 below presents different estimates for the debt to GDP ratio depending on the arguments in the literature. Figure 4: Gross and Net Debt % GDP (OECD Estimate) (Top Left) Gross and Net Debt (Broda and Weinstein (24) % GDP (Top Right) Gross and Net Debt (excl. SS Surpluses) % GDP (Bottom Left) Gross Debt excl. Foreign Exchange Reserves and BoJ Holding of Debt % GDP (Bottom Right) 5 This system of accounting is different from the consolidated or unified accounting system in which all tax revenue of different government accounts is pooled together and bonds are issued to cover any aggregate difference between expenditures and revenues. See Broda and Weinstein (24) for more details 9

11 The OECD s calculation of gross and net liabilities is presented in Figure 4 (top left) that includes the net liabilities of the central government, local governments and the social security funds. According to OECD calculation, the net debt in September, 217 stands at 129 % of GDP which is still highest among the OECD economies. Broda and Weinstein (24) calculates the debt burden by summing together the net debts of the Japanese government (including central government, local government and social security funds), postal savings, and government financial institutions (which are considered to be a part of the government). Due to the lack of data on postal savings, Figure 4 (top right) presents estimates of net debt using calculations of Broda and Weinstein (24). The net liabilities reduce marginally to 126% from 129% of GDP (OECD calculation) in September, 217 since government financial institutions are overall in surplus. Since 212, Bank of Japan has substantially increased its holding of government debt. The BoJ holding of government debt increased from about 2% of GDP in June, 212 to a little over 8% of GDP in September, 217 which has led to a dramatic decline in the debt holding by the private sector. If the debt holding of the BoJ is excluded from the net debt estimated using the net debt reduces to barely 45% which is very low 6 Another argument concerns the treatment of social security funds in calculating the debt burden.the social security funds in Japan are in net surplus and there is some literature that claims that social security funds are illiquid and thus should not be excluded from the calculation of net debt since they cannot be used to pay off debt immediately. As of September, 217, the social security funds had net assets of about 44% of GDP excluding which the net debt as % of GDP would increase from 129 % of GDP (estimated using OECD calculation) to about 172% of GDP which inflates the debt burden (See figure 4 6 Broda and Weinstein (24) also specify the cases when BoJ holding of debt should or shouldn t be excluded from the calculation of the actual debt burden: How to treat holdings by the BOJ depends on what ones forecast of monetary policy is. If one assumes that the money base will not change as a share of GDP, then debt held by the central bank never needs to be repaid and one does not need to count the bonds held by the central bank as a liability of the consolidated government. 1

12 (bottom left)). Excluding the debt holding of the BoJ brings this down to 91% of GDP as of September, 217 which is much lower than the gross debt to GDP ratio that seem to present a risky position of the Japanese debt. The last case is the pessimistic case (Figure 4 (bottom right)) that deducts foreign exchange reserves (that stands at 37% of GDP as of September, 217) from the gross liabilities of the General government. The gross debt as a % GDP excluding the foreign exchange reserves is 196% of GDP and once BoJ holding of debt and the loans made by government financial institutions (that stand at 16% of GDP) are also excluded since they are not a liability of the government, it is reduced to 99% of GDP. The purpose of presenting a detailed review of different scenarios of Japanese debt is to argue that it may not be as risky as it is believed to be. This is possibly the reason for high appetite of the private sector for the government debt (although it has declined in recent years due to a massive increase in BoJ holding of debt). The fact that the 1 year bond yield is as low as close to % (shown in Table 2 below) is an evidence to the fact that the debt situation in Japan is not considered to be risky. While it is true that the overall debt burden in Japan is still very high when compared with other countries but the concerns of default of Japan may be overrated. However, since the majority of the government is held by the private sector, the debt plays a major role in affecting the private sector expectations. As the empirical evidence will indicate in the later sections, fiscal policy actions aimed at stabilizing debt were welcomed by the private sector. Table 2: Fiscal and Monetary Policy Year 1-Year Govt. Call Rate Government primary Exchange rate Current account Bond Yield balance (% GDP) (Yen/USD) balance(%gdp)

13 3 Fiscal Policy and the Role of Household Expectations Given the above scenario of fiscal policy in the backdrop, the main objective of this paper is to explain the economic slowdown in Japan through the impact that fiscal policy had on public expectations. Since the role of monetary policy was limited especially before 212, the primary focus of the paper will be on the fiscal policy which appears to have been tried although with not much success over the last two decades. The private sector expectations regarding the preference for excessive holding of government debt and lack of economic response to much debated fiscal stimulus can help us understand the macroeconomic condition of Japan over the course of last two decades. Following Woodford (2) the equation (1) below provides the link between the fiscal policy and the fiscal expectations of the private sector. The equation (1) presents the debt valuation equation which relates the real value of net government liabilities to the present value of expected future primary budget surpluses that includes interest saved on the part of its liabilities that the public is willing to hold in monetary form (or seigniorage revenue)- a relationship that necessarily holds in the rational expectations equilibrium. W s t P t = T =t [ β T t U c (y T, m T ) E t s T + i ] T MT s U c (y t, m t ) 1 + i T P T ( ) where s T denotes the real primary government budget surplus and 1 R t = β Uc y t+1, Ms t+1 P t+1 is the stochastic discount factor. Thus, equation (1) can be rewritten as follows: ) U c (y t, Ms t P t (1) P t P t+1 W s t P t = T =t [ q t,t s T + i ] T MT s 1 + i T P T (2) where q t,t+1 = 1 R t s t = T t P t g t 12

14 Therefore, price level can be determined as follows: P t = T =t q t,t [ s T + W s t ] i T MT s 1+i T P T (3) where s T denotes the real primary government budget surplus M t is the end-of-period money balances, P t is the price level, g t is the (real) government purchases, c t is the (real) private consumption expenditure, T t is the (nominal) tax obligations net of any government transfers, W t is the financial wealth of household that includes money balances and bonds, < β < 1 is the discount factor, i t is the nominal interest rate it on a one-period riskless claim, i t 1+i t is the effective cost of holding wealth in monetary form. The equation (2) above is not a constraint on the government s fiscal policy but is an equilibrium condition that results from the private sector optimization problem. It provides link between fiscal policy and household expectations. If households believe that the prevailing regime is Regime F, then news of lower primary surplus stemming from either lower taxes or higher government purchases or transfers, when there is no expectation of higher taxes in future, will lead to decline in the market value of debt since the fiscal backing that supports debt declines. The decline in the households wealth as a result of lower market value of debt will raise households demand for goods causing the price level to increase. The monetary policy coordinates in Regime F by preventing interest payments on the debt from exploding and permitting surprise inflation to revalue government debt. 13

15 If however, households believe they are in Regime M, a fiscal shock that lowers primary surplus today is initially financed by higher real debt today since the price level is pinned down by the monetary policy. Thus, higher real value of debt today creates the expectation that future primary surplus would increase to ensure that the debt valuation equation holds which in turn delivers the Ricardian equivalence. Thus, the impact of fiscal policy on the economy depends a great deal on the household expectations about the prevailing regime. In case of Japan, however, it is difficult to distinguish between the two regimes. The government s efforts to reinvigorate the economy through fiscal stimulus coincides with an announcement of target of fiscal consolidation within a specified period of time. The government expects the households to believe that the prevailing regime is regime F which will eventually be reversed (through fiscal consolidation) within a certain time given the high debt burden. In order to guide the household expectations about fiscal policy, government announces the target periods for fiscal consolidation in the budget (See footnote 3 above). If households expectations are aligned with that of government, they should increase spending at least in the near future believing the current regime to be regime F. However, one of the reasons for the failure of fiscal policy in anchoring fiscal expectations is the continuous failure of the government in achieving fiscal consolidation in accordance with its earlier announcements thus further increasing the debt burden. Thus, the automatic response of the households to a fiscal stimulus is to in turn increase their debt holdings thus delivering the Ricardian outcome. The expectation is that at some point in the future, due to the failure of the government s fiscal consolidation effort, the debt burden would be so high that the debt will hit the ceiling of the private sector financial assets (Hoshi and Ito, 212) when the government will have no choice but to consolidate debt. Also, as Hoshi and Ito (212) claim that the expectations can change without warning. Failure in passing the bill to raise the consumption tax, for example, may change the public perception on realization of tax increases. When the crisis happens, the Japanese financial institutions that holds large 14

16 amount of government bonds sustain losses and the economy will suffer from fiscal austerity and financial instability. The next discusses the model and the results. 4 Econometric Methodology In order to understand how fiscal policy failed to stimulate the Japanese economy over the past 2 decades, we need to understand the impact of fiscal policy on the household expectations. For this purpose, this paper uses a Bayesian Structural Vector Autoregressive (B-SVAR) model to examine the impact of fiscal policy shocks on the economy through their impact on the household expectations. 4.1 Bayesian SVAR The VAR model in its structural form can be written as: p y ta = y t la l + e t, for1 t T, (4) l=1 where y t is an n 1 vector of endogenous variables, e t is an n 1 vector of structural shocks, A l is an n n matrix of structural parameters for l p with A is invertible, p is the lag length and T is the sample size. The vector e t is Gaussian with mean and covariance matrix I n. The equation (4) can be written in the compact form as follows: y ta = x ta l + e t, for1 t T, (5) where x t = [y t 1,...y t p] and A + = [A 1,...A p]. The model can be written in reduced form as: y t = x ta + u t, u t N(, Σ) (6) 15

17 The reduced-form and structural representations of the model are linked through Σ = (A A ) 1, A = A + A 1 (7) Thus, the relationship between the reduced-form residuals and the structural parameters is: u t = (A 1 ) e t (8) We ll estimate the reduced form VAR model using heirarchical modeling approach introduced by Giannone, Lenza and Primiceri (215) (henceforth GLP). One of the advantages of using this approach is that it allows us to use informative priors that shrinks the richly parameterized unrestricted VAR model towards a parsimonious naive benchmark that in turn reduces estimation uncertainty and improves out-of-sample forecasting and the accuracy in the estimation of impulse response functions. Following GLP (215), the optimal choice of the informativeness of these priors will be treated as additional parameters. Bayesian inference on the model in Eq. (6) amounts to updating prior beliefs about the reduced form VAR parameters (A, Σ) after having observed a sample y 1 p:t = {y 1 p,..., y,...y t 1, y t }. Rewriting the model (6) in vectorized form: Y = (I n X)α + U where u N(, Σ I T p ), y = [y p+1,..., y T ], Y = vec(y), x = [1, y t 1,...y t p], X = I N x, u = [u p+1,..., u T ], U = vec(u), A = [A 1,..., A p ], α = vec(a) The number of regressors for each equation is given by k = np+1. We choose the prior on the reduced form VAR parameters (A, Σ) to be Normal-Inverse-Wishart distribution given 16

18 as follows: Σ IW (Ψ; d) α Σ N(a, Σ Ω) (9) The conditional likelihood of the VAR model described above can be written as p(y 1:T α, Σ, y 1 p, ) = 1 (2π) 1 Σ T n/2 ( 2 exp 1 ) ( 2 tr(σ 1 Ŝ) exp 1 ) 2 (α ˆα) (Σ 1 (x x))(α ˆα) The joint posterior of the reduced form VAR parameters (A, Σ) can be obtained by multiplying the prior density by the likelihood function. The marginal posterior of the variance covariance matrix - Σ and the conditional posterior of the coefficients - A Σ can be given by: Σ Y IW (Ψ + û û + (Â â) Ω 1 (Â â, T p + d) α Σ, Y N(ˆα, Σ u (x x + Ω 1 ) 1 ) where Ψ, d, a, Ω are the function of a lower dimensional vector of hyperparameters - γ - the coefficients that parameterize the prior distribution. The choice of these hyperparameters can be made by evaluating their posterior (p(γ y)): p(γ y) p(y γ)p(γ) where p(γ) is the hyperprior and p(y γ) is called the marginal likelihood (ML) given by p(y γ) = p(y θ, γ)p(θ γ)dθ where θ = (A, Σ). Maximizing the posterior of the hyperparameters corresponds to maximizing the one-step-ahead out-of-sample forecasting ability of the model. Once we have maximized the posterior of the hyperparmeters, the hierarchical structure also implies that the unconditional prior for the parameters θ has a mixed distribution: 17

19 p(θ) = p(θ γ)p(γ)dγ That is conditional on the value of the posterior mode of γ, we can then draw the VAR coefficients α, Σ from their posterior which is Normal-Inverse-Wishart as given above. Choice of priors for the VAR coefficients in equation (9) can be as follows: We set the degrees of freedom of the Inverse-Wishart distribution to d = n + 2, which is the minimum value that guarantees the existence of the prior mean of Σ and Ψ is a diagonal matrix with an n 1 vector on its main diagonal which is equal to the diagonal of the reduced form model estimated in AR(1) form. Again, following GLP (215), the conditional Gaussian prior for α can be chosen as a combination of the following prior densities: The baseline prior is a version of the so-called Minnesota prior which is centered on the assumption that each variable follows a random walk process - a reasonable approximation of the behavior of an economic variable 1, if i = j and s = 1 (E)[(A s ) i,j Σ] =, otherwise λ 2 1 Σ ui,h s cov((a i,s ) i,j, (A r ) h,m Σ u ) =, 2 ψ j if m = j and r = s /(d n 1), otherwise The hyperparameter λ, controls the scale of all the variances and covariances, and effectively determines the overall tightness of this prior. Sum of coefficients prior (favors unit root) which can be implemented with a set of n 18

20 artificial observationsone for each variable. This prior states that a no-change forecast is a good forecast at the beginning of the sample. We construct the following set of dummy observations: y + n n = diag ( y µ ) x + n (1+n p) = [ n 1, y +,..., y + ] where y is an n 1 vector containing the average of the first p observations for each variable. The hyperparameter µ controls the variance of these prior beliefs: as µ the prior becomes uninformative, while µ implies the presence of a unit root in each equation and rules out cointegration. The third prior is the dummy-initial-observation prior which can be implemented using the following dummy observation y ++ 1 n = diag ( y x ++ n (1 (1+n p) = [ 1 δ, y++,..., y ++ ] δ ) This prior states that a no-change forecast for all variables is a good forecast at the beginning of the sample. The hyperparameter δcontrols the tightness of the prior implied by this artificial observation. As δ the prior becomes uninformative. if δ, all the variables of the VAR are forced to be at their unconditional mean, or the system is characterized by the presence of an unspecified number of unit roots. These priors λ, µ, δ are treated as additional parameters in the model. The choice of densities for these hyperpriors is as follows: Table 3: Priors Parameter Density Mode Standard Deviation λ Gamma.2.4 µ Gamma.8 1 δ Gamma

21 4.2 Recursive Identification In order to estimate the impact of fiscal policy, we need to identify fiscal policy shocks. We identify fiscal policy shocks using a recursive identification scheme and estimate a Bayesian Structural Vector Autoregression (B-SVAR) model, therefore, accounting for the posterior uncertainty of the impulse-response functions. The recursive assumptions can be summarized by y t = [g t, z t, p t, τ t, mb t, r1, er t, r t, d t ] and A is the lower triangular matrix such that A = where g t is the real government purchases, z t is the real GDP, p t is the price level, τ t is the real net tax revenues, mb t is the real monetary base, r1 is the 1 year government bond yield, er t is the real broad effective exchange rate, r t is the overnight uncollateralized call rate, d t is the real market value of debt held by private sector. All the variables are in log levels except for overnight call rate and 1 year government bond yield. The identification scheme corresponds to the assumption that fiscal variables are sluggish in nature relative to monetary policy variables, that is, the fiscal variables do not respond contemporaneously to monetary policy variables. The ranking of government spending at the top implies that government spending does not respond contemporaneously to the rest of the variables whereas it impacts all the variables. 2

22 5 Results 5.1 Government Purchases Shock The results can be seen in figures 5 through 9, where we have plotted the impulse responses of our 9 variables along with annualized inflation, consolidated debt of the government and the central bank and the sum of primary surplus and the seigniorage revenue (labeled as total primary surplus) to the fiscal and monetary policy shocks, identified according to the methodology above. Impulse responses are in the log levels of the variable shown. We start by examining the impact of fiscal policy on the household expectations and in turn on the Japanese economy. Following GLP (215), the Bayesian SVAR model has been estimated using the MCMC algorithm to simulate the posterior of coefficients of the BSVAR including the hyperparameters with Metropolis algorithm used to approximate the posterior of the hyperparameters. Figure 5 provides the impulse responses to 1 standard deviation government purchases shock (expansionary fiscal policy). The government spending on impact is positive and then declines for about 4 quarters after which it stays persistently at a higher level. The tax revenues increase by a small amount on impact, however, they continue to increase for about 2 years and then stays persistently at a higher level. The impulse response of the primary surplus shows the net impact of the expansionary government spending shock. The primary surplus is initially negative, however, it immediately (in only 2 quarters) becomes positive and stays high for about 16 quarters or about 4 years. With no significant response of either price level or the real interest rate, the decline in the real market value of debt stabilizes the debt valuation equation in response to lower primary surplus initially. The real value of debt becomes insignificant after only two quarters. Thus debt valuation is stabilized by higher primary surpluses in the future. This sudden reversal of the expansionary fiscal policy suggests the inconsistency in the fiscal policy discussed above. However, it could also be a natural fiscal 21

23 policy response to the lack of coordination from the monetary policy. Had monetary policy responded by reducing the policy rate in response to fiscal expansion, the fiscal policy could have stayed expansionary for a longer period of time eventually succeeding in stimulating the economy. Bernanke (217) explained the lack of a persistent fiscal stimulus in Japan through the lack of coordination from the monetary policy in stabilizing debt. Thus, fearing higher debt burden government immediately reverses the initial stimulus. The GDP, however, increases in response to higher government spending. This is in line with the finding of Miyamoto, Nguyen and Sergeyev (217) who found a positive output multiplier in case of Japan in response to higher government spending. Thus, it appears that the government spending fails to stimulate the economy. The immediate reversal of higher government spending failed to anchor private sector expectations. 5.2 Tax Revenue Shock Figure 6 presents the responses to 1 standard deviation shock to the tax revenues. Higher tax revenues implies a contractionary fiscal policy. A positive shock to the tax revenues is followed by an even higher tax revenues for about 4 quarters after the shock. It then declines slightly but stays persistently at a higher level. The government spending doesn t respond on impact (because of the identification assumption). It then declines for about 8 quarters and then becomes statistically insignificant. The net impact of this contractionary tax revenue shock is higher primary surplus and higher total primary surplus (sum of seigniorage revenue and primary surplus) over the entire time period in consideration. in response to higher primary surplus, the price level and in turn the annualized inflation rise to stabilize the debt valuation equation. One of the explanations behind higher inflation in response to a contractionary fiscal policy shock is that given the high debt burden, higher primary surplus is seen as the effort by the government to consolidate debt thus reducing the riskiness of the government debt held by the private sector. The decline in the riskiness of their debt holdings enables the private sector to respond with a higher spending. 22

24 The monetary policy responds by increasing the policy rate which seems appropriate given higher inflation. Thus, higher tax revenues (or a contractionary fiscal policy) evidently do a better job in anchoring inflation expectations and thus stimulating the economy. This is due to the higher existing debt burden in Japan due to which a fiscal consolidation effort by the government is viewed as an effort to stabilize the economy by reducing the riskiness of the debt. 5.3 Monetary Base Shock Figure 7 presents the impulse responses to 1 standard deviation shock to the monetary base. As discussed earlier, monetary base was almost constant until 21 after which it increased primarily due to the quantitative easing by the BoJ as a part of which the BoJ maintained reserve balances at a high level and simultaneously increased its holding of government bonds. Thus, the monetary base shock can also be viewed as a shock to the debt held by the private sector. The larger the debt held by the BoJ, less is the debt held by the private sector. One surprising result is that the higher debt holding by the BoJ did nothing to reduce the debt holding by the private sector since the impulse response of the real market value of debt is insignificant. The response of consolidated debt (sum of debt held by the private sector plus the monetary base) however matches to that of the monetary base. This shows that even after the increased debt holding by the BoJ, the overall debt level was increasing so much so that there was no decline in the debt held by the private sector. The response of government spending is insignificant for about 12 quarters after the shock to the monetary base however it gradually declines to support higher debt. The response of tax revenues is also insignificant for about 8 quarters after the shock however they also start rising to support the increase in debt. The net impact of fiscal policy can be seen from the higher primary surpluses as well higher total primary surplus (sum of primary surplus and 23

25 seigniorage revenues). The monetary policy appropriately responds by decreasing interest rate to allow higher inflation to reduce the higher real market value of debt as a result of monetary base shock. The price level and therefore the annualized inflation rises slightly for a few quarters, however, higher debt level affects household expectations by increasing the riskiness of the debt. The households respond by decreasing their expenditure and therefore price level becomes insignificant. Thus, an expansionary monetary base shock also fails to stimulate the economy in the presence of higher debt burden. 6 Conclusion The purpose of this paper is to examine the impact of fiscal policy on the Japanese economy. The expectations of household play an important role in determining the impact of fiscal policy in Japan. Fiscal consolidation efforts by the government were welcomed by the private sector as an effort by the government to reduce the riskiness of the debt and thus stabilize the economy. The empirical evidence indicates that while government tried to anchor the household expectations, it failed consistently because of its inability to achieve the objectives announced to the private sector. As discussed above, the government announced fiscal stimulus which over time was to be followed by fiscal consolidation, however, governments failure to achieve consolidation of debt was viewed negatively by the households and thus fiscal policy failed to pull the economy out of stagnation. 24

26 References Bernanke, B.S., 217. Some Reflections on Japanese Monetary Policy. Brookings Institution Ben Bernankes Blog. Broda, C. and Weinstein, D.E., 24. Happy news from the dismal science: reassessing the Japanese fiscal policy and sustainability (No. w1988). National Bureau of Economic Research. Giannone, D., Lenza, M. and Primiceri, G.E., 215. Prior selection for vector autoregressions. Review of Economics and Statistics, 97(2), pp Hoshi, T. and Ito, T., 212. Defying Gravity: How Long Will Japanese Government Bond Prices Remain High? (No. w18287). National Bureau of Economic Research. Ito, T. and Mishkin, F.S., 26. Two decades of Japanese monetary policy and the deflation problem. In Monetary Policy with Very Low Inflation in the Pacific Rim, NBER-EASE, Volume 15 (pp ). University of Chicago Press. Ito, T., 214. We are all QE-sians now. Institute for Monetary and Economic Studies, Bank of Japan, Discussion Paper, (214-E), p.5. Posen, A., Fiscal Policy Works When It Is Tried. Restoring Japans economic growth. Institute of International Economics: Washington DC, pp Miyamoto, W., Nguyen, T.L. and Sergeyev, D., 216. Government spending multipliers under the zero lower bound: Evidence from Japan. 25

27 Woodford, M., 21. Fiscal requirements for price stability (No. w872). National Bureau of Economic Research. 26

28 Data Table 4 Variables Definition Source Real government purchases Government Final Consumption Exp. + Government Gross fixed capital formation + Government consumption of fixed capital + Other current outlays + Property Income Paid by government - Gross Interest payments OECD Real net tax revenues Tax on production and imports + Total direct taxes + Property income received by government + Social Security contribution receipts + Other current receipts - Gross interest receipts - Social security benefits paid by government OECD Real GDP Nominal GDP deflated by GDP deflator OECD Real monetary base Banknotes and coins in circulation + Current account (Reserve) balances BoJ, (deflated by GDP deflator) Flow of Funds Real market value of debt Market value of gross liabilities of BoJ, held by the private sector central, local & social security funds Flow of Funds - debt held by government - debt held by BoJ Call rate Uncollateralized overnight call rate (Main monetary policy instrument) OECD 1 Year govt. bond yield OECD Real broad effective Weighted averages of bilateral FRED, Federal Reserve exchange rate exchange rates adjusted by Bank of St. Louis relative consumer prices Price Level Core Inflation Index OECD 27

29 Figures Figure 5: Impulse responses to Government Spending Shock Real Government Purchases 1-3 GDP 1-4 Price Level Real Net Tax Receipts Real Monetary base Exchange Rate Call Rate Annualized Inflation Seigniorage Consolidated Debt yr interest rate Real Market Value of Debt Real Interest Rate Primary Surplus Total Primary Surplus The figure reports the median (solid line) and the 16th and 84th percentiles (grey area) of the distribution of the impulse response functions of the large BVAR to a one standard deviation monetary policy shock. The x-axis is the number of quarters after the shock and the y-axis measures the size of the impulse response on levels. The impulse responses are a response to higher government spending shock (expansionary fiscal policy) 28

30 Figure 6: Impulse responses to Tax Revenue Shock 1-3 Real Government Purchases 1-3 GDP 1-4 Price Level Real Net Tax Receipts Exchange Rate Annualized Inflation Real Interest Rate Seigniorage 2.1 Consolidated Debt Real Monetary base 1 yr interest rate Call Rate Real Market Value of Debt Primary Surplus Total Primary Surplus The figure reports the median (solid line) and the 16th and 84th percentiles (grey area) of the distribution of the impulse response functions of the large BVAR to a one standard deviation monetary policy shock. The x-axis is the number of quarters after the shock and the y-axis measures the size of the impulse response on levels. The impulse responses are a response to higher tax revenue shock (contractionary fiscal policy) 29

31 Figure 7: Impulse responses to Monetary Base Shock 1-3 Real Government Purchases 1-3 GDP 1-4 Price Level Real Net Tax Receipts Real Monetary base 1 yr interest rate Exchange Rate Call Rate Real Market Value of Debt Annualized Inflation Real Interest Rate Seigniorage Consolidated Debt Primary Surplus Total Primary Surplus The figure reports the median (solid line) and the 16th and 84th percentiles (grey area) of the distribution of the impulse response functions of the large BVAR to a one standard deviation monetary policy shock. The x-axis is the number of quarters after the shock and the y-axis measures the size of the impulse response on levels. The impulse responses are a response to higher monetary base (expansionary monetary policy). 3