In response to the prolonged economic downturn, the Japanese government adopted an expansionary fiscal policy stance for most of the 1990s. With a brief interruption in 1997, when growth temporarily improved, the general government’s fiscal position (excluding social security) steadily deteriorated from balance in FY1990 to an estimated deficit of 8 percent of GDP in FY1998. In particular, the government tried to revive activity through a series of stimulus packages between 1992 and 1995, and again through substantive packages in 1998 and 1999 (Table 6.1). These packages have involved significant “headline” figures (often above 2 percent of GDP), capped by a ¥24 trillion package (5 percent of GDP) in November 1998.

In response to the prolonged economic downturn, the Japanese government adopted an expansionary fiscal policy stance for most of the 1990s. With a brief interruption in 1997, when growth temporarily improved, the general government’s fiscal position (excluding social security) steadily deteriorated from balance in FY1990 to an estimated deficit of 8 percent of GDP in FY1998. In particular, the government tried to revive activity through a series of stimulus packages between 1992 and 1995, and again through substantive packages in 1998 and 1999 (Table 6.1). These packages have involved significant “headline” figures (often above 2 percent of GDP), capped by a ¥24 trillion package (5 percent of GDP) in November 1998.

Table 6.1.

Japanese Stimulus Packages

(Trillions of yen)

article image
Source: Data provided by the authorities.

Partly based on IMF staff estimates.

However, doubts have been raised over the amount of real economic stimulus provided by the government (e.g., Posen, 1998). The demand impact of the stimulus packages has largely been limited to the so-called “real water” measures (items that lead directly to government expenditure and thus boost economic activity), such as public works spending and tax cuts. Significant portions of the stimulus packages’ headline amounts, however, have consisted of financial measures—such as loans provided to government financial institutions (GFIs) by the Fiscal Investment and Loan Program (FILP)—which have not directly led to higher aggregate demand as long as the economy was not severely credit-constrained. For example, the November 1998 package included ¥8 trillion in public investment and ¥6 trillion in tax cuts (later raised to ¥9 trillion) as real water components, while financial measures accounted for about ¥8 trillion (Box 6.1).1

The size and composition of public works measures have also attracted criticism. Public works measures constituted the bulk of most stimulus programs. However, budgetary allocations for public works projects were frequently below the amounts announced in the packages because these included a significant portion for land purchases (15 percent or more). In addition, a recent study by Ishii and Wada (1998) suggested that the actual increase of public works between 1992 and 1996 fell ¥10 trillion (2 percent of GDP) short of the amount contained in the stimulus packages, mainly as a consequence of poor implementation at the local government level. Questions have also been raised over whether diminishing returns to scale from public construction projects (owing to a weak project selection process, cost overruns, and a bias toward projects in rural areas) may have contributed to declining fiscal multipliers (Bayoumi, Towe, and Oishi, 1998).

A breakdown of the shift in the general government’s structural balance indeed shows a surprisingly small role played by public investment in recent years.2 While the general government’s structural balance fell from a 2½ percent of GDP surplus in 1990 to a deficit of 3½ percent of GDP in 1998, the increase in public investment only contributed 1 percentage point to that decline (Figure 6.1). A more significant factor was the decline in (structural) tax revenue of more than 4 percent of GDP between 1990 and 1998.

Figure 6.1.
Figure 6.1.

Components of General Government Structural Balance, 1980–98

Source: IMF staff calculations.

How can the large increase in the fiscal deficit be reconciled with questions about the effectiveness of fiscal stimulus measures, and what has been the role of public investment and tax cuts? Deciding on the true level of economic stimulus provided by the government is clearly important for future policy analysis. This chapter seeks to address these questions by focusing on the following issues:3

  • The limited rise in public investment. The first section of the chapter explores why the increase in public investment has been relatively small in recent years, compared to the prominent role of public works in the stimulus packages.

  • Why was there a sharp drop in tax revenue? The revenue decline went well beyond the scope of the government’s tax cuts in recent years. As laid out in the second section, this has reflected a sharp unexplained fall in tax elasticity in the early 1990s.

Stimulus Programs, the Budget Process, and the Implementation of Public Works

Fiscal operations in Japan are conducted through complex institutional arrangements, the focal point of which is the central government’s General Account, which receives almost all national tax and bond revenues. There are also 38 Special Accounts, largely under the control of individual ministries, which are partly cofinanced by borrowing, e.g., from the Fiscal Investment and Loan Program (FILP). Local governments comprise 47 prefectures and 3,200 municipalities, whose fiscal operations (especially taxation and borrowing) are to a substantial extent controlled by the central government. The FILP provides loans for government financial institutions (GFIs) and other public agencies. Although not formally part of general government, it operates under the authority of the Ministry of Finance, and its investment plan is often referred to as the “second budget.”

Stimulus programs typically affect all layers of the public sector. Real water measures (tax cuts and public works) are implemented through supplementary budgets and, in some cases, initial budgets for the following fiscal year. Past stimulus programs have also contained substantial measures to be carried out by GFIs and other public sector agencies. For example, the November 1998 package took several steps to counteract credit shortages, including through capital injections and funds for GFIs worth ¥5.9 trillion (intended to promote additional loans and loan guarantees of up to ¥27 trillion), an extension of the ceiling for loans by the Housing Loan Corporation (at a cost of ¥1.2 trillion), and additional export credits for Asia (¥1 trillion). Although these measures have been partly financed by the central government budget, they have been implemented to a substantial extent through the FILP, drawing on deposits from public savings and pension schemes in the Trust Fund Bureau, and requiring no additional government borrowing.

The stimulus packages have relied on the substantial participation of local governments. Although stimulus packages are decided at the central government level, local authorities share in the revenue losses for most tax categories, and carry out the bulk of public works projects. The central government implements some public works on its own, but most of the investment projects have consisted of either jointly financed projects (with roughly equal shares) or projects carried out under the sole responsibilities of local governments. Local authorities have received additional financial support from the central government and the FILP but, recognizing the difficult financial situation of many local governments, the latest stimulus package in November 1998 no longer contained any provisions for fully self-financed local government public works.

The timing of the public works measures is difficult to predict. A typical stimulus program requires some 3-6 months between the date of announcement and commencement of the related construction projects. As for the central government, the coordination process required to pass a supplementary budget takes about a month, followed by one more month during which orders and contracts are completed. Another month typically passes before construction starts. The process for local governments depends on the passage of supplementary budgets in local councils that may be delayed, depending on the prevailing local political situation. In the case of large construction projects (exceeding ¥500 billion), progress is delayed further by adherence to WTO government procurement regulations (which require an interval of 40 days for the bidding process).

Official statistics do not provide a comprehensive picture on the progress of implementation. Survey-based construction indicators (on orders and public works starts) published by the Ministry of Construction are available 4-6 weeks after the end of a calendar month. These data cover up to 50 percent of GDP-based public investment and precede preliminary national income data with a short lead of at most one quarter. However, information provided by the indicators does not correspond to the fiscal accounts, comprehensive information for which is only available after 12 months from the end of the fiscal year, owing largely to a delay in consolidating the accounts of local authorities.

At the time of writing, the following data were available to evaluate the impact of stimulus packages: (i) Monthly information for central government accounts through May 1999; (ii) preliminary national income accounts for the first quarter of 1999; (iii) full national income accounts (including general government) for FY1997; and (iv) central and local government settlement data for FY1997.

This chapter also addresses the question why Japan has had to rely to a relatively large extent on discretionary policy measures. The generally low cyclical variability of unemployment and social welfare benefits in Japan imply a low level of automatic stabilizers in Japan compared to other major industrial countries. Without active policy measures, such limited automatic stabilizers would have constrained the fiscal support to the economy. The third section thus analyzes the extent to which stimulus measures have been used to offset the relative lack of automatic stabilizers in Japan.

The limited transparency of the Japanese fiscal system hampers much of the following analysis. Japanese fiscal accounts offer only a partial picture of government finance operations, mainly because of insufficient account consolidation across various layers of government (see Box 6.1). In particular, detailed consolidated accounts for the general government are only provided in the context of the National Income Accounts. These data become available electronically with a lag of 9 months from the end of the fiscal year, owing to difficulties in compiling the accounts of some 3,300 local authorities that carry out the bulk of public expenditure, and many details are further delayed. As a result, most of this paper focuses on stimulus measures that took place before 1998.

Stimulus Packages and Public Investment

All of the stimulus packages in the last ten years have included significant additional public works spending. Public works (typically consisting of infrastructure-related construction projects) have traditionally played an important role in Japan, partly reflecting the infrastructure needs of an expanding economy and partly for political reasons (Schlesinger, 1997). They have been seen as an instrument to quickly boost aggregate demand, given the swift impact of a rise in public orders on employment and incomes in the construction sector.4

Consequently, overall public investment has seen a sizable increase in recent years.5 The share of public investment in GDP rose from 6½ percent in fiscal year 1990 to 8¾ percent in 1995 (see Figure 6.2), with particularly strong increases in some of the years with large stimulus packages. For example, in fiscal years 1992 and 1993, the government implemented four successive packages with a combined worth of ¥20 trillion in public works, leading to record investment increases in both years. Similarly, a ¥6 trillion package in 1995 provided a strong boost to investment in that year. Investment fell in 1996 and 1997 as stimulus measures were phased out, but rose again strongly in the wake of the 1998 packages.

Figure 6.2.
Figure 6.2.

Japanese Public Investment

(Percent of GDP)

Nevertheless, investment seems to have risen by much less than might have been expected from the stimulus measures. As shown in Figure 6.3, the nominal increase in public investment (on a national accounts basis) has indeed been much smaller than the public works components of the stimulus packages.6 The increase in public investment in 1993, for example, was only around ¥4 trillion (panel A), or less than a third of the packages’ public works measures in that year (panel C). Similarly, the investment increase in 1995 was about ¥3 trillion, compared to public works measures worth ¥6 trillion.

Figure 6.3.
Figure 6.3.

Changes in Public Investment1

(Trillions of yen)

Sources: National income accounts (A); central government general account budget (A, B); White Book on Local Government Finance (A, B); and data provided by the Ministry of Finance (C).Note: NIA = National Income Accounts.1 Fiscal year basis (April-March).2 Contains double-counted expenditure items.

What has caused these disappointing results? There appear to be three main factors behind the comparatively small investment increase, namely the contractionary nature of initial budgets throughout the 1990s, financial difficulties at the local government level, and shortfalls in the implementation of the stimulus measures.

Contractionary Initial Budgets

Stimulus packages have served in part to offset expenditure cuts in initial budget plans. Reflecting the government’s determination to maintain fiscal discipline, initial budgets were typically contractionary throughout the 1990s, containing substantial expenditure cuts compared to actual spending levels of the previous year. The restrictive stance is reflected by the fact that initial budgets are planned (and reported) with reference to the previous year’s initial budget, rather than the higher level of spending implied by supplementary budgets. Although a useful device for maintaining budget discipline, this mechanism obscures the level of contraction frequently implied by initial budgets. In many years, the fiscal stance switched to contraction early in the fiscal year before expansionary stimulus packages were enacted in response to worsening economic conditions.

This process is illustrated for the sum of central and local government public works expenditure in Panel B in Figure 6.3.7 The dark bars—the difference between initial budget plans and actual expenditure in the previous year—reflect the contractionary nature of initial budgets, which provided for expenditure cuts of around ¥4 trillion (0.8 percent of GDP) per year between 1993 and 1996. Actual spending, however, increased in most of these years as supplementary budgets more than offset cuts planned in the initial budgets. This is reflected in the chart, where the difference between initial budgets and final outcomes (indicated by white bars) roughly corresponds to the public works components of the stimulus packages.

Data for the central government alone allow a closer analysis of changes in actual public works spending (Figure 6.4). The availability of the central government’s revised budget data facilitates a more detailed breakdown of changes in public works expenditure (the data do not exist for local governments as a whole). It holds that:

(PWt - PWt-1) = (IBt - PWt-1) + (RBt - IBt) + (PWt - RBt)

where RB is the allocation for public works in the revised budget. The second term on the right hand side shows the change in the public works budget due to supplementary budgets, while the third term identifies expenditure shortfalls or overruns. The chart shows that central government projects are more or less fully implemented, and that the implementation occurs relatively fast. Central government funds for public works projects are generally spent in the budget year and small expenditure shortfalls (such as in 1993 and 1995) have been made up by additional expenditures in the following year.8

Figure 6.4.
Figure 6.4.

Public Works Spending of Central and Local Government1

(Trillions of yen)

Sources: Central government general account budget; and Local Government Finance White Book.1 Fiscal year (April-March). Central government data for 1998 based on information provided in the budget documents.

Financial Difficulties of Local Governments

A second important factor explaining the muted impact of stimulus measures on public investment relates to the role of local governments. During the 1990s, the central government has assigned roughly two-thirds of the increased public works spending to local authorities (without providing a commensurate increase in funding). As a result, local government investment rose by more than ¥2 trillion per year in the early 1990s, with most of the increase accounted for by midyear supplementary budgets implementing the central government’s stimulus packages (see Figure 6.4). The large stimulus programs of 1995, however, did not lead to a corresponding increase in local public works, and these have also dropped sharply in 1996 and 1997, consistent with the observations by Ishii and Wada (1998).

The capacity of some local authorities to expand public investment has been affected by their increasingly precarious financial situation. As tax revenues have been stagnant due to weak activity, particularly in larger urban prefectures that rely heavily on corporate tax payments, the continued rise in public investment has increasingly been financed through local bond issues. The amount of outstanding local government bonds shot up from 12 percent of GDP in 1990 to 22 percent of GDP by the end of fiscal year 1997, and some local authorities have been on the verge of being put under direct administrative rule by the central government.9

As a result, more recent stimulus packages have served to offset an underlying decline in local government investment. For example, indications are that independent public works (which account for 60 percent of all local public works) were reduced on a large scale in 1999, more than offsetting increases in projects co-financed by the central government. FY1999 budget data for 25 prefectures showed a decline in locally funded public works by 16 percent compared to FY1998 initial budgets. Although public works funded by the central government increased by 5 percent, overall local public works were still projected to drop by 11 percent (Okue, 1999). Similarly, a Nikkei survey of 670 municipal governments found that cuts in self-financed investment projects were used to offset rising burdens in other areas.

Implementation of Stimulus Measures

The extent to which stimulus packages have directly added to overall investment demand is difficult to estimate. Besides possible offsets through cuts in central or local expenditure elsewhere, some of the funds provided by the stimulus packages may have been allocated for use by public enterprises or have remained unused. However, the limited availability of relevant data makes it difficult to assess exactly how much of the government’s stimulus plans have effectively translated into additional spending.

One approach is to relate spending in excess of initial budget plans to the size of the stimulus measures (Ishii and Wada, 1998). For example, using fiscal data available for the central and local governments, a crude implementation ratio can be calculated by comparing excess public works spending (over initial budgets) to the amount of public works in that year’s stimulus packages. To account for some possible delay in the execution of projects, any budgetary over-shooting of expenditure over and above revised budget estimates in the following year is added to excess spending of the current year. An implementation ratio below unity reflects either unspent supplementary budget funds or offsetting cuts in investment expenditure elsewhere.10

The calculation is complicated by two data-related issues. First, parts of the stimulus programs have been carried out by public enterprises, for which the concept of initial and supplementary budgets does not apply. However, changes in investment by public enterprises have largely mirrored central government investment, and have not been of a dimension that would fundamentally change the results of this exercise. Second, central and local government expenditure data contain double-counted items (essentially equal to the central government’s contribution to jointly financed investment projects). Based on a comparison with national accounts data, the overlap is estimated to be around 20 percent, but the calculations also allow for a smaller or larger percentage.11

Based on this rough estimate, project implementation appears to have improved through the 1990s.Table 6.2 contains a range of estimates for the implementation ratio, depending on the assumed amount of double-counting between central and local government public works spending. Using the central case of a 20 percent expenditure overlap (Case II), as much as 90 percent of the stimulus packages’ public works measures translated into additional demand in 1995 (the latest year for which data was available), compared to only around 60-70 percent in earlier years. While this most likely reflects the more urgent need for public stimulus given the depth of recession in the early 1990s, it is possible that the ratio again declined in 1998 as financial constraints on local governments began to bite.

Table 6.2.

Implementation of Public Work Projects

(Billions of yen unless otherwise noted)

article image
Sources: Data provided by the Japanese authorities; and IMF staff estimates.

Staff estimate for actual public works expenditure. Note that the November 1998 package included public works worth Y1.4 trillion for inclusion in the initial budget for 1999.

Including excess spending (over revised budget estimates) in following fiscal year.

The Decline in Tax Revenue

The rising budget deficit in the 1990s has reflected a strong drop in tax revenue, particularly in the immediate post-bubble years. General government tax revenue dropped from a peak of 22 percent of GDP in FY1990 to 18 percent of GDP in FY1994, a level from which it has not yet recovered (Figure 6.5). The revenue fall was mainly seen in taxes collected by the central government (which are partly shared with local governments), while own-tax revenues of local authorities remained relatively steady at 7 percent of GDP. In the following paragraphs, the focus is thus on the main components of central government revenue.

Figure 6.5.
Figure 6.5.

General Government Tax Revenue Developments

Sources: Data provided by the Japanese authorities; and IMF staff calculations.1Data for 1997 (local government only), 1998, and 1999 are based on projections by the Japanese authorities.2Tax measures only relate to taxes collected by the central government.

Tax revenue losses were partly caused by a series of tax cuts since 1994, although the government’s shift to fiscal consolidation in 1997 offset much of the earlier measures. Following the collapse of the bubble, the government’s initial strategy was to support the economy through public works projects, and tax policy instruments were not used in a substantial way before early 1994. Since then, steps taken by the government included the following:

  • A temporary tax cut of ¥6 trillion (1¼ percent of GDP) was announced in FY1994, consisting mainly of a 20 percent reduction in personal income tax payments across the board (affecting both the central and local government level). In the following year, a more substantial tax reform package included an upward shift in income tax brackets that resulted in permanent tax relief of ¥3½ trillion, which was to be offset from FY1997 by an increase in the consumption tax rate from 3 to 5 percent. In addition, the government granted a temporary income tax rebate of ¥2 trillion, which was also maintained in FY1996.

  • In FY1997, the combination of the consumption tax rate hike with the expiration of temporary tax cuts resulted in an effective increase in the tax burden of about ¥6 trillion—roughly equal to the tax reductions implemented three years earlier.

  • Personal income tax reductions resumed, however, as the economy entered into recession in late 1997. Personal income tax rebates granted in FY1998 amounted to around ¥4 trillion, of which ¥1 trillion was retroactive to taxes paid in FY1997. These temporary measures were replaced by permanent tax cuts in FY1999, effected through the lowering of the top marginal tax rate from 65 percent to 50 percent, and proportional reductions in other income tax brackets.

  • The government also took steps to reduce corporate taxation. In particular, two rate cuts in FY1998 and FY1999—partly offset by base-broadening measures—brought the corporate tax rate down to 40 percent from 50 percent in 1990 (to which it had already been reduced from 55 percent during the late 1980s).

Indications are, however, that tax policy measures have played only a secondary role in the revenue drop. The cumulative revenue impact of tax measures remained limited. Between 1994 and 1998, permanent tax measures accounted for a cumulative reduction in revenues of ¾ percent of GDP, while temporary measures led to losses of about ¼ percent of GDP each year (see Figure 6.5). Overall, tax cuts accounted for only a 1 percent of GDP decline in revenue by 1998, compared to a much larger overall revenue decline of 4½ percent of GDP. Moreover, the revenue losses were largely concentrated in the period 1990-94, that is before the government embarked on its policy of tax cuts. This suggests that changes in tax revenue elasticity have played a major role in recent revenue developments.

Tax Elasticity

Tax elasticity measures how tax revenue would respond to changes in economic growth under an unchanged tax system. Most industrial economies have elastic tax systems (defined as elasticity being greater than one), where tax revenue grows at a higher rate than GDP even in the absence of tax policy measures. This reflects the influence of a variety of factors, including increases in the number of taxpayers, bracket creep, and improvements in tax administration.

To define elasticity, first consider the concept of tax buoyancy, to which the elasticity concept is closely related. Buoyancy compares the change in overall tax revenue to the change in the tax base (nominal GDP for the most part of this study):


However, part of the change in revenue may be due to tax policy measures taken by the government. For example, if it is estimated by the Tax Bureau that a tax rate increase yields an additional revenue of M in the current year, adjusted tax revenue would equal Ta = T - M, and tax elasticity would be defined as:


Trends in Tax Elasticity

The decline in tax revenue has indeed been associated with a fall in tax elasticity during years with a sharp slowdown in growth: 12

  • During the late 1970s and early 1980s, tax elasticity was relatively constant around an average of 1¼, before rising to 1¾ between 1986 and 1990, the peak years of the asset price bubble (Table 6.3).

  • As the economy entered into recession in the aftermath of the bubble years, tax elasticity also fell and became negative in 1991–94. Severe income and wealth losses appear to have had a depressing effect on tax collections (see below).

  • In the mid-1990s, tax elasticity levels briefly recovered before dropping to an unprecedented low in the recession year of 1997 (although measurement problems associated with the revenue impact of the consumption tax hike are likely to exaggerate the dimension of the drop in that particular year).13 However, elasticity appears to have again returned to more normal levels in 1998 and—if 1997 is excluded—has been close to its historic average in recent years.

Table 6.3.

GDP Growth, Tax Revenues, and Tax Elasticities

(Average in percent)

article image
Source: IMF staff calculations.

An analysis of individual central government tax categories provides further details on elasticity developments. Despite obvious advantages of using GDP as a broad measure for the overall tax base, individual revenue components are more closely related to narrower macroeconomic aggregates (such as corporate tax revenue to corporate profits, etc.). Therefore, while tax elasticity of a revenue component may remain constant relative to a narrow base, a decline in that base relative to GDP could result in a drop in elasticity vis-à-vis GDP.14 Alternative tax bases have therefore been used to analyze individual revenue components.

In the case of corporate taxes, elasticity has been partly driven by the rise and fall of profits in connection with the asset price bubble.15 As shown in Figure 6.6, corporate tax elasticity vis-à-vis profits was practically constant in recent years. The swings in elasticity relative to GDP are largely explained by profit developments. Having soared in the late 1980s, corporate profits fell by almost half between 1989 and 1993. As a result, corporate tax revenue—one quarter of total tax revenue—fell sharply in the early 1990s, but recovered between 1994 and 1996 when profits rebounded. However, corporate tax elasticity vis-a-vis profits has been below unity in recent years, contributing to a decline in the effective corporate tax rate from 50 percent in 1990 to 42 percent in 1997. It is still unclear to what extent this inelasticity reflects temporary factors (such as loss carry-forwards) that had yet to be unwound.

Figure 6.6.
Figure 6.6.

Elasticity of Central Government Tax Revenue Components Using Alternative Tax Bases1

Sources: Japanese authorities (including fiscal estimates for 1998); and IMF staff calculations.1Except for total revenue, all charts show elasticity computed using GDP and an alternative tax base.

Reasons for the elasticity decline in other tax categories have been less evident. Indications are, however, that tax revenues have been affected particularly by a sharp slowdown in growth:

  • Personal income tax elasticity dropped sharply in 1992 and 1997. Personal income remained broadly unchanged in relation to GDP (at around 70 percent), and tax elasticity relative to personal income therefore shows similar fluctuations to the elasticity relative to GDP (see Figure 6.6). Possible reasons for the weakening of the revenue collections in a recession are income cuts resulting in significantly lower tax claims due to a highly progressive tax system with high tax thresholds, deductions of real estate and financial asset losses, and increased noncompliance and slackening collection efforts.16

  • Similarly, indirect tax elasticity has not been significantly affected by the choice of a particular tax base. Indirect tax elasticity has generally been more volatile, however, than the elasticity of other tax categories, and the impact of the 1992 recession has thus been of a lesser order than on the income tax side. As for the exceptional elasticity drop in 1997, this may well reflect difficulties in estimating the revenue impact of the consumption tax hike, given the sensitivity of the elasticity estimate to that particular variable.

The relative impact of tax cuts and elasticity changes on tax revenue is highlighted by means of two tax revenue simulations:

  • Unchanged tax policies: On the assumption that there were no tax measures by the government since 1990, tax revenue has been simulated on the basis of actual GDP growth and actual tax elasticities.

  • Unchanged tax policies and elasticities: Assuming that tax elasticity would have remained at its average level for the 1970s and 1980s (around 1¼), tax revenue has been simulated on the basis of actual GDP growth.

The results illustrate that the effects of tax cuts were small, and that the tax increase in 1997 offset a large part of earlier reductions (Figure 6.7). By 1996, the cumulative impact of tax cuts on general government revenue accounted for 1½ percent of GDP, compared to an elasticity-related loss of around 3 percentage points (relative to what revenue would have been under pre-bubble elasticities). However, earlier tax cuts were substantially offset by fiscal tightening in 1997, including through the consumption tax increase that alone yielded revenue gains of about ½ percent of GDP. The 1998 tax measures led to a partial reversal of that stance with the result that, between 1990 and 1998, tax measures accounted for slightly more than 1 percentage point in an overall revenue loss of 5 percent of GDP.

Figure 6.7.
Figure 6.7.

Relative Impact of Tax Measures

(Percent of GDP)

Sources: Japanese authorities; and IMF staff calculations.

The Role of Discretionary Policies

Macroeconomic analysis of fiscal policy is frequently based on a decomposition of the fiscal deficit into a structural and a cyclical component. Movement in the structural balance is generally associated with discretionary fiscal policy measures, while the impact of automatic stabilizers (movements in the deficit reflecting cyclical fluctuations in economic activity) is given less prominence. As shown in Figure 6.8, movements in both the actual and structural fiscal balance and the output gap have been closely correlated in Japan in recent years. Compared to other major industrial countries, however, changes in the cyclical component appear to have been relatively small, particularly if compared to countries such as Canada, Germany, and the United Kingdom.

Figure 6.8.
Figure 6.8.

G7 Countries Fiscal Balance and Output Gap

Sources: IMF, World Economic Outlook database; Japanese authorities; and IMF staff estimates.

In Japan, the large deterioration in the structural deficit over the 1990s in part substituted for the relative lack of automatic stabilizers. Given the smaller impact of the economic cycle on the deficit, the need to use discretionary measures to adjust the fiscal stance puts greater weight on active fiscal policy management and results in a less smooth adjustment process:

  • Compared to automatic stabilizers, discretionary policies suffer from implementation lags caused by procedural delays, and are more affected by forecast errors.

  • In Japan, the perceived need to quickly return to fiscal balance in the face of increasing public debt levels may have contributed to an asymmetric reaction to economic shocks. Compared to automatic stabilizers, discretionary policies may have tightened early in the recovery, but loosened late in a downturn. The experience of 1997, when a sharp fiscal contraction contributed to the economic downturn, may have been a case in point.

An econometric approach has been used to quantify differences in automatic stabilizers across G-7 countries.17 The regressions relate the change in the structural and cyclical balance in each country to changes in the output gap:

ΔBalance = β0ΔOutput gap + β1ΔOutput gap-1 + ε.

The lagged output gap is included to reflect lags in the collection of taxes. The sample period was 1970–1997, with data taken from the IMF’s databank. There is an element of circularity in this exercise, as the IMF’s estimate of the structural deficit is derived from the estimated output gap. To help check the robustness of the results, the regressions were repeated using OECD estimates for the output gap (drawn from the Organization for Economic Cooperation and Development’s analytical database) and IMF estimates of the structural balance.

The results of the estimation can be summarized as follows (Table 6.4):

  • Japan has the lowest automatic stabilizers among the major industrial countries. An increase in the output gap of one percentage point translates into an increase of the cyclical deficit of about a third of 1 percent of GDP (even less when social security is excluded). This is only about half of the deficit response in countries with strong automatic stabilizers, for example, the U.K.18 Other countries with relatively small automatic stabilizers include the U.S. (with similarly low social benefit levels) and Italy, where high deficit and debt levels may have prompted the suppression of automatic stabilizers (Buti, Franco, and Orgena 1997).

  • By contrast, Japan’s structural balance has adjusted strongly to fluctuations in the business cycle in the past two decades. Historically, a change of 1 percentage point in the output gap has translated into an increase in the structural deficit of around half a percent of GDP, compared to a fifth of 1 percent or less in most other G-7 economies.

  • Overall, Japan’s fiscal stance has been strongly countercyclical. By adding up the coefficients of both equations, the estimates measure the overall responsiveness of fiscal policy to changes in the economic cycle. With the exception of the U.K., Japan’s fiscal position has reacted more strongly (by a margin of 20–30 basis points) to changes in the output gap than that of other economies.

  • The use of OECD output gap estimates produces some reduction in estimated automatic stabilizers but, in most cases, larger responses to the output gap. However, the differences are not large, and hence endogeneity does not appear to significantly bias the overall result.

Table 6.4.

Comparison of Automatic Stabilizers Across G-7 Countries

article image
Source: IMF staff calculations, based on World Economic Outlook and OECD Analytical Database.

Based on OLS regressions of the change in the cyclical deficit (the difference between the actual and the structural deficit ratio) on the change in the output gap (using observations from 1970 to 1997). A star indicates significance at the 5 percent level. First row results are derived from WEO data, second row results are based on OECD data.

Based on OLS regressions of the change in the structural deficit on the change in the output gap (using observations from 1970 to 1997). Star indicates significance at the 10 percent level.

Coefficients are based on a restricted sample from 1983–95.

Comparison with Other Major Economies

The econometric results have been used to simulate a deficit path for Japan by applying other countries’ reaction functions to Japanese output data. In the upper panel of Figure 6.9, average coefficients from other G-7 countries have been used to simulate both cyclical and structural deficit components for Japan, yielding a simulated overall fiscal balance. Since the model has been estimated in differences, the simulations are based on a dynamic method—with each year’s outcome depending on the previous year’s outcome—starting in 1981, when the actual and structural deficits were roughly equal. The simulation equation is:


with the β’s representing the average coefficients estimated for other G-7 economies, and residuals from the original Japan equation reflecting past exogenous shocks.

Figure 6.9.
Figure 6.9.

Simulation of Structural Balance: Comparison with Major G-7 Economies

Source: IMF staff calculations.1Using average of regression coefficients for G-7 economies (excluding Japan).

The simulations illustrate the difference between Japan and other industrial economies:

  • Japan’s fiscal system achieved a larger fiscal surplus during the bubble years than would have been true of other large economies in comparable circumstances (at the peak, the difference amounted to 1½ percent of GDP), and the deficit also reacted more sharply to the economic downturn through 1996. This has been achieved by large changes in the structural balance, which declined by 6½ percent of GDP between 1990 and 1996, compared to about 2½ percent of GDP using G-7 coefficients.

  • The impact of the automatic stabilizers is illustrated in the lower panel of Figure 6.9, where structural balances were calculated by subtracting simulated automatic stabilizers from Japan’s actual fiscal balance. By that measure, the turnaround in the structural balance between 1990 and 1996 implied by average G-7 coefficients would have been around 5½ percent of GDP, or 1 percentage point less than what actually occurred.

  • The adjustment implied by coefficients for the U.K.—which has the largest automatic stabilizers among the major economies—is still smaller at 4½ percentage points. Such a difference reflects a significant portion of the impact of the stimulus packages, which raised the deficit by about 2½ percent of GDP over the same period (Table 6.5). The path using U.K. data also illustrates that Japan’s large fiscal easing in 1993 might have come earlier and could have been less sharp if automatic stabilizers had played a larger role. Similarly, the deficit reduction in 1997 and subsequent expansion would have been achieved by automatic stabilizers alone.

Table 6.5.

Change in General Japan’s Government Structural Balance, 1990–98

(Percent of potential GDP)

article image
Source: IMF staff calculations.


This chapter has identified the major components contributing to fiscal expansion in Japan in recent years. Contrary to the notion that fiscal policy has mainly been driven by the government’s various stimulus packages over the past years, a large drop in tax revenue elasticity in the early 1990s has been identified as the major force behind the widening of the structural deficit (see Table 6.5). Public works spending and tax cuts played an important role in subsequent years, yet their contribution to fiscal expansion was comparatively small.

The large stimulus packages of the past decade have partly been necessitated by the small size of automatic stabilizers in the Japanese economy. However, discretionary policies tend to adjust less efficiently to cyclical swings in the economy than automatic stabilizers, and the relatively larger need for discretionary measures has contributed to two adverse developments in Japan’s budgetary process:

  • Given a potentially large room for discretionary measures, the budgetary system has evolved toward a structure that seeks to protect fiscal policy from an expansionary bias, resulting in a cycle of contractionary initial budgets, followed by expansionary supplementary budgets in midyear. This structure has contributed to uncertainty and limited the effectiveness of fiscal policy measures.

  • The same motive has also led to an overreliance on public works spending (which generally have low rates of return), as well as temporary tax cuts whose impact on demand is likely to have been limited.19

As economic activity recovers, a major concern for fiscal policy relates to the timing of renewed fiscal consolidation. Indeed, with Japan’s fiscal position adjusting less to changes in the business cycle than in other countries, comparatively large improvements in the structural balance will be needed to achieve fiscal adjustment. The results of this chapter suggest, however, that discretionary polices would need to continue mirroring the gradual workings of automatic stabilizers during the anticipated upswing, as they did during the recession years of the past.

Appendix. Calculating Tax Elasticities

Choosing a base

The choice of a macroeconomic tax base constitutes the first step in calculating tax elasticities. The actual tax base (total taxable incomes) is not easily available in historical form and is also not commonly projected in economic models. However, the actual base is often closely related to a macroeconomic aggregate, such as GDP, personal income, or corporate profits. For this chapter, the aggregate that produced the smallest variation in tax elasticity was chosen as tax base (for example, GDP for total tax revenue, personal disposable income (before tax) for personal income tax, etc.). In the following, “tax base” refers to this macroeconomic aggregate.

Basic definitions

Define tax revenue in year 0 by T0 = t0 B0, with t being the effective tax collection rate and B the tax base. The budget revenue estimate for year 1 is given by


where * denotes budget estimates. The first term of the right hand side is tax revenue at unchanged rates (given a projected rate of base growth). The second term is the expected revenue impact of tax policy measures, defined as M=Δt1*B1*, i.e., the change in the collection rate resulting from policy measures, times the new base. Estimates for M are contained in Japanese budget documents.

This information can be used to compute tax elasticities and construct a time series for econometric analysis. Define tax revenue elasticity as


where T1-M is the part of tax revenue that is unaffected by tax changes. Unfortunately, however, this formula gives only an approximation for the true elasticity because M is just an estimate for the budgetary impact of tax measures. The actual impact cannot be calculated for lack of observable data. This is illustrated by a breakdown of the deviation between actual and budgeted tax revenue:


The right-hand side decomposes the forecast error into an unanticipated change in the tax base and a deviation of the effective tax collection rate from its budget projection. The discrepancy in the tax collection rate reflects both under- or overestimation of the revenue impact of tax policy measures on the collection rate and unforeseen changes in tax elasticity. Without further information, the impact of both effects cannot be disentangled.

If the base projection in the budget is close to the final outcome (i.e., B1* is approximately equal to B1), M is likely to be a good approximation and the elasticity can be calculated reasonably well. Therefore, the quality of elasticity calculations is closely linked to the quality of base projections and revenue forecasts in the budget.

Possible sources for measurement error

Elasticity calculations for Japan are complicated by two shortcomings of Japanese fiscal data. First, estimates for the revenue impact of tax measures are on an appropriations basis, and may date back several years. For example, the estimate for the revenue impact of the consumption tax increase in 1997 dates from 1994, when the decision on tax increase was made, and may thus be based on inaccurate macroeconomic assumptions.

Second, estimates for general government tax elasticity are hampered by the lack of aggregate information on independent local government tax measures. The central government produces estimates for the impact of changes in shared taxes on local government revenue, but these taxes account for only a third of total local government tax revenue. Since fluctuations in local revenue have been small compared to central revenue, the impact of this lack of information may be limited, and has been ignored for the purpose of this paper. For the same reason, however, calculations for individual tax categories have been restricted to central government revenue only.

Constructing adjusted time series

To calculate tax elasticities over a period of time, actual revenue data need to be converted into a series that shows what the revenues would have been had there been no changes in the tax system. One method would be to apply current tax rates to the bases of earlier years, thus simulating a revenue series that corresponds to the current tax structure. However, this method imposes heavy data requirements, including detailed information on the distribution of the base by brackets or rate categories (Chand, 1975).

A more readily useable procedure requires only information on the revenue impact of tax measures. The current year (year 0) is set as the reference year. In the most simple case, tax revenues of previous years are adjusted according to:


While this remains a practical way of obtaining adjusted time series, the disadvantages of this method are clear. Forecasting errors contained in M accumulate over time and the proportional adjustment is a rather crude way of taking changes in the tax system into account. These shortcomings should be kept in mind when analyzing the results.

In the case of Japan, the method is slightly more complex because the authorities provide two estimates for the revenue impact of tax measures—a first-year impact and a full-year impact. The full-year impact refers to the hypothetical revenue gain or loss that would result if the tax measure would apply to the full fiscal year, whereas the first-year impact takes the timing of measures into account to estimate the actual revenue impact. Moreover, the authorities have also introduced temporary tax measures that need to be subtracted from tax revenue before calculating elasticities.

Again setting the current year as the reference year 0, define M¯i as the first-year impact of a tax change becoming effective in year i. With Mi denoting the full-year impact, adjusted tax revenue is defined as:


where Ti denotes revenue net of temporary measures.Table 6.6 provides the data and calculations for the general government tax revenue elasticity.

Table 6.6.

Central Government Tax Revenues (General Account)1

(Billions of yen, unless otherwise noted)

article image
article image
Source: Data provided by the Japanese authorities; and IMF staff calculations.

Fiscal year (April-March).


  • Bayoumi, T., 1998, “The Japanese Fiscal System and Fiscal Transparency,” in B.B. Aghevli, T. Bayoumi, and G. Meredith (eds.), Structural Change in Japan: Macroeconomic Impact and Policy Challenges (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bayoumi, T., C. Towe, and I. Oishi, 1998, “Fiscal Policy Issues, in Japan—Selected Issues, IMF Staff Country Report No. 98/113 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Buti, M., D. Franco, and H. Ongena, 1997, Budgetary Policies During Recession: Retrospective Application of the “Stability and Growth Pact” to the Post-War Period, Commission of the European Communities, Economic Papers No. 121.

    • Search Google Scholar
    • Export Citation
  • Chand, S.K., 1975, “Some Procedures for Forecasting Tax Revenue in Developing Countries,” International Monetary Fund, Fiscal Affairs Department, Departmental Memorandum 75/91.

    • Search Google Scholar
    • Export Citation
  • Ishii, H., and E. Wada, 1998, Local Government Spending: Solving the Mystery of Japanese Fiscal Packages, Institute for International Economics, Working Paper 98-5.

    • Search Google Scholar
    • Export Citation
  • Lipworth, G., and G., Meredith, 1998, “A Reexamination of Indicators of Monetary and Financial Conditions,” in B.B. Aghevli, T. Bayoumi, and G. Meredith (eds.), Structural Change in Japan: Macroeconomic Impact and Policy Challenges (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Okue, K., 1999, “Filling the Vacuum: Japan Policy Monitor,” Dresdner Kleinwort Benson Research Paper, Tokyo (April).

  • Posen, A., 1998, Restoring Japan’s Economic Growth (Washington: Institute for International Economics).

  • Schlesinger, J. M., 1997, Shadow Shoguns: The Rise and Fall of Japan’s Postwar Political Machine (New York: Simon & Schuster).

  • Virley, S., and M. Hurst, 1995, Public Finances and the Cycle, Treasury Occasional Paper No. 4 (London: H.M. Treasury).

  • Watanabe, K., T. Watanabe, and T. Watanabe, 1998, Tax Policy and Consumer Spending: Evidence from Japanese Fiscal Experiments, unpublished manuscript, Bank of Japan.

    • Search Google Scholar
    • Export Citation

Recent financial measures may have had larger effects. For example, small and medium-sized enterprises (SMEs) were reported to suffer from insufficient access to bank credit in 1998, and the subsequent extension of public loan guarantees for SME borrowing (outside any stimulus program) has led to a substantial increase in SME investment.


The structural balance is defined as the fiscal balance under a closed output gap. By eliminating the effects of cyclical swings on the fiscal position, the structural balance is a widely used measure for the fiscal policy stance.


An analysis of the impact of fiscal policy on activity, which is beyond the scope of this paper, is contained in Lipworth and Meredith (1998) and in the chapters by Bayoumi and Ramaswamy and Rendu in this book. Lipworth and Meredith estimated a multiplier of public expenditure of 1-1¼, while VAR approaches employed by the other authors yielded an estimate about half the size.


Japan’s construction sector accounts for 10 percent of total employment, compared to an average of 6–7 percent in other major industrial countries.


Public works spending accounts for roughly three-quarters of general government investment. Fiscal data on public works spending are not fully available on a timely basis, but national income accounts provide an early estimate for the demand impact of the stimulus packages.


For ease of comparison, the scales in panels A-C of Figure 6.3 have been kept the same, so that an investment change of, say, ¥10 trillion is identical in all three panels.


The figure shows the breakdown of changes in public works spending into two components:

(PWt - PWt-1) = (IBt - PWt-1) + (PWt - IBt)

where PW is actual public works spending (budget settlement basis), and IB the allocation for public works in the initial budget. For the purpose of this chart, budgetary spending by central and local governments has simply been added up, ignoring double-counted items (some 20 percent of total; see below). Time series data for consolidated general government spending have only been available on a settlement basis and do not allow for the breakdown shown in this chart.


A larger expenditure shortfall expected for FY1998 owes to the announcement of the package late in the fiscal year.


The central government is required to intervene in local government finances if debt or deficit-related thresholds are exceeded. Bond issuance restrictions come into effect once the ratio of interest payments to revenues exceeds 20 percent. Moreover, prefectures fall under the financial restructuring authority of the central government if an adjusted deficit measure exceeds 5 percent of budgetary expenditures (20 percent for municipalities).


This measure does not allow for stimulus measures being implemented through initial budgets in subsequent years. However, the degree to which such measures were “additional” can be questioned, given that initial budgets were largely contractionary in recent years.


An estimate for the overlap is obtained by adding central and local government expenditure (excluding social security and tax transfers) as reported in the respective fiscal accounts, and comparing this sum to general government expenditure (excluding social security) from the national income accounts.


A discussion of the concept of tax elasticity and its application to Japanese fiscal data, including measurement problems, is contained in Box 6.2 and in the Appendix.


Moreover, nominal GDP growth in 1997 was close to zero, inflating the absolute value of the elasticity measure.


If ηT, Base is the elasticity of tax revenue relative to a particular tax base, it holds that ηT, GDP = ηT, Base · ηBase, GDP.


Corporate profits are taken from the Financial Statements of Incorporated Enterprises, published by the Ministry of Finance.


An econometric analysis of tax income found a significant impact of land prices (and somewhat less of stock prices) on revenue. However, the effect was too small to fully account for the elasticity decline.


Automatic stabilizers would a priori be expected to play a lesser role in Japan than in other countries. While Japanese tax elasticities have on average been comparable to that of other industrial economies, the low level and cyclical variation of both registered unemployment and social welfare benefits would imply that business cycle fluctuations have less direct implications for public spending than in many other countries.


The estimated adjustment coefficients for the U.K. are similar to those obtained by Virley and Hurst (1995).


The impact of permanent tax cuts on consumer spending has been found to be about three times the size as that of temporary cuts (Watanabe and others. 1998).

How Japan Responded to Asset Price Collapse