Fiscal Sustainability of Japanese Prefectural Government Debt

This paper investigates whether the Japanese local government debt is sustainable. We apply the fiscal stabilization rule that each local government improves their primary balance in the current year when they issue additional debt the previous year. We check it using panel data of the Japanese prefectural governments. We find that Japanese prefectural governments were not sustainable across 1956-2007, though were sustainable across 1956-1989.


Introduction
In Japan, as in other countries, the recession caused by the subprime loan crisis has reduced tax revenues and necessitated additional government expenditures.Meanwhile, debt crises in Greece, Ireland, and Portugal have focused on financial reform in the EU.Japan's ratio of central and local government debt to GDP is relatively high compared to other OECD countries (Figure 1), and national financial reform has become a major debate in Japan.Furthermore, fiscal problems among municipalities and prefectures have been a topic of sustained discussion in Japan.In Kochi Prefecture in July 2004, the Prefectural Governor Daijiro Hashimoto declared a financial crisis during an address to the regular prefectural assembly and asserted that Kochi would face fiscal reconstruction in 2007 if conditions remained unchanged.In March 2007, the designated organization for fiscal secondary reconstruction in Yubari City, Hokkaido, highlighted the increased need to reform finances.More recently, it was said that Okayama and Miyagi Prefectures would be organized for fiscal reconstruction until 2011 in 2008 1 .Given these discussions, it is timely to consider the fiscal sustainability of Japan's prefectural governments.This paper investigates whether the Japanese prefectural government debt is sustainable or not, applying the empirical analysis of the debt stabilization rule á la Bohn (1998) [2] and Mendoza and Ostry (2008) [7].We find that Japanese prefectural governments were not sustainable across 1956-2007, though were sustainable across 1956-1989.
There are several literatures that discusses about the sustainability of the Japanese prefectural government debt, such as Doi (2004) [4], Sumi and Kawase (2007) [10], and Saito (2010) [9].These studies, however, adjudge that government debt is sustainable if the current debt is less than the discounted present value of the primary balance.Moreover, these studies do not use long-term data and are not framed on the basis of fiscal stability rules articulated in seminal articles on fiscal solvency such as Bohn (1998 The rest of this paper is structured as follow.Section 2 reviews the previous literatures.Section 3 presents the empirical method and estimates and interprets the results.Finally, Section 4 concludes.

Method for Fiscal Sustainability Test
We use the method of Mendoza and Ostry (2008) [7] that has applied in a cross-country data test by Bohn (1998) [2].We use 1955-2007 annual data for Japan's 47 prefectures (i = 1, 2, ⋅⋅⋅, 47), (t = 1955, 1956, ⋅⋅⋅, 2007).The entire period can be divided into the period before the bubble burst (prefectural budgets were relatively balanced) and after the burst (reduced tax receipts necessitated additional fiscal expenditures).For each sample period, we estimate the equation , 0 where , i t s is the primary surplus per gross prefectural products (GPP) in prefecture i, , i t y  and , i t g  denote temporary fluctuations in government spending and GDP, and , i t b is the government debt-to-GPP ratio.Results of the Hausman-Wu test supported use of a fixed effects model with specific effects for each prefecture.This model allows constant terms that vary from prefecture to prefecture.Since there may be first-order serial correlation that varies across prefectures, we follow Mendoza and Ostry (2008) [7] in adjusting our estimate for first-order serial correlation by specifying that , , 1 , In addition, we make the White adjustment for heteroscedasticity using estimated prefectural standard errors and covariance.
The regressions use alternative measures of temporary fluctuations in government outlays and GDP.One is obtained by detrending the data using the Hodrick−Prescott filter.The resulting detrended series are labeled "output gap": ( ) The second set follows Bohn (1998) [3] to construct measures of temporary fluctuations in output and government purchases that enter in the closed-form solution of Barro's (1986) [1] tax-smoothing model.These measures are defined as GVAR and YVAR for government purchases and output, respectively: In these expressions, a subscript T denotes the trend value of the corresponding variable.

Result
Table 1 shows the estimation results of Equation ( 1), for the period before the Japanese housing and stock market bubble4 burst (until 1989), and after the bubble burst (since 1990) 5 .Values in row I, column II represent the entire period.Those in row III, column IV pertain to pre bubble.Values in row V, column VI are post bubble.
However, estimated values of the constant terms 0,i β in Equation ( 1) are not reported 6 .As seen in Table 1, φ takes a positive and significant valueonly for the period before the bubble burst.The estimated value of φ is insignificant for the entire period and for the post-bubble period.
We derive the following two points.First, before the bubble burst the value of φ is positive and significant, meaning that Bohn's (1998) [2] sufficient condition for sustainability was fulfilled.Second, post-bubble tax revenues fell until the early-to-mid-1990s, and public works were initiated as an economic stimulus.Fiscal discipline was absent, and Bohn's (1998) [2] conditions were not met.Finally, the Bohn (1998) [2] criteria were not met for the entire period because of the large post-bubble fiscal stimulus and increases in the chronic budget deficit.Thus, we see that the condition for sustainability of the fiscal policy reaction function under the assumption of linearity did not hold.η ～i.i.d) was corrected by estimating.In addition, heteroscedasticity is fixed by White cross-section standard errors and covariance.Output gap and government spending gap are made by real GPP and real government expenditure extracted from trends of the Hodrick−Prescott filter, YVAR, GVAR is computed from the definition of Barro (1986) [1].