¹ There are examples of as many as three supplementary budgets being passed in a single calendar year. For a discussion of supplementary budgeting practices in Japan, see Campbell, John C., Contem-porary Japanese Budget Politics (Berkeley and Los Angeles: University of California Press, 1977)Google Scholar.

² While there have been exceptions, most supplementary budgets have been enacted in the last three months of a calendar year or in January of the following year.

³ The key phrase here is keizai taisaku (economic counterpolicies). Sometimes this phrase has been used alone in reporting on such policies, but more often it has been prefaced with the words sogo (comprehensive) or kinkyu (emergency).

⁴ The best examples were the four supplementary budgets enacted to carry out economic stimulus and other activities in 1977 and 1978. See Shinbun, Asahi, Asahi Nenkan (Japan Yearbook) (Tokyo: Asahi Shinbunsha, 1978 and 1979)Google Scholar.

⁵ Although Table 1 contains only those stimulus packages announced since May 1987 inclusive, our attempt to explain the timing of the packages also includes the four packages enacted in the late 1970s.

⁶ When we say popular business press, we are generally referring to such publications as the economic journal Nihon Keizai Shinbun and Japan's three principal daily newspapers—Asahi Shinbun, Yomiuri Shinbun, and Maimcht Shinbun.

⁷ The FILP, Fiscal Investment Loan Plan, is essentially an extrabudgetary system of finance available to the government that utilizes funds collected through the country's Postal Savings System. For a discussion of the FILP and other aspects of the Japanese financial system, see Suzuki, Yoshio, TheJapa-nese Financial System (Oxford: Clarendon Press, 1987)Google Scholar.

⁸ While this methodology produces numbers comparable to the “consensus” understanding of the value of the economic stimulus packages, it still somewhat overstates the magnitude of many of them. Perhaps the best illustration of this is the September 1995 package. When loan allocations and previously budgeted land purchases are netted from the ¥14 trillion total, ¥9 trillion of expenditure remains. Of this amount, ¥1.1 trillion was earmarked for Uruguay Round countermeasures, but it was not made clear whether this would be allocated to expenditure or loan programs. Depending upon the outcome, the calculated total of “hard” expenditure would be between ¥7.9 and ¥9.0 trillion.

⁹ For an illustrative discussion of efforts by the Ministry of Finance to reduce budget deficits in the 1970s, see, for example, Noguchi, Yukio, “Public Finance,” in Yamamura, Kozo and Yasuba, Yasukichi, eds., The Political Economy ofJapan, vol. 1 (Stanford, Calif: Stanford University Press, 1987)Google Scholar.

¹⁰ See Louise do Rosario, “Giving It Back: Declining Economy Sparks Debate on Tax Cuts,” Far Eastern Economic Review (March 11, 1993), 13.

¹¹ For evidence of the neutrality of Japan's economic stimulus policies, see Beason, Dick, “Neutrality of Fiscal Packages during Japan's Heisei Recession” (Manuscript, Department of Business and Economics, University of Alberta, 1996)Google Scholar. The only amendment we should add to this concerns the 1998 package, which some bureaucrats in the Ministry of International Trade and Industry publicly supported. See Peter Landers and Trish Soywell, “The Big Gamble: Will Keynesian Medicine Work for Japan and China?,” Far Eastern Economic Review (November 26, 1998), 98-99.

¹² See, for example, Henry Sender, “More of the Same,” Far Eastern Economic Review (October 5, 1995).

¹³ For the revisionist view of economic policy-making in Japan, see Johnson, Chalmers, MITI and the Japanese Miracle (Stanford, Calif.: Stanford University Press, 1982)Google Scholar. Perhaps the best general statement of the influence of electoral politics on policy-making in Japan can be found in Calder, Kent, Crisis and Compensation: Public Policy and Political Stability in Japan, 1949—1986 (Princeton: Princeton Univer sity Press, 1988)Google Scholar. Other scholarly work we refer to includes, on budgeting, McCubbins, Matthew and Noble, Greg, “The Appearance of Power: Legislators, Bureaucrats, and the Budget Process in the United States and Japan,” in Cowhey, Peter and McCubbins, Matthew, eds., Structure and Policy in Japan and the United States (New York: Cambridge University Press, 1995)Google Scholar; and more generally on policy-making in Japan, Ramseyer, Mark and Rosenbluth, Frances, Japan's Political Marketplace (Cambridge: Harvard University Press, 1993)Google Scholar.

¹⁴ See Schoppa, Leonard, “Gaiatsu and Economic Bargaining Outcomes,” International Organization 47 (Summer 1993)Google Scholar.

¹⁵ The reason that most scholars accept the “political surfing” view is that while it conforms to the dominant politician view, it does not at the same time conflict with the view that Japanese bureaucrats are autonomous in the area of economic policy-making.

¹⁶ See Kohno, Masaru and Nishizawa, Yoshitaka, “A Study of the Electoral Business Cycle in Japan,” Comparative Politics 22 (January 1990)CrossRefGoogle Scholar. It is true that Alesina and Roubini also found evidence of the “political business cycle” in Japan that went beyond political surfing. See Alesina, Alberto and Roubini, Nouriel, “Political Cycles in OECD Economies,” Review of Economic Studies 59 (October 1992)CrossRefGoogle Scholar. However, their model tested the impact of elections on economic outcomes—namely, GDP growth and unemployment—not on the impact that elections had on the probability that a certain economic policy would be announced.

¹⁷ Kohno and Nishizawa also pointed out that the lack of a statistically significant relationship may be due in part to the difficulties of specifying such a model correctly. See Kohno and Nishizawa (fn. 16), 159-90.

¹⁸ In the early 1960s the leadership of the LDP made it known that it would soon face a support crisis because of demographic changes that were taking place in the country. In a 1963 Chuo Koron (Central Review) article, LDP member Ishida Hirohide stated that socioeconomic changes brought about by Liberal Democratic Party policies would lead to conservative-progressive electoral parity unless the LDP began to attract new supporters. See Horie, Fukashi and Umemura, Masahiko, Toyo Kodc to Seiji Ishiki (Voting behavior and political consciousness) (Tokyo: Keio Tsushinsha, 1986), chap. 1Google Scholar.

¹⁹ Kohno and Nishizawa (fn. 16), 163.

²⁰ To serve its electoral interests the LDP used not only fiscal policy but also other policy tools. For a general discussion of how the ruling party used the policy process to solve the numerous electoral crises it faced throughout the postwar period, see Calder (fn. 13).

²¹ The New Liberal Club was formed in 1976 by several reform-minded members of the LDP. It put up candidates in the next five lower-house elections but rejoined the LDP after its disappointing performance in the double election of 1986.

²² The LDP returned to power less than a year later, either as a minority party or in coalition with some other party or individuals. In terms of its underlying electoral strength and the strength and unity of the opposition it faces, it is not the same party it was earlier in the postwar period.

²³ The LDP is currently the governing party, but in coalition with other parties, as in the recent past. It still has not recovered from its loss in 1993.

²⁴ This is not to suggest that a governing party with a majority is uninterested in expanding its level of support. Rather, it is to emphasize that governing parties with sufficient support to obtain a majority have to be concerned more with their core supporters turning out and voting than with attracting new supporters.

²⁵ The motivation for this would simply be more portfolios or a greater share of the most important portfolios, given that the distribution of portfolios tends to reflect the party's distribution of seats. See, for example, Browne, Eric and Franklin, Mark, “Coalition Payoffs and European Parliamentary Democracies,” American Political Science Review 67 (June 1973)CrossRefGoogle Scholar.

²⁶ This division may be somewhat misleading because governing parties with a weak or nonexistent majority actually face both imperatives simultaneously. However, parties that need to attract additional support must still be sure that their core supporters turn out and vote.

²⁷ These were prepared by the Advisory Group on Economic Structural Adjustment for International Harmony, a group appointed by Prime Minister Nakasone and chaired by former Bank ofJapan governor Maekawa Haruo. For a discussion of specific proposals contained in the reports, see Balassa, Bela and Noland, Marcus, Japan in the World Economy (Washington, D.C.: Institute for International Economics, 1988)Google Scholar.

²⁸ As previously stated, U.S. policymakers originally encouraged economic stimulus policies to increase domestic demand in Japan and thereby increase the flow of U.S. exports. While U.S. policy-makers have long desired domestic economic stimulus policies, they want them now even more because of the depressed state of the Japanese economy.

²⁹ See Schoppa (fn. 14).

³⁰ The economic stimulus package corresponding to the April 1993 entry in Table 1 was announced by Prime Minister Miyazawa at his meeting with President Clinton.

³¹ By salient issues, we mean those problems that define the agenda of an election. This concept helps define the critical distinction that exists between a party's core supporters, who generally vote in response to partisan or long-term cues, and those potential supporters it wishes to add to its rolls (peripheral voters), who vote more in response to election-specific, short-term cues. For a detailed discussion and empirical tests of this idea, see Budge, Ian and Farlie, Denis, Explaining and Predicting Elections (London: George Allen and Unwin, 1983)Google Scholar; and Petrocik, John, “Issue Ownership in Presidential Elections with a 1980 Case Study,” American Journal of Political Science 40 (August 1996)CrossRefGoogle Scholar.

³² For the purposes of our amended model, both lower-house (House of Representatives) and upper-house (House of Councillors) contests are included. We realize that the House of Councillors is less powerful than its counterpart but note that its control has nonetheless been extremely important if not essential for the success of any governing party's legislative program during much of the period we considered. It is difficult for us to imagine that it would be ignored electorally, and for this reason we included such elections in our analysis. See the appendix for the dates of all general elections held during the period under consideration, and see also fn. 48 for a discussion of the statistical significance of this variable at different lags.

³³ The circumstances described here assume that options such as calling an election or forming a new cabinet are neither politically desirable nor procedurally feasible.

³⁴ See fn. 51 for a discussion of the substantive and statistical significance of this variable at different lags.

³⁵ During the twenty-four-year period from 1975 to 1998 inclusive, twenty-six cabinet reshuffles occurred.

³⁶ Cabinet support data collected byJijitsushinsha after 1992 were provided by the Chuo Chosasha news organization.

³⁷ In our discussion of model results, we report the statistical and substantive significance of this variable at different lags. See fn. 49.

³⁸ Meetings of U.S. presidents and Japanese prime ministers are listed in the appendix.

³⁹ For the period under consideration, the mean value of the year-on-year changes in the production index was 2.79 while minimum and maximum values were −19.2 and 14.0 respectively.

⁴⁰ See fn. 51 for a discussion of this variable's estimates at different lags.

⁴¹ It may seem reasonable to argue that a successive package would not be announced for two or three months at a minimum and that our assumption of two consecutive months is not restrictive enough. In response, we point out that our data contain examples of package announcements two and four months apart, confirming our belief that it is nearly impossible that packages will be announced two months in a row.

⁴² Beck, Nathaniel, Katz, Jonathan, and Tucker, Richard, “Taking Time Seriously: Time-Series Cross-Section Analysis with a Binary Dependent Variable,” American Journal of Political Science 42 (October 1998)CrossRefGoogle Scholar.

⁴³ We are grateful to Tse-Min Lin, Robert Lowery, and Mark Jones for helpful suggestions on this problem. For treatments of this problem generally, see Bye, Barry and Riley, Gerald, “Model Estimation When Observations Are Not Independent,” Sociological Research and Methods 17 (November 1989)Google Scholar; Zegler, Scott L., Liang, Kung-Yee, and Self, Steven G., “The Analysis of Binary Longitude Data with Time Independent Covariates,” Biometrika 72 (April 1985)Google Scholar; and Simon Jackman, “Times Series Models for Discrete Data: Solutions to a Problem with Quantitative Studies of International Conflict,” (Manuscript, Department of Political Science, Stanford University, 1998).

⁴⁴ The plots we examined were those pairing predicted probabilities with (1) changes in the Pearson X², (2) changes in deviance, and (3) changes in influence.

⁴⁵ For this case, cabinet support rates were positive at lags four and five, and no other independent variables took on values that would have indicated the announcement of an economic stimulus package was a high probability.

⁴⁶ These were the cases for December 1977, September 1978, and May 1995.

⁴⁷ Cabinet support rates were low at times t, t-1, and t-2. It is therefore possible that our lag structure is somewhat misspecified. See the discussion in fn. 51.

⁴⁸ The decision to include national elections at a two-month lag was based on our understanding of how incumbent governments like to time the packages; to receive credit for taking such an action, but to escape punishment for being overtly political. In the interest of robustness, however, we ran the model with this variable at different lags. The results were statistically insignificant coefficients on this variable at time t (a lag that predicted failure perfectly), time t-1 (coefficient = 0.832 and significance = 0.352), and time t-3 (coefficient = 0.325 and significance = 0.775).

⁴⁹ As in the case of national elections, our determination that a one-month lag on the “new government” variable was appropriate was theoretically based. To determine if this was the only statistically significant lag, however, we ran the model with the “new government” variable at lags t(0), t-2, and t-3. In each case, the new government variable predicted failure perfectly.

⁵⁰ Because foreign pressure was not operative until after 1986, we ran the model with a structural break variable that produced no change in the overall fit of the model and no substantive or statistical change in its coefficients. Moreover, the log odds ratio for the structural break variable was small (0.402) and well out of the range of statistical significance (p < .55). When we ran the model for the post-1986 period, results were somewhat different. The overall statistical significance of the model did not change, but the substantive and statistical significance of some of the variables did move. In the post-1986 period elections were much more important than for the whole series (a log odds ratio of 2.718) while gaiatsu dropped but only very slightly (a log odds ratio of 2.219). The log odds ratio on the new government variable represented the most important change; although it remained virtually unchanged in size, it moved just out of the range of statistical significance.

⁵¹ Running the model at different lags for cabinet support and year-on-year changes in the production index resulted in differences of neither substantive nor statistical significance for these two control variables.

⁵² This proved to be a statistically insignificant difference.