1. Throughout the book, I use the terms cabinet, ministers, and government interchangeably to refer to the executive.
2. Parties in a coalition government divide the portfolios. The party that controls the portfolio on a particular dimension (here, the monetary-policy dimension) is referred to as the portfolio party or the cabinet minister. The terms coalition partners or nonportfolio parties refer to parties that participate in the governing coalition but do not hold the portfolio on the monetary-policy dimension.
3. The term backbench legislators refers to legislators in the governing party(ies) who are not members of the cabinet.
1. This contracting approach has engendered critics of its own, who note that politicians may be unwilling to punish the central banker (de Haan 1997; McCallum 1997; Waller 1995). That is, the decision to enforce the contract also reflects the strategic interaction of central bankers and politicians. Defenders argue that the announcement of specific goals and targets changes this interaction by increasing the transparency of monetary policy, which may in turn increase the public costs to politicians of reneging on the contract.
2. Research suggests that ex-post measures may be a less-effective method of controlling bureaucrats. Bureaucrats can defy the enacting coalition’s preferences by pursuing policies that appeal to a new majority of legislators (McCubbins, Noll, and Weingast 1989). Bureaucrats can also exploit the imperfect relationship between their policy choices and policy outcomes to avoid sanctions (Banks 1989). Finally, politicians may have difficulty punishing bureaucrats without hurting affected constituents; extensive budget cuts or dismantlement of the agency may destroy politically desirable programs.
3. Political economists have also attempted to gauge a bank’s independence using behavioral indicators. Cukierman (1992; Cukierman, Webb, and Neyapti 1992), for example, measures independence using the turnover of central bank governors, arguing that high turnover indicates that the government retains control over monetary policy. A long tenure, however, could reflect the governor’s compliance with the policy demands of his political principals. The Bank of France, for example, has had only a few governors in the post-World War II period but is usually classified as dependent by scales of independence based on formal indicators. Behavioral measures of independence are ex-post indicators, themselves a product of the strategic interaction between politicians and central bankers in the policy process—interactions that depend on the institutional environment.
4. All legal rankings of central bank independence share a number of shortcomings (Forder 1996, 1998a; Mangano 1998). Most rankings simply assign equal weight to each component and add them together to form the index. Some factors, however, may carry more weight in the determination of central banks’ overall independence. Additionally, some components may be interactive, becoming important only in the presence of other variables. Further, the scales consider only the banks’ formal structures, ignoring some behavioral features that may affect the banks’ performance (Forder 1996).
5. Ratification of the Maastricht Treaty, however, has changed the constitutional status of central banks not only in Germany but throughout the EU. The treaty calls for each member-state government to grant its central bank formal independence. The constitutional status of the treaty therefore makes the institutional organization of the central bank a constitutional issue.
6. Leftist opponents of the reform challenged its legality in court. The Constitutional Council, France’s highest constitutional body, actually struck down the reform in August 1993. The Council argued that under the French Constitution, only the government has the right to determine monetary policy. The Council did acknowledge that the reform would become legal once the Maastricht Treaty came into effect because the treaty commitment would override the constitution (Wall Street Journal, 5 August 1993; Financial Times, 13 August 1993).
7. Examples of the delegation literature include Banks (1989), Banks and Weingast (1992), Calvert, McCubbins, and Weingast (1989), Lupia and McCubbins (1993), McCubbins, Noll, and Weingast (1987, 1989), McCubbins and Schwartz (1983), Moe (1987), and Weingast and Moran (1983).
8. Although the time-consistency framework dominated macroeconomic approaches to monetary policy in the 1980s, the approach has some critics, who contend that the setup does not accurately characterize the policy process (Blinder 1997; Freedman 1993; McCallum 1995, 1997).
9. The effect of central bank independence on real economic variables remains a point of controversy. One line of research investigates how central bank independence affects the costs of disinflation in terms of lost output and increased unemployment. In theory, the credibility of an independent central bank should lower these costs because economic agents will adjust their inflation expectations more rapidly than with a dependent central bank. Empirical studies, however, show that the costs of disinflation are actually higher under an independent central bank (Posen 1998; Walsh 1995a). A second set of political economists has investigated the relationship between central bank independence and labor-market organization. This work demonstrates that systems with highly unionized workforces and coordinated wage bargaining can better respond to the policy actions of an independent central bank, helping to keep unemployment low (Franzese 2000; Hall and Franzese 1998; Iversen 1998, 1999).
10. Moe (1989, 1990) recognizes that political institutions can influence politicians’ incentives over bureaucratic structure. In his discussion of the U.S. case, he argues that the president has specific preferences over bureaucratic institutions, preferring a hierarchical top-down structure to enhance the president’s ability to control policy. In contrast, Moe assumes that legislators usually act as conduits for interest-group pressures.
1. In a coalition government, cabinet ministers typically act as agents of their party, rather than as individuals or representatives of factions (Laver and Shepsle 1994).
2. More than one nonportfolio party may exist. For stylistic convenience, however, I refer only to a single nonportfolio party.
3. Backbench legislators and coalition partners need not have perfect information about the cabinet’s preferences in order to limit the range of its policy discretion. Under the assumption that cabinet ministers have a lexicographic preference for office, the threat of a veto or dismissal should provide incentive for any government—even one with policy preferences opposed to its supporters—to remain within the range of acceptable policies.
4. Economists have proposed both noncontingent rules and contingent policy rules (Keech 1995). Noncontingent rules specify the policy response regardless of the economic situation. They establish clear standards to evaluate performance, although their inflexibility limits their feasibility. Contingent rules specify different policy responses depending on the economic situation. It may be impossible, however, to specify rules for every possible situation. Further, policymakers might have discretion in defining which situation applied to the rule. Finally, contingent rules may become so complex that it is difficult to determine whether the government has complied with the rule.
5. Typically, the financial sector closely monitors the government’s monetary policy choices. But politicians may be reluctant to depend on it for information because the financial sector lacks a broad political constituency.
6. I model the government’s and central bank’s actions as simultaneous moves. The bureaucrat makes a comment on the government’s policy proposal, either endorsing or criticizing potential policy actions. Simultaneous actions are plausible because monetary policy-making is a continuous process.
7. I examined allowing the government to punish the bureaucrat for other reasons, including sending a message that led the legislator to veto the government. The results are qualitatively similar.
1. Case accounts of the interaction between the German government and the Bundesbank abound. See, for example, Heisenberg (1999), Henning (1994), Goodman (1992), Marsh (1992), Kennedy (1991), and Scharpf (1987).
2. Marsh (1992) suggests that the Bundesbank’s actions contributed to the downfall of the Erhard (1966), Keisinger (1969), and Schmidt (1982) governments.
3. In spring 1993, I conducted a number of off-the-record interviews with bank officials regarding the Bundesbank, some of which I draw upon in this chapter.
4. Approximately 78 percent reported that the recommendation by the European Monetary Institute, the forerunner of the European Central Bank, would also be an important source of information.
5. A Belgian central banker noted that as his country’s bank had become de facto more independent in the 1980s, the governor had increased his political role, appearing more frequently in newspapers and television.
1. If the issue does not affect vital Lander interests and the Bundesrat fails to pass the legislation by a simple majority, then a simple Bundestag majority can overturn the Bundesrat veto. In this case, the Bundesrat veto only delays the bill’s implementation. If the Bundesrat vetoes the legislation by a two-thirds majority, however, the Bundestag can overturn the veto only with a two-thirds majority. For issues affecting Lander interests, however, a simple majority in the Bundesrat can veto legislation, unless the government can muster a two-thirds majority in the Bundestag to override the veto.
2. The Economist noted that it would “take a very nervous heart to register a flutter at what is contained in the Bill. Nothing could be more moderate” (quoted in Fforde 1992, 27).
3. The Alford index subtracts the percentage of voters from non-working classes voting for the Left party(ies) from the percentage of blue-collar workers voting for the Left party(ies). To create an Alford index that covers the same period as the dependent variable, I averaged two indices, one from the 1960s–early 1970s and the other from 1990. For the 1960s–early 1970s Alford index, I employed data from Powell (1982). His data were collected from a variety of sources. Next, I constructed an index using the 1990 World Values Survey. Respondents identified both which party they would support in a hypothetical general election and their occupation. Manual workers (skilled, semiskilled, unskilled), foremen/supervisors, and agricultural workers were coded as blue collar. All others were included as non–working class. Left parties included Socialist, Labor, Communist, and other far Left parties. For each country, I calculated the percentage of voters from non–working classes voting for the Left party(ies) and subtracted that from the proportion of blue-collar workers voting for the Left party(ies). The correlation between the World Values index and Powell’s data is r = .52, significant at the 0.05 level.
Although the two sources had many countries in common, the World Values Survey did not include Australia or New Zealand. I imputed the missing values by regressing the World Values Survey index on Powell’s data. Using the coefficients from the equation and the values of the missing countries on Powell’s data, I computed the missing values for the World Values Survey index. Additionally, Powell does not include data for Spain because it was not a democracy until the mid-1970s. Therefore, I included only the data from the World Values Survey for the Spanish case. Finally, I ran each model using Powell’s data and the index created by the World Values Survey separately (results not reported). The results for each index were in the predicted direction and significant (p < 0.05), although at lower levels of confidence than the averaged index.
4. According to Powell (1982), extremist parties exhibit one of the following characteristics: (1) a well-developed nondemocratic ideology; (2) a proposal to break up or fundamentally alter the boundaries of the nation; or (3) diffuse protest, alienation, and distrust of the existing political system. I updated Powell’s measure to include the average polarization for 1960 to 1990. I follow Powell’s classifications with the following exceptions: France (include National Front) and Spain (include Communist, Separatist parties).
5. Following Strom (1990a), Powell and Whitten (1993) define a committee system as strong if it shares two of the three following characteristics: size (over 10 committees), specialization to fit the government bureaucracy, and limitations in the number of committeeships held by individual legislators. Inclusive committee systems require that committee chairmanships be distributed proportionally among all parties, regardless of their participation in government. Countries with strong and inclusive committee systems include Austria, Belgium, Denmark, Germany, the Netherlands, Norway, Sweden, and Switzerland.
6. I calculated the proportion of time a country was governed by a coalition or minority government from 1960 to 1990 using data from Woldendrop, Kernan, and Budge (1993). I counted instances of “divided government” in the United States as minority government because the president’s party lacks an overall majority in Congress.
7. I also ran the models using the index from Cukierman, Webb, and Neyapti (1992) as the dependent variable. The results were similar (results not reported).
8. The Alford index had a significant negative relationship with the other scales, except for the Cukierman index. There, the coefficient was in the predicted direction but failed to attain conventional levels of significance (t-statistic = −1.53, critical value for two-tailed test = −2.16).
9. I also ran each of these variables individually. Both committee strength and the coalition/minority government variables were significant and in the predicted direction. Polarization was in the predicted (negative) direction but did not attain conventional levels of significance (t-statistic = −1.46, critical value for two-tailed test = −2.16). Results not reported.
10. This test is a check for spatial correlation between errors of cases located next to one another.
11. RESET is the regression specification error test proposed by Ramsey (1969). The test uses the predicted values of the dependent variable from the original equation, raised to the second, third, and fourth powers, as additional regressors. The significance of these additional regressors is evaluated with a standard F-test. A significant test indicates that the model is misspecified.
12. I also used a population-adjusted measure of trade openness, based on Taagepera and Hayes (1977). I created the population-weighted variable by regressing the (logged) proportion of trade on the (logged) population for the sample of countries. I then subtracted the predicted proportion of trade from the actual values to form the measure. This measure also failed to attain significance.
13. It can be argued, somewhat counterintuitively, that Left parties will favor independent central banks. Left parties recognize that they possess little anti-inflation credibility with financial and capital markets, contributing to higher risk premia and the possibility of capital flight (Garrett 1998a). To demonstrate their commitment to responsible economic policies, Left parties may institute an independent central bank. Right parties have more credibility with the markets and consequently, less incentive to choose an independent central bank.
14. I employed average measures for other time periods, including 1970–79 and 1980–89. These measures were not significant.
15. Posen’s data set did not include Norway and Sweden. Following Posen (1995), I constructed measures for these two countries using a variety of sources. Additionally, Posen’s measure of financial sector strength can be broken into two dimensions—a financial-sector dimension, composed of the presence of universal banking and the regulatory role of the central bank, and a political dimension, composed of federalism and party-system fractionalization. The financial dimension is highly correlated with the punishment index (r = 0.85). The political dimension is correlated with the presence of strong bicameralism (r = 0.83). Both correlations are significant at the 0.001 level.
16. I performed similar analyses for models with other measures of central bank independence as the dependent variable. The other models exhibited some evidence of influential outliers, although Britain was the only outlier common to all four indices.
17. Belsley, Kuh, and Welsch (1980) suggest observations with IDFBETAil > 2/(n½) exert undue influence. This cutoff is stronger than the standard recommended by Bollen and Jackman. For the Alford index, observations failing to meet this standard include Canada (−0.79), Denmark (0.69), and Britain (1.19). For strong bicameralism, Australia (−0.89) exceeds the cutoff. Finally, Australia (0.63), Belgium (−0.49), and Britain (−0.99) exceed the cutoff for the punishment index.
18. I also tried individual dummy variables for Britain, Denmark, and Australia. The Britain dummy variable attained statistical significance. The Denmark dummy was not significant on its own but became significant when used in a model with the Britain dummy. The dummy for Australia was not statistically significant at any point.
19. If the absolute value of DFFITS is less than or equal to 0.34, then the weight is equal to one. If IDFFITSI > 0.34, then the weight is 0.34/IDFFITSI. Weights for the bounded regression are contained in table 5.
1. Following King, Burns, and Laver (1990) and Warwick (1994), I estimate how different cabinet, party system, and economic system attributes affect expected, rather than actual, cabinet duration. That is, I use the attributes to estimate how long the cabinet is expected to survive on the day it takes office. Actual cabinet duration is, instead, a function of a stochastic process.
2. In alternative specifications, I used (1) a dummy variable for capital controls from Leblang (1997) and (2) a measure of overall capital and current account openness developed by Quinn (1997). The results from all three measures were qualitatively similar, although at varying levels of statistical significance.
3. The source for this data is Golden, Lange, and Wallerstein (1998). In alternative specifications, I also used a measure of the centralization of wage bargaining from Iversen (1999). The results were similar, but I lost almost 25 percent of the sample when this variable was included.
4. I excluded Switzerland due to the permanent oversize status of their executive council. I excluded other countries due to limitations in the availability of data.
5. In alternative specifications, I ran the analysis without this control variable. The results were similar.
6. Ordinary robust estimators are also vulnerable to outlying or influential cases. To overcome this problem, this robust estimator begins by performing an initial screening for high-leverage cases based on Cook’s Distance. Cases that have values of Cook’s Distance greater than one are dropped from the analysis. The robust estimator then proceeds iteratively and downweights observations based on both Tukey’s and Huber’s weighting scheme (Hamilton 1992). The results are thus not only robust to departures from normality on the part of the dependent variable but are also shielded from being overly influenced by high-leverage observations.
7. Using the OLS estimates, the mean predicted value of cabinet durability in an open economy increases from 0.93 to 1.12 as central bank independence moves from low (0.2) to high (0.7). Using the estimates obtained via robust regression, the mean predicted value of cabinet durability in an open economy increases from 0.58 to 1.06 over the same range of central bank independence.
8. For clarity of presentation, I omitted the confidence intervals. The graph was simulated using the results obtained via OLS (model III in table 8).
9. To calculate these values, I used the range of central bank independence. For a “dependent” central bank, I restricted the variable to be its lowest level in the sample (0.17). To capture the effect of an independent central bank, I restricted the variable to be its highest value in the sample (0.69). To measure economic openness, I employed different levels of the capital account and trade openness variables. For a “closed” economy, I restricted capital account openness to be at its lowest level in the sample (1.5) and trade openness to be one standard deviation less than the mean (30 percent of GDP). For an open economy, I restricted capital account openness to be at its highest level in the sample (4) and trade openness to be one standard deviation more than the mean (100 percent of GDP).
10. For the OLS results, the differences in predicted cabinet durability are not statistically significant at the 95 percent confidence level at any point, although the confidence intervals (not shown) do approach significance at lower levels of unionization.
11. For Left government, I restricted the partisanship variable to take on the value of 1, indicating a Left cabinet with a bare majority of legislative seats. For Right government, the partisanship variable takes on a value of 0, indicating no Left parties in the cabinet. The levels of openness were computed as before. The graph was simulated using the results obtained via OLS (model III in table 8).
12. The difference between predicted cabinet durability for Left and Right governments under conditions of economic openness is statistically significant. The difference between Left and Right under a closed economy is not.
1. Between 1969 and 1972, unions negotiated substantial wage increases, pushing up nominal wages more than 10 percent a year (Goodman 1992). Between 1970 and 1974, unit labor costs increased over 64 percent (Locke 1995).
2. Additionally, in 1975, unions bargained for the scala mobile, which indexed wages to price increases, helping to fuel inflation (Lange 1986; Locke 1995).
3. Revenues from the North Sea oil began to flow in 1977, easing the balance-of-payments problems.
4. The Conservatives had previously considered an independent central bank. As a member of the opposition in 1978, future Chancellor Nigel Lawson called for an independent Bank of England. Other Conservatives recalled that in the late 1970s, proposals for an independent central bank had received serious debate within the party (Financial Times, 2 November 1989).
5. See, for example, John Maples, former Conservative MP and economic secretary to the Treasury, in the Guardian (30 May 1994) and Bryan Gould, Labour MP, in the Guardian (5 January 1994).
6. Note that the Westminster system protected the Conservative Party’s position in office, even though intraparty conflicts cost Thatcher her job.
7. If the bank missed the target range (if inflation was either above or below the target), it would have to write an open letter to the chancellor, explaining the reasons for missing the target and the actions that the bank would take to correct the problem (Financial Times 12 June 1997; 13 June 1997).
8. Abbott had served on the Treasury Select Committee in 1993–94 that had advocated a more independent Bank of England. She was the only member of that committee to vote against that recommendation. She argued that the Labour government’s bill destroyed democratic accountability and that academic research had established only correlation, rather than causation, between central bank independence and inflation.
9. Interestingly, in Britain, many proponents of central bank independence also advocated sweeping reforms to Britain’s political and electoral institutions. The Liberal Democrats, who supported an independent bank, demanded the adoption of a proportional representation electoral system. Additionally, The Economist strongly advocated independence for the Bank of England (see, for example, The Economist, 28 October 1989), while endorsing political and electoral reform (for example, in its series, “Britain’s Constitution,” beginning 14 October 1995). Will Hutton, a columnist for the Guardian and an opponent of independence, argued that an independent Bank of England would be ineffective and unaccountable within the current political system. Instead, an independent central bank could work only under a “wider constitutional settlement … with reconstructed Houses of Parliament and [a] system of regional government. The Bank of Britain [sic] would then become federally constituted like the Bundesbank and the Fed with members from Scotland, Wales, and the English regions, and made accountable to a two-chamber parliament in which, hopefully, the committee system could be more effective” (Guardian, 30 August 1993).
10. Institutional reform can also occur if the traditional governing parties do not adjust their policies adequately enough to political and economic shocks. Political entrepreneurs can attempt to capitalize on voter dissatisfaction and appeal to groups unrepresented by the traditional governing party(ies) to press for political reforms that are likely to produce party systems more responsive to the changing demands of voters.
11. I am not arguing that central bank reform and electoral reform necessarily occur together. Both events are far too contingent on a variety of other factors for that to be the case. But if they share a similar underlying cause, then these events are likely to coincide in at least a number of cases.
12. The economic reforms instituted by the Labour government had exacerbated existing divisions within the National Party. Consequently, the new National government had a “somewhat unwieldy caucus of backbenchers with widely differing views on the economy” (Keesing’s Archives 1991, 37 (11): 38643).
13. In 1993, the Citizen Initiated Referenda Act also gave voters the right to petition for a (nonbinding) referendum to be held on a particular issue, another departure from traditional Westminster political institutions (Mulgan 1994).
1. Accounts of the Maastricht Treaty abound. Some useful discussions include Eichengreen (1993a), Eichengreen and Frieden (1994), Gros and Thygesen (1998), McNamara (1998), Moravscik (1998), Sandholtz (1993).
2. The debates surrounding the economic effects of EMU have produced a voluminous literature. For helpful reviews, see Eichengreen (1993a; 1997), Gros and Thygesen (1998), and Wyplosz (1997), as well as special issues of Economic Policy (EMU 1998) and the Oxford Review of Economic Policy (Allsopp and Vines 1998).
3. Prior to the advent of the single currency, both the European Monetary Institute (the forerunner of the European Central Bank) and the European Commission would have to issue reports to the Council of Ministers recommending countries for participation based on their compliance with the convergance criteria.
4. For individual level analyses of support for the euro, see Gabel (1999), Kaltenthaler and Anderson (1999) and Gartner (1997).
5. The treaty could have come into effect if it had been approved by a five-sixths majority of the Danish parliament. Although the major parties supported the treaty, they could not assemble such a majority, sending it to a referendum.
6. In October 1992, the Danish government renegotiated the treaty, securing guarantees for Danish autonomy in the areas of defense, citizenship, and the single currency.
7. The possibility of currency speculation prevented member-state governments from justifying a currency realignment to their publics in advance of any action. If governments had attempted to prepare public opinion for a general realignment, their statements would have unleashed an overwhelming wave of speculation in the exchange markets.
8. In an anti-EMU article, Martin Feldstein draws a similar conclusion: “What is clear to me is that the decision [to adopt the euro in 1999] will not depend on the economic advantages and disadvantages of a single currency. The decision of whether or not to form a monetary union will reflect deeply held political views about the appropriate future for Europe and the political advantages and disadvantages to the individual countries and even to the individual political decisionmakers themselves” (Feldstein 1997, 23).