Political economists have long noted the association between federalism and central bank independence (e.g., Banaian, Laney, and Willett 1983; Lohmann 1998a; Posen 1995). Federal systems, in which political authority is divided between the central government and constituent units, tend to have independent central banks. Germany, Switzerland, and the United States—all federal systems—have independent central banks. Unitary countries, in which political power is centralized in the national government, were more likely to have dependent central banks during the 1970s and 1980s. Although the association between federalism and central bank independence appears to be strong, the mechanism linking the two institutions remains unclear. Why do we observe a correlation between federalism and independence? How does federalism affect the choice of central bank institutions?
I contend that the association between federalism and central bank independence reflects the potential for intraparty conflicts over monetary policy. Federal systems increase the potential for intraparty conflict over monetary policy, giving politicians incentives to choose an independent central bank. In federal systems, party politicians are likely to face different monetary-policy pressures. Constituents will make a variety of economic policy demands, reflecting regional differences in economic activity. Additionally, the political institutions of federalism—powerful regional offices, bicameral legislatures, and so on—force party politicians to appeal to different sets of constituents and to run for office at different times. Because the possibility of intraparty conflict over monetary policy is high in federal systems, the political benefits of an independent central bank are higher as well.
In this chapter, I provide empirical support for this argument. The first section identifies factors that condition the nature of the relationship between backbench legislators, coalition partners, and cabinet ministers in the area of monetary policy. The second section examines episodes of central bank choice and reform in Germany and Britain to illustrate how federalism shaped the potential for intraparty conflict and, in turn, the choice of central bank institutions. The third section tests the argument statistically against the cross-national variation of central bank independence in the 1970s and 1980s.
The choice of central bank institutions reflects the potential for intraparty conflict over monetary policy. An independent central bank can prevent some of these potential conflicts by acting as a check on the cabinet’s discretion over policy. An independent central bank not only insulates monetary policy from day-to-day political manipulation, but it can also publicize its policy conflicts with the government, providing backbench legislators and coalition partners with information about the cabinet’s policy choices and the consequences of those choices. In systems where conflicts over monetary policy are likely, therefore, politicians will choose an independent central bank. In systems where conflict between backbenchers, coalition partners, and government ministers is less likely, politicians will opt for a dependent central bank.
What factors contribute to the potential for conflict over monetary policy? First, party legislators and coalition partners will be more suspicious of the cabinet if cabinet ministers possess policy incentives that differ from their legislative supporters and coalition partners. In this situation, backbench legislators and coalition partners recognize that the government is likely to pursue policies that do not reflect their interests. Second, political conflict is likely where backbench legislators and coalition partners can credibly threaten to punish the government for its monetary-policy performance.
The factors that shape the relationship between backbench legislators, coalition partners, and government ministers reflect the configuration of electoral, legislative, and government institutions, as well as the distribution of constituent preferences between and across political parties. In particular, many of the institutions of federalism increase the possibility of conflict over monetary policy.
The divergence of monetary-policy incentives between legislators, parties, and ministers contributes to the potential for conflict. As policy incentives increasingly diverge, backbench legislators and coalition partners become less trustful that government ministers will pursue policies in their interests. Consequently, they will prefer an independent central bank. If legislators, coalition partners, and government ministers face similar policy incentives, however, conflict over monetary policy is less likely. Backbench legislators and coalition partners will trust the cabinet and consequently will see no need for an independent central bank to check the government’s policy discretion.
In the area of monetary policy, the distribution of constituent preferences and electoral institutions conditions the degree of incentive divergence between legislators, parties, and ministers.
Constituent Preferences
The distribution of monetary-policy preferences among party supporters may force candidates from the same party to make different appeals to win votes. Manipulating monetary policy can help some politicians in the party. But if the governing party(ies) has constituents with diverse preferences over monetary policy, changes in monetary policy may generate externalities among other constituents (e.g., an interest-rate increase may help creditors but hurt debtors), damaging the electoral fortunes of other candidates within the party. This problem may be particularly acute for large catch-all parties or parties not based on economic appeals (Budge and Farlie 1983; Kirchheimer 1966).
For example, higher interest rates help exporters and net creditors by keeping inflation low but increase the load on net debtors. If the governing party appeals to both exporters and net debtors, raising interest rates may improve the party’s popularity in some areas but hurt its electoral chances in others. This tension occurred in Britain during the early 1990s, damaging the popularity of the incumbent Conservative government. During the 1980s, Prime Minister Thatcher sold public housing to the current occupants, creating a new class of home owners and potential Conservative voters (Garrett 1992). Most of these mortgages were variable rate; that is, the interest payments fluctuated according to the current interest rate. During the early 1990s, the Conservative government increased interest rates in order to preserve the pound’s position in the EMS. These increases squeezed private debtors, particularly working-class individuals who had recently purchased their own homes. These new home owners, once grateful to the Conservatives, soon turned hostile as the government’s monetary policy saddled them with exorbitant mortgage payments.
Electoral institutions may also provide politicians from the same party with divergent monetary-policy incentives. Parties that compete across differently sized electoral districts are likely to have conflicts over monetary policy. Candidates in smaller districts may want to use monetary policy to appeal to particular interests, while candidates in larger districts must balance demands from a variety of constituents (Rogowski 1987).
Additionally, nonconcurrent elections for the government and part of the legislature provide politicians within the governing party(ies) with different incentives over the sequencing of monetary policy. For instance, the government may manipulate monetary policy in order to engineer an economic boom for its election period. The eventual increase in inflation resulting from a monetary surprise, however, may adversely affect legislative elections in the next period (Alesina and Rosenthal 1995).
Finally, electoral institutions may encourage the formation of coalition governments. The existence of coalition governments creates a higher potential for monetary-policy conflicts within the governing parties. If coalition partners face constituents with different preferences over monetary policy, they will disagree about the proper course for policy. Cabinet ministers must therefore make policy choices that will not alienate either their own legislative party or their coalition partner.
If the government’s legislative and coalition supporters can punish the government—if, for example, a policy consensus exists within the governing party(ies)—then the government has an incentive to demonstrate that its policy choices reflect the interests of its party legislators and coalition partners. It does not want backbench legislators or coalition partners to withdraw their support even if its policy choices unexpectedly lead to unfavorable outcomes. In this situation, the government has incentives to support an independent central bank.
If the government’s principals cannot levy a punishment on the government, however, the government has little incentive to support institutions that enhance the transparency of its policy choices. Legislators and coalition partners, for instance, may lack the capacity to organize themselves to punish the government. Or they may accept the government’s policy choices because the expected outcome of challenging the government is worse than the current policy. Finally, government ministers may possess the (institutional and persuasive) ability to build new majorities in favor of the outcomes they produced (McCubbins, Noll, and Weingast 1989). At the very least, they may have the capability to prevent their legislative and coalition supporters from organizing themselves to punish the government. If the government’s principals cannot punish the government, the government will not want to limit its policy flexibility by supporting institutions that supply information to its legislative and coalition supporters. Instead, the government will support a dependent central bank.
The credibility of backbench legislators’ and coalition partners’ threats to punish the government reflects the polarization of the political system, the organization of legislative institutions, and the existence of coalition and minority governments.
Polarization
The polarization of the system affects the ability of backbench legislators and coalition partners to punish the government over a policy dispute. Dissatisfied backbenchers and coalition partners must cooperate with at least part of the opposition to challenge the government’s policy. Highly polarized systems present few opportunities to cooperate with the opposition because of the large policy distance between parties (Sartori 1976). Consequently, backbench legislators and coalition partners will be less likely to challenge the government’s policy. In systems with low polarization, however, backbenchers and coalition partners will be more willing to craft a temporary legislative coalition with the opposition to defeat the government.
Legislative Institutions
Legislative institutions influence the possibility of punishment by affecting the cost to backbench legislators and coalition partners of challenging the government’s policy. In systems with endogenous electoral timing, dissatisfied party legislators and coalition partners must consider the possibility that a challenge to the government’s policy could precipitate early elections (Huber 1996). The possibility of losing their seats in a general election may provide them with an incentive not to challenge the government. In systems with strict party discipline, backbenchers face another disincentive to challenge the government because party leaders possess the institutional authority to punish legislators who fail to support the government. Finally, if the political fortunes of legislators are closely tied to their party’s participation in the government, party legislators may be reluctant to risk their status by challenging government policy. In contrast, in systems where the benefits of serving in the governing party are low—if, for example, opposition legislators can still influence policy through a strong committee system—then party legislators may be more willing to veto government policy (Strom 1990a, 1990b).
Coalition and Minority Governments
In multiparty parliamentary systems, the government faces the threat of punishment not only from party legislators but also from coalition partners. The existence of coalition or minority governments increases the potential for punishment. Multiparty coalition governments are more likely to fall over a policy dispute than are single-party majority governments.
The political systems of Germany and Britain illustrate how legislative-cabinet relations and the potential for intraparty conflict over monetary policy shape the choice of central bank institutions. Although specific events may precipitate an episode of central bank reform, the relationship between legislators, coalition partners, and government ministers provides the underlying conditions that shape the choice and stability of central bank institutions.
The divergence of policy incentives among German politicians and the possibility that the government will be punished over a policy dispute help ensure the commitment to an independent Bundesbank. The government’s legislative and coalition supporters depend on the bank to help them check the government’s informational advantages in monetary policy. The government relies on the bank to help soothe potential conflicts with its principals—backbench legislators, coalition partners, and voters—allowing the government to retain its position in office.
Backbench legislators and coalition partners possess monetary-policy incentives that differ from those of the government, due in part to Germany’s federal system. As a consequence, the government’s legislative and coalition principals cannot trust the government to pursue the monetary policies they desire. First, each party’s constituents possess a range of interests over monetary policy, reflecting the regionalization of the German economy. Because Land party organizations control the party list for elections to the lower house, members of parliament must balance Land interests with the demands of the government to ensure their renomination.
Second, the bicameral legislature creates the potential for conflict between the government, selected in the lower house (Bundestag), and party legislators in the upper house, the Bundesrat. Legislators in the Bundesrat are appointed representatives of the Land governments. Under certain circumstances, the Bundesrat may veto legislation (Conradt 1989; Lohmann 1998a).1 These veto powers prevent the government from proposing legislation that will enhance its own authority at the expense of the Lander. Because the government needs a majority in the Bundesrat to pass partisan legislation, it cannot enact policies that could damage the party’s electoral fortunes at the Land level. In particular, nonconcurrent elections for the Bundestag and Land governments provide party politicians with different incentives over monetary-policy sequencing. If the government attempts to manipulate monetary policy to engineer an economic boom for its own election, it may produce outcomes that hurt the party’s electoral fortunes at subsequent Lander elections.
Finally, the proportional representation electoral system strongly discourages the formation of single-party majority governments. Throughout the postwar period, Germany has been governed by a coalition government. Coalition partners often have constituents with different and sometimes conflicting preferences, as illustrated by the tension between the Social Democrats and the Free Democrats in the late 1970s and early 1980s (Goodman 1992; Scharpf 1987). An independent Bundesbank helps prevent the government from pursuing policies that will hurt their party legislators and coalition partners.
An independent Bundesbank can also help the government stay in office by bolstering the credibility of its economic policies. Because monetary policy remains a high-salience issue in Germany, the government’s principals—backbench legislators, coalition partners, and ultimately, voters—use it to gauge the government’s competence. The nature of the political system enhances the government’s vulnerability on this dimension. First, the relative proximity of the major parties on the economic and monetary-policy dimension contributes to the possibility of punishment. Although each party espouses different policy priorities, their economic programs reflect a German consensus about the proper balance between market and state. The Christian Democrats and the Free Democrats share a procapitalist orientation. But the Christian Democrats also implemented the “social market” economy, designed to ameliorate the less-desirable aspects of capitalism. Unlike other Socialist parties in Europe, the Social Democrats moderated their economic policies relatively early after World War II. Frozen out of national office for over a decade, the Social Democrats rejected radical Marxism in the Bad Godesberg program of 1959, attempting to make themselves a more palatable alternative to Christian Democratic rule. Participation in the Grand Coalition in the late 1960s gave the Social Democrats a reputation as a stable center-left party capable of governing Germany.
Second, the coalition dynamics of the German party system also increase the possibility that the government will lose office over a policy dispute. As the pivot party in the party system for much of the postwar period, the small Free Democrats have participated in coalitions with both the Christian Democrats and the Social Democrats. Consequently, the Free Democrats could threaten to leave the government if their senior partner did not comply with the coalition agreement.
The Bundesbank’s independence can prevent party legislators, the Free Democrats, and voters from defecting from the governing coalition. During the 1970s, the Social Democrats relied on the policy credibility of the Bundesbank to forge a coalition with the Free Democrats, despite their differences on economic policy. Although many blame the Bundesbank for exacerbating tensions between the two parties in the early 1980s (e.g., Scharpf 1987), the bank contributed to the coalition’s longevity by helping assuage the Free Democrats’ concerns about the influence of the Social Democrats’ left wing. Social Democratic ministers therefore had an incentive to tolerate an independent central bank in the 1970s. In contrast, Christian Democratic governments have less incentive to support an independent central bank. Because they share similar economic priorities with the Free Democrats, the Free Democrats are less likely to withdraw their support over an economic policy dispute. Consequently, the Christian Democratic governments have less need to rely on an independent central bank to verify their policy performance.
Indeed, Christian Democratic governments have twice attempted to increase the central government’s authority over the Bundesbank. During the founding of the Bundesbank in the 1950s and the reorganization of the Bundesbank in the 1990s, the Christian Democratic–led government tried to enhance its influence over the composition of the central bank’s main decision-making body. In both instances, mistrust between the government and its party supporters in the Bundesrat helped ensure the Bundesbank’s independence.
The Founding of the Bundesbank
After World War II, the Allies created a decentralized central banking structure consisting of legally autonomous Land central banks coordinated by the Bank deutscher Lander (BdL). Article 88 of the 1949 German Basic Law required the new German government to reorganize this central bank system, touching off a prolonged debate about its institutional design. The appointments procedure to the new central bank emerged as a key issue (Berger 1997; Goodman 1992; Lohmann 1994). One side called for a centralized system, where a directorate composed only of central government appointees would determine monetary policy. Proposals for a more decentralized system sought to retain the BdL’s federal structure and increase the influence of the Land central bank presidents, appointed by the Land governments, in the policy process.
The issue divided the Christian Democratic–led coalition government. After heated internal debate, the cabinet eventually agreed to a decentralized plan, sending it to the legislature (Lohmann 1994). The Bundesrat, composed of Land representatives, easily endorsed this decentralized proposal, with support from both Christian Democratic representatives and the Social Democratic opposition. The Social Democrats’ support for the plan reflected in part their lack of strength at the national level. Frozen out of power at the national level but with significant representation in the state governments, the Social Democrats perceived a decentralized structure as the best way for them to influence monetary policy (Goodman 1992). Opposition to the bill, however, came from the Free Democrats, who had weak support at the state level. They believed that a centralized structure would afford them the best opportunity to affect policy. Consequently, the Free Democrats introduced a counterproposal for a centralized bank to the Bundestag. Legislators referred the competing bills to committee, where the legislation stalled (Lohmann 1994).
In 1955–56, the BdL pursued a restrictive monetary policy, initiating a conflict with the government (Berger 1997). Upset with the BdL’s policies, Chancellor Adenauer publicly criticized the bank. The government called for a new centralized structure, which would integrate the bank with other policymaking institutions. The Bundestag passed the government’s motion, but the Bundesrat rejected it and reintroduced the proposal for a decentralized structure. With a majority in the Bundesrat, the Christian Democrats may have been expected to support a centralized plan that would increase the influence of the national party over monetary policy. Christian Democratic representatives in the Bundesrat, however, were motivated both by the desire to maintain state influence over monetary policy and by the need to place an institutional check on their own party’s government because they had different incentives over monetary policy.
The competing proposals went again to a Bundestag committee. This time, the committee produced a compromise, outlining a Central Bank Council composed of up to 10 central government appointees and 11 Land central bank presidents, chosen by their respective state governments. Proponents argued that the majority status of Land central bank presidents on the council provided a vital safeguard against the ability of the central government to manipulate policy. The proposal unanimously passed the Bundestag and overwhelmingly passed the Bundesrat, with support from states dominated by both the Christian Democrats and the Social Democrats (Lohmann 1994). The divergence of policy incentives between the government and party legislators in the Bundesrat therefore helped to guarantee the Bundesbank’s independence.
The Reorganization of the Bundesbank
In the early 1990s, the Bundesbank was reorganized to incorporate the eastern Lander into the system. Again, the composition of the Central Bank Council surfaced as one of the central issues. Advocates of centralization, including the Christian Democratic–led government and central government appointees to the Bundesbank, argued that the addition of new Land central bank presidents would render decision making cumbersome and unwieldy. Opponents of centralization, including most Land central bank presidents, argued that the majority status of Land central bank presidents on the council ensured the Bundesbank’s independence (Henning 1994; Kaltenthaler 1996; Lohmann 1994, 1998a).
The Bundesbank president, Karl Otto Pohl, originally proposed a Central Bank Council composed of seven government appointees and seven Land central bank presidents, with the president (appointed by the central government) as a tiebreaker. The Bundesbank Council rejected this proposal as too centralized. Pohl next proposed reducing the number of Land central banks to eight, with no change in the number of central government appointees (up to 10). By specifying which states would maintain representation on the council, Pohl was able to buy off some Land central bank presidents and forge a winning coalition within the council. In an unusual move, Land central bank presidents who voted against the proposal protested by sending a letter to Chancellor Kohl requesting that each Land retain a seat on the Central Bank Council. This opposition included Land central bank presidents who had been nominated by both Social Democratic and Christian Democratic state governments.
Anticipating opposition from the Bundesrat, the government proposed a Central Bank Council composed of nine Land central bank presidents and eight government appointees, increasing the relative influence of central government appointees but preserving the majority status of the Land central bank presidents. The Bundesrat, however, rejected that proposal by a two-thirds majority, which included some representatives from Christian Democratic states, and reiterated its preference for the “one Land, one Land Central Bank” principle. The government countered by promising key states that they would retain a Land central bank, eventually reducing the opposition in the Bundesrat to a simple majority. Under German law, this simple majority could not block the Bundestag’s passage of the bill. Although the reform increased the relative influence of central government appointees, the opposition from party legislators in the Bundesrat forced compromises that preserved the Bundesbank’s independence.
Germany’s federal institutions therefore strongly shaped the Bundesbank’s independence. In particular, the Bundesrat has consistently opposed any increase in the central government’s control over the central bank. Although Bundesrat legislators may come from the same party as the cabinet, they have different monetary-policy incentives than their lower house counterparts. Consequently, they cannot trust the cabinet to pursue policies in their interests. As a result, they supported an independent central bank to act as a check on the government’s discretion.
Throughout the postwar period, Britain possessed a dependent central bank, firmly under government control. The Bank of England’s dependent status reflected the similarity of monetary-policy incentives between party legislators and government ministers. That similarity in turn stems from the relative homogeneity of intraparty constituent demands and Britain’s majoritarian institutions.
Until the 1970s, the Conservative and Labour parties each appealed to constituents with fairly homogenous preferences over monetary policy (Hall 1986). The Conservatives represented business and middle-class interests. Labour’s main constituents included the trade unions and the working class. Although for much of the postwar period British economic policy reflected a consensus of the two parties, each party had different economic priorities. The Conservatives stressed controlling inflation. Labour emphasized unemployment.
The Westminster political system reinforced the ideological coherence between party legislators and the government (Lijphart 1984). The majoritarian electoral system helps to ensure single-party majority governments, preventing coalition governments. Because Parliament is essentially a unicameral legislature, party politicians do not face nonconcurrent elections, giving them similar incentives over policy sequencing. Additionally, the unitary nature of the British system meant that the governing party did not have to win regional and local elections to implement its legislative agenda. Backbench legislators and government ministers, therefore, had similar incentives over monetary policy. Consequently, backbenchers had no reason to support an independent central bank. Instead, they could trust their government ministers to pursue policies that reflected their interests.
Government ministers also had little incentive to support a more independent central bank. The legislative and electoral institutions of a Westminster system make it highly unlikely that party legislators would withdraw their support over a policy dispute. Although party legislators possess the formal authority to dismiss the government over policy disputes and conflicts, backbenchers face a number of institutional disincentives to prevent them from doing so. First, the benefits of serving in the governing party are high. Because the legislative committee system is weak, opposition legislators possess little influence over policy or distributive benefits (Strøm 1990b). Second, governments typically have the authority to call for new elections, forcing rebellious legislators to risk their own seats at an early election (Huber 1996). Third, party leaders control a variety of electoral and party resources, allowing them to punish legislators who fail to support the government’s agenda. Because of these costs, governments rarely have to face challenges from within their own party, regardless of the opposition’s position. Consequently, government ministers had little incentive to support an independent central bank.
The 1946 Bank of England Bill
The postwar institutions of the Bank of England were largely determined in 1946. During the immediate postwar period, the Labour government, under Clement Attlee, sought to reshape the relationship between market and state. With planning and government management of the economy, the Labour Party hoped to achieve full employment and greater social equality. Their program involved more active use of macroeconomic policy; expansion of social services, including establishment of the National Health Service and the Family Allowance; and nationalization of vital economic institutions, including the coal, electric, gas, rail, and steel industries. Of these nationalization proposals, only the steel sector proved controversial.
As part of this program, the new Labour government proposed nationalizing the Bank of England in 1945. Since its founding in 1694, the bank’s stock had been held privately. Its governor and the Court of Directors had been selected from within the bank or especially recruited by the bank. Although there had been policy cooperation between the Bank of England and the Treasury in the twentieth century, these relations were informal and uncodified. In particular, the bank’s performance relied heavily on the governor’s executive ability. During the interwar years, the arrangement worked well under the forceful leadership of Montagu Norman. Under Norman, the bank accepted the Treasury’s control over policy. Norman once declared, “I am an instrument of the Treasury” (Fforde 1992, 15). In the management of its affairs, however, the bank considered itself “operationally and institutionally distinct from the Government” (Fforde 1992, 15).
The Labour Party had already espoused nationalizing the Bank of England during the 1930s. Many Labourites blamed Norman for failing to cooperate with MacDonald’s Labour government and contributing to Labour’s defeat in the 1931 elections. The 1937 manifesto Labour’s Immediate Programme argued that the bank must be nationalized in order to integrate monetary policy with plans for a national investment policy. Labour also discussed proposals to nationalize joint stock banks to help rationalize investment. The 1945 manifesto simply stated that “The Bank of England with its financial powers must be brought under public ownership, and the operations of other banks harmonised with industrial needs.”
The actual Bank of England Bill was limited and technical in nature, reflecting extensive consultation between the new Labour government and the bank during its preparation. The bill established a new Court of Directors, appointed by the chancellor, to direct monetary policy. This Court of Directors consisted of 16 individuals, including the governor and deputy governor. The court would serve five-year terms, although bank officials had requested seven-year terms (Fforde 1992). Clause 4(1) of the bill authorized the government to give directives to the bank, as “they think necessary in the public interest.” Further, Clause 4(3) stated that the bank had the power to direct commercial banks, if it considered it “necessary in the public interest” and the action was “authorised by the Treasury.” Finally, the bill discussed plans to compensate owners of the private stock. It made no attempt to nationalize the joint stock banks.
Discussion of the bill centered largely on two issues: Clause 4(3) and the terms of compensation for shareholders. Clause 4(3) proved most contentious. Critics charged that it was vague and gave too much authority to the Bank of England over commercial banks. They contended that the wording of the clause might allow the government to compel banks to reveal information about the private business of their clients. The government countered with amendments designed to clarify the purpose of the amendment and protect the rights of bank customers (Fforde 1992).
The terms of compensation for private shareholders also generated some challenges. Although none of the stockholders petitioned against the terms of compensation, a Commons Select Committee held hearings on their fairness. Appearing before the Select Committee on the Bank of England Bill, Treasury counsel Cyril Radcliffe went to great lengths to defend the compensation scheme (Commons 1945). This task was made more difficult because neither the chancellor nor the bank governor wanted to disclose the exact value of the bank’s assets. Some members of the Labour Party complained that shareholders were overcompensated, but a party conference motion critical of the compensation terms was defeated (Craig 1982).
Overall, the plans for nationalization were relatively uncontroversial and moved through Parliament quickly (see, for example, Fforde 1992; Morgan 1984; Thompson 1996).2 Most saw the proposed bill as formalizing an already close working relationship between the Treasury and the Bank of England, a working relationship that had become even closer during the war effort. Hugh Dalton, Labour’s Chancellor of the Exchequer, quipped that “the Old Man of the Treasury and the Old Lady of Threadneedle Street [the Bank of England], who had for some time been living in sin, were now to be married” (Pearce 1994).
Interestingly, the appointments procedure for the bank aroused little discussion. Radcliffe argued that although arrangement between the bank and the Treasury had worked well, “this is the time when what commonsense would require as the satisfactory structure as between the Bank and the Government should be put into legal form, and not left to working out by what I will call ‘satisfactory chance.’ . . . The time has come when the Government should avow its right to appoint the Governor, the Deputy-Governor, and the Court of Directors of this institution” (Commons 1945). During debate in the House of Lords, Lord Pethick-Lawrence pointed out that under the bill, policy decisions would be subject to parliamentary questioning, whereas in the current system, Parliament’s authority over the bank was unclear.
The bill received only token opposition from the Conservatives. Winston Churchill, leader of the opposition, declared that he would not in principle oppose the bill. Most Conservative critics maintained that the present system had worked well in the past and that there was little reason to change it. These arguments did little to challenge Labour’s bill. Indeed, Robert Boothby, a Conservative who had regularly criticized Norman, actually voted with the Labour government in support of the measure (Morgan 1984).
The ease with which the bill passed reflected the similarity of policy incentives faced by party politicians in Britain and the ideological coherence of the Labour Party at that time. Party legislators could trust their ministers to pursue monetary policy that benefited the party as a whole. Consequently, they had little incentive to limit the government’s authority over the bank.
Because the relationship between backbench legislators, coalition partners, and government ministers strongly influences the choice of central bank institutions, variations in that relationship across systems can explain cross-national differences in central bank independence. This section statistically tests the relationship between incentive divergence, the threat of punishment, and central bank independence.
Incentive Divergence
The degree of incentive divergence between legislators, parties, and ministers reflects both the distribution of constituent preferences and the configuration of electoral institutions.
Constituent Preferences. To measure the heterogeneity of constituent preferences over monetary policy, I employ the Alford index (Alford 1963). 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). High values reflect a high incidence of class voting. Low values indicate that class is a less-salient predictor of vote. Because low values imply that party politicians are likely to face constituents with a variety of preferences over monetary policy, the Alford index should have a negative relationship with central bank independence.3
Electoral Institutions. Politicians from the governing party(ies) are likely to compete in differently sized districts or at different times in systems with “strong” bicameral legislatures. Lijphart (1984) defines strong bicameralism as a situation in which both houses have roughly equal legislative powers and are selected according to different electoral rules. I use a dummy variable for countries with strong bicameralism: Australia, Germany, Switzerland, and the United States. Systems with strong bicameral legislatures should have more independent central banks.
The Threat of Punishment
The credibility of the threat of backbench legislators and coalition partners to punish the government reflects the polarization of the political system, legislative institutions, and the existence of coalition and minority governments.
Polarization. Following Powell (1982), I measure polarization as the average percentage of electoral support for extremist parties for the years 1960 to 1990. Because high levels of polarization imply that party legislators and coalition partners will be less willing to punish the government, polarization should be negatively related to central bank independence.4
Legislative Institutions. To capture the influence of legislative institutions, I use a dummy variable for systems with “strong and inclusive” committee systems (Powell and Whitten 1993; Strom 1990a). Because party legislators and coalition partners are more likely to withdraw their support from the government in systems with strong and inclusive legislative committees, the committee variable should have a positive relationship with central bank independence.5
Coalition and Minority Government. I calculated the proportion of time a country was governed by a coalition or minority government from 1960 to 1990. In systems with a high proportion of time under a coalition or minority government, the central bank will be more independent.6
Table 2 summarizes the independent variables and their sources. I expect independent central banks in systems with moderate Left parties (low Alford index), strong bicameralism, low polarization, strong committees, and regular coalition or minority governments.
The sample includes the 18 countries listed in table 1. Data availability and reliability limited the sample.
For the dependent variable, central bank independence, I used all four indices of central bank independence listed in table 1.7 I report the results from only the average measure. Unless noted, results for all indices were similar. I used ordinary least squares (OLS) regression. I performed all calculations using Stata 5.0.
The results of model I, reported in table 3, are partially consistent with the incentive divergence and potential punishment hypotheses. The Alford index has a negative, significant effect on central bank independence. Systems in which the Left parties have heterogeneous class support possess more independent central banks.8 Systems with strong bicameral legislatures also possess more independent central banks. None of the indicators associated with potential punishment, however, is significant, although each coefficient is in the predicted direction.9 Polarization, committee strength, and coalition/minority government are highly correlated. Although each is not individually significant, a Wald test indicates that the three variables are jointly significant at the 0.01 level of significance (F(3, 12) − 5.17). Model I also passes all diagnostic tests, indicating that the residuals are not serially correlated10 (Durbin-Watson), are normally distributed (skewness and kurtosis), and are homoscedastic (Cook and Weisberg 1983). Additionally, there is no evidence of omitted variable bias (RESET).11
Given the joint significance of the polarization, committee strength, and coalition/minority government variables, I created a composite indicator for the threat of punishment by adding (1–polarization), committee strength, and coalition/minority government. I expect this indicator to have a positive relationship with central bank independence. The results from this equation, model II, support the hypothesis. The punishment index is significant and in the predicted direction. The Alford index and strong bicameralism variables remain significant and in the predicted direction. Model II explains 74 percent of the variance in the dependent variable for this sample. The model also passes all diagnostic tests.
TABLE 2. Independent Variables
|
|
|
|
|
Coalition/ |
|
Australia |
0.25 |
1 |
0.00 |
0 |
0.00 |
1.00 |
Austria |
0.14 |
0 |
0.01 |
1 |
0.55 |
2.54 |
Belgium |
0.20 |
0 |
0.17 |
1 |
1.00 |
2.83 |
Britain |
0.33 |
0 |
0.02 |
0 |
0.02 |
1.00 |
Canada |
0.01 |
0 |
0.00 |
0 |
0.32 |
1.32 |
Denmark |
0.34 |
0 |
0.19 |
1 |
1.00 |
2.81 |
France |
0.16 |
0 |
0.23 |
0 |
0.93 |
1.70 |
Germany |
0.17 |
1 |
0.02 |
1 |
1.00 |
2.98 |
Ireland |
0.13 |
0 |
0.02 |
0 |
0.61 |
1.59 |
Italy |
0.15 |
0 |
0.36 |
0 |
1.00 |
1.64 |
Japan |
0.15 |
0 |
0.15 |
0 |
0.20 |
1.06 |
Netherlands |
0.15 |
0 |
0.06 |
1 |
1.00 |
2.94 |
New Zealand |
0.32 |
0 |
0.01 |
0 |
0.00 |
0.99 |
Norway |
0.28 |
0 |
0.09 |
1 |
1.00 |
2.91 |
Spain |
0.20 |
0 |
0.08 |
0 |
0.00 |
0.92 |
Sweden |
0.33 |
0 |
0.05 |
1 |
1.00 |
2.88 |
Switzerland |
0.19 |
1 |
0.07 |
1 |
1.00 |
2.93 |
United States |
0.13 |
1 |
0.00 |
0 |
0.60 |
1.60 |
aComputed as: (1 – polarization) + committee strength + proportion of coalition/minority government. The calculation of Austria’s punishment index, for example, is: (1 – 0.01) + 1 + 0.55 = 2.54.
Given the importance of central banks for economic performance, political economists have offered several other explanations of the variation of central bank independence.
Tying the Hands
McCubbins, Noll, and Weingast (1989) argue that politicians structure bureaucracies to guarantee that their policy goals will not be overturned in the future. Similarly, Goodman (1991) claims that governments that prefer low inflation and that expect a short tenure in office will adopt an independent central bank to limit the ability of future governments to manipulate economic policy. The “tying-the-hands” argument suggests that highly polarized systems or systems with highly antagonistic parties should have independent central banks. Systems with moderate centrist parties should have dependent banks because politicians can trust the opposition to pursue similar policies. The results from models I and II shown in table 3 contradict this hypothesis. Broad-based moderate Left parties (low Alford index) and low levels of polarization are associated with high levels of central bank independence.
TABLE 3. OLS Estimation of Central Bank Independence
(Dependent Variable: Mean Central Bank Independence)
A second group argues that economic openness influences the level of central bank independence (Maxfield 1997). An open economy, dependent on international trade and foreign capital, requires an independent central bank. An independent central bank can help prevent inflation from eroding the competitive position of the economy, especially in countries that do not have floating exchange rates. An independent central bank also helps attract foreign capital and investment by demonstrating the host government’s commitment to price stability (Maxfield 1997).
I tested the effect of economic openness in two ways. First, I included trade as proportion of GDP for the years 1970 to 1989. Model III in table 3 indicates that trade has no statistically significant effect on central bank independence.12 Second, I included a measure of capital account openness based on Quinn (1997), computed by averaging his measure for the period 1973 to 1988. This measure has no statistically significant effect on central bank independence (model IV).
Partisanship
Political economists have also argued for a link between government partisanship and central bank structure. Right parties are traditionally more concerned with controlling inflation, whereas Left parties place more emphasis on employment and wealth redistribution (Alesina 1989; Alesina and Sachs 1988; Havrilesky 1987; Hibbs 1987). Consequently, Right parties will institute an independent central bank to counter inflationary pressures in the economy. Left parties will prefer a dependent bank, which allows them to manipulate monetary policy to enhance growth and employment.13 This argument, however, suggests that new governments would reform the bank to suit their economic policy goals, implying more cross-temporal variation in central bank independence than is empirically observed.
To test the hypothesis, I created a measure of Left government strength based on Cameron (1984). The measure multiplies the percentage of cabinet seats held by Left parties by the percentage of a legislative majority held by Left parties in the legislature for each year and then averages those measures across the time period. Model V in table 3 reports the results of including the partisanship measure for the years 1970 to 1989. No statistically significant relationship between partisanship and central bank independence exists.14
Interest Groups
A fourth set of explanations centers on interest-group pressure as a determinant of bureaucratic structure (Moe 1989, 1990). Both Posen (1993, 1995) and Maxfield (1994) argue that the organization and political influence of a nation’s financial sector strongly determine the central bank’s institutional structure. Posen argues that systems in which the financial sector can organize an effective opposition to inflationary policies will have independent central banks.
To test this hypothesis, I used Posen’s measure of financial sector strength, which is based on the presence of universal banking, the central bank’s role in the regulation of the financial sector, federalism, and party fractionalization at the national level (model VI in table 3).15 This measure of financial sector strength does have a positive, significant relationship with central bank independence. This model also passes all diagnostic tests. When financial sector strength was included with the variables from model II, however, the punishment index and financial sector index failed to attain significance because those two variables are correlated (r = 0.67; results not reported). However, the residual variance (SEE) is smaller in model II than in model VI (0.0893 versus 0.1191), suggesting that model II provides a more efficient prediction of central bank independence.
I also compared models II and VI with a parameter encompassing test (Granato, Inglehart, and Leblang 1996; Granato and Suzuki 1996). The parameter encompassing test indicates whether the substitution of one model’s parameters for another’s is statistically distinguishable. I used a joint F-test to test whether model II encompasses model VI (symbolically, model II ξ model VI). An insignificant statistic indicates that zero restrictions on model VI has no statistical effect on a joint model, composed of indicators from both models (i.e., model II ξ model VI). In contrast, a significant statistic signals that model II does not encompass model VI, meaning that model VI contains information that model II fails to capture. I also examined whether model VI encompasses model II. The results of the F-tests indicate that model II encompasses model VI and that model VI does not encompass model II (see table 4). Consequently, it is possible to conclude that the variables in model II possess more predictive power than does the strength of the financial sector.
Small N studies may suffer from a variety of statistical problems, including the possibility that parameter estimates are strongly affected by an individual case (Granato, Inglehart, and Leblang 1996; Jackman 1987). An observation’s influence reflects both its discrepancy and leverage. Discrepancy occurs when an observation is associated with a large residual. Standardized and studentized residuals indicate which observations possess unusually large residuals. In an examination of the residuals of model II, the United Kingdom appears to be an outlier because it has a standardized residual of 2.22, outside the usual cutoff of plus or minus 2.0 (see table 5). The standardized residuals also suggest that Belgium (–1.64) and Denmark (1.65) may be outliers, although neither exceeds the cutoff point.16
Leverage refers to the potential for the model as a whole to be influenced by a few cases. Two statistics, Cook’s Distance (Di) and DFFITS, measure leverage (Belsley, Kuh, and Welsch 1980). A Cook’s Distance (Di) greater than 4/n indicates an influential observation. According to this criterion, Britain (0.49), Australia (0.30), and Denmark (0.23) are influential. A related diagnostic, DFFITS, also points to the influence of Britain and Denmark. Both Britain (1.69) and Denmark (1.02) had DFFITS values greater than the 2*(k/n)1/2 cutoff recommended by Bollen and Jackman (1990).
Another diagnostic, DFBETAi, measures the effect of deleting one observation on each parameter estimate. An absolute value of DFBETAi greater than one indicates that the observation shifts the coefficient at least one standard error (Bollen and Jackman 1990).17 According to this criterion, Britain exerts undue influence on the parameter estimate for the Alford index (Alford DFBETAU.K. = 1.19). None of the DFBETAs for the other independent variables exceeds the cutoff, although the punishment index DFBETA for Britain is extremely close (punishment DFBETAU.K. = –0.99). The direction of these DFBETAs, however, indicates that Britain actually dampens the size of the coefficient for both the Alford index and the punishment index (i.e., pulls them both toward zero). That both these variables are significant even when Britain is in the sample lends credence to the results.
Overall, the diagnostics suggest that Britain and possibly Denmark are problematic cases. Does the existence of these outliers invalidate the parameter estimates of model II? Following Granato, Inglehart, and Leblang (1996), I handled these cases in a variety of ways. Table 6 compares the OLS results from model II in table 3 with a variety of estimation strategies. First, I have included a dummy variable for the influential cases (Britain and Denmark) in model 6a.18 Second, I deleted Britain and Denmark from the sample (model 6b). In both models, the independent variables continue to be significant and in the predicted direction. As expected, the parameter estimate of the Alford index becomes even more negative, while the coefficient of the punishment index increases slightly.
Model VI ξ Model II |
Form |
Test |
Form |
Model II ξ Model VI |
5.31* |
F(3, 13) |
Joint model |
F(l, 13) |
1.24 |
*p < 0.05
I also re-estimated model II using three nonparametric techniques. The first technique employs a variant of robust regression. Robust regression produces parameter estimates that perform well even if the data slightly violate basic OLS assumptions concerning normality (Western 1995). The method I employ begins by dropping any observation that has a Cook’s Distance Di > 1 (no observations in my sample meet this criterion). The remaining observations are weighted according to their residuals, using Huber weights and subsequently, biweights. Weighted least squares is then used to generate new estimates. The process is iterated until the maximum change in weights drops below 0.01. Interestingly, the robust regression places zero weight on Britain and a very small weight on the Danish case (table 5). Consequently, the estimate in model 6c is similar to model 6b.
As with OLS, however, robust regression is also susceptible to highly influential observations (Western 1995). To constrain the influence of such cases, I used a simple one-bounded influence estimator suggested by Welsch (1980) and employed by Granato, Inglehart, and Leblang (1996), which uses the DFITTS values from the original OLS regression to down-weight observations.19 The estimates, shown in model 6d, are similar to the estimates in the original OLS regressions.
Finally, I used a bootstrap technique to determine confidence intervals around the parameter estimates for each of the independent variables. Bootstrap resampling estimates the sampling distribution of a statistic, treating the sample as a population (Mooney 1996; Mooney and Duval 1993). I resampled with replacement the residuals of model II one thousand times. Model 6e presents the bootstrapped standard errors. Using the bootstrapped standard errors, I calculated the confidence intervals using the normal approximation method. For each independent variable, the 95 percent confidence interval does not span zero. The bootstrap confidence intervals therefore support the conclusions of the original OLS estimates in model II.
TABLE 5. Diagnostics and Case Weights
|
|
|
|
|
DFBETA |
DFBETA |
Australia |
–1.35 |
–1.39 |
0.30 |
–1.14 |
–0.39 |
–0.89 |
Austria |
0.73 |
0.71 |
0.02 |
0.29 |
–0.16 |
–0.12 |
Belgium |
–1.64 |
–1.75 |
0.10 |
–0.70 |
0.08 |
0.29 |
Britain |
2.22 |
2.66 |
0.49 |
0.16 |
1.19 |
–0.17 |
Canada |
1.21 |
1.23 |
0.21 |
0.93 |
–0.79 |
–0.25 |
Denmark |
1.65 |
1.77 |
0.23 |
1.02 |
0.69 |
–0.22 |
France |
–0.38 |
–0.37 |
0.00 |
–0.11 |
0.04 |
0.05 |
Germany |
0.48 |
0.47 |
0.02 |
0.32 |
–0.03 |
0.23 |
Ireland |
0.44 |
0.43 |
0.00 |
0.15 |
–0.08 |
–0.06 |
Italy |
–1.31 |
–1.35 |
0.04 |
–0.44 |
0.19 |
0.18 |
Japan |
0.00 |
0.00 |
0.00 |
0.00 |
–0.00 |
–0.00 |
Netherlands |
–0.02 |
–0.02 |
0.00 |
–0.01 |
0.00 |
0.00 |
New Zealand |
–0.86 |
–0.85 |
0.06 |
–0.50 |
–0.33 |
0.05 |
Norway |
–0.86 |
–0.85 |
0.04 |
–0.39 |
–0.15 |
0.12 |
Spain |
–0.41 |
–0.40 |
0.00 |
–0.17 |
–0.01 |
0.04 |
Sweden |
–0.34 |
–0.32 |
0.00 |
–0.18 |
–0.11 |
0.04 |
Switzerland |
0.47 |
0.46 |
0.02 |
0.30 |
–0.01 |
0.23 |
United States |
0.28 |
0.28 |
0.00 |
0.17 |
–0.42 |
0.14 |
aCutoff for influential observation: |Standardized Residuall > 2.0 (Bollen and Jackman 1990).
bCutoff for influential observation: |Studentized Residuall > 2.0 (Bollen and Jackman 1990).
cCutoff for influential observation: Di > 4/n. When N = 18, Di > 0.22 (Granato, Inglehart, and Leblang 1996).
dCutoff for influential observation: DFFITSi > 2*(k/n)½. When N = 18, DFFITSi > 0.94 (Bollen and Jackman 1990).
eCutoff for influential observation: DFBETAi > 1.0 (Bollen and Jackson 1990).
fRobust regression weights (Western 1995).
gBounded-influence weights (Welsch 1980).
TABLE 6. Diagnostics on Model II Dependent Variable: Central Bank Independence (Average)
Each of these techniques demonstrates that the parameter estimates obtained in model II are robust. The Alford index, strong bicameralism, and the punishment index retain their significance and remain in the predicted direction.
Political parties provide a mechanism to account for the correlation between federalism and central bank independence. In large federal systems, party politicians face a variety of incentives over monetary policy, increasing the potential for intraparty conflicts over economic and monetary policy. The policy demands of party constituents will likely reflect regional differences in the economy, creating differences in the policy preferences of party politicians. Additionally, federal institutions force party politicians to satisfy different sets of constituents or compete for office at different times. The existence of federal veto points in the policy process—such as an upper chamber to represent the interests of regional units or powerful regional offices—means that parties must enjoy electoral success across the country, even in regions with very different economic activity, in order to control the policy process.
The information asymmetries of the monetary-policy process, however, create potential conflicts between backbench legislators, coalition partners, and cabinet ministers. The severity of these conflicts conditions politicians’ incentives over the choice of central bank institutions. In systems in which politicians face different incentives over monetary policy—as in federal systems—and in which backbench legislators and coalition partners can credibly threaten to withdraw their support from the government, these potential conflicts can hurt the government’s ability to remain in office. Consequently, parties in federal systems must create institutions that will help them balance the different interests and policy demands of party politicians and their constituents—institutions that can help them stay in office longer as well as build and maintain a viable electoral coalition. An independent central bank can help prevent these potential conflicts by acting as a check on the cabinet’s discretion. As a result, party politicians have an incentive to choose an independent central bank.
Systems in which legislators, coalition partners, and government ministers share similar incentives over policy or in which the government’s position in office is secure have a low potential for intraparty conflict over monetary policy. Party politicians have less incentive to limit the cabinet’s policy discretion with an independent central bank.
The case accounts of central bank choice in Germany and Britain illustrate how these potential intraparty conflicts conditioned the choice of central bank institutions. In Germany, where federal institutions provided politicians with a variety of incentives over monetary policy, the Bundesrat vetoed plans to decrease the Bundesbank’s independence in the 1950s and the 1990s. In Britain, on the other hand, where party politicians shared similar policy incentives, the political control of the Bank of England was not an issue in the 1940s. The statistical results also demonstrate that differences in political systems can account for cross-national variations in central bank independence.
The potential for intraparty conflict over monetary policy can therefore explain the cross-national variation of central bank independence in the 1970s and 1980s. While the potential for intraparty conflict reflects the configuration of political institutions (including federalism), it also is a product of the distribution of constituent preferences over economic and monetary policy. These preferences may change over time, altering the monetary-policy incentives faced by party politicians and in turn increasing the potential for intraparty conflict over monetary policy. As the potential for conflict increases, the political benefits of an independent central bank increase and pressure for central bank reform will grow. The next two chapters examine this process of central bank reform in the 1990s.