explaining the dissent voting behavior of Bank of ... - Springer Link

4 downloads 0 Views 785KB Size Report
Jan 29, 2010 - affected the appointment of Mervyn King to Governor of the Bank, who ..... winning side of a decision), four members (excluding Davies who ...
Public Choice (2011) 146: 413–442 DOI 10.1007/s11127-010-9597-6

A decade of dissent: explaining the dissent voting behavior of Bank of England MPC members Mark N. Harris · Paul Levine · Christopher Spencer

Received: 4 June 2008 / Accepted: 7 January 2010 / Published online: 29 January 2010 © Springer Science+Business Media, LLC 2010

Abstract We examine the dissent voting record of the Bank of England Monetary Policy Committee. Contrary to findings in the FOMC literature (for example Havrilesky and Schweitzer in The Political Economy of American Monetary Policy, pp. 197–210, 1990; Chappell et al. in Q. J. Econ. 108(1):185–218, 1993), the effects of members’ career backgrounds and the political channel of appointment on voting behavior are negligible, reflecting the distinct institutional constraints and incentives associated with UK monetary policy. Our findings also suggest that literature which characterizes voting behavior as being predominantly determined by members’ internal or external status is overly simplistic. This view is supported by econometric results appertaining to the introduction of memberspecific fixed-effects, which account for possible unobserved heterogeneity. Keywords Bank of England · Monetary Policy Committee · Career background effects · Dissent voting · Appointment channels · Unobserved heterogeneity

1 Introduction A neglected aspect in the growing literature on interest rate setting by the Bank of England Monetary Policy Committee (MPC), is the effect of political influence and constituent

M.N. Harris Department of Econometrics and Business Statistics, Monash University, Clayton, Victoria 3800, Australia e-mail: [email protected] P. Levine Department of Economics, University of Surrey, Guildford, GU2 7XJ, UK e-mail: [email protected] C. Spencer () Department of Economics, Loughborough University, Leicestershire, LE11 3TU, UK e-mail: [email protected]

414

Public Choice (2011) 146: 413–442

groups on voting behavior, arising either through the appointments procedure or as a consequence of members’ career backgrounds. This is in contrast to the literature on Federal Open Market Committee (FOMC) decision making, where the policy choices of its members, including dissents, are not only rigorously modeled as a function of these characteristics, but found to be associated with such factors in econometric estimation and statistical analysis (see amongst others, Gildea 1990; Havrilesky and Schweitzer 1990; Havrilesky and Gildea 1992; Chappell et al. 1993; Keech and Morris 1997; Krause 1994, 1996; Chappell and McGregor 2000, and more recently, Adolph 2005; Chappell et al. 2005; McGregor 2007).1 The fact that such effects have hitherto remained uninvestigated is surprising, particularly when viewed against the backdrop of major changes to the UK monetary policy framework, which were introduced in May 1997 by the newly elected Labour Party. These changes not only witnessed the establishment of the nine-member MPC, but coincided with a shift to so-called operational independence whereby interest-rate setting was no longer under the direct control of politicians. As this legislative innovation was clearly introduced with a view to improving macroeconomic outcomes by isolating monetary policy decisions from political pressure, the question of whether votes on the policy rate may be subject to political or constituent group influence is not an insignificant one.2 In aiming to assess the effects outlined above, this paper extends the existing literature on MPC decision making in a number of ways. Using ordered probit analysis, we test the hypothesis that the political and non-political channels through which MPC members are appointed affect the decision to dissent, as do differences in members’ career backgrounds: specifically, the impact of these characteristics on dissenting is estimated as part of a wider estimation strategy which also controls for the corresponding impact of a member’s type (i.e., internal versus external member), the MPC’s inflation and output projections, and different chairmen. Our analysis adds to the strand of partisanship literature, as exemplified by (Chappell et al. 1993), that investigates the role for political influence in monetary policymaking through the appointments channel. We also build on related work by Havrilesky and Schweitzer (1990), who predict that different career backgrounds influence an FOMC member’s propensity to dissent on the side of monetary ease or tightness. Here, the appointment of members with particular career backgrounds can be envisaged as meeting the representational demands of constituents or organized interests, who may have specific policy preferences. In the context of this paper these backgrounds are readily quantifiable, and correspond to years spent working in the following areas or organizations: academia; the Bank of England; banking and finance; government; industry; and non-governmental organizations. Unobserved heterogeneity in the form of member-specific fixed-effects is also introduced into the estimation strategy. This innovation is motivated by the increasingly accepted view that the Bank of England MPC is ‘individualistic’ (Blinder 2007)—that is, compared to 1 The reasons as to why FOMC voting behavior, and by implication monetary policy decisions at the Fed,

might be affected via the channels described above are well known. Many explanations appeal to the economic theory of bureaucracy, drawing on the Fed’s symbiotic relationships with Congress and the US President. 2 This is especially the case when one considers how the nuances of the UK political and parliamentary

system, and the particular institutional framework enjoyed by the Bank of England, may shape monetary policy by affecting voting decisions on the policy rate. For example, Santoni (1984) draws on the history of the Bank of England from its establishment in 1694 until 1932, illustrating that monetary policy actions are a function of the set of incentives confronting policy makers. As incentives differ not only across time, but across institutions, the incentives facing UK monetary policy makers may differ substantially from those facing the Fed.

Public Choice (2011) 146: 413–442

415

‘collegial’ MPCs, such as the FOMC and ECB Governing Council, the Chairman enjoys less influence in shaping decision outcomes, and cannot so easily induce members to ‘fall into line’.3 Accordingly, we propose that if MPC members share diverse views, and vote in accordance with them, then such an estimation strategy is warranted.4 Indeed, this strategy may be loosely viewed as a test of the extent to which the MPC is individualistic, as reflected by the magnitude and number of statistically significant fixed-effects parameters. However, prior to the empirics, we first relate our contribution to previous studies of MPC voting behavior, placing particular emphasis on the institutional and political factors that characterize monetary policy making in the UK.

2 Why dissent? The internal-external distinction and beyond The majority of contributions that examine the policy choices of MPC members are geared towards explaining the voting patterns of internal and external committee members (see for example Gerlach-Kristen 2003, 2009; Bhattacharjee and Holly 2005; Besley et al. 2008; Brooks et al. 2007; Harris and Spencer 2009); and, even where features such as career backgrounds are accounted for (as in Besley et al. 2008 and Riboni and Ruge-Murcia 2008), they are introduced in the context of reaction function estimation. Given its notoriety as a topic of scrutiny in the literature, we first outline the nature of the internal-external distinction, accounting for why the behavior of both of these groups might be expected to differ. Yet, while this institutional feature is clearly important in explaining member’s votes, focusing almost exclusively on it potentially omits other factors which may drive voting behavior. To this end, a number of additional considerations—most notably the anticipated effects emanating from the channel of appointment and career backgrounds—are examined. 2.1 The internal-external distinction A fundamental difference between the five internal and four external members comprising the MPC is that “internal members have full-time executive positions in the bank” whereas external members “work mostly part time, and have no executive responsibilities” (Besley et al. 2008: 218).5 The internal members, for example, comprise the Governor of the Bank, 3 To account for how different consensual norms may influence voting behavior, the reader is referred to

Blinder’s (2007) typology of monetary policy committees, which partly explains the significantly greater number of dissents cast by Bank of England MPC members in relation to FOMC members. 4 Coincident with the above assertion is the observation that not all MPC members act in accordance with

a view prevalent in the literature, that compared to MPC members appointed from within the ranks of bank staff, external MPC members choose lower interest rates (Gerlach-Kristen 2003), and are even characterized by loss functions that are more sympathetic to deviations of output from potential (Gerlach-Kristen 2009). Both of these findings suggest that external (internal) members may be more predisposed to dissent on the side of monetary ease (tightness), something which in practice is clearly not observed across all members. While this argument provides support for the presence of voter heterogeneity—and thus a fixed-effect approach—it more generally highlights an area in which the MPC literature is deficient, namely that the insider-outsider distinction is overly simplistic, and as is discussed later, does not capture other potential influences on voting behavior such as channel of appointment effects. 5 The Bank’s position on whether external MPC positions can be full or part-time is found in a recent doc-

ument by the Treasury Committee (a committee comprised of MPs from a cross-section of political parties, which has responsibility on behalf of Parliament, for scrutinizing the conduct of monetary policy and MPC appointments at regular public hearings), which published a report entitled The Monetary Policy Committee of the Bank of England: ten years on (18th September 2007, House of Commons Paper No. 299). It states that

416

Public Choice (2011) 146: 413–442

the two Deputy Governors for Monetary Policy and Financial Stability respectively, and two senior members with executive responsibilities for the Bank’s operations, one of whom is the Bank’s Chief Economist. External members are appointed mainly from academia and the financial sector. Explanations as to why these groups should be distinct in their voting behavior are numerous. As noted in Buiter (1999), Gerlach-Kristen (2003) and Harris and Spencer (2009), differences may emerge if internals share an organizational consensus or ‘Bank view’, that is distinct from the view(s) of external members: in Harris and Spencer (2009), for example, such a device is potentially responsible for the finding that internal members are more likely to vote as a bloc. Moreover, as internals enjoy a five-four majority over externals, this increases their chances of being on the winning side of a decision. Relatedly, externals may be easily out-voted at each meeting if rigorous research to back up their arguments is lacking. Such a situation may arise if externals are denied access to adequate research resources at the Bank: for example, if external members are to provide analysis questioning the findings of the Bank’s standard suite of models (or indeed, to conduct policy exercises using Bank models but conditioned on different assumptions), resources need to be provided to cater to these requirements. As there is strong evidence of a resources dispute developing in the early years of the MPC, the resource issue, and its proposed implications for voting, is not moot. Due to space constraints, a more detailed account of this episode is expounded in Appendix A.6 There are, as might be anticipated, additional reasons why the voting behavior associated with internals and externals may be expected to differ. Internals who wish to build a career within the Bank of England may face career concerns, and thus be incentivized to ‘fall in line’ with each other: here, disagreeing too much with one’s colleagues may be perceived by members as being detrimental to career progression within their organization.7 In contrast, because the prospect of career progression within the Bank is not an issue for externals, there

external members will be given part-time appointments unless individual circumstances, such as a conflict of interest, warrant a full-time position being offered. For example, Dr. Sushil Wadhwani was employed by a private investment organization, so became a full-time external MPC member on that basis; on the other hand, Charles Goodhart, a prominent UK monetary economist worked for the Bank as a part-time MPC member whilst maintaining university-based responsibilities. However, even then, a full time position should be a last resort: that is, as long as an external member is offered an MPC position in good time, it should be possible for additional non-Bank employment to be arranged, such that a member should not require a full-time position. 6 We are grateful to an anonymous referee for highlighting the potential role of resource control, given that it

is hereunto unexplored in the Bank of England MPC literature. Nevertheless, the notion of resource control as used here refers to control by Bank executives over research facilities for external members, who may wish to conduct their own research at the Bank. This is distinct from usage referring to how the Fed is incentivized to conduct policy so as to maximize the total amount resources under its control (Shughart and Tollison 1983), or how staff size and monetary partisanism are shaped by a Bank’s statutory independence (Vaubel 1997). Our example perhaps comes closest to Havrilesky and Gildea (1992), who emphasize how FOMC chairmen exert resource control over fellow FOMC members sitting on the Board of Governors, and exploit this to constrain their “independent” behavior. 7 Meade and Stasavage (2008) introduce a role for career concerns, suggesting that members of a monetary

policy committee are less likely to voice dissent when the transcripts of meetings are placed in the public domain. In relation to the above arguments, there may also be a role for career concerns, particularly if MPC members perceive a link between voting behavior and reappointment. This prospect has been suggested by members of the UK political establishment—Howard (2002) suggests that if one is seeking reappointment, then voting for lower, as opposed to higher interest rates may secure a second term. Preliminary econometric results (not presented here), however, do not to support this conjecture. Interestingly, it does not seem to have affected the appointment of Mervyn King to Governor of the Bank, who cast more dissenting votes on the side of monetary tightness than any other MPC member up to the time of his (re)appointment. This observation is seemingly at odds with Howard’s suggestion.

Public Choice (2011) 146: 413–442

417

is less incentive to vote as a bloc or fall into line with other members. Indeed, externals may even be incentivized to dissent to gain media attention, thereby increasing their public profile (Gerlach-Kristen 2003).8 2.2 Political and non-political appointments to the MPC While the nature of the internal-external distinction may incentivize members to vote in different ways, it does not, by construction, explicitly capture fundamental differences in the way members are appointed. This is unlike the FOMC, where Bank Presidents are appointed by their Reserve Bank’s Board of Directors, and Board Governors by the US President. Internal members are appointed in one of two different ways. The Governor and two Deputy Governors are all Crown appointments, which means that they are appointed by the Queen on the advice of the Prime Minister. In practice however, the Chancellor of the Exchequer may play a significant role in selection.9 With respect to the two remaining internals, although statutes prescribe that these members should be appointed by the Governor only after the Chancellor of the Exchequer has been consulted, in practice, the Chancellor has little say in the matter.10 All four external MPC members are appointed directly by the Chancellor. For the purposes of subsequent analysis, we refer to members appointed by either the Crown or directly by the Chancellor as political appointments, where the adjective ‘political’ signifies that the individual is chosen directly by a member of the Executive Branch of UK government.11 In contrast, the two internal members chosen by the Governor are classed as non-political appointments. We note that this type of distinction—between political and non-political appointees—is also developed in the contributions of Krause (1994, 1996), who examines FOMC voting. The observation that the overwhelming majority of MPC members are political appointments raises a number of issues. On one argument, the notion that governmental influence shapes interest rate decisions through the political appointments channel is highly plausible. Cobham (2000) for instance, argues that the MPC appointments procedure is opaque and

8 The central bank independence literature also suggests that longer term lengths increase a central bank’s

political independence from the government (Grilli et al. 1991). This is of interest insofar as both the Governor and two aforementioned Deputy Governors are appointed for five year renewable terms, whereas all other MPC members serve shorter three year renewable terms. In enjoying longer term lengths, the Governor and the two Deputy Governors may be less susceptible to governmental pressure to reduce interest rates; this avenue of research is not, however, pursued in this study. 9 In previous correspondence with the Secretary of the Bank of England, it was suggested that the relative

influence of the Prime Minister and Chancellor in selecting the Governor and two Deputy Governors changes from administration to administration, depending on personalities and circumstances. For instance, while newspaper reports from the 1980s suggest that Governor Robert Leigh-Pemberton was chosen solely by Margaret Thatcher, the Chancellor is said to have made the decision in other cases. We thank Peter Rogers for shedding light on this issue for us. These informal aspects of the appointments procedure are not seemingly captured by the 1998 Bank of England Act, and differ from the case of the Fed, where the US President enjoys total autonomy in appointing the Chairman. 10 We thank Charles Goodhart for clarifying this matter. 11 This of course assumes either that the Chancellor and the Prime Minister share common preferences re-

garding whom to appoint. As there is no evidence to suggest that Prime Minister Blair and Chancellor Brown were at loggerheads over Bank appointments, the above assumption is not entirely unreasonable.

418

Public Choice (2011) 146: 413–442

open to political opportunism.12 Similarly, a number of witnesses at Treasury Committee13 hearings, including ex-MPC members, have indicated that while they do not believe that the Chancellor behaved in an opportunistic manner when choosing externals, the institutional framework does not preclude future Chancellors from abusing their position (Treasury Select Committee 2007a: 33–35). Relatedly, the fact that political appointments to the MPC are subject to fewer legislative constraints than US presidential appointments to the FOMC,14 may make it relatively easier for the Executive Branch of UK government to ‘pack’ the MPC with “eminently reliable agents whose views align with their own” (Havrilesky and Gildea 1992: 398). This line of reasoning suggests that political appointees might exhibit strong partisan behavior, or vote in a way which reflects any time-consistent inflationary bias on the part of the government. However, as discussed next, there are compelling arguments to suggest that channel of appointment effects may in fact be small—especially when one considers how the framework for UK monetary policy mitigates the so-named “democratic deficit” associated with monetary policy making. 2.2.1 The democratic deficit of monetary policy and the inflation target As noted by Tucker (2008), prior to operational decisions on monetary policy being delegated to the MPC in May 1997, the UK enjoyed no history of non-elected officials assuming responsibility for major policy decisions: monetary policy did not suffer from a ‘democratic deficit’, as it was under the aegis of elected politicians, who were accountable for their actions on the floor of the House of Commons. Goodhart (2006) proposes that the setting of an inflation target by the executive branch of UK government effectively mitigates the democratic deficit that arose following the shift to operational independence. This is unlike the United States, where the corresponding deficit “is mitigated by the politicisation of the appointments process” (Goodhart and Meade 2003). On this argument, political appointments to the MPC might not be expected to exhibit behavior consistent with notions of partisanship, as the process of selection itself is considerably less politicized. Accordingly, we propose that political and constituency group pressure is more likely to influence monetary policy through the choice of the inflation target itself, as opposed to through the choice of individuals selected to set policy. This conjecture is broadly consistent with the well known view of McCallum (1995), who argues that by retaining goal independence, the government effectively controls monetary policy through its power over the goal: in the case of an inflation targeting institution, this amounts to control over the inflation target itself (Du Plessis 2005). Indeed, as has been consistently reiterated by the Chancellor, the UK inflation 12 In one Treasury Select Committee hearing, Richard Lambert (an external member who joined the MPC in

May 2003) was asked whether ‘a couple of calls to Japan was all there was to becoming a member of the MPC.’ (Oral evidence of Mr. Richard Lambert, taken before the Treasury Committee on Monday 16 June 2003). 13 The Treasury Select Committee has no power to veto appointments to the MPC, even if it does not approve

of them. 14 Whereas in the US, Senate approval is required for appointments to the Board of Governors and the position

of Fed Chairman and Deputy Chairman, political appointments to the MPC are not subject to similar Parliamentary approval: the hands of the Chancellor and Prime Minister are thus significantly less constrained with respect to the choice of appointees. This facet is important insofar as even though the institutional provision requiring Senate approval for nominees impedes the ability of the US President for ‘packing the Fed’ (Morris 2002: 82), there is strong evidence that a large number presidential appointees behave in partisan fashion. Therefore, one might infer that if a US president bound by Congressional constraints can still appoint partisans, a relatively unrestrained UK Executive Branch can easily ‘pack the MPC’.

Public Choice (2011) 146: 413–442

419

target is not cast in stone, and can be changed it if it is deemed expedient to do so:15 since 1997, the target has been changed once from 2.5% Retail Prices Index excluding Mortgage Interest Payments (RPIX) inflation to 2% Consumer Prices Index (CPI) inflation.16 Building further on the above arguments, we propose that while in practice it may be easy to appoint partisans to the MPC, it would at the same time make little sense for the Executive Branch to appoint (i) members lacking the required economic expertise to hit the primary objective of UK monetary policy—an explicitly defined and transparent inflation target of 2% CPI inflation—or (ii) individuals who would systematically undermine the Government’s own inflation targeting framework, through pursuing, for example, short term output gain at the cost of significantly overshooting the target itself. This is attributable to the fact that the inflation target ‘is 2 per cent at all times’ and represents ‘the rate at which the MPC is required to achieve and for which it is accountable’.17 The ability to hit the inflation target thus represents the benchmark against which the performance of the MPC and its members—who are all individually accountable for their decisions to Parliament—is measured.18 In this sense, the choice of, and behavior of political appointees should not only be perceived as reflecting the (very few) political constraints on selection, but the nature of the inflation targeting framework into which MPC members are appointed. 2.3 Career background effects The notion that career backgrounds play a role in driving voting behavior is well explored in the FOMC literature. For instance, Chappell et al. (1995) find that partisan behavior is partially attributable to the career backgrounds of FOMC members, noting that “experience in government, particularly at the Federal Reserve Board, is associated with significantly stronger preferences for monetary ease” (Chappell et al. 1995: 130). Similar findings are also uncovered in Havrilesky and Schweitzer (1990), Gildea (1990) and Adolph (2005). Of particular relevance to this study is the work of Havrilesky and Schweitzer (1990), who

15 “The inflation target will be confirmed in each Budget. There is a value in continuity and I will have

proper regard to that. But I will also need to consider the case for a revised target at these times on its merits.” Remit for the Monetary Policy Committee, letter from the Rt. Hon. Gordon Brown to Eddie George, 12th June 1997. Available at http://www.bankofengland.co.uk/monetarypolicy/pdf/chancellorletter970506.pdf, accessed 15 June 2009. 16 Moving to CPI inflation may provide a more objective yardstick against which possible future UK entry

into the Eurozone might be judged—this is because the CPI measure is similar to the HICP measure of inflation targeted by the European Central Bank. The decision to switch might thus be viewed as being both economically and politically expedient. We further note that while the primary objective of UK monetary assumes the form of a symmetric inflation target, there is no reason why the Chancellor cannot, should he wish to do so, re-define ‘price stability’ as a range of prices, or even in qualitative (as opposed to explicitly quantitative) terms. For a public choice oriented discussion of the democratic deficit in the context of the European Central Bank, the interested reader is referred to Mueller (2005). 17 Remit for the Monetary Policy Committee, letter from the Rt. Hon. Gordon Brown to Mervyn King, 21st

March 2007. Available at: http://www.bankofengland.co.uk/monetarypolicy/pdf/chancellorletter080311.pdf, accessed 5th May 2009. 18 This is unlike the Fed, whose (dual) targets are only qualitatively, and not quantitatively defined (Belongia

2007), rendering monetary policy far less understood by the public. Building on this point, Buiter (2008) suggests that bureaucratic considerations and Congressional constraints may prevent the Fed from announcing an explicit inflation target, even though it is in principle, able to do to. This is because the opaque and non-quantified objectives of the Fed benefits Congress, the whims of whom the Fed has to bend to in order to maintain its autonomy.

420

Public Choice (2011) 146: 413–442

introduce a general theoretical framework for dissent voting.19 While we do not formally present their model in this section (see Appendix B for a fuller treatment), we note that our empirical treatment of career backgrounds, especially with respect to variable construction in Sect. 4, draws heavily on their approach, which is premised on the notion that the government has a time-consistent inflationary bias: accordingly, MPC members whose career backgrounds lie relatively closer to central government (experience in government and industry) are deemed more liable to conform to such bias, and the more one conforms to it, the greater the propensity to dissent on the side of monetary ease. Conversely, members whose experiences are relatively further from central government (experience in academia, financial organizations and at the central bank) are considered more prone to dissent on the side of monetary tightness. As is shown later on, MPC members’ career backgrounds are marked by considerable diversity, facilitating a comprehensive analysis of career background effects on voting behavior. We do, however, note that just as in Havrilesky and Schweitzer (1990), a number of judgement calls were made with respect to the impact of particular career backgrounds on voting behavior. However, these effects might be expected to be small, given the nature of the democratic deficit and the inflation targeting framework within which MPC members take interest rate decisions. In addition to the effects discussed above, a number of other factors may inform the decision to dissent, which further to being shaped by career experience, may be determined by member-specific factors that are difficult to observe, and thus difficult to proxy for: these include members’ views about the underlying structure of the economy, and members’ information sets. Different MPC members will invariably hold different views about the underlying structure of the economy, and be exposed and receptive to different sources of economic information. As Blinder (2007) notes, even when faced with the same information (for example, as presented at the Bank’s monthly ‘pre-MPC’ meeting), members of a monetary policy committee may have contrasting views as to the appropriate policy stance. As all of these factors represent potential sources of disagreement regarding interest rate policy, this may lead to dissent voting.20 A natural way to capture unobserved member-specific factors, 19 Havrilesky and Schweitzer also assume that there is a utility and disutility associated with dissenting.

Utility arises from a member believing it is morally right to dissent. This explains why, even in the face of potential pressure to vote with the majority of MPC members, some individuals may choose to make their differing opinion known through recording a dissenting vote. On the other hand, disutility arises precisely because of the need for committee members to fall into line with each other. There is some evidence to suggest this may in practice happen: recent work by Harris and Spencer (2009) shows that compared to external members, internals are significantly more prone to vote as a bloc. While this tendency may reflect an institutional consensus amongst internal MPC members, such members may desist from voting against each other because they work for the same organization: the conjecture here is that voting against one’s peers too often will be viewed in a dim light, and may affect career prospects, as already discussed in Sect. 2.1. Building on this argument, it is noted that while dissent voting is not actively encouraged on the MPC, its individualistic nature (see, for example, Blinder and Wyplosz 2005 and Blinder 2007) may lessen the stigma attached to dissenting: Nakahara (2001), for instance, cites former external MPC member DeAnne Julius as attributing dissent voting to individually accountability, without which members would “lose the incentive to make public their position at the voting stage even if they had voiced opposing views during the debate.” We propose that while individual accountability may mitigate the stigma attached to dissenting—as might the individualistic nature of the MPC (Blinder 2007)—it does not eliminate in altogether. 20 In May 2002, Sushil Wadhwani, an external member who served on the MPC between June 1999 and May

2002 claimed that monetary policy was ‘held too tight because of a biased forecast’ (Wadhwani 2002: 1), the main inflation projection in the May 2002 Inflation Report assumed a ‘higher pass-through into prices’ than was probable (Wadhwani 2002: 3), and the level of potential output was ‘too pessimistic’ (Wadhwani 2002: 10). Such sentiment is perhaps unsurprising: as Charles Goodhart notes, the introduction of external MPC members into a committee forecasting process such as the one underlying the Bank’s quarterly Inflation

Public Choice (2011) 146: 413–442

421

is to use a fixed-effects approach, an innovation which is pursued as part of the estimation strategy in Sect. 4. Prior to this, we address the empirics.

3 Dissenting votes Voting data are obtained from the Minutes of MPC Meetings, and identifies who the dissenting voters are at each meeting, and whether they dissented on the side of ease or tightness. We define two types of dissent voting: dissent for tighter policy and dissent for looser policy. There are important caveats to these definitions, which are expounded as follows: (i) Dissent for tighter policy: Defined as where a member votes for a higher short-term interest-rate than the rate chosen by the winning majority of MPC members. A member may vote for no change or a decrease in the interest-rate but still be classed as dissenting for tighter policy if the rate chosen by the MPC is lower than their chosen rate. (ii) Dissent for looser policy: Defined as where a member votes for a lower short-term interest-rate than the rate chosen by the winning majority of MPC members. A member may vote for no change or an increase in the interest-rate but still be classed as dissenting for looser policy if the rate chosen by the MPC is higher than their chosen rate. We also find it useful to define as assenting vote as one that is cast in agreement with the winning majority of MPC members at each meeting. Table 1 documents the dissent voting behavior corresponding to different groups of MPC members. In addition to identifying members according to their internal-external status (as in Gerlach-Kristen 2003; Harris and Spencer 2009), we also take into account whether members are appointed by political or non-political means. This yields three groups: (i) politically appointed internals; (ii) non-politically appointed internals; and (iii) politically appointed externals.21 The first three columns show statistics pertaining to the number of assents, and dissents cast on the Table 1 Number of dissenting votes cast by the MPC, August 1997–May 2007a Assents

Ease

Tightness

All

dissents

dissents

dissents

All Members

892(85.4)

83(7.9)b

70(6.7)

153(14.6)

Internal MembersP

318(92.2)

3(0.9)

24(7.0)

27(7.8)

Internal MembersNP

214(89.9)

7(2.9)

17(7.1)

24(10.1)

External MembersP

360(78.3)

73(15.5)

29(6.2)

102(21.7)

a Results based on data from 119 meetings b Figures in parentheses (·) express the number of assenting and dissenting votes cast as a percentage of votes cast within each group P Denotes political NP Denotes non-political appointment

Report projections, is ‘inevitably likely to generate some tension and disagreements’ (Goodhart 2001: 62). Also see Blanchflower (2006). 21 All externals are political appointments.

422

Public Choice (2011) 146: 413–442

side of ease and tightness respectively. The final column shows the total number of dissenting votes cast, irrespective of direction.22 Figures in parentheses (.) express the number of dissenting votes cast as a percentage of all votes cast within each group. When votes across all members irrespective of sub-groupings are considered, dissenting votes cast for monetary ease and tightness are split relatively evenly (83 vs. 70 votes, All Members). Further, irrespective of a member’s internal-external status or the channel of appointment, all MPC members are most likely to cast assenting votes. These broad similarities are, however, deceptive. External members are more than twice as likely to dissent than politically and non-politically appointed internal members respectively (102 vs. 27 and 24 votes).23 Whereas internal members—irrespective of the channel of appointment—are more prone to dissenting on the side of tightness, external members do so on the side of monetary ease. Chi-squared tests were also conducted to test for differences between members based on their internal-external status and channel of appointment. Our findings, full details of which are provided in Appendix C, suggest that internal and external members differ significantly in their voting behavior. However, while political and non-political appointees were found to vote in a significantly different way, no significant difference was found in the voting behavior of politically and non-politically appointed internal members. This suggests that differences which at first appeared to be associated with appointment channel effects are in fact attributable to internal-external factors. Finally, Table 2 shows the dissent voting behavior of MPC members at an individual level. Out of the fourteen external members in the sample, six members (Buiter, Goodhart, Budd, Walton, Besley and Sentance) cast more dissenting votes on the side of tightness than ease. At this level of disaggregation, the reputation for externals to cast dissenting votes on the side of monetary ease seems to be driven by just four individuals: Julius, Wadhwani, Nickell and Allsopp. Excluding Buiter, whose numerous tightness dissents are balanced by a significant number of ease dissents, and Lambert, who did not dissent, the remaining members cast only a modest amount of ease dissents. In short, externals vary considerably in their individual dissent voting behavior. Turning to internal members, two individuals— Mervyn King and Andrew Large—might be viewed as driving tightness dissents. Moreover, once King is omitted from the sample, the total number of tightness dissents cast by internals fall to a figure below that of externals (28 vs. 29 votes). Further, while internals are less prone to cast dissenting votes per se (put another way, they are far more likely to be on the winning side of a decision), four members (excluding Davies who cast only two votes) buck 22 In the first decade of the Bank of England MPC almost 15% of votes cast by its members were dissents. This

figure is markedly higher than the portion of dissenting votes associated with monetary policy committees at comparable institutions such as the US Federal Reserve, the Bank of Japan and the Swedish Riksbank. 23 On average, external members dissented approximately once every five votes. This figure is around one in

ten for internal members. Interestingly, exact permutation tests were also used extensively to test for differences in the dissent voting behavior associated with internal and external members at each MPC meeting. For each individual meeting nine hypotheses were examined, and in all 121 sets of test statistics were generated: specifically, for each type of dissenting vote defined in Sect. 3 (dissent irrespective of direction, dissent for tighter policy, dissent for looser policy), we tested for whether (1) internal members cast such votes more often than external members (one-tailed test), (2) external members cast such votes more often than internal members (one-tailed test), and (3) whether there was any general difference between groups (two-tailed test). Each hypothesis was tested against a null of no difference in voting behavior. When judged against conventional levels of significance, members do not differ in their dissent voting behavior. However, the very small sample size (MPC membership ranged from five to nine individuals) raised problems of statistical power, and specifically, the prospect of failing to reject the null hypothesis when it is false (i.e., a type II error). Moreover, it was found that for a test to have any statistical power at conventional levels (power = 0.8), all internal members would have to dissent while all external members assented.

Public Choice (2011) 146: 413–442

423

Table 2 Number of dissenting votes cast by individual MPC members, August 1997–May 2007 Votesa

Assents

Ease dissents

Tightness dissents

All dissents

Internal membersP, Eddie George Mervyn King

72

72

0

0

0

119

106

0

13

13

David Clementi

61

57

1

3

4

Andrew Large

40

31

0

9

9

Rachel Lomax

47

43

2

2

4

John Gieve

16

16

0

0

0

Ian Plenderleith

59

54

2

3

5

John Vickers

28

23

0

5

5

Charles Bean

81

77

4

0

4

Paul Tucker

60

53

1

6

7 17

Internal membersNP

External membersP Willem Buiter

34

17

8

9

Charles Goodhart

34

31

0

3

3

DeAnne Julius

45

31

14

0

14

Sir Alan Budd

18

14

0

4

4

Sushil Wadhwani

37

24

13

0

13

Stephen Nickell

73

56

13

4

17

Christopher Allsopp

37

26

11

0

11

Kate Barker

73

68

4

1

5

Marian Bell

36

36

5

0

5

Richard Lambert

34

34

0

0

0

David Walton

12

9

1

2

3

David Blanchflower

12

8

4

0

4

Tim Besley

9

6

0

3

3

Andrew Sentance

8

5

0

3

3

P Political appointment NP Non-political appointment a Denotes total number of votes cast by each member  Howard Davies, a Deputy Governor who only served in the first two meetings, lies outside the sample used in estimation and is thus omitted from the table  Mervyn King served as an Executive Director at the Bank before becoming a Deputy Governor

the trend that internals are more likely to dissent on the side of tightness—namely Bean, Lomax, Gieve and George. Not all internals are the same. The reason for George casting no dissents may be precisely put down to his role as MPC Chairman between June 1997–May 2003: he was never on the losing side of a decision, arguably due to his power as to make the policy proposal at each meeting, coupled with his more general influence as Governor of the Bank.24

424

Public Choice (2011) 146: 413–442

4 Econometric analysis 4.1 Model specification and variables When voting on the policy proposal, MPC members are viewed as being faced with three mutually exclusive choices: to dissent on the side of ease, to assent, or to dissent on the side of monetary tightness.25 A natural candidate for modeling such behavior is the ordered probit (OP) model, Z∗gt = xgt β + εgt , Zgt = −1

if

Z∗gt

(1) ≤ γ1 ,

Zgt = 0

if γ1 < Z∗gt ≤ γ2 ,

Zgt = 1

if Z∗gt > γ2

(2)

where Z∗gt is a stacked m × 1 vector of −1s, 0s and 1s corresponding to members’ votes to dissent for looser policy, assent, or dissent for tighter policy. xgt is a m × h matrix containing h independent variables, and β is a 1 × h vector of parameter estimates. As the xgt matrix contains those variables which form the basis of our estimations, we now turn attention to their construction. 4.1.1 Channel of appointment and career background variables To account for the different member types introduced in Table 1, three indicator variables were created: IntPol equals one for a politically appointed internal member and zero otherwise; IntNPol equals one for a non-politically appointed internal member and zero otherwise; lastly, ExtPol is assigned a one for an external member and a zero otherwise. Unlike internal members, externals are all politically appointed. To avoid the dummy variable trap, IntNPol was omitted during estimation. With respect to parameter estimates, a negative sign for ExtPol was expected. This reflects the stylized findings in the literature— rationalizations for which are discussed in Sects. 2 and 3—that internal (external) members choose higher (lower) rates, and are more likely to dissent on the side of monetary tightness (ease). Political-appointment effects, if present, might also be expected to run in the same direction: this is because ceteris paribus, we assume that partisans appointed by a left-ofcenter party are more predisposed to dissent on the side on monetary ease than non-political 24 Significantly, it is possible for the Chairman to lose a vote. This happened for the first time during the

100th meeting of the MPC in August 2005, when Governor King was on the losing side of a 5–4 split. Having ascertained the policy stances of all committee members, King tabled a policy proposal which he then proceeded to vote against. 25 Our estimation strategy extends the original approach of Havrilesky and Schweitzer (1990) in a number of

important respects. First, in addition to using dissenting votes, we utilize all votes cast in agreement with the policy proposal (assenting votes). As such, information contained in the voting record is no longer wasted, making our approach consistent with the U.S. FOMC approaches of Chappell et al. (1993, 2004, 2005). Second, in addition to career characteristics data, our estimations control for the presence of other factors which determine the decision to dissent, thus addressing the issue of omitted variable bias. Third, and building on the first point, through utilizing assenting votes, the larger sample size permits us to pursue estimation strategies which condition on unobserved heterogeneity.

Public Choice (2011) 146: 413–442

425

appointees.26 Of course, the assertion that political appointees might be anticipated to dissent on the side of monetary ease, rests on the observation that all such appointments in our sample have been made by the Labour Party. The election of a right-of-center Conservative government may change this behavioral assumption. Yet given there are strong arguments to suppose that the selection process for MPC members is not politicized (Goodhart 2005; Goodhart and Meade 2003), any such effect is anticipated to be at most small. Moreover, if differences between MPC members are not determined by the appointment channel, it implies that the IntPol parameter should be statistically no different from zero: like the omitted group, its members are all internals, with its members only differing in how they are appointed. To capture career background effects, a series of covariates proxying members’ career characteristics are also constructed. Career backgrounds are categorized according to years spent working in six broadly defined categories: (i) (ii) (iii) (iv)

Academia—refers to years working at a university in an academic capacity. Bank—denotes the number of years employed at the Bank of England. Finance—refers to positions held in banking and finance. Government—denotes years spent working in the civil service or for the UK Govern-

ment. (v) Industry—refers to years spent an economist in industry. (vi) NGO—refers to non-governmental organizations. This covers both national and international independent research organizations such the Organization for Economic Cooperation and Development (OECD), and transnational institutions such as the International Monetary Fund (IMF), World Trade Organization (WTO) and Bank for International Settlements (BIS). Our classification system covers only full time positions and secondments held by MPC members up to but not including time working on the MPC; excluded from the criteria are all part-time positions, special advisory roles and academic consulting. Consequently, all time served on the MPC is purposely neglected. While the effects of career backgrounds have required making some judgement calls, voting behavior is anticipated to be influenced in the follow ways. We assume that backgrounds in academia, finance, at the Bank of England and NGOs promote tightness dissents: in the case of academia, this reflects the large impact of the literature on time-consistent monetary policy, and a view that experience in academia promotes independent thinking, hence lowering members’ susceptibility to yield to short-run political pressures. Experience at the Bank of England is assumed to engender an acute awareness of the inflationary consequences of activist monetary policy, thus promoting tightness dissents. For ‘career’ central bankers, 26 Empirical support for the conjecture that left-wing parties are more likely to pursue economic policies

which boost output is exemplified in the work of Alesina and Roubini (1992). We also note that as the MPC was not only established by a Labour government, but one which held power for the duration of the sample, it is not possible to test for voting differences associated with political members chosen by different UK political parties. This is unlike studies of FOMC voting, where the monetary policy preferences of individual members are modeled as a function of the party-political affiliation of the individuals who appointed them. Such studies typically cover periods encompassing different political administrations, Republican and Democrat. For example, Chappell et al. (1993) find that the power of appointment provides ‘an important channel of systematic partisan influence’ (Chappell et al. 1993: 209), such that Democrat appointees exhibit significantly different voting behavior than Republican appointees. While our data preclude us from testing for such effects, it does not prevent testing for differences between political and non-political appointments; further, as the Labour Party is politically left-of-center, we posit that such partisan influence, if present, will manifest itself in MPC members dissenting on the side of monetary ease.

426

Public Choice (2011) 146: 413–442

Table 3 Career backgrounds of individual MPC members, June 1997–May 2007 Career experience (years) Academia

Bank

Finance

Industry

Govt.

NGO

Internal members Eddie GeorgeP

0

29

0

0

0

6

18

7

0

0

0

0

David ClementiP

0

22

0

0

0

0

Andrew LargeP

0

0

26

8

5

0

Rachel LomaxP

0

0

0

0

36

1

John GieveP

0

0

0

0

28

0

Ian PlenderleithNP

0

32

0

0

0

3

John VickersNP

17

0

0

0

0

0

Charles BeanNP

18

0

0

0

7

0

Paul TuckerNP

0

12

10

0

0

0

Mervyn King,NP,P

External members Willem BuiterP

23

0

0

0

0

0

Charles GoodhartP

15

17

0

0

3

0

DeAnne JuliusP

4

0

0

9

0

7

Sir Alan BuddP

22

0

4

0

10

0

Sushil WadhwaniP

8

0

10

0

0

0

Stephen NickellP

30

0

0

0

0

0

Christopher AllsoppP

0

31

4

0

0

2

Kate BarkerP

0

0

0

16

0

0

Marian BellP

0

0

15

0

3

0

Richard LambertP

0

0

35

0

0

0

David WaltonP

0

0

19

0

2

0

David BlanchflowerP

21

0

0

0

0

0

Tim BesleyP

18

0

0

0

0

0

5

0

0

15

0

0

Andrew SentanceP P Political appointment NP Non-political appointment

 Howard Davies, a Deputy Governor who only served in the first two meetings lies outside the sample used in estimation and is thus omitted from the table  Mervyn King served as an Executive Director at the Bank (N P ) before becoming a Deputy Governor (P )

dissenting on the side of monetary tightness may also be used to signal their credentials as being ‘conservative’ or ‘inflation-averse’. Finally, the inclusion of prior experience in finance and NGOs reflects a view that such careers are removed from governmental power and influence. We also propose that time spent in industry and government will promote ease dissents. In the case of industry, while rising prices may imply higher wage claims and thus rising costs for the firm (prompting calls for the monetary authorities to bring inflation under control through tightening interest rates), ease dissents are more likely to be promoted as higher interest rates hit the ability of firms to invest and borrow, reduce consumer expenditure, and reduce the international competitiveness of products for export through exchange rate effects. Table 3 documents the career experiences of all individuals who served on the

Public Choice (2011) 146: 413–442

427

Fig. 1 Mean career experience of MPC members, June 1997–May 2007

MPC between June 1997 and May 2007, measured in the number of years spent in each category prior to joining the MPC. For estimation purposes, and in line with Havrilesky and Schweitzer (1990), these variables are subsequently manipulated: for each MPC member, experience within each career category is expressed as the difference between the number of years spent working in that category and the committee mean for that category.27 These new variables, which we label AcadD , BankD , FinD , GovtD , IndD and NGOD —where the D subscript denotes ‘deviation from the committee mean’—vary across time due to members’ overlapping terms. Figure 1 shows how the committee mean for each career category has changed over the period under scrutiny: for instance, mean experience at the Bank of England has fallen considerably since the establishment of the MPC, while experience in government rose markedly during the first eight years, before falling back to mid-1997 levels at the start of 2006. By contrast, the average level of academic experience remained relatively high and constant across the whole sample period. To complement Fig. 1, Table 4 shows the nature of members’ overlapping terms, and captures how membership of the MPC has changed over time. On average, the composition of the MPC changed approximately every six months, raising the prospect that over the sample period, as different personalities both entered and left the group, the decision making dynamics of the committee underwent considerable change.

27 Career characteristics should not be interpreted as career ‘fixed’ effects as a member’s career experience

for a given characteristic is not detrended by its mean. As the committee mean for a given characteristic changes (i.e., with the turnover of new members with different career backgrounds) so too does a member’s given career characteristic. More formally, (xgt − x g ) is usual, but in our case we have (xgt − x t ), where x represents a given career characteristic, x t its mean value in period t , x g is the mean of that characteristic for member g, and xgt is the period t value of that characteristic for member g.

2 1 3 6 12 12 4 9 12 1 2 1 8 1 24 7 2 2 1 2 1 8

Jun 97–Jul 97: Aug 97: Sep 97–Nov 97: Dec 97–May 98: Jun 98–May 99: Jun 99–May 00: Jun 00–Sep 00: Oct 00–May 01: Jun 01–May 02:c Jun 02: Jul 02–Aug 02: Sep 02: Oct 02–May 03: Jun 03: Jul 03–Jun 05: Jul 05–Jan 06: Feb 06–Mar 06: Apr 06–May 06: Jun 06: Jul 06–Aug 06: Sep 06: Oct 06–May 07:

6 5 7 8 9 9 9 9 9 8 9 8 9 9 9 9 9 8 8 7 8 9

MPC Size 0 −1 +2 +1 +1 −1, +1 −2, +2 −1, +1 −1, +1 −2, +1 +1 −1 +1 −1, +1 −1, +1 −1, +1 −1, +1 −1 −1, +1 −1 +1 +1

Change in membership Internalb Kingd , Plenderleith Plenderleith Plenderleith Plenderleith Plenderleith, Vickers Plenderleith, Vickers Plenderleith, Vickers Plenderleith, Bean Plenderleith, Bean Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker Bean, Tucker

Members Internala GeorgeG , DaviesDG GeorgeG , KingDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG , ClementiDG GeorgeG , KingDG GeorgeG , KingDG , LargeDG GeorgeG , KingDG , LargeDG KingG , LargeDG , LomaxDG KingG , LargeDG , LomaxDG KingG , LomaxDG , GieveDG KingG , LomaxDG , GieveDG KingG , LomaxDG , GieveDG KingG , LomaxDG , GieveDG KingG , LomaxDG , GieveDG KingG , LomaxDG , GieveDG

DG Denotes deputy Governor

G Denotes Governor (and hence MPC Chairman)

d Mervyn King served as an Executive Director at the Bank before becoming a Deputy Governor

c Includes the emergency MPC meeting of 18th September 2001

b Non-political appointment

a Political appointment

Meetings

Period

Table 4 The changing composition of the MPC, June 1997–May 2007a

Buiter, Goodhart Buiter, Goodhart Buiter, Goodhart, Julius Buiter, Goodhart, Julius, Budd Buiter, Goodhart, Julius, Budd Buiter, Goodhart, Julius, Wadhwani Julius, Wadhwani, Allsopp, Nickell Julius, Wadhwani, Allsopp, Nickell Wadhwani, Allsopp, Nickell, Barker Allsopp, Nickell, Barker Allsopp, Nickell, Barker, Bell Allsopp, Nickell, Barker, Bell Allsopp, Nickell, Barker, Bell Nickell, Barker, Bell, Lambert Nickell, Barker, Bell, Lambert Nickell, Barker, Lambert, Walton Nickell, Barker, Lambert, Walton Nickell, Barker, Walton Barker, Walton, Blanchflower Barker, Blanchflower Barker, Blanchflower, Besley Barker, Blanchflower, Besley, Sentance

Externala

428 Public Choice (2011) 146: 413–442

Public Choice (2011) 146: 413–442

429

4.1.2 Additional variables To control for economic conditions, we also construct Taylor-rule type covariates based on the MPC’s in-house inflation and output growth forecasts as published in the Bank’s quarterly Inflation Report. Forecast horizons in line with views expressed by the MPC (Bank of England 1999) that interest-rate changes take two years to maximally impact inflation, and approximately one year for output are chosen. The published modal projections are also adjusted following the method suggested by Goodhart (2005),28 and are expressed in deviation form: specifically, output growth minus potential (assumed to be 2.4% per annum) and the deviation of inflation from target. We denote these forecasts πG and GDPG .29 While it is clearly important to control for economic conditions, we also note that it is not immediately obvious what the signs for πG and GDPG should be.30 To account for the possibility that different types of members might not respond equally to changes in macroeconomic conditions (based on the discussion of members’ information sets and economic models in Sect. 2), IntPol, IntNPol and ExtPol were also interacted with the Taylor-type variables. In all, six interaction terms were created: (Intpol × GDPG ) and (Intpol × πG ); (IntNpol × GDPG ) and (IntNpol × πG ); and (Extpol × GDPG ) and (Extpol × πG ). Lastly, to capture the impact under different MPC chairmen (Governors George and King), we include the binary variable Chair, which assumes a value of one (zero) if George (King) occupied the post.31 4.1.3 Fixed-effects We now address the potential role for fixed-effects. As there are repeated observations for individual MPC members, it is possible to condition on unobserved individual heterogeneity 28 Goodhart argues that the ex-post nature of the Bank’s published forecasts diminishes their importance in

explaining the MPC’s policy decisions: this is because in practice MPC members react to ex-ante forecasts (i.e., conditioned on the interest-rate set by the MPC in the previous month). We use the ex-post forecasts to construct proxy ex-ante forecasts. These forecasts are potentially much closer to those on which individual voting decisions are based. 29 Estimations were also performed using Taylor-rule type variables constructed from real-time consensus

forecasts of GDP growth and inflation, obtained from HM Treasury’s Forecasts for the UK Economy. Published monthly, this is a compendium of forecasts produced by city and independent forecasters. We note that whilst decisions on UK interest rates are in part a function of the Bank’s central inflation and GDP projections, their quarterly nature makes their incorporation into our econometric framework difficult—this is because the MPC takes interest rate decisions on a monthly basis. For this reason, and to take into account new forecast information available to MPC members at each monthly meeting, we also ran regressions using these new variables. Results consistently placed a smaller parameter on the inflation term, although the output term was still negative and often insignificant. All other parameters were robust to the inclusion of these alternative forecasts. However, we opted for the results based on the MPC’s Bank forecasts due to their purported importance in informing the MPC’s decisions. Full consensus estimation results are available from the authors on request. 30 We thank an anonymous referee for raising this issue. In the monetary policy rules literature (for example,

Taylor 1993), it is usual to expect the signs on the output and inflation variables to be positive: this would imply that UK interest rates should be higher (lower) when forecast inflation is above (below) its target level, and output is above (below) potential. However, in this paper, we are modeling dissenting votes, and not members’ interest rate choices as in Besley et al. (2008) and Harris and Spencer (2009). We are not estimating a monetary policy rule. For this reason, higher values of πG and GDPG —which should in theory be associated with members choosing higher interest rates—might not necessarily imply a simultaneous desire to dissent for greater tightness. 31 We note that two additional variables—a dummy to capture the impact of gender, and a reappointment

dummy to proxy for the role of career concerns—were introduced, but subsequently dropped from estimations as they proved to be consistently insignificant.

430

Public Choice (2011) 146: 413–442

by augmenting equation (1) to include an unobserved effect, αg : Z∗gt = xgt β + αg + εgt .

(3)

This begs the question of how to treat the αg . Whilst non-linear panel data estimation has traditionally focussed on treating unobserved heterogeneity as random due to the ‘incidental parameters’ problem (Neyman and Scott 1948), recent developments suggest that the nature of our sample permits a fixed-effects estimation strategy. Specifically, we are in a position contrary to that typically observed in the panel data literature, with small cross-sectional component relative to the sample (i.e., large T and small N ). The overwhelming majority of the 25 MPC members in the sample are observed over a relatively long time period (t = 1, . . . , Ti ): other than Davies (Ti = 2)—who was removed from the sample due to an insufficient number of observations—the number of time periods ranged from Ti = 8 (Besley) to Ti = 121 (King). Heckman (1981) suggests that a temporal sample size of T = 8 is sufficient for any significant fixed T bias to have essentially disappeared. Greene (2004) provides further evidence, citing a significant reduction in biases from T = 3 onwards. In light of these arguments, we include fixed-effects dummies for all MPC members bar Davies. This amounts to relaxing the commonly maintained assumption of   (4) E xgt αg = 0, ∀g, t. The baseline equation on which estimation is based hence becomes Z∗gt = xgt β + αg Dg + εgt ,

(5)

where Dg represents a dummy variable for member g. 4.2 Estimation results Estimates for the ordered probit model are shown in Table 5. Standard errors are given in parentheses (.), with corresponding levels of significance, where ***,**,* denote 1, 5 and 10% levels respectively. AIC and BIC denote the Akaike and Bayesian information criteria, where a smaller value suggests a better specification. Model 1, which models dissent voting using career characteristics alone, can be viewed as a ‘baseline’ specification, and is analogous to the estimation strategy in Havrilesky and Schweitzer (1990); Models 2 and 3 augment the baseline specification by controlling for different member types (IntPol, IntNPol and ExtPol), the impact of economic conditions, and Chairman effects; Model 4 introduces member-specific fixed-effects. While in Models 1, 2, and 3 the joint hypothesis of career backgrounds having no significant effect was rejected outright (p < 0.01 in every case), the direction of some career background effect parameters did not conform to previously stated priors. For example, whereas experience at the Bank of England was found to promote tightness dissents (positive, significant, and consistent with our predictions), time spent working in industry also promotes dissent on the side of monetary tightness (significant, positively signed, not consistent with predictions). Experience in government, finance and academia generally has no significant effect on dissent voting behavior. However, even where career variables are highly significant, the impact on dissent voting is often negligible.32 These findings are somewhat at odds 32 This assertion is also based on calculating marginal effects, which are not reported here due to space

constraints.

Public Choice (2011) 146: 413–442

431

Table 5 Determinants of dissent: ordered probit estimates Model 1

Model 2

Model 3

Model 4

πG



3.167 (0.470)∗ ∗ ∗



GDPG



−0.390 (0.101)∗ ∗ ∗

3.285 (0.512)∗ ∗ ∗



IntPol



0.283 (0.144)∗∗

ExtPol



−0.320 (0.154)∗∗

−0.424 (0.161)∗ ∗ ∗

IntPol ×πG





1.214 (0.833)



IntNPol ×πG





1.445 (0.972)



ExtPol ×πG





5.213 (0.661)∗ ∗ ∗

IntPol × GDPG





−0.231 (0.185)



IntNPol × GDPG





−0.228 (0.200)



ExtPol × GDPG





−0.544 (0.132)∗ ∗ ∗

Chair



−0.113 (0.117)

−0.109 (0.119)

0.265 (0.150)∗

−0.276 (0.129)∗∗ – –



– 0.389 (0.165)∗∗

0.001 (0.009)

0.007 (0.009)

0.009 (0.009)

BankD

0.048 (0.008)∗ ∗ ∗

0.043 (0.009)∗ ∗ ∗

0.043 (0.009)∗ ∗ ∗



FinD

0.005 (0.009)

0.005 (0.010)

0.007 (0.010)



0.019 (0.008)∗∗

0.010 (0.009)

0.012 (0.009)



0.118 (0.018)∗ ∗ ∗

0.128 (0.019)∗ ∗ ∗

0.132 (0.020)∗ ∗ ∗



−0.280 (0.036)∗ ∗ ∗

−0.272 (0.041)∗ ∗ ∗

−0.273 (0.042)∗ ∗ ∗







Plenderleith,NP







Clementi,P







Vickers,NP

−0.519 (0.264)∗∗







Bean,NP

−0.004 (0.319)







Tucker,NP







0.045 (0.261)

Large,P







0.669 (0.274)∗ ∗ ∗

Lomax,P







Gieve,P

−0.380 (0.304)







Buiter,P

−0.431 (0.486)







Goodhart,P

−0.735 (0.287)∗ ∗ ∗







Julius,P

−0.359 (0.324)







Budd,P







Wadhwani,P







Nickell,P







Allsopp,P







Barker,P







Bell,P







Lambert,P

−1.433 (0.314)∗ ∗ ∗







Walton,P

−0.356 (0.355)







Blanchflower,P







Besley,P







0.862 (0.471)∗

Sentance,P







0.996 (0.491)∗∗

−1.523 (0.062)∗ ∗ ∗

−1.726 (0.147)∗ ∗ ∗

−1.811 (0.154)∗ ∗ ∗

AcadD

GovD IndD NGOD

George,P

γ1 γ2

−1.615 (0.066)∗ ∗ ∗

1.660 (0.146)∗ ∗ ∗



– −0.671 (0.261)∗∗

−0.558 (0.264)∗ ∗ ∗

−0.729 (0.294)∗ ∗ ∗

−2.145 (0.278)∗ ∗ ∗ 0.237 (0.390) −2.040 (0.285)∗ ∗ ∗ −1.032 (0.238)∗ ∗ ∗ −1.763 (0.288)∗ ∗ ∗ −0.635 (0.266)∗∗

0.171 (0.467) −1.897 (0.427)∗ ∗ ∗

1.622 (0.149)∗ ∗ ∗

−2.075 (0.203)∗ ∗ ∗ 1.538 (0.187)∗ ∗ ∗

432

Public Choice (2011) 146: 413–442

Table 5 (Continued) S UMMARY S TATISTICS AIC

1014.159

952.434

940.889

916.531

BIC

1053.773

1016.807

1025.069

1055.180

† King is the omitted variable, Davies dropped due to insufficient observations ‡ Standard errors in parentheses (·)  Denotes internal member  Denotes external member ∗ ∗ ∗ Denotes two-tailed significance at the 1% level. No of obs = 1045 ∗∗ Denotes two-tailed significance at the 5% level. No of obs = 1045 ∗ Denotes two-tailed significance at the 10% level. No of obs = 1045 P Denotes politically appointed member NP Denotes non-politically appointed member

with the FOMC literature, where career backgrounds play an important role in determining voting behavior.33 We further note that the career background parameters are very robust to specification change, as evidenced by the magnitudes and signs of corresponding parameter estimates across models 1, 2 and 3. The decision to augment the baseline specification in Model 1 with additional covariates is also clearly desirable based on AIC and BIC grounds. The most noteworthy innovation in models 2 and 3 is to control for the effects of different member types. Our findings support the conjecture that compared to channel of appointment effects, the internal-external distinction plays a significantly greater role in driving dissent voting behavior. Model 2 suggests that whereas politically appointed internals (IntPol) are more likely to dissent on the side of tightness than non-politically appointed internals (IntNPol), external members (ExtPol) are relatively more prone to dissent on the side of monetary ease. However, the parameter estimate for IntPol is smaller in absolute size than for ExtPol. Further, the fact that IntPol is positively signed suggests that being appointed by the executive branch of government promotes tightness dissents. This result stands at odds with our prediction that ceteris paribus, political appointees should either dissent on the side of ease, or behave in a way that is not significantly different from non-politically appointed internals. Any positive appointment channel effects, however, appear to be outweighed by effects appertaining to the insider-outsider distinction, and more specifically, the negative impact associated with a member’s external status: here, the confluence of being politically appointed and an external MPC member results a negatively signed parameter for ExtPol. 33 As an anonymous referee has suggested, one interpretation of our findings is that our priors were at best

speculative, and moreover, that certain member characteristics might not hold the same behavioral predictions for the Fed and the Bank of England. On the one hand, this observation is not entirely unreasonable given the particular nuances of each institution. Indeed, a general explanation for such differences may be attributable to the role of institutional design, and more specifically, how the associated democratic deficit is mitigated. Saying this, while our choice of priors involved making some judgement calls, it seemed reasonable to suppose that priors similar to those used in FOMC papers might be applicable to the MPC. Independent of this point, it was also highlighted that because the time frame used in Havrilesky and Schweitzer (1990) differs to ours, it may have affected our findings. Although not examined here, one way to address this problem would be to undertake a comparative study of both institutions using FOMC and MPC data over a comparable period.

Public Choice (2011) 146: 413–442

433

The impact of economic conditions is also notable. While πG and GDPG are generally highly significant in models 2 and 4, GDPG is negatively signed. This result is in line with other findings in the literature, namely Besley et al. (2008) and Harris and Spencer (2009). The magnitude of the πG parameter is substantially greater than that corresponding to all other model parameters, implying that deviations from target inflation play an important role in driving dissent. Model 3, which introduces interaction terms, reports similar findings, but the statistical significance of the estimated parameters is restricted to external members only: economic conditions assume no role in driving dissent for both politically and nonpolitically appointed internal members, for whom the associated model parameters are not significantly different to zero. This evidence suggests that the factors determining dissent are different for internals and externals. The fixed-effects specification in Model 4 excludes career characteristics on the grounds that because career experience is by definition specific to each member, it represents an individual characteristic, which in addition to factors such as a member’s information set and model of the economy, is implicitly captured by a member’s fixed-effects dummy.34 Crucially, most member dummies exhibited significance at the 5% level or less, and a test of the hypothesis that fixed-effects dummies are jointly insignificant was rejected 2 = 144.46, p = 0.000), as was a test for the joint equality of coefficients outright, (χ23 2 (χ22 = 137.14, p = 0.000). For completeness, tests for the joint equality of coefficients within different member subgroups were also conducted, and the null of no difference across rejected in every case (χ42 = 20.18, p = 0.0005 (IntPol); χ32 = 9.77, p = 0.0206 (IntNPol); 2 = 102.60, p = 0.000 (ExtPol)). Most member dummies are also negatively signed. This χ13 reflects the fact that King (the omitted dummy) cast a substantial number of tightness dissents relative to other members. Coupled with the tests of equality and joint significance, the number of statistically significant dummies is indicative of considerable voter heterogeneity, thus providing support for Blinder’s contention that the Bank of England MPC is individualistic.35 Finally, the effect under different Chairman is ambiguous—parameter es34 The inclusion of career covariates generated fixed-effects estimates which were considered nonsensical.

A possible explanation for this finding outcome is that career covariates are highly correlated with member dummies, and although they vary across time for each member, may lack sufficient variation to avoid collinearity. Results available from the authors on request. 35 For completeness, we also performed estimations treating the α in (3) as random. Wooldridge (2002), for g

instance, states that ‘it almost always make sense to treat the unobserved effects as random’. Using Stata, estimation in the presence of AR(1) errors was also implemented, using the aroprobit command of Heiss (2007). This latter strategy allows for the fact that ‘just as observed covariates can change over time, so too can unobserved influences and determinants of the outcomes’ (Heiss 2007: 1). The random effects model (3) thus becomes Z∗gt = xgt β + αgt + εgt ,

εgt ∼ iid(0, σε2 )

(3 )

where αgt = ραg,t−1 + vit ,

vit ∼ iid(0, (1 − ρ 2 )σ 2 ).

(3 )

Here, αgt is an i.i.d. random variable with zero mean and variance σ 2 , only now the unobserved heterogeneity obeys a stationary AR(1) process over time with correlation parameter ρ. It is noteworthy that if αgt = αg ∀t , (3 ) reduces to the standard random effects model (see Heiss 2007 for full details). Results are shown in Table 6, where we note that for Models 5 and 6 (random effects) ρ is interpreted as the proportion of variance explained by the panel-level variance, and for Models 7 and 8, ρ represents the correlation parameter in (3 ). We restrict our attention to career background covariates, the estimated parameters of which are highly similar to those in Models 1, 2 and 3: experience in academia and finance is not statistically significant, while industry and NGO experience exert the greatest impact (while still having incorrect signs). Interestingly, AIC

434

Public Choice (2011) 146: 413–442

Table 6 Ordered probit estimates: random effects and AR(1) errors Random effects Model 5 πG GDPG Chair

3.273 (0.508)∗ ∗ ∗ −0.313 (0.121)∗ ∗ ∗ 0.326 (0.163)∗

AR(1) errors Model 6 3.085 (0.495)∗ ∗ ∗ −0.234 (0.123)∗

Model 7

Model 8

4.492 (1.085)∗ ∗ ∗

4.627 (1.082)∗ ∗ ∗

−0.901 (0.307)∗ ∗ ∗

−0.888 (0.292)∗ ∗ ∗

−0.015 (0.135)

−0.693 (0.400)∗

−0.629 (0.353)∗

−0.006 (0.010)



0.003 (0.028)



0.092 (0.031)∗ ∗ ∗

AcadD



BankD



FinD





0.007 (0.027)

GovD



0.029 (0.010)∗ ∗ ∗



0.035 (0.030)

IndD



0.147 (0.024)∗ ∗ ∗



0.217 (0.067)∗ ∗ ∗

NGOD



−0.336 (0.050)∗ ∗ ∗



−0.535 (0.133)∗ ∗ ∗

γ1

−1.699 (0.149)∗ ∗ ∗

γ2

1.824 (0.151)∗ ∗ ∗

ρ

0.352 (0.057)∗ ∗ ∗

0.048 (0.010)∗ ∗ ∗

−0.016 (0.013)

−1.745 (0.127)∗ ∗ ∗

−3.210 (0.530)∗ ∗ ∗

0.426 (0.074)∗ ∗ ∗

0.868 (0.035)∗ ∗ ∗

1.881 (0.132)∗ ∗ ∗

2.426 (0.390)∗ ∗ ∗

−3.297 (0.553)∗ ∗ ∗ 2.657 (0.411)∗ ∗ ∗ 0.807 (0.046)∗ ∗ ∗

Summary statistics AIC

951.273

936.480

864.268

853.755

BIC

980.984

995.901

898.931

918.128

Standard errors in parentheses (·). No of obs = 1045 ∗ ∗ ∗ Denotes two-tailed significance at the 1% level ∗∗ Denotes two-tailed significance at the 5% level ∗ Denotes two-tailed significance at the 10% level

timates range from negative and insignificant (Models 2 and 3), to positive and significant (Model 4). In terms of model selection, the information criteria do not select any model unanimously: according to the BIC, Model 2 performs best, while Model 4 is preferred by AIC. In some respects, this result is unsurprising: due to its asymptotic consistency and its heavy penalty on complexity, BIC typically selects more parsimonious specifications. Conversely, AIC often chooses less parsimonious specifications as complexity is not so heavily penalized, especially for small or moderate sample sizes. However, it is notable that alternative goodness of fit measures not reported here (McFadden’s R2 and Adjusted R2 , Cox-Snell Maximum Likelihood R2 , Cragg-Uhler R2 and McKelvey and Zavoina’s R2 ) all identify Model 4 as having the best explanatory power.

5 Conclusion At the outset of this paper, we set out to explain the type and frequency of dissenting votes cast by MPC members, by appealing to career backgrounds, the channel of appointment, and unobserved heterogeneity. The findings of the FOMC literature, which establishes a and BIC identify the AR(1) model, which confirms the presence of high degree first order serial correlation, as superior to the RE specifications.

Public Choice (2011) 146: 413–442

435

prominent role for such factors in determining voting behavior raised the hereunto neglected question of whether votes on the policy rate at monthly Bank of England MPC meetings were similarly influenced. To motivate the issue, we rationalized why MPC members might be incentivized to vote in different ways, by proposing that the internal-external distinction was too simplistic, and paying particular attention to the institutional features governing UK monetary policy. While it was found that voting patterns exhibit considerable heterogeneity across all members, the econometric evidence initially presented something of a ‘career background puzzle’—unlike the FOMC literature, career experience plays a very weak role in determining a member’s decision to dissent; moreover, where career backgrounds are significant, they are often counter-intuitively signed. A credible resolution to this puzzle is found by appealing to how the democratic deficit associated with UK monetary policy is mitigated: constituent group pressure may more plausibly manifest itself through the choice of the inflation target itself, as opposed to the individuals chosen to set policy. We also proposed that the channel of appointment for the MPC is considerably less politicized than for its US counterpart, conjecturing that it would play a very limited role in determining voting behavior. Our empirical results broadly support this view, and serve to highlight the relationship between the UK political and parliamentary system, and the associated monetary policy framework. Again, central to this relationship is the mitigation of the so-called democratic deficit of monetary policy making, which is achieved through the institution of an explicit inflation target. We did find, however, that the internal-external distinction plays an important role, which may be attributable to career concerns and organizational factors. While the role of different chairmen was found to be ambiguous, the deviation of the MPC’s inflation forecast from target was found to heavily influence external dissents, but not those associated with internals, whose parameter estimates were statistically no different to zero. An alternative approach was to assume that the determinants of dissent are best captured by member-specific fixed-effects. Econometric estimates provide strong encouragement for the use of such an innovation, and suggest that the MPC is indeed individualistic (Blinder 2007). Integral to this approach is the view that although demarcating between different MPC member types during estimation may be fruitful, such a strategy fails to capture the true extent of member heterogeneity. Finally, future work in this area might systematically compare the determinants of dissent voting in different monetary institutions, such as the US Federal Reserve, Swedish Riksbank and Bank of Japan. Our findings may also have important implications for the institutional design of policy institutions such as the US Federal Reserve: if one’s aim is to reduce the role of politics and partisanship in shaping votes on the FOMC—and thereby US monetary policy decisions—a potential mechanism for doing this may be to adopt a transparent and explicit inflation targeting regime, as is current practice in the UK. This may, however, simply relocate the problem of governmental and constituent group influence to the choice of target itself. Moreover, as more central banks delegate monetary policy to committees, and place the voting record in the public domain, investigation of this issue can only heighten our understanding of monetary policy, and of the individuals and institutions that shape it. Acknowledgements Spencer acknowledges financial support from the Economic and Social Research Council (ESRC Postgraduate Studentship R42200134224). This paper is a substantially re-worked version of “The Dissent Voting Behaviour of Bank of England MPC Members”, University of Surrey Department of Economics Working Paper No. 03/06. Authors are listed alphabetically. We would like to thank Ali Choudhary, Andy Dickerson, Stephen Drinkwater, Huw Edwards, Carré Emmanuel, Alessandro Flamini, Anthony Glass, Karligash Kenjegalieva, Akira Murayama, Tom Weyman-Jones, two anonymous referees, the Editorin-Chief of this journal, and seminar participants at the Department of Economics, University of Sheffield for helpful comments and suggestions. The authors also extend special thanks to Nicoletta Batini and Willem Buiter for additional clarification of the MPC resources dispute in 1999. The usual disclaimer applies.

436

Public Choice (2011) 146: 413–442

Appendix A: Resource control In 1999 a dispute over MPC resources emerged, which subsequently became the subject of a House of Commons Treasury Select Committee report. The report, entitled Report on Research Assistance for Monetary Policy Committee Members (Treasury Select Committee 1999), centered on the experience of a group of external committee members (Willem Buiter, Charles Goodhart, DeAnne Julius, and Sushil Wadhwani) who were refused research resources that were requested from the executive leadership at the Bank: the Governor (Eddie George) and two Deputy Governors (Mervyn King and David Clementi) were all united in their opposition to the request.36 On becoming public, the dispute received coverage in the financial press and news media. Newspapers reported that some externals believed higher productivity was easing long-term inflationary pressure, but felt they had insufficient resources to support their view.37 Whilst the Governor publicly described the dispute as a “storm in a teacup”,38 the Treasury Select Committee report concluded that this was not the case, and warned the Bank to better manage future disagreements. Although the dispute was ultimately resolved—such that external members requiring extra resources are now allocated one post-graduate and one graduate economist each—the process by which it was settled is of particular interest. Prior to becoming public, the matter was taken to the Court of the Bank of England for arbitration. At the time of the dispute, the Court was comprised of the Governor, the two Deputy Governors, and sixteen independent non-executive Directors, with the latter sixteen members comprising a sub-committee in its own right called NedCo. Although neither body plays a role in the formulation of monetary policy, both the Court and Nedco enjoy important roles with respect to the governance of the Bank. Whereas the Court has overall responsibility for management of the Bank’s affairs (this includes determining Bank strategy and ensuring the most efficient use of the Bank’s resources), NedCo plays an important monitoring role, which includes ensuring that the MPC’s processes and procedures work effectively. The role of the Court in attempting to settle the dispute has been heavily criticized by Willem Buiter: “During my three years on the MPC, the external MPC members once tried to involve the Court in a matter that ought to have been of interest to the Court, given its role in the governance of the institution. This was during the conflict between the external members of the MPC and the Bank’s executive leadership, about the creation of a dedicated research unit to support the external MPC members, independent of the Bank’s Executive. The Court was utterly ineffective and resolved nothing.” (Buiter 2008) The Treasury Committee singles out NedCo for similar criticism, asserting that: 36 In e-mail correspondence with the authors, Willem Buiter indicated that while Charles Goodhart wanted to

prevent a confrontation, for other members this was not the case. For instance, Buiter himself claims to have had strong arguments with the Governor and David Clementi about the issue. 37 See “Bank says MPC research row has been settled”, The Guardian, Wednesday November 24, 1999; also

see further articles in The Independent on October 30th 1999, November 22th 1999 and November 24th 1999. While we are not in a position to substantiate all of the claims made in the newspaper reports cited here (i.e., most of the events described therein took place behind closed doors, and may have been based on the testimony of anonymous sources from within the Bank), it is clear from the Treasury Committee report and personal statements made by MPC members, that the reports are not without some foundation. 38 Select Committee on Treasury, Minutes of Evidence, Examination of Witnesses, Question 81, Tuesday 23rd November 1999.

Public Choice (2011) 146: 413–442

437

“The issue was only fully resolved after it had been made public and became a focus of one of our public hearings. We believe that the Non-Executive Directors [of the Court], need to be much more pro-active in ensuring that the procedures of the MPC operate fairly with respect to both internal and external members.” (Select Committee on Treasury, Ninth Report (paragraph 40), prepared 28th March 2001. Words in square brackets added by the authors). Although this matter is not pursued any further in this paper, we suggest that the aspects of resource control covered here, and their implications for voting, are clearly deserving of further attention in the literature. For further details relating to this affair, and the role of the Treasury Select Committee in relation to the MPC and UK monetary policy formulation, the reader is referred to Treasury Select Committee (2007a) and Treasury Select Committee (2007b).

Appendix B: The Havrilesky-Schweitzer (H-S) model In setting out the formal model, we envisage a MPC composed of g members, each of whom may have amassed career experiences, for different durations, in j different fields. These experiences are referred to by H-S as a member’s career characteristics. The practical analogue of these j characteristics would be experiences in different sectors or areas of the economy, such as private industry and finance. These measures are developed in Sect. 4.1.1. For member g, denote her j th career characteristic as Xgj , such that X j represents the MPC’s mean for that characteristic. So-called ‘career proximity’ to central government is increasing in Xgj − Xj such that Xgj − Xj > 0 (< 0) promotes dissents on the side of monetary ease (tightness). As Xgj − Xj > 0 (< 0) becomes larger (smaller), so too does the propensity to dissent on the side of ease (tightness). However, given there are j characteristics, the extent to which a given MPC member dissents is ultimately a function of how each characteristic is weighted. We are now in a position to write an expression for member g’s utility, namely Ug (Dg ) = U (Dg | Xgj − X j , j = 1, 2, . . . , N ).

(B.1)

It is further assumed that (B.1) is characterized by a unique global maximum that defines the optimal number of dissenting votes, D g , such that ∂U ∂(Dg − D g ) ∂ 2U ∂(Dg − D g )2

< 0,

(B.2)

< 0.

(B.3)

In (B.1), the utility achieved by member g is a function of the number of dissenting votes cast, the direction and number of which is conditioned by career proximity parameters, Xgj − X j , j = 1, 2, . . . , N . It turns out that the actual number of dissents cast by member g, Dg , will not necessarily equal the number of dissents which maximize utility. This is because members also experience disutility, an expression for which is given by Vg (Dg ) = V (Dg | Xgj − Xj , j = 1, 2, . . . , N )

(B.4)

438

Public Choice (2011) 146: 413–442

which has a unique global minimum at Dg = 0. Here, the assumed properties of Vg (Dg ) ∂V > 0, ∂Dg

(B.5)

∂ 2V >0 ∂Dg2

(B.6)

imply that as the number of dissents moves further away from zero, the disutility felt by member g increases at an increasing rate. It is easily shown that when the marginal utility of increasing dissent equals the marginal disutility of increasing dissent, member g’s net utility will satisfy an unconstrained maximum where: ∂V ∂U = . ∂Dg ∂Dg

(B.7)

Put another way, marginal net utility must be zero. To glean the normative implications of the model, H-S consider the conditions required to ensure a monotonic transformation from the weighted career characteristic differences, Xgj − X j , to the actual number of dissents, Dg . Due to the nature of the first order conditions for utility and disutility in (B.1) and (B.4), the actual number of dissents is not guaranteed to map monotonically onto career characteristic differences. Ensuring such a transformation requires the restriction that the marginal net utility of the j th member increasing dissent towards her global optimum is strictly less than that pertaining to the (j + kth) member: as Havrilesky and Schweitzer state, this holds the implication that “a member with marginally stronger moral convictions in favor of dissenting cannot be marginally more easily cowed by group. . . disapproval”. This can be formally proved as follows. In order to partially differentiate, express the net welfare of member g as Wg = Ug − Vg = Wi (Dg , Xgj − X j , j = 1, 2, . . . , N ). The first order condition for a maximum is then ∂Wg (Dg , Xgj − X j , j = 1, 2, . . . , N ) = 0 ∂Dg

(B.8)

which results in an optimal choice Di = Di∗ (Xgj − X j , j = 1, 2, . . . , N ). From (B.3) and (B.6) ∂ 2 Wg 0 and there is a monotonic transforma-

tion from the each of the career characteristic differences, Xgj − X j , to the actual number of dissents Dg , iff

∂ 2 Wg ∂ Xˆ gj ∂Dg

=

∂ ∂ Xˆ gj

∂W

( ∂Dgg ) > 0. That is, the marginal net utility of the j th mem-

ber increasing dissent towards her global optimum is strictly less than that pertaining to the (j + kth) member. This proves the result.

Appendix C: Testing for group differences based on members’ internal-external status and the channel of appointment Building on the analysis in Sect. 3, Chi-squared tests were extensively used to test for differences in members’ voting behavior, based on the criteria used to construct the three sub-groups examined in Table 1. Results are displayed in Table 7. Specifically, differences associated with three alternative, but highly related criteria were examined: first, that the pattern of votes for assent, ease and tightness were no different between groups; second, that the pattern of votes for ease and tightness dissents were not significantly different; and third, that the pattern of votes for assenting and dissenting votes per se did not significantly differ. We initially proceeded with a joint test of the null hypothesis that the three subgroups introduced in Table 1 were not significantly different in their voting behavior (H0 : Internal MembersP = Internal MembersNP = External MembersP ). While the null of no difference was rejected at the 1% level for each criterion, subsequent tests between different pairs of sub-groups revealed a more complex picture. Most interestingly, political and nonpolitically appointed internal members (InternalP vs. InternalNP ) did not significantly differ in their behavior (p > 0.1 in every case, indicating acceptance of the null of no difference between these groups). There were, however, statistically significant differences between internal members— whether politically appointed, not-politically appointed, or collectively—and external members. This finding provides support for the notion that a member’s internal-external status has a significant bearing on voting behavior. Moreover, the results at the foot of Table 7—which indicate highly significant differences between political and non-political appointments— should be interpreted with caution: such differences are arguably driven by internal-external distinction. Indeed, while politically appointed internal members are generally more likely to assent and cast the majority of dissenting votes on the side of tightness—summing across votes for these two groups resulted in a group which retains the dissent voting characteristics of external members (note that Political appointments = Internal MembersP + External MembersP , whereas Non-political appointments = Internal MembersNP only). As an example, the sheer number of dissenting votes cast by external members led to an overwhelming tendency for politically appointed members to dissent on the side of monetary ease.

References Adolph, C. (2005). Three simple tests of career influences on monetary policy. Paper presented at the annual meeting of the American Political Science Association, Marriott Wardman Park, Omni Shoreham, Washington Hilton, Washington, DC, September 1st, 2005. Alesina, A., & Roubini, N. (1992). Political cycles in OECD economies. Review of Economic Studies, 59(4), 663–688. Bank of England. (1999). The transmission mechanism of monetary policy. Bank of England. Available from www.bankofengland.co.uk.

Public Choice (2011) 146: 413–442

441

Belongia, M. T. (2007). Opaque rather than transparent: why the public cannot monitor monetary policy. Public Choice, 133, 259–267. Besley, T., Meads, N., & Surico, P. (2008). Insiders versus outsiders in monetary policymaking. American Economic Review (Papers and Proceedings), 98(2), 218–223. Bhattacharjee, A., & Holly, S. (2005). Inflation targeting, committee decision making and uncertainty: the case of the Bank of England’s MPC. Cambridge Working Papers in Economics 05/30, University of Cambridge. Blanchflower, D. (2006). Reflections on my first four votes on the MPC. Speech given at the Bank of England’s Agency for Wales, Wednesday 27th September 2006. Blinder, A. (2007). Monetary policy by committee: why and how? European Journal of Political Economy, 23(1), 106–123. Blinder, A., & Wyplosz, C. (2005). Central bank talk: committee structure and communication policy. Conference paper, Central Bank Communication, at the ASSA meetings, Philadelphia. Brooks, R., Harris, M., & Spencer, C. (2007). An inflated ordered probit model of monetary policy: evidence from MPC voting data. MPRA Paper 8509, University Library of Munich, Germany. Buiter, W. (1999). Answers to the questionnaire ‘Bank of England commission; issues for consideration’. December 17th 1999. Buiter, W. (2008). A financial stability committee for the Bank of England? http://blogs.ft.com/maverecon/ 2008/06/a-financial-stability-committee-for-the-bank-of-england/, June 11, 2008. Chappell, H. W., & McGregor, R. R. (2000). A long history of FOMC voting behavior. Southern Economic Journal, 66(4), 906–922. Chappell, H. W. Jr., Havrilesky, T. M., & McGregor, R. R. (1993). Partisan monetary policies: Presidential influence through the power of appointment. Quarterly Journal of Economics, 108(1), 185–218. Chappell, H. W. Jr., Havrilesky, T. M., & McGregor, R. R. (1995). Policymakers, institutions, and central bank decisions. Journal of Economics and Business, 47(2), 113–136. Chappell, H. W. Jr., McGregor, R. R., & Vermilyea, T. (2004). Majority rule, consensus building, and the power of the Chairman. Journal of Money, Credit, and Banking, 36(3), 408–422. Chappell, H. W. Jr., McGregor, R. R., & Vermilyea, T. (2005). Committee decisions on monetary policy. Cambridge: MIT Press. Cobham, D. (2000). Restoring trust at the Treasury. Financial Times. 10th August 2000. Du Plessis, S. A. (2005). The democratic deficit and inflation targeting. South African Journal of Economics, 73(1), 93–104. Gerlach-Kristen, P. (2003). Insiders and outsiders at the Bank of England. Central Banking, 14(1), 96–102. Gerlach-Kristen, P. (2009). Outsiders at the Bank of England’s MPC. Journal of Money, Credit and Banking, 41(6), 1099–1115. Gildea, J. A. (1990). Explaining FOMC members’ votes. In T. Mayer (Ed.), The political economy of American monetary policy (pp. 211–228). Cambridge: Cambridge University Press. Goodhart, C. A. E. (2001). The inflation forecast. National Institute Economic Review, 175, 59–66. Goodhart, C. A. E. (2005). The Monetary Policy Committee’s reaction function: an exercise in estimation. Topics in Macroeconomics 5(1), Article 18. Goodhart, C. A. E. (2006). Letter. In The state of monetary policy. Select Committee on Economic Affairs, House of Lords, First Report, December 18th. Goodhart, C. A. E., & Meade, E. (2003). Central banks and supreme courts. Mimeo, London School of Economics. Greene, W. (2004). The behaviour of the maximum likelihood estimator of limited dependent variable models in the presence of fixed effects. Econometrics Journal, 7(1), 98–119. Grilli, V., Masciandaro, D., & Tabellini, G. (1991). Political and monetary institutions and public financial policies in the industrial countries. Economic Policy, 6, 341–392. Harris, M. N., & Spencer, C. (2009). The policy choices and reaction functions of Bank of England MPC members. Southern Economic Journal, 76(2), 482–499. Havrilesky, T., & Gildea, J. A. (1992). Reliable and unreliable partisan appointees to the Board of Governors. Public Choice, 73(4), 397–417. Havrilesky, T., & Schweitzer, R. (1990). A theory of FOMC dissent voting with evidence from the time series. In T. Mayer (Ed.), The political economy of American monetary policy (pp. 197–210). Cambridge: Cambridge University Press. Heckman, J. (1981). Statistical models for discrete panel data. In C. Manski, & D. McFadden (Eds.), Structural analysis of discrete data with econometric applications. Cambridge: MIT Press. Heiss, F. (2007). Limited dependent panel data models with AR(1) error components in Stata. Mimeo, University of Munich, Department of Economics, December 2007. Howard, M. (2002). Speech to the Institute for Public Policy Research. Monday March 11, 2002.

442

Public Choice (2011) 146: 413–442

Keech, W. R., & Morris, I. L. (1997). Appointments, presidential power, and the Federal Reserve. Journal of Macroeconomics, 19(2), 253–267. Krause, G. A. (1994). Federal Reserve policy decision making: political and bureuacratic influences. American Journal of Political Science, 38(1), 124–144. Krause, G. A. (1996). Agent heterogeneity and consensual decision making on the Federal Open Market Committee. Public Choice, 88(1), 83–101. McCallum, B. T. (1995). Two fallacies concerning central bank independence. American Economic Review (Papers and Proceedings), 85, 207–211. McGregor, R. R. (2007). Federal Reserve transparency: the more things change, the more they stay the same? Public Choice, 133, 269–273. Meade, E. E., & Stasavage, D. (2008). Publicity of debate and the incentive to dissent: evidence from the US Federal Reserve. Economic Journal, 118(525), 695–717. Morris, I. L. (2002). Congress, the President, and the Federal Reserve: the politics of American monetary policy-making. Ann Arbor: University of Michigan Press. Mueller, D. C. (2005). Constitutional political economy in the European Union. In W. F. Shughart II, & R. D. Tollison (Eds.), Policy changes and political responses: public choice perspectives on the post 9/11 world (pp. 57–73). Berlin: Springer. Nakahara, N. (2001). Speech given by Nobuyuki Nakahara, member of the Policy Board of the Bank of Japan, at the Capital Markets Research Institute in Tokyo on December 11, 2001. Neyman, J., & Scott, E. (1948). Consistent estimates based on partially consistent observations. Econometrica, 39, 359–382. Riboni, A., & Ruge-Murcia, F. J. (2008). Preference heterogeneity in monetary policy committees. International Journal of Central Banking, 4(1), 213–233. Santoni, G. J. (1984). A private central bank: some olde English lessons. Federal Reserve Bank of St. Louis Review, 66, 12–22. Shughart, W. F. II, & Tollison, R. D. (1983). Preliminary evidence on the use of inputs by the Federal Reserve system. American Economic Review, 73(3), 291–304. Taylor, J. (1993). Discretion versus policy rules in practice. Carnegie-Rochester Conference Series on Public Policy, 38, 195–214. Treasury Select Committee (1999). Research assistance for Monetary Policy Committee members. Treasury Committee Report HC 43 290, 9 December 1999. Treasury Select Committee (2007a). The Monetary Policy Committee of the Bank of England: ten years on, Vol. I. Treasury Committee Report HC 299-I, published on 18 September 2007. Treasury Select Committee (2007b). The Monetary Policy Committee of the Bank of England: ten years on, Vol. II. Treasury Committee Report HC 299-II, published on 19 February 2007. Tucker, P. (2008). Remarks on ‘Making monetary policy by committee’. Discussant remarks by Mr Paul Tucker, Executive Director and Member of the Monetary Policy Committee of the Bank of England, on “Making Monetary Policy by Committee” by Alan Blinder, at the Bank of Canada Conference “International Experience with the conduct of Monetary Policy Under Inflation Targeting”, Ottawa, 22–23 July 2008. Vaubel, R. (1997). The bureaucratic and partisan behavior of independent central banks: German and international evidence. European Journal of Political Economy, 13(2), 201–224. Wadhwani, S. B. (2002). The MPC: some further challenges. National Institute of Economic and Social Research Governors’ Seminar. National Institute of Economic and Social Research. Wooldridge, J. (2002). Econometric analysis of cross section and panel data. Cambridge: MIT Press.