Monetary technocracy and democratic accountability: how central bank independence conditions economic voting

Abstract

Central bank independence (CBI) implies that elected governments delegate monetary policy to technocrats in central banks. I argue that given the substantial influence of monetary policy on consumption, investments, exchange rates, capital flows and government spending, all of which critically determine the performance of the economy, CBI can blur the lines of responsibility for economic performance between elected governments and central banks. It can thereby weaken voters’ ability and willingness to electorally punish (or reward) governments on the basis of economic outcomes. Utilizing data from the Comparative Study of Electoral Systems, I test how CBI conditions the effects of macroeconomic performance on electoral support for incumbents in 38 countries from 1996 to 2016. The results show that CBI significantly attenuates the reward and punishment mechanism of elections based on economic records. The finding of this article sheds new lights on how the problem of democratic accountability caused by the rise of CBI actually materializes in elections, the most important sanctioning mechanism of democracy. Further, it identifies CBI as another crucial condition that can explain variations in the magnitude of economic voting across countries, as yet unexplored in the election literature.

Keywords:

• Central bank independence
• economic voting
• monetary policy
• clarity of responsibility
• democratic accountability

Acknowledgments

I would like to thank the RIPE editors, four anonymous reviewers, Eric Chang, Christian Houle, and Shahryar Minhas for insightful comments and suggestions on this article. Participants at the 2019 American Political Science Association Annual Meeting also provided helpful feedback. Finally, I would like to express my special thanks to Cristina Bodea for her invaluable advice for this article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data that support the findings of this study are openly available here: https://www.dropbox.com/sh/y195g4y6p2302nl/AADenGXFcjMYvfBddCHvh5fMa?dl=0

Notes

1 “Democratic accountability” refers here to "the electorate's capacity to reward or sanction incumbent politicians" (Samuels, 2004, p. 425).

2 Freeman, 2002) defines room to maneuver as "the degree to which officials in one country can choose a distinctive mix of welfare outcomes for their citizens" (p. 890).

3 Carlin and Hellwig’s recent study on Latin America (2020) addresses the conditional effects of “neoliberal reforms” on economic voting. Yet they do not include CBI as one of those reforms in their theoretical discussion. Rather, they focus on the influence of trade, financial markets, tax reform, private sector ownership, labor markets, and minimum wage.

4 Often, the public finds the problem of high unemployment and low growth under CBI so serious that citizens come to demand that the banks improve those conditions rather than focusing too much on low inflation. For example, in 2017 New Zealanders elected a coalition government that included the populist New Zealand First Party and the Labour Party, which had campaigned on a promise to add maximum sustainable employment to the objectives of the Reserve Bank of New Zealand (the RBNZ). As a result, the RBNZ, once considered one of the most independent central banks in the world, now has dual mandates of price stability and full employment (Binder, 2020).

5 For more evidence on the negative effects of CBI on economic growth and employment, see Fischer & Capie, 1994) and Soskice & Iversen, 2000).

6 See Lewis-Beck & Stegmaier, 2000) for a review of the economic voting literature.

7 I am grateful to Ju Yeon Park for generously sharing her study’s replication dataset (Park, 2019).

8 Respondents’ votes cast for other parties that constitute a coalition government are coded 0.

9 Unemployment data was retrieved from the World Bank and real GDP growth data from the Penn World Table.

10 As an additional robustness check, I add another analysis using a dichotomous CBI variable (CBI-High) coded 1 if a country’s CBI value is greater than 0.5 and coded 0 otherwise. This is included in the Appendix (Table A.7). The main findings are robust to the use of this CBI dummy variable. Further, I replace the aggregate CBI measure with each of its four components in separate models (Table A.4 and A.5 in the Appendix). The results show that those components also reduced the magnitude of economic voting in six models out of eight.

11 One may argue that because ‘within-country’ variations in CBI values tend to be relatively small, only a weak inference can be drawn from country fixed effects models. This is a valid argument. Yet, it must be noted that important studies of CBI (e.g., Aklin & Kern, 2021; Bodea & Hicks, 2015a, Bodea & Hicks, 2015b; Clark & Hallerberg, 2000; Eijffinger et al. , 1996; and Garriga, 2016) have relied on country fixed effects to estimate the influence of CBI on various outcome variables despite the aforementioned problem. Presumably this is because the authors deem the benefits of the strong control on unobserved country heterogeneity to be greater than the drawbacks. Still, to demonstrate the robustness of my findings, I add another analysis. In Table A.8 in the Appendix, I re-estimate my main models using random effects as a robustness check. The outcomes in the table show that CBI still significantly reduces the effects of both unemployment and growth on respondents’ votes for incumbents.

12 Angrist & Pischke, 2008; Steenbergen & Jones, 2002.

13 Pregibon’s Delta-Beta statistic measures the standardized change in coefficients “due to the deletion of the observation along with all others that share the same covariate pattern.” (Maor et al., 2013, p. 597) Studies using a dichotomous dependent variable often employ this diagnostic test to identify unusually influential observations or groups of them because “maximum likelihood estimation of binary choice models tends to be more sensitive to outlier observations” (Dettrey & Palmer, 2013, p. 725). Generally, an observation is considered problematic when its Pregibon Delta-Beta statistic is above 1.0 (Dettrey & Palmer, 2013). According to my estimation, no observation in my sample has a Pregibon Delta-Beta statistic larger than 1.0. In fact, no observation in my sample has a Pregibon Delta-Beta statistic larger than 0.4. Some scholars suggest that observations with Pregibon Delta-Beta statistics three times larger than the average can be considered influential observations (Thachil, 2014). Based on this definition of influential observations, I re-estimate my main models excluding those observations with a Pregibon Delta-Beta value three times larger than the average across all the observations and present the outcomes in Table A.3 in the Appendix.

Additional information

Notes on contributors

Hyunwoo Kim

Hyunwoo Kim is Visiting Assistant Professor at the Department of Political Science at Michigan State University. He studies the political economy of macroeconomic policy and international finance.