INTRODUCTION
Transparency in political decision-making is often proposed as a remedy to failures of the political agency. Transparent individual decisions are thought to foster accountability and align the interests of politicians/agents with voters/principals. Many countries have implemented open government reforms (Carey Reference Carey2008) and grant public access to recorded individual votes, i.e. votes with the names of politicians voting for and against a proposal.
However, a small number of recent theoretical contributions (e.g. Fox Reference Fox2007; Prat Reference Prat2005) discuss potential negative consequences of transparency in political decision-making on the quality of policy outcomes and representation of voter preferences, particularly when actions of politicians such as individual votes are recorded and made publicly available. Empirical evidence of the causal effects of full transparency on political decision-making is currently scarce at best. Therefore, the present paper attempts to fill this gap using an especially informative natural setting.
Voters in Switzerland reveal their policy preferences in referenda. Parlia- mentarians vote on identical policy issues as voters in referenda. Thus, we can directly confront referendum results with parliamentary decisions in the Lower and Upper House of the Swiss Parliament to obtain a direct measure of divergence, i.e. an inverse measure for the quality of politicians’ decisions with respect to revealed voter preferences. Since the winter session of 2006, publicly available video streams of the Upper House’s sessions allow identifying votes of individual politicians, instead of only aggregate decisions prior to 2006. In the Lower House, final individual votes of politicians have been recorded electronically and published before and after the year 2006. Thus, we exploit a difference-in-difference setting to identify the effect of increased transparency of individual votes on the quality of political decisions with respect to voter preferences. Empirical results indicate that making it easier to identify individual votes and actions of politicians does not necessarily improve the quality of representation in terms of aligning political decisions and voter preferences more closely.
We present the empirical identification strategy in Section 2, assess the causal effect of transparency on the quality of political representation in Section 3, and conclude in Section 4.
INSTITUTIONAL SETTING AND IDENTIFICATION STRATEGY
We exploit a unique quasi-natural institutional setting in Switzerland to identify the effect of transparency of politicians’ votes in Parliament on the quality of political decisions with respect to voter preferences.
Two houses, the Lower House (National Council/Nationalrat) and the Upper House (Council of States/Ständerat) make up the Swiss Parliament. Both houses have the same amount of political power and decide on the same legislative and constitutional amendments. Parliamentary proposals have to be approved by the majority of members in each house in order to be adopted. Members of the Lower House are elected under a system of proportional representation with open party lists and the members of the Upper House are elected by plurality rule.Footnote 1 The aggregated results of the final policy decisions of both houses are published.
In an effort toward open government reforms, video streams of the parliamentary sessions have been made publicly available on the internet since the winter session of 2006. Thus, both the discussion and individual final votes of members of the Upper House can be identified as politicians vote by raising their hands. Before 2006, identifying individual votes in the Upper House after the session was impossible. Importantly, parliamentarians did not decide themselves on the publication of the recordings but neither did they impede the Parliamentary Services from releasing them. Greater transparency in the Upper House was constantly demanded by different proponents, including certain party officials, the media, and other groups.Footnote 2 The individual final votes of politicians in the Lower House have been recorded electronically and have been made publicly available since 1995.
The legislative and constitutional proposals of Parliament are not directly enacted. Before proposals are enacted, Swiss voters are given the opportunity to demand a referendum and they can propose constitutional amendments by initiatives (Frey Reference Frey1994; Portmann et al., Reference Portmann, Stadelmann and Eichenberger2012).Footnote 3 Referendum results reflect voter preferences with regard to constitutional and legislative proposals (Schneider et al., Reference Schneider, Pommerehne, Frey, Hibbs and Fassbender1981). Voters and parliamentarians decide on identical proposals with exactly the same wording. Thus, the parliamentary decisions can be matched and directly compared with revealed voter preferences in referenda (Carey and Hix Reference Carey and Hix2013; Hersch and McDougall Reference Hersch and McDougall1988; Matsusaka Reference Matsusaka2010).
We take the absolute difference between the voter yes-share in a referendum and the yes-share of politicians in their parliamentary decisions in the respective house as a natural and inverse measure of the quality of political decisions, i.e. we observe divergence from voter preferences in the two houses. The absolute difference to the voter yes-share of each house represents a meaningful measure of preference representation (Golder and Stramski Reference Golder and Stramski2010). Different levels of divergence for the two houses are due to the electoral system, which remained unchanged over the period analyzed, and thus makes the Lower House a control group for the Upper House, where transparency of individual decisions changed. Divergence from voter preferences is associated with adverse consequences, particularly in relation to candidates’ election prospects (Stadelmann et al., Reference Stadelmann, Portmann and Eichenberger2013b).
Similar to other countries, Swiss parliamentarians rely on common tools (e.g. surveys, personal experience, etc.) to infer voter preferences as they are only revealed after parliamentary decisions which is essential when evaluating legislative shirking (Garret Reference Garrett1999). Therefore, our results are likely to generalize to decisions in countries where policy preferences of voters are not directly observable from referendum results.
Our dataset consists of 91 referenda that cover a broad range of different topics. The corresponding, identical parliamentary decisions of each of the two houses took place during the three legislatures (46th, 47th, 48th) from 1999 to 2011, leading to a total of 182 observations. The data are available from the Swiss Parliamentary Services.
The identification strategy follows directly from the institutional setting under examination: For the entire period of analysis, we observe actual divergence from revealed voter preferences of the Lower and the Upper House. From 2006 onwards (during the third year of the 47th legislature, over a year prior to elections), the individual votes of members of the Upper House were made transparent by the introduction of video streams. Individual votes in the Lower House were made public throughout the period of analysis. Thus, we employ a standard difference-in-difference strategy to identify the causal impact of increased transparency on the quality of political decisions with respect to voter preferences.
RESULTS
Figure 1 and its accompanying Table 1 convey the central effect of increased transparency on divergence from voter preferences. Before the public availability of video streams (prior to the winter session of 2006), the average divergence from voter preferences was 15.4%-points in the Lower House and 27.9%-points in the Upper House. After the public availability of video streams, the average divergence was 14.7%-points in the Lower House and 26.9%-points in the Upper House. Thus, the reduction in divergence corresponds to 0.7%-points for the Lower and 1.0%-points for the Upper House. The difference-in-difference identifies the causal effect of the increase in transparency which corresponds to a meager and insignificant 0.3%-points lower divergence. These results do not provide systematic support to the common claims that increased transparency generally improves the quality of political representation.
Descriptive Results—The Effect of Full Transparency on the Divergence From Voter Preferences
Notes: ***, **, and * indicate significance levels of below 1%, between 1 and 5%, and between 5 and 10%, respectively.
Table 2, specification (1) confirms the baseline results from Figure 1 with a dummy for the Upper House, a dummy indicating the public availability of video streams, and an interaction between the two to identify the effect of increased transparency. Specification (2) shows that even in the short term (i.e. the 30 matched parliamentary/referenda decisions made directly before and after the introduction of full transparency) no significant impact due to transparency is observed, and the point estimate of the interaction effect turns positive.Footnote 4
Difference-in-Difference—Full Transparency of Individual Votes Due to Video Stream Availability Does Not Reduce Divergence of Politicians From Voter Preferences
Notes: The dependent variable for all estimations is “Divergence from voter preferences”. Robust standard errors are reported throughout the table. ***, **, and * indicate significance levels of below 1%, between 1% and 5%, and between 5% and 10%, respectively.
Transparency in political decisions may be particularly important when conflicts of interest arise. Therefore, we focus on a subsample where business associations and trade unions issue different voting recommendations. However, we still do not observe any significant impact due to the existence of public video stream availability (specification 3). Similarly, when left and right political parties issue different recommendations (specification 4), greater transparency does not play any significant role in decreasing divergence either.Footnote 5
Making the decision processes in politics more transparent may have its disadvantages when voters have priors for controversial issues. Initiatives are usually more controversial than proposals brought forward by Parliament. In specification (5), which focuses on initiatives only, we find a positive but not significant difference-in-difference effect. The importance of a policy issue in a referendum is partly reflected by turnout. Specification (6) considers a subset of referenda which received a relatively high turnout. While we do not find any significant effect of increased transparency, the point estimate is with 3.5%-points relatively large and again positive which suggests, if anything, higher divergence for potentially controversial proposal due to transparency.
Finally, specification (7) looks at the full sample and includes additional control variables. Again, we do not find any significant interaction effect of video availability on divergence of politicians from voter preferences.
CONCLUSIONS
Pundits of fully transparent political processes suggest that transparency fosters greater accountability of politicians towards their voters. We exploit a natural setting to test this claim and analyze the effect of transparency of political processes on the quality of political decisions. We use the observed divergence of politicians from voter preferences as an inverse and direct measure of quality. The introduction of publicly available video streams of the debates and votes of the Swiss Parliament in 2006 allows identifying individual votes of politicians in the Upper House. Final individual votes have been published before and after 2006 for the Lower House of Parliament. Thus, we use a difference-in-difference setting to identify the causal effect of transparency on divergence from voter preferences.
Empirical results show that full transparency does not necessarily increase the quality of political representation in the Upper House compared to the Lower House. The point estimates of the impact of increased transparency of individual actions on divergence from voter preferences are sometimes even positive. Thus, there is no clear support for the common belief that transparency always increases accountability. We highlight, however, that the way transparency was achieved, may not prove powerful enough to change the behavior of politicians towards voters. While, in principle, all decisions were transparent after 2006 and could be analyzed ex-post, gathering information from videos might represent an obstacle to the simple and fast dissemination into political debates. Nevertheless, politicians might be expected to adapt their behavior towards voters given that they can be evaluated prior to elections. We recognize that we cannot provide direct evidence on the power of the change in transparency such that there remains the potential that the increase in transparency was too weak to affect the quality of representation. However, our results may generalize to countries where greater transparency is publicly declared but relevant information for voters is not easily available.
While our empirical results seem unexpected, they are generally consistent with predictions of recent and elaborated models (Fox Reference Fox2007; Prat Reference Prat2005) which highlight the potential negative impacts of transparency of individual votes on the quality of political representation. Studies from other areas where video streams/surveillance was introduced also cast doubt on the effectiveness of such measures (see Stutzer and Zehnder Reference Stutzer and Zehnder2013 for a review of video surveillance).
