Political Institutions and the Opinion–Policy Link

Abstract

Abstract The link between public opinion and policy is of special importance in representative democracies. Policymakers’ responsiveness to public opinion is critical. Public responsiveness to policy itself is as well. Only a small number of studies compare either policy or public responsiveness across political systems, however. Previous research has focused on a handful of countries – mostly the US, UK and Canada – that share similar cultures and electoral systems. It remains, then, for scholars to assess the opinion–policy connection across a broad range of contexts. This paper takes a first step in this direction, drawing on data from two sources: (1) public preferences for spending from the International Social Survey Program (ISSP) and (2) measures of government spending from OECD spending datasets. These data permit a panel analysis of 17 countries. The article tests theories about the effects of federalism, executive–legislative imbalance, and the proportionality of electoral systems. The results provide evidence of the robustness of the ‘thermostatic’ model of opinion and policy but also the importance of political institutions as moderators of the connections between them. Acknowledgements Previous versions were presented at the European Consortium for Political Research Joint Sessions of Workshops, St. Gallen, Switzerland, 12–17 April 2011, the Annual Meeting of the Elections, Public Opinion and Parties specialist group of the Political Studies Association, Exeter, UK, 9–11 September 2011, the Ash Center at the Kennedy School, Harvard University, 19 October 2011, Koc University, Istanbul, 26 January 2012, and the Annual Meeting of the Midwest Political Science Association, 12–15 April 2012. We are grateful to participants at each presentation, especially Christine Arnold, Susan Banducci, Shaun Bowler, Ali Carkuglo, Nicole Doerr, Mark Franklin, Archon Fung, Brigitte Geissel, Armen Hakhverdian, Will Jennings, Quinton Mayne, Francisca Rojas, Hollie Russon-Gilman, Belgin San Akca, Eser Sekercioglu, Paul Warwick and Kamil Yilmaz. We also are grateful to Kevin Arceneaux and Luke Keele for helpful advice with aspects of the statistical analysis. Notes 1. Indeed, we have little sense for the connection between opinion and policy in very low salience domains, i.e. domains in which we do not even gather opinion data (Burstein 2003). A lack of a connection between opinion and policy in non-salient domains would be neither surprising nor a bad thing, since public preferences in these domains surely are less informed. 2. We have in addition developed an argument about the effects of electoral systems (Soroka and Wlezien 2010), which we summarise and test below. 3. Unlike the thermostat that governs the heating (and/or air conditioning) units in our homes, which sends a dichotomous signal, R is a continuous variable, and captures both direction and magnitude. 4. There some exceptions such as abortion or desegregation. Consider, e.g. the GSS question on ‘whether or not you think it should be possible for a pregnant woman to obtain a legal abortion’ under six different conditions. These instances are rare, however – very few policy domains lend themselves to questioning about specific levels of policy. 5. There is other evidence of thermostatic public responsiveness as well, including Franklin and Wlezien (1997), Erikson et al. (2002), and Jennings (2009). 6. Note that this dovetails with public responsiveness to spending. Public opinion in year t reacts (negatively) to policy for year t and policymakers adjust policy (positively) in year t + 1 based on current (year t) opinion. Now, if studying policy that, unlike budgetary policy, is not lagged, then policy change could represent year t public opinion, which in turn responds to lagged (year t – 1) policy. That is, the model can be adjusted to reflect the reality of the policy process. 7. Note that different economic variables, including unemployment, inflation and business expectations, were included in the model though to little effect. 8. For a more extended (and nuanced) discussion, see Wlezien and Soroka (2011). 9. They do, however, show that proportional systems produce more representative legislatures. 10. Powell finds that what mostly matters is the polarisation of the party system. 11. For related work considering dynamic representation in majoritarian (and non-majoritarian) systems, albeit using MIP responses or ideology rather than policy, see Hakhverdian (2010) and Hobolt and Klemmensen (2005, 2008). 12. This generalises Rogowski and Kayser's (2002) argument relating to the comparatively higher seats–votes elasticities in majoritarian systems. 13. Also note that Hobolt and Klemmensen do not actually assess responsiveness to public preferences for policy and focus instead on public mentions of the ‘most important problem’ facing the country. 14. The file is distributed by GESIS, ZA file #4747/4748. 15. Note that the question used here is about ‘cuts’ rather than ‘spending’ (‘more’, ‘less’, or ‘about the same’), the latter of which is the focus of our own past work. The ISSP asks questions about spending ‘more’ or ‘less’ in various specific policy domains (e.g. health, defence), but the only question capturing overall spending preferences asks about ‘cuts’. Results for the question used here nevertheless move alongside the average across all eight specific-domain ‘spending’ questions. 16. Converting to a constant currency, such as US dollars, presents other difficulties – namely, it means that trends in spending are affected by shifts in exchange rates over time. 17. We use ENPP instead of the effective number of electoral parties (ENEP) because of our conceptual focus on how electoral systems influences parliamentary control. Using ENEP makes little difference, however, as does using (logged) median district magnitude. 18. That is, state and local spending includes funds from the national government that are difficult to disentangle. Note also that there are other alternative measures of federal revenues, including in particular Vatter's (2009), but the one used here has the practical advantage of being available for more of the countries in our sample. 19. Employing generalised least squares (GLS) techniques makes no substantive difference. 20. This is not surprising given the difficulty of comparing responses to the items across countries, that is, they may not be registering the same things. 21. The measure is from the International Monetary Fund's (IMF) World Economic Outlook Database (2010). We have described this in the data section, above. 22. Our previous research on preferences for spending has included a linear time trend to account for increases in the underlying preferred levels of spending over time that would reflect growing economic capacity over time. We do not include such a variable here because our measure of spending takes into account GDP. When included, trend is not a robust predictor and it makes little difference for the estimated effects of the other variables. 23. Note that Rodden's measure is not available for all 17 countries, so these estimations rely on 13 only. The excluded countries are: Czech Republic, Hungary, Japan and Slovenia. 24. There is reason to suppose that it does not matter as much as our analysis implies – i.e. that the results overstate the size of federalism's effect. Consider that they imply little public responsiveness in highly federalised countries, such as the US and Canada, where our previous research of time-serial dynamics has demonstrated a significant amount. 25. Recall that Powell and Lijphart focus on the congruence between the median voter and government position in the wake of elections, not the responsiveness of governments to shocks to preferences in between elections. 26. Note that this model relies on 16 countries – Poland is excluded from models of representation because the field date for the final ISSP survey was late, in 2008, and our spending data end in that year. 27. Consider findings in other work exploring the impact of partisanship with more data, e.g. Blais et al. (1996). 28. This would be in line with past work, including, e.g. Blais et al. (1996) and Soroka and Wlezien (2004). 29. The same is true when we use the median district magnitude. This again points towards the possibility that more proportional electoral systems decrease rather than increase dynamic representation, but the pattern also is too unreliable to credit. 30. Again, some countries are excluded due to missing data: in addition to Poland (see Note 26), Hungary, Japan and Slovenia. 31. The US is an extreme case on both dimensions, and so it is worth noting that the effects are larger and more reliable when it is excluded from the analysis.