When Voters Matter
The Limits of Local Government Responsiveness
Michael W. Sances (Temple University)
V.O. Key famously wrote, “Unless mass views have some place in the shaping of policy, all the talk about democracy is nonsense.” While formally democratic, governed as they are by elected representatives, whether local governments meet Key’s standard of democratic responsiveness is unclear. In recent years, several scholars have found correlations between the policy outputs of large cities and the views of their publics, with more liberal cities producing more liberal policies. Yet these patterns could emerge even if cities were not democratically responsive. Additionally, local government is much more than large cities; yet we know little about how the thousands of smaller municipal, township, and special district governments represent their voters. In a recent article in Urban Affairs Review, I find evidence that local governments are indeed responsive in some areas, but not at all responsive in others.
A correlation between local policy and local opinion could reflect policy responding to voter opinion, voter opinion responding to policy, or policy responding to some other factor that correlates with voter opinion. As is commonly understood, correlation is not causation, and it is only public opinion causing public policy that counts as responsiveness. To tease out causation, we can use the fact that local areas in the US have seen dramatic changes in public preferences over the past few decades, with some areas becoming more more liberal and others much more conservative. We can then compare changes in local preferences, measured using the Democratic share of the two-party vote, against changes in local fiscal policy. In my article, I measure both opinion and policy at the county level, and local policy includes the total spending and revenue of all local governments in a given county, helpfully collected by the federal Census of Governments.
The figure below summarizes the data used in this project. Each point in each panel represents a group of counties in a one-percentile bin of Democratic vote share. For instance, in 1957 the least Democratic counties in the Non-South spent 1.5 thousand dollars (in real terms) per capita, while the most Democratic spent about 700 dollars. In 1957 there was barely any relationship between these two variables, while in 1982 and 2012 there are strong and positive relationships.
Because prior to 1982, Democratic vote likely captured something other than ideological liberalism, I use the data from 1982 forward. While the correlations in the plots suggest a positive relationship, as discussed above this could be misleading. Looking instead at changes over time, however, yields more evidence for the responsiveness hypothesis. In particular, as counties increase their Democratic voting by one standard deviation, they increase their local government spending by about 230 dollars per capita. Thus, in the aggregate local governments do seem to give their voters what they want.
However, digging further into the data reveals a large swath of local government apparently fails to meet Key’s standard of responsiveness. An advantage of the Census of Governments data is we can differentiate spending by policy area, as well as by type of government. Doing so by policy area, we see that police, fire, and housing spending do respond to public preferences. However, these areas combined represent just under 10% of total local spending in the average county. For the largest spending area, public education (45% of total spending), there is no evidence of responsiveness.
Additionally, of the 90,000 local governments in the US, almost 40,000 are “special district” bodies that perform a single function such as fire protection, while over 10,000 are school districts. These special purpose governments are granted considerable autonomy from other governments, presumably because it is thought they can better respond to voters in a given area. Despite making up such a large share of local government in our democracy, however, there is no evidence – at least using the same standards by which we judge responsiveness at other levels – that these governments respond to local preferences. In particular, imagine a county increases its Democratic vote share by one standard deviation. My estimates suggest that in this county, the change in preferences would lead to 250 dollars more spending per person in cities and other general purpose governments, but would not change special purpose district spending at all.
Overall, then, local governments partly meet Key’s standard of democratic responsiveness, but they partly do not. It is important to note that these findings do not mean local officials do not invest considerable time and effort attempting to respond to their constituents. Likely, they do try to respond, but are just prevented from doing so by state laws, federal mandates, and economic forces outside of their control. Identifying precisely why some local governments respond while others do not – and how policy makers might assist local governments in responding – should be an important focus of future research.
Michael W. Sances, an assistant professor of political science at Temple University, was assistant professor at the University of Memphis when this article was submitted. He studies representation and accountability through the lens of U.S. state and local government, and received his PhD in political science from the Massachusetts Institute of Technology.