State Policy Ideology in 2 Dimensions

As many readers already know, the Mercatus Center at George Mason University just released a new study I’ve coauthored with Texas State political scientist William Ruger, Freedom in the 50 States 2011: An Index of Personal and Economic Freedom. It’s the second edition of a study first published in 2009. The new edition updates and expands the data, ranks the states, and examines some correlates of economic and personal freedom.

I’ll try to blog a few of the findings from the study here in the near future, but in this post I wanted to explore something more in the vein of traditional academic political science, using the same data. (The data for the study, and to which I refer here, are available to the public at statepolicyindex.com.) When we try to understand how states differ systematically from each other in the mixes of public policies they have chosen, we have at our disposal a statistical method known as factor analysis. Factor analysis allows us to extract from a large number of variables a small number of dimensions – new variables that express meaningful correlation across policies. For instance, if states that have strict gun control laws also tend to have high cigarette taxes and lax abortion laws (which is true by the way), factor analysis could tell us that because of the high correlations across these three variables, a single variable can explain much of the variance in all three policies.

The dimensions that emerge from factor analysis of state policies are referred to as “state policy ideology.” We could think of them as regime orientations, ways in which different state governments have typically approached a wide range of public policies. The most important dimension of policy ideology is, unsurprisingly, the liberal-conservative, left-right dimension. Policy-liberal states tend to have, for instance, strict gun control laws, high cigarette taxes, and lax abortion laws, while policy-conservative states tend to have lax gun control laws, low or moderate cigarette taxes, and strict abortion laws. If every state approached all these issues “pragmatically” or idiosyncratically, then there would be no reason to expect policy choices over these extremely different policies to correlate together.

Previously, most of the state politics literature focused solely on the left-right dimension of public policy. However, with our large dataset of over 200 policies, we have been able to discover a second dimension: civil libertarianism-authoritarianism. Exploratory factor analysis on all 200 policies revealed a second, significant dimension that, by construction, is uncorrelated with the left-right dimension.

Now, I’m going to get a little more technical. Because I’ve just done exploratory factor analysis, I’ve taken all policy variables into account, and all policy variables contribute somewhat to the scores on left-right and civil-libertarian ideology, even when those variables relate to these dimensions in an insignificant or nonsensical way. For instance, civil-libertarian states also tend to have state liquor stores. That’s a bit odd. Maybe it has to do with the fact that historically rural states used to be more prohibitionist, but nowadays state liquor stores often sell liquor at lower prices than private stores in other states. At any rate, it’s a variable that we probably don’t want contributing to our civil-libertarianism scores.

So what I do next is to choose policy variables that a) should be theoretically related to left-right and civil-libertarian ideologies and b) were properly so related in the initial, exploratory analysis. Then I run separate factor analyses on those variables that should be most clearly related to left-right ideology and to civil-libertarian ideology and extract the primary factors. These are the “purified,” if you will, indicators of state policy ideology.

These are the policies that I used to create the civil libertarianism-authoritarianism variable: initial term of concealed carry licenses, index of open-carry ease, index of concealed-carry ease, restrictions on multiple purchases of guns, gun show regulations, medical marijuana laws, maximum possible penalty for a single marijuana offense, Salvia divinorum bans, existence of sobriety checkpoints, ability of police to take DNA from felony arrestees, existence of death penalty, incarceration rates per capita (controlling for violent and property crime rates), police per capita (controlling for violent and property crime rates), weapons arrests as a percentage of total arrests, prostitution arrests as a percentage of total arrests, gambling arrests as a percentage of total arrests, and “drug enforcement rate” (roughly, the average drug user’s likelihood of being arrested in a year). Civil libertarian states tend to have lax gun control laws, lax marijuana and Salvia laws, lower arrest rates for nonviolent drug offenses and other victimless crimes, and fewer police and incarcerated citizens per capita (controlling for violent and property crime rates), and are on average less likely to have the death penalty, authorization for taking DNA from felony suspects, or sobriety checkpoints, while civil-authoritarian states tend to take the opposite approach on these issues.

Now, since I’ve run factor analyses on separate sets of variables to create the indices of liberalism-conservatism and civil libertarianism-authoritarianism, it’s not necessarily the case that the two indices are uncorrelated with each other. In fact, conservatism and civil libertarianism correlate at r=0.25. Conservative states tend to be a bit more civil-libertarian, perhaps because conservative states are also less likely to have strict gun control laws, and because abortion laws load simply onto conservatism, not onto civil libertarianism. The most policy-conservative states in 2008 were North Dakota, Idaho, Mississippi, Wyoming, Utah, Texas, and South Dakota, in that order. The most policy-liberal states in 2008 were California, Massachusetts, New Jersey, New York, Connecticut, Maryland, Rhode Island, and Hawaii, in that order. The most civil-libertarian states in 2008 were Vermont, Alaska, Maine, Montana, New Hampshire, Washington, New Mexico, and Idaho, in that order. The most civil-authoritarian states were Illinois (by far), California, Maryland, Virginia, New York, Kentucky, and Mississippi, in that order. The following figure is a scatterplot of state policy ideology in two dimensions:

Rural, Western, and overwhelmingly white states are more likely to be civil-libertarian, while urban, Eastern or Southern, and racially diverse states are more likely to be civil-authoritarian. It remains for further research to discover some of the reasons for these relationships. For instance, do tough-on-crime policies reflect black preferences (polls show black voters are overwhelmingly supportive of gun control) or white responses to a perceived “racial threat” (tight drug laws and aggressive policing seem more likely to reflect this dynamic).

4 thoughts on “State Policy Ideology in 2 Dimensions

  1. Ok, first your data tell you (objectively) that the two main axes correspond to what we might (subjectively) call conservatism and civil libertarianism. Then you edit them to come up with axes that better correspond, subjectively, to those two concepts and say:


    Now, since I’ve run factor analyses on separate sets of variables …., it’s not necessarily the case that the two indices are uncorrelated with each other. In fact, conservatism and civil libertarianism correlate at r=0.25.

    But it was you, a libertarian writer, who subjectively edited the axes, so few left-liberals would believe even a small correlation. It would be nice to see the full list of tweaks you had to make in order to get this result.

    The rest of the stuff, e.g. about authoritarianism and diversity is also interesting, but you might be better off just using your original “raw” axes. Dumber is usually better.

  2. Although it is not unexpected to find a strong correlation for multiple aspects of public policy along the traditional liberal-conservative spectrum, I think it undermines the more anecdotal evidence from pundits and the media that there is a wide group of people undeserved by the traditional two parties. I noticed that Keith Poole’s blog has a post (http://voteview.spia.uga.edu/blog/?p=1908) scaling voters as legislators like DW Nominate scores do for legislators and found their first dimension correlated with the traditional left-right spectrum, with issues like raising taxes or regulation abortion correlated.

    Groups like Cato seem to argue that there’s this mass of libertarian voters left out of the two party system, but I’m not seeing much in the way of academic research indicating that.

    I’m fascinated by the high ranking of Virginia as a civil authoritarian state despite its initial high rankings in your study. I am assuming that it’s because Virginia’s rankings were primarily driven by its economic rankings, which end up on the liberal-conservative side, and Virginia actually has some average/poor policies in the sphere that ends up on the civil-authoritarian side.

    Being both a resident of Virginia and a former resident of Mississippi I can tell you your findings fit with my own experiences, as unrepresentative as they may be.

    On trying to find explanatory factors for the rankings on the civil-authoritarian, have you thought of the age of the state? Essentially taking some observations from Mancur Olson and more recently Jonathan Rauch in “Demosclerosis” that older governments, be they countries or states, accumulate more and more rules and regulations that are always tweaked and modified but never really repealed?

    1. The left-right dimension is definitely a far more powerful explanatory factor in state (or federal) policies than any other ideological dimension. Even here, “civil libertarianism” is not the same thing as “libertarianism,” pur et simple. One reason is that most people think of politics in a one-dimensional way. I think the highest figure Boaz and collaborators have gotten for libertarians in the American electorate is about 15%. This brings me to the second, more important reason why we don’t see a true libertarian-populist dimension in public policies: the views of libertarians (and populists, to the extent there really are any) can be ignored within the context of electoral politics, firstly, and legislative organization, secondly. It’s unlikely that libertarian legislators will get elected, but even if one does get elected, it’s unlikely that (s)he will be able to buck party leadership. It will be interesting to see how this changes, if at all, with a large libertarian bloc in the NH House (and one libertarian NH senator).

      With regard to civil libertarianism & personal freedom, it’s important to note that they’re far from identical. However, there is a decently strong correlation between the two (r=0.49). I think the reason for that is that conservative and liberal lifestyle freedoms tend to cancel out (guns, homeschooling, tobacco, motorists vs. alcohol, gambling, marijuana, same-sex partnerships), leaving the remaining variance to be determined by law enforcement & civil liberties policies. Virginia looks like a bit of an outlier here – #22 on personal freedom, but near the bottom on civil libertarianism.

      1. There are a lot of factors pushing everyone, from states to representatives to voters, into a single ideological dimension. Some of it may be a little chicken and the egg. I know there was a study of voters during the 2008 election that found that voters with a negative opinion of Obama but the view that taxes should be raised on the rich were more likely to change their view to opposing new taxes on the rich during the campaign as they learned that Obama supported repeal of the Bush tax cuts for the rich.

        I see that New Hampshire is close to 0 on the first dimension. Since the dimension is the standard liberal-conservatism, and once you get too far to the right on conservatism you start losing freedoms on the personal freedom, I assume that “peak freedom” would be somewhat toward the center, but more to the right given the correlations you found in the study between conservatism and the two types of freedom.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s