Construction of Index

Construction of Index


We started by collecting data on state and local public policies affecting individual freedom as defined above.49 All of the statutory policies are coded as of January 1, 2009, the fiscal data are coded for the fiscal year 2007–2008, the law-enforcement data cover the entire year of 2008, and all data are also back-coded consistently to January 1, 2007 (FY 2006–2007). We omit federal territories. The database covers fiscal policy, gun control, alcohol regulation, marijuana policies, tobacco and smoking laws, automobile regulations, law-enforcement data, education policies, land-use and environmental laws, labor-market regulations, health-insurance policies, utilities deregulation, occupational licensing, asset-forfeiture rules, eminent-domain reform, court systems, marriage and domestic-partnership regulations, campaign-finance laws, and sundry mala prohibita and civil-liberty issues.

In many cases, we directly code statutes with dichotomous or simple ordinal variables. In some cases, we code continuous statistical variables that capture both the relevant statutory framework and the manner in which legislated policies are administered (e.g., expenditure and revenue levels, arrest rates for victimless crimes, etc.). Although we went directly to the statutes and legislative-session data for many of our variables (which are now available online for all 50 states), we also collected fiscal data from the Census Bureau and Bureau of Economic Analysis, law-enforcement data from the Federal Bureau of Investigation, health-insurance-policy data from the Henry J. Kaiser Family Foundation, labor-market-regulations data from the Department of Labor, Occupational Safety and Health Administration, and National Academy of Social Insurance, and so on.

In some cases, more complex ordinal scales are created from the simpler variables, but the disaggregated data are available in separate spreadsheets at for researchers to create their own scales. For instance, we create an index of eminent-domain reform by taking into account four dimensions of reform: whether any reform has been enacted (binary yes/no variable); standards for private takings (simple ordinal variable, coded “1” if all takings for private use are prohibited, “0.5” if only certain private-to-private transfers are prohibited, and “0” if there are no effective restrictions on this type of eminent-domain use); blight definitions (simple ordinal variable, coded “1” if a stricter definition of blight has been implemented either implicitly or explicitly, “0.5” if a vague definition of blight has been retained but the standard of proof for proving blight has been raised, and “0” otherwise); and whether the constitution enshrines additional restrictions on eminent domain (simple ordinal variable, coded “1” if all additional restrictions have been thus enshrined, “0.5” if only some have, and “0” if none have). Another example is our creation of an index of difficulty of asset forfeiture from three variables: standard of proof for showing property subject to forfeiture, innocent owner burden, and percentage of proceeds going to law enforcement. We employ these ordinal variables to capture unified policy concepts whose individual elements are dependent on each other and thus should not be treated independently.

The spreadsheet with all the variables included in our freedom index is available in Microsoft Excel 97–2003 format at 50-states-2011. To find the sources and formulas for constructed variables, interested readers can also download individual spreadsheets for each policy area. Given that individual readers may ultimately have a different view of how to weight each variable, we invite them to apply different weights to each variable and come up with their own state freedom rankings. One of the benefits of our construction and ranking of freedom is that it initiates a discussion of what it means to be free and focuses attention on the proper relationship between the government and the people in a free society. Just starting that conversation is worthwhile since it draws people away from— or at least challenges—the notion that they should only think in terms of some standard like “justice as fairness” or equality or any number of competing values in tension with individual liberty.

We do not wish to claim that our database is fully comprehensive in terms of policy coverage. In a few cases we found that coding state law directly would have been an exceedingly complex endeavor resulting in abstruse measures unlikely to illuminate the issue. Tort reform is the most important example. States have implemented a wide variety of measures to counteract abuse of the tort system, and many of these highly technical and frequently idiosyncratic reforms are not strictly comparable across states. The relative importance of these features was also unclear to us, making the construction of a summary index of tort reform virtually impossible. Furthermore, a fundamental problem with this approach to coding tort reform is that the states with the most flawed tort systems, from a business perspective, have implemented the most reforms. We have instead chosen to present a single variable capturing the quality of states’ tort systems: the percentage of respondents in the U.S. Chamber of Commerce survey indicating satisfaction with that state’s liability system. This continuous variable seems to capture the concept we want quite well: West Virginia, Louisiana, and Mississippi come out at the bottom, while Delaware scores first.

Some policies seem to have minor importance for freedom, such as unenforceable bans on adult toy sales. We did not think it worth the effort to code for all 50 states laws that would not much affect the final freedom scores in any case.

Finally, the database does not include any policies for which there was no state variation. For example, because all states license medical doctors, licensing of medical doctors was not included in the measure of occupational and professional licensing

We have also carefully calculated the numerators of the state tax and spending ratios. For taxes, we count all tax revenues except motor fuel and mineral severance taxes, and we exclude “current charges” (mostly user fees: university tuition, highway tolls, airport fees, and the like), “miscellaneous general revenue” (interest earnings, special assessments, sale of property, and “other general revenue”), utility revenue, liquor-store revenue, and insurance-trust revenue. For spending, we actually include two sets of two variables. The first variable, weighted more heavily, includes all state and local government expenditures. The second variable subtracts current charges from expenditures on the theory that it is better for government spending to be paid for out of user fees than general taxes. The two central problems with the second measure are that (1) states may receive windfall revenues from contracting out public services, causing their current charges figures to jump dramatically (for example, Indiana’s 2006 highway revenue), and (2) government spending can crowd out private alternatives, even when funded through user fees, and especially when guaranteed by a legal monopoly. Thus, we do not wish to reward states with high current charges excessively, but we also want to take the possible advantages of the user-fee approach into account. These two variables are measured in two different ways, with and without adjustment for federal grants.

A final adjustment worth mentioning is that performed on fiscal decentralization (local own-source revenues divided by total state and local government spending). This variable captures the devolution of taxing powers to local governments. In order to attract mobile taxpayers and businesses, local governments with taxing authority should seek to eliminate rents in order to keep taxes low (“market preserving federalism”).50 Another advantage of fiscal decentralization is that it allows individuals to choose to live in jurisdictions that provide a preferred mix of public goods.51 Fiscal federalism, rightly understood, can thus promote individual freedom. However, states that have larger populations are more likely to be fiscally decentralized because they typically have some local jurisdictions with large populations that enjoy economies of scale. Compare Texas to Vermont: is Vermont really less decentralized than Texas, as the data indicate? After all, Texas has local jurisdictions that are larger in population than the entire state of Vermont! We decide to adjust fiscal decentralization for state population in order to capture the true range of choice that citizens enjoy among jurisdictions in a state. In Vermont’s case, the relevant tax competition occurs perhaps more among New England states than among the towns of Vermont. We perform the adjustment by regressing fiscal decentralization on the natural log of state population and taking the residuals (once the natural log is controlled, raw population has no effect on fiscal decentralization).

Weighting the standardized variables to create overall measures of economic and personal freedom has elements of both art and science. We decided to weight economic and personal freedom equally to create an “overall freedom” score. Fiscal and regulatory policies are weighted equally to create the economic freedom score. Fiscal policies have to do with taxing, spending, and government employment and wages. Regulatory policies include government regulations intended to effect particular economic outcomes, such as higher productivity, redistribution among interest groups, or resolution of externalities, as well as miscellaneous features of the economic system, such as the quality of the tort system. The personal freedom/paternalism category focuses on regulatory policies whose justification seems to be regulation of individual choice in the alleged interests of the individual or the “public.”52 The dividing line between economic and paternalist regulation is often unclear; how should homeschooling laws count, for instance? In general, we have erred on the side of placing regulations that are not directly related to economic issues in the paternalist category.

Within these three categories—fiscal policy, regulatory policy, and paternalism—the rule of thumb we use to weight particular issues is the salience of the issue (i.e., the substantive importance of state policy variation) and the number of people affected by it. We use the existence of explicit constitutional protections at either the federal or state level as prima facie evidence of high salience. We do this because the fact that representatives of the people have chosen to incorporate a right into a constitution apparently reflects a widespread belief that such a right is too fundamental to leave to the discretion of transitory legislative majorities. Our choices of weights may certainly be challenged, but we have tried a number of different weights vectors and have found the results to be quite robust. We recommend that reviewers therefore treat with skepticism very small differences between states on overall freedom scores.

Figure 6 gives the weights for the categories and issue subcategories. For the individual variables’ weights, consult the data appendix. Discussions of changes from the first index and of variable codings and weighting justifications follow.

Figure 6

Changes from the First Index

We have made a few changes from the first edition as we attempt to hone our measures of freedom. Of course, in order to maintain consistency over time, all of the new and revised variables have been backcoded to 2007.

In fiscal policy, we measure government spending, employment, and taxation somewhat differently from the first edition. Variables that previously used “corrected GSP” in the denominator now use personal income. The spending and employment variables were previously also adjusted for federal grants received by each state, but now we include and weight equally variables that are not so adjusted. The rationale is that states do have a choice whether or not to take such grants, and federal grants are associated with future increases in state spending and taxation.53 Finally, the tax variables no longer include mineral severance taxes, the onus of which is largely on consumers around the world rather than state residents. This change drastically improves the scores of states like Alaska and Wyoming, although the state spending financed by such taxes still counts against them.

Under regulatory policy, we added health-insurance and labor-regulation variables, such as the individualhealth-insurance mandate in Massachusetts, employer verification of legal-resident status, and mandated family leave. Therefore, we increased the weights for these issue subcategories slightly. We also dropped some environmental regulations from the index on the grounds that there may be a legitimate government role in dealing with the environmental externalities concerned: state endangeredspecies acts and state wetland statutes and programs. Therefore, we decreased the weight of the land and environment regulation subcategory substantially. We also decreased the importance of the eminent domain reform subcategory somewhat. Recent evidence suggests that state eminent-domain reforms have often been ineffectual, as governments have used quasi-legal end-runs around the traditional eminent-domain process to seize private property.54 Finally, we adjusted other regulatory policy weights to accommodate these changes.

Another significant change to the regulatory policy category is the inclusion of a new, more accurate occupational-licensing indicator. The new indicator includes not just licensing in selected occupations, but a comprehensive coding of licensure requirements in each state, with an overall index of licensure that estimates the percentage of each state’s workforce that is subject to licensure

Under paternalism, we added several new variables and replaced the asset-forfeiture measure. The new variables are mostly concentrated in the issue subcategories mala prohibita and civil liberties, marijuana laws, and gun control, with the subcategories reweighted accordingly. The asset-forfeiture measure is a much more accurate representation of the rules that actually affect the incentives and ability of police to seize the property of innocent owners.


49. The following is adapted from Sorens, Muedini, and Ruger, “U.S. State and Local Public Policies in 2006.”

50. Yingyi Qian and Barry Weingast, “Federalism as a Commitment to Preserving Market Incentives,” Journal of Economic Perspectives 11, no. 4 (1997): 83–92; and Rodden, “Reviving Leviathan.”

51. Tiebout, “A Pure Theory of Local Expenditures.

52. For a critique of the term “the public” and its uses, see Lysander Spooner, A Letter to Grover Cleveland On His False Inaugural Address; The Usurpations and Crimes of Lawmakers and Judges and the Consequent Poverty, Ignorance, and Servitude of the People (1886; republished by Kessinger Publishing, n.d.), 7.

53. Sobel and Crowley, “Do Intergovernmental Grants Create Ratchets?

54. David T. Beito, “Fox News on Eminent Domain Through the Back Door,” History News Network, August 3, 2010, entries/129846.html, accessed December 1, 2010.

' '