It identified numerous measures of personal freedom and combined them with the Economic Freedom of the World index to create the Human Freedom Index as an all-inclusive measure of liberty in the negative sense. I realized I was lucky to be in that room and that this was important work.

Subscribe now

My thoughts lingered on one thing, however.

I despise international travel.

My own proclivities aside, the project launched with the Cato Institute, as far as I could tell, as an unqualified success. I saw it as a sister index to Economic Freedom of the World and was perfectly happy to recommend it and support it as best I could.

On my short list of “best thing ever” in my narrow world of measuring institutions was the time that Bernie Sanders was trying to express during the first you-know-who administration how disgusted he was with the United States for forming closer ties to Saudi Arabia. To decry Saudi Arabia, he cited the Human Freedom Index, apparently oblivious to what it measured. On C-SPAN2. On the Senate floor.

In 2020, right before COVID-19, I was invited to be the data collection person for the index and to be a co-author. I accepted. Along with the other things I already mentioned on Substack, this was how I spent much of my time during the pandemic.

To recap, the Economic Freedom of the World index has five areas: the size of government, the quality of the legal system and property rights, sound money, the freedom to trade internationally, and regulation. To that, the Human Freedom Index added the rule of law, security and safety, movement, religion, association, assembly, and civil society, expression and information, and relationships. That brings the number of areas to 12, which are then equally weighted.

Before continuing, a complaint about introductory statistics pedagogy. Undergrads are taught about a series of issues with linear regression, which, for the uninitiated, means “drawing a line of best fit through some dots, with style.” One of these problems is called “multicollinearity,” which means that two variables you are considering are closely related to each other, so it is difficult to tell what is causing what when you are then looking at a third variable.

For each of these issues with linear regression, the undergraduate is then taught, “here’s a thing that fixes that, tell the software to use that and it’s probably fine.” The thing the undergraduate is taught for multicollinearity is called Principal Component Analysis.

Except Principal Component Analysis doesn’t really do that. What Principal Component Analysis does is take a group of variables and boil them down to combinations of them that are clearly distinguishable from one another.

You can then take the principal components and use them in a regression. You won’t have issues with multicollinearity at that point, because you can tell them apart by how they were constructed. How do you interpret the principal components? It’s not always clear, and the correct interpretation may not relate to your original research question.

Something closely related to principal component analysis is factor analysis. A factor analysis assumes that there is some number of concepts in the background of a bunch of data, and it hopes to sus them out.

Both principal component analysis and factor analysis had been used for decades to analyze the Economic Freedom of the World index. You would always end up with the same conclusion, that all the areas of economic freedom except the size of government were mostly doing the same thing, and then the size of government was off in its own world doing something else. That this result was already known was part of what made Jan Ott’s argument so befuddling.

No one had yet done a factor analysis of the Human Freedom Index. So, after haphazardly learning what the best practices were, I applied it. What are the latent factors behind whether or not a country is free in the classical liberal sense? They are,

1. Embracing markets as an organizing tool of society.
2. Whether the government engages in social repression, such as repressing religious groups or perceived threats to its power.
3. Whether the society is organized along “European” lines of strong property rights enforcement and large welfare states, or weak property rights and small welfare states.

In literally the last post on this Substack I described the importance of state building, which gets you more property rights and a bigger welfare state. I am not denying any of that. But both (1) and (2) are clearly about the level of freedom in a society. Then, (3) is about the nature of that level of freedom.

Let’s call the first the “Trappings of Capitalism,” the second the “Absence of Social Repression,” and the third “Northern European-ness.” You can then rapidly summarize the nature of negative freedom in any given country using the factor analysis:

A tool like factor analysis cannot tell us what belongs in a formative construct, but it can give us a sense of what we are looking at.

The world is a complicated, interesting place.

I still don’t like traveling.

Further reading:
Ryan H. Murphy. 2025. “Social Repression and Variation in the Human Freedom Index.” Social Indicators Research 177: 1193-1212.

Share

There are dozens of book-length apologias for anarcho-capitalism. If you want a deep dive, ask your favorite LLM. But anarcho-capitalism is the endpoint of going further and further with capitalism until you run out of freedom to free. One famous exchange goes along these lines:

“You libertarians are the type of people that would allow fornication in public parks!”

“What do you mean, public parks??”

We get to the freest of the free markets. Is that an open access order? I actually do not know the answer to that question.

Subscribe now

One path to anarcho-capitalism mirrors competitive autocracy. All land is “private,” but you can at least imagine a world where the importance of stable rules leads to land concentrating in the hands of landlords who amount to rulers. Now we are back to being ruled by competing stationary bandits. That is not an open access order.

In a more general model, suppose there is someone without money or property. There would presumably be places willing to take this person in and feed and shelter them. But the definition of an open access order is whether the ability to compete and start profit-seeking, political, and civil society organizations is open to all. Formally, yes, this person without means is free to make use of the same institutions that anyone else is.

Private property means the power to exclude. In most cases we prefer to be able to exclude someone ranting about how pigeons put a computer chip in his brain from our homes and property. Within certain limits, it is okay for that person to do that in the public square. But unless someone sets aside private property specifically for that purpose, then there is no public square.

The libertarian theorist would say that the person isn’t being coerced by not being allowed to rant about pigeon conspiracy theories on private property. That’s fine. But open access orders are not defined in terms of coercion, so I lean toward saying that such a society does not count as an open access order.

At the very beginning of this series of posts, I described the conflict between classical liberals’ opposition to the state and the apparent and obvious desirability of improving the quality of state institutions. Do you see where we are headed?

One way of getting a general measure of institutions is to shove together four of the Worldwide Governance Indicators and call the result “state capability,” or, in another framing, institutional “effectiveness.” When you throw all sorts of institutional measures together and stir them together, you generally get this “effectiveness” as being what explains most of the variation among countries’ institutions.

Unsurprisingly, that effectiveness is positively related to everything in Economic Freedom of the World, except the variables related to the welfare state, i.e., more effectiveness corresponds to larger welfare states.

I described at length how economic freedom becomes the better answer for rulers as state capacity increases, if the ruler wants growth. That’s the political side of state capacity and economic freedom. The social side of state capacity relates to trust. Better institutions mean people are more trusting. More trust facilitates better institutions. Societies with more social trust are more willing to accept expert opinion that policies like deregulation are safe.

The primary path of institutional development runs through this tunnel of institutional effectiveness, social trust, and economic liberalization. This is the same tunnel that leads countries to transition from limited access orders to open access orders. So, instead of conceiving anarcho-capitalism in limited access terms, what would open access anarchy look like?

What follows is speculative, though I do not think it is any more speculative than existing models of anarcho-capitalism. I believe this model is more plausible and more consistent with New Institutional Economics, while maintaining society as an open access order. But it means that we reach something approaching anarchy via increasing state capacity, not destroying it.

Gradually, different bureaucracies could be privatized. Moving down this path is not an entirely radical policy proposal. What differs is the extent of the process and the need to show that such privatizations are worthy of the public’s trust.

Sweden is an extremely high-trust society. Through school vouchers, it has quasi-privatized its school system. Private ownership and operation are widespread among European airports. Japan’s passenger rail system is largely privately operated. The technology and mechanisms for privatizing vast swaths of the economy are already in place. The question is whether we can build states competent and trustworthy enough to gain the public’s approval for using them.

The last question is where the money comes from. Much of the trust in question concerns “social insurance.” If I fall on hard times or face illness or disability, will I be supported or will I die in a ditch? Elites and the upper-middle class would need to take responsibility for funding such a system and have sufficient trust from the rest of the population to do so.

If that sounds ridiculous, and it probably is, then we need to build our institutions and societal bonds until it no longer sounds ridiculous.

That raises another question: if we are replacing a system in which the state does nothing but tax 30% of the income of the upper-middle and upper classes and efficiently transfer the money or pay for social goods, does that differ in any meaningful way from a system in which that tax instead becomes a strongly enforced social obligation?

In terms of economic incentives, I do not know whether it does. At some point, you are getting rid of the government for the sake of saying you got rid of the government.

But this society remains an open access order, as far as I can tell. And it sure seems like a better path than competitive autocracy.

These thoughts were pulled from an unpublished paper that I never put online. Where I likely drew the most inspiration from in writing it was: Scott Sumner. 2010. “The Great Danes: Cultural Values and Neoliberal Reforms.” Working paper.

Share

Subscribe now

What if you can choose among several Stationary Bandits? Each has their own fiefdom, and you are free to leave. That puts a constraint on their power over you. They need to attract people to live in their fiefdom, so they cannot act arbitrarily, and they need to establish some kind of record of protecting rights and not pillaging everyone living in their territory, because that will harm their bottom line in the long run. Is the first best solution to good institutions competitive Stationary Bandits?

The boring set of answers to how countries develop is the package of stuff that pretty much all countries do. You free your markets, you educate the masses, you create a functioning bureaucracy, you solve basic public health issues, and you democratize. The sexier, edgier answer is that this is all wrong, democracy is kinda screwed, and what we want is a polity operating under the logic of the Stationary Bandit. This is essentially the argument of Curtis Yarvin and, with some additional moving parts, is known as neocameralism.

The prime example of economic development heterodoxy is, of course, China. China did not democratize. Its party leaders are wiser than the bumbling politicians of the West, and they are able to profit from the extraordinary success since the reforms under Deng Xiaoping. These reforms led to the greatest, most rapid event of poverty reduction in human history. It is also not an open access order. Their data is below As a reminder, the cutoff for being classified as an open access order is a score of 80. China is nowhere close.

There are actually other examples. A country I have mentioned a few times here is Singapore. Singapore runs elections and they are genuine elections. They are also elections that the People’s Action Party has never lost. The party maintains control through a combination of giving headaches (or worse) to any opposition parties while co-opting any true grievances the opposition parties uncover.

Singapore offers its public sector employees very high salaries and provides an extremely strong deterrent against corruption. It even ties political officeholders’ compensation to GDP growth. Singapore isn’t run by a Stationary Bandit. It is run by a professionalized class of Stationary Bandits, all well-incentivized and educated. Imagine a country that was nothing but Manhattan, but four times Manhattan’s population and all the infrastructure was gleaming and everything worked and there was no crime. That is Singapore.

In 2025, the GDP per capita (PPP) of the United States was $76,931. In Singapore, it was $139,593.

Singapore is a limited access order.

The other popular example of modern successful autocracy is Rwanda. The country under Paul Kagame has done remarkably well following a horrific genocide in the 1990s. It is not yet actually a wealthy country (its own GDP per capita PPP figure is $3,559), but it seems to be for real and aspires to be the Singapore of Africa. Kagame is also an autocrat. Rwanda is not an open access order.

Finally, if you wish to be a hipster about it, the tiny European city state of Liechtenstein may qualify as a successful limited access order, but it is far too small for me to have data on it.

There are other countries that went through periods of autocratic (or worse) transition where the country ended up as a modern, functioning, wealthy, democratic society. South Korea, Taiwan, Chile, Portugal, and Spain are examples that come to mind. I do not think those are very positive examples in terms of a process of development, except that they seem no better or worse than other high performing countries in 2026. F.A. Hayek seemed to have had positive things to say about this brand of regime, but my understanding is that this was a clumsy old man saying silly things, not anything that we should be taking seriously.

It just isn’t clear why we want to look away from the standard path of institutional development to these exotic stories of superior economic performance for a randomly chosen developing country, let alone shift a functioning (or quasi-functioning) open access order to this model of politics.

China had a remarkable 40 year run, which was in part less surprising given how educated its population already was. With a modicum of economic freedom, it got very far and lifted its population out of poverty. While it is making progress with AI, its economic performance since COVID-19 is somewhat unclear. Who knows where Rwanda will end up, though I wish them the best. Singapore apparently runs on magic pixie dust that allows central planning to work and makes government bureaucrats more trustworthy than your spouse. Good luck repeating that.

Where I am headed here is that we have a general path for economic freedom, which involves the full constellation of economic development: educated, trusting, efficient, wealthy societies propelling us towards greater freedom and liberalization, because that is what is true and beneficial for humanity. The alternative to that path is the model of competitive autocracy, which centers on a small handful of case studies. Shouldn’t our model be the standard model of economic development, only we continue extending its logic?

Shouldn’t we be asking how to achieve open access anarchy?

An older paper of mine that relates to these questions is:
Ryan H. Murphy. 2019. “Governance and the Dimensions of Autocracy.” Constitutional Political Economy 30: 131-148.

Share

Neither narrative is really correct.

Subscribe now

I have no interest in talking about the partisan politics of the issue.

If you look at all the major advanced economies, the problem is Social Security and pensions. There are no medium or long run trends in government consumption, government investment, or government ownership of the economy. The area where there are systematic increases is in transfer payments, and that is driven largely by aging populations.

Serious commentators across the political spectrum understand this is a problem. It is also politically radioactive. A great irony is that people in public policy see the problem of a growing state and understand its nature, while many of those demanding action against the growing size of government would prefer no action be taken on the dimension of public policy that is actually the problem.

We discussed previously that the size of government can be boiled down to two-ish concepts that are negatively correlated with one another. The first is government consumption, transfers and subsidies, and the top marginal tax rate. Call that “the welfare state.” The second is government investment and government ownership over the economy. Call that “socialism.” State capacity is positively correlated with the welfare state and negatively correlated with socialism. Yes, let me repeat: across countries, the welfare state is negatively correlated with socialism.

There are two ways of having a small size of government. The first is that the country is too poor to have a large size of government. That is the normal way. The second way is something no one really understands. The second possibly relates to Spanish colonialism.

There are three ways of having a large size of government. The first is that a country is wealthy and can afford to. This is called “Wagner’s Law.” That is the most normal way. The second most common way is the natural resource curse, where all sorts of money (often but not always from oil) flows into the coffers of the state without the state needing to do anything to get it. The third consists of governments that did not observe that socialism’s use-by date was 35 years ago, and they’re still trying to make it work.

There is a lot of overlap between categories two and three. There is almost no overlap between the Wagner’s Law countries and the other two groups, except Norway. Maybe Canada. Norway is in constant existential crisis because of all the money they get from oil while being ideologically opposed to fossil fuels. Canada’s oil money comes from the province of Alberta, which is presently considering secession.

If you look at the Size of Government rankings in Economic Freedom of the World, what you will see is an incoherent mess. Several wealthy countries have among the lowest index scores (largest governments), but they’ll be standing next to, um, Algeria. At the top of the list are many impoverished countries, but then there is a stray Hong Kong here and there.

An exercise I did a few years ago was to say, what if we controlled for Wagner’s Law? Well let me more formally introduce it as a concept. The regularity, which was first observed back in the 19^th century, is that as a country’s GDP grows, almost inevitably the public sector share of GDP grows. We can argue about why. But it happens.

There are three pieces of the Size of Government I identify as the welfare state. What if I take each of these pieces, and see how much of each I can explain using Wagner’s Law, that is, GDP per capita.^[1] I can then take the part that is unexplained by Wagner’s Law for each of the variables, turn them back into index values on the same scale as the other Size of Government variables, and recalculate the Size of Government, now stripped of Wagner’s Law.

It was straightforward to do this with government consumption and transfers and subsidies. There was also a relationship with the top marginal tax rate, but it was too negligible to bother with (probably because having a very high top marginal tax rate suppresses your GDP).

When you make that adjustment, the rankings for the Size of Government suddenly look far more coherent. Here are the top 15 and bottom 10 countries, first unadjusted and then adjusted:

Some of the most impoverished countries in the world are near the top of the Size of Government score before the adjustment. But once they are adjusted downward, the pattern that emerges is quite clear: the countries with the smallest states in the world are in Latin America. That, in conjunction with the Philippines ranking eleventh (and maybe even the presence of the United States) makes Spanish colonialism a prime candidate for why states stay small.

Hong Kong, Switzerland, and the United States are all in the top ten. We can interpret this as evidence that a true commitment to free markets genuinely prevents governments from becoming large. We can interpret this as federalism and decentralized politics serving as a check on big government. Or we can interpret this as nothing because the sample size is three.

I believe the second column better reflects what the institution of a limited size of government looks like. I would not recommend putting this into Economic Freedom of the World. I mechanically stripped out the parts of the data that are positively related to GDP, which would make the index dishonest for the purposes of what we want it to be. I did check how it would affect the overall economic freedom rankings, and it affected them less than you would think, but that just means it would be pointless.

I also tried to look at some of the obvious relationships between the adjusted data and possible underlying causes, but didn’t get very far. It was also clear that simply controlling for GDP per capita in the statistical analysis itself accomplished the same thing as adjusting the Size of Government for Wagner’s Law, so the technique I developed is unnecessary for further formal statistical analysis.

Well what did we learn then? Governments stay small either because the country is poor or for mysterious reasons possibly related to Spanish colonialism. Governments get really big either because the country is rich enough to afford it, because of the presence of natural resources, or because they are live action roleplaying communism.

For a country like the United States to actually cut the size of government in some meaningful way that would show up in the index (rather than just stop its trajectory) is incredibly difficult. In some preliminary research, I looked at the biggest spending-side fiscal consolidations (when someone looked at the budget, said “oh no,” and actually got politicians on board with cutting stuff) in modern history and found essentially no effect on future index scores. Part of that is because of how Economic Freedom of the World and its Size of Government area are constructed. But part of it is also because people don’t have the stomach for it.

If something is both popular with experts and the public, we would already have done it. Proactively cutting Social Security benefits (rather than the automatic haircut that may happen in a few years) is incredibly unpopular, even though it would almost certainly reduce pain down the road. An attempt at enacting the popular idea of “reducing waste” just happened and it didn’t seem to amount to much in a positive direction.

I mean, what do you want to do? Close down public elementary schools? The world is a difficult place to make better.

Further Reading:
Ryan H. Murphy. 2022. “Breaking Wagner’s Law: Which Countries Have the Most Limited Government?” Public Finance Review 50, no. 4: 484-509.
Ryan H. Murphy. 2022. “Sizing Up the Size of Government.” In Bridwell Institute Annual Report 2021/2022, p. 2-8.
Ryan H. Murphy. 2024. “Accounting for Large Fiscal Government Size.” Economic Affairs 44, no. 3: 589-600.

Share

^[1] Actually, the natural log of GDP per capita.

It is useful to think of two broad approaches to applied research in social science, “thin” and “thick.” “Thin” means you turn everything into a number. Critics of thin methods will say they oversimplify and misunderstand how the real world works. Supporters will say thin methods are the only way to do science seriously. Modern economics is mostly in the “thin” camp. “Thick” means that when you look at an event or a country or a company, you gather all the historical, social, and circumstantial details and try to give a complete account of it. North, Wallis, and Weingast operate in the “thick” framework. I wanted to allow it to work in the “thin” framework as well. Tell me a year and a country, and I will give you a number.

Subscribe now

Shortly after reading Violence and Social Orders, a point that resonated with me was that when you read the institutional specifics of what they described as going into an Open Access Order, they amounted to a combination of economic freedom, state capacity, and democracy. That is not at all the focus of the book. The book is about the process and prerequisites for a country to become like a Denmark or a United States, where the institutions which allow for starting a business or a political organization are available to all and not merely dominated by the elite.

I did what seemed natural and put measures of economic freedom, democracy, and state capacity on the same scale, and then took their average (mean). Singapore was near the top of the list. I immediately discarded the project. Singapore is not an open access order. You can hypothetically start a competing political party, but not really. Of all wealthy countries, it is one that is ruled by the elites, for the elites, and they just happen to share and listen just enough to prevent the masses from revolting. They are extremely effective at this, but nevertheless.

Several months later I was standing in an indoor waterpark with my family. I realized that the problem was that if a country scored poorly on any of the three characteristics (economic freedom, democracy, or state capacity) it really wasn’t what North, Wallis, and Weingast were describing. So I needed to have more “weight” on whichever of the three scored the lowest; I needed to penalize imbalance. One way of doing that would be to take the harmonic mean. The harmonic mean is much more elegant than having weights somehow be a function of which variable is lowest, and it also has the benefit of annoying pedants who don’t like it when you use the harmonic mean “incorrectly,” as this would entail.

The choices of which data to use are somewhat obvious. I have my own measures of state capacity and economic freedom. Varieties of Democracy has its own democracy data, but I could even use its Liberal Democracy Index that best fits in with Violence and Social Orders. The official economic freedom data begins in 1970. I have additional data that goes back to 1950. But I would have democracy and state capacity data going all the way back to the early 19th century. Fortunately, the economic historian Leandro Prados De La Escosura has more rudimentary (but perfectly acceptable) data for a select set of countries going back to 1850.^[1] I would be able to score countries on an open access order measure back a very, very long time.

The result is a continuous variable running from zero to one hundred. Where to draw the line for what counts as an open access order is theoretically arbitrary. In practice, a cutoff of 80 seems to work rather well. It strongly passed face validity and it matched the conclusions of a later book that applied Violence and Social Orders to several case studies.

There are also gaps in the data during World War I and World War II for obvious reasons. For simplicity, we can report the countries classified as open access orders as of 2020, to give a sense of the data.

Coming out of World War II, Switzerland was the only country considered an open access order as of 1950, largely because many economies remained closed. The United States was not considered an open access order until 1965, not because of economic freedom but because the V-Dem Liberal Democracy captures the civil rights issues that were rectified in that decade (this is the data doing what we want it to do). Here are the complete time series from 1950-2020 for eight countries of interest:

Finally, we can consider the select countries for which we have data before 1950. Prior to World War II, there were several countries classified as open access orders: the UK (in 1925), Australia (1905-1910), Denmark (1910), New Zealand (1910), Switzerland (1875-1910), the Netherlands (1925-1930), Norway (1930), and Sweden (1925-1935). The historical variable that changed around this period, causing countries to move across the threshold, is actually women’s suffrage.

The women’s suffrage issue isn’t mentioned at any point by North, Wallis, and Weingast. Raising it as an issue to consider here isn’t “woke;” it is something that fell out of the data, and once you see it, it seems like a good point. Violence and Social Orders cites the UK as likely the first open access order, but the data point rather strongly to Switzerland instead. The issue of women’s suffrage is actually a red herring there, though, as Switzerland just scores so well in everything else that it surpasses the threshold before it extends the franchise to women.

The general pattern in history is that many of these countries were building up welfare states for the first time as they were putting the finishing touches on their democracies. For some, that meant there was a brief window when they all came together at once, and the country was scored as an open access order. France is one country highlighted by North, Wallis, and Weingast as another early open access order, but the data indicate that its economic freedom had receded before it became a full democracy. Here are the time series for Switzerland, the United States, the United Kingdom, and Greece.

I have read in contexts I can’t exactly cite that some economists see a sharp divide between “thick” and “thin” economic history, and they are not very happy with the thin (data-centric and conducive to statistical methods) approach. I don’t know. These results look just about right to me, and where there is conflict between this data and what was concluded via thick economic history (e.g., regarding the UK and France), I am inclined to side with this data. But the two approaches support one another regardless. I would not have thought to create this construct if I had not read Violence and Social Orders.

If there are doubts about how putting a number on something oversimplifies it, here is a chart found in the paper summarizing how much goes into the measure of open access orders:

Further Reading: Ryan H. Murphy. 2023. “Measuring Open Access Orders.” Journal of Institutional Economics 19: 618-634.

Share

^[1] Under the influence of Jan Ott, De La Escosura unfortunately removed the size of government from a later updated edition of his project. I used the earlier version of the data.

He is also famous enough to be the subject of memes. In a deeply unfunny era of the internet, “Chuck Norris Facts” were ubiquitous. I refuse to link to them or to explain what they are. Someone eventually got the idea to do the same thing with Daron Acemoglu.

Hence Daron Acemoglu Facts. I take no joy in reporting this.

Subscribe now

Acemoglu, like me, is a New Institutional Economist. In Why Nations Fail, the central distinction is between “inclusive” and “extractive” institutions. That sounds very similar to open access orders versus limited access orders from Violence and Social Orders by North, Wallis, and Weingast, which I discussed a few days ago. The resemblance is real. Violence and Social Orders is a very boring academic book that, in my view, specifies its mechanisms clearly and is basically correct. Why Nations Fail is 50% longer and told through beautifully scripted historical vignettes. It offers little guidance on how to actually achieve inclusive institutions, aside from the contingencies of history and what certainly feel like platitudes. I am exaggerating the similarities somewhat between open/limited access and inclusive/extractive institutions, but you can tell which book I prefer.

Despite being heavily cited and discussed, Why Nations Fail drew heavy criticism. In 2019, Acemoglu and Robinson published The Narrow Corridor, which specifies a clearer mechanism for arriving at inclusive institutions. If the state becomes too strong relative to society (for example, civil-society organizations), it crushes society. If society becomes too strong relative to the state, it prevents the state from ever really getting off the ground. State and society therefore need to remain roughly in balance. If they do, they push one another forward; once both advance far enough, you get those essential inclusive institutions.

Let me pause to say that I take the most important living scholar in my narrow subfield very seriously. Right before COVID-19 hit, I had the bright idea of organizing an undergraduate reading group for summer 2020 where I provided the books and students read them over break. The Narrow Corridor was the centerpiece. I got eight undergraduates to read it, along with a second book chosen from a list of books on economic development. We met in person, masked, just before the semester began in August 2020. The event is burned into my memory because it was the first time I did pretty much anything after everything shut down in March 2020. The undergraduates seemed to really like it, but, in retrospect, I reckon it succeeded partly because the books gave them something to do while they were stuck at home for all those months.

As I was reading The Narrow Corridor that summer, I was working on my paper about measuring state capacity. My co-author and I realized that Acemoglu and Robinson did not offer a formal empirical test of their theory, instead relying on theoretical models and rich historical vignettes. That made sense: at that point, there was still no long run dataset on state capacity of the sort needed to test their claims. Originally, we thought one application of our data in that paper could be a test of The Narrow Corridor, but it became clear that the task was too large for one part of that project, so we temporarily shelved it.

We eventually returned to the project and mapped our data onto the framework presented in The Narrow Corridor. To represent the strength of the state, we used our state capacity measure. To represent the strength of society, Varieties of Democracy offers numerous measures of civil society strength, because of course it does. We therefore had the two axes needed to see whether the pattern described in The Narrow Corridor held. Here is the book’s central figure:

What we found instead was that, between 1855 and 2015, most countries moved upward and rightward, regardless of whether state and society began in anything resembling balance.

We highlighted four countries as illustrative cases, the United States, Japan, Honduras, and Russia:

We also asked whether a country’s initial position predicted its likelihood of moving upward and rightward. The figure below explores that. Combinations of state power and societal power that are more likely to correspond to moving upwards or rightwards take on bluer hues, while those less likely are red. Along the middle of the diagonal of blocks (i.e., the narrow corridor), we should expect bluer blocks. Instead, we observe:

Finally, we tested whether distance from the middle of the corridor was negatively related to the increases in the sum of civil society strength and state capacity. That is to say, we tested whether being closer to the narrow corridor meant the country moved more upward and rightward. If anything, we found a positive relationship. Being distant from the narrow corridor on average, meant the country was about to move more in the northeasterly direction.

Taken together, these results give me little confidence in the central empirical claim of The Narrow Corridor. We had every professional incentive to find support for the theory, as it would have been useful to produce a series of charts that lined up perfectly with a massive political economy tome that everyone was reading. We did not find the opposite of every claim in the book, but we found no support for its core dynamic pattern.

The books written by Acemoglu and his co-authors since The Narrow Corridor have been both more explicitly political and topical. I have not read them, but I have not felt as obligated to do so. I gave them a thousand pages of runway to convince me what they were saying was novel and true, and they did not do so.

Further Reading:
Ryan H. Murphy and Colin O’Reilly. 2023. “The Expansive Corridor: Testing Acemoglu and Robinson.” Journal of Development Studies 59, no. 7: 1060-1075.

Share

The first exercise was joint work with my co-author, Colin O’Reilly. Two elements of our familiar state capacity data spoke to the question. One was the Rule of Law subindex, which nearly everyone can agree is “good,” whether or not one interprets it as economic freedom. The other was fiscal capacity, which is the sophistication of a country’s tax system.

Subscribe now

Our work on state capacity leaned heavily on a 2009 paper by Besley and Persson. Their theoretical model proposed complementarities between government investment in efficient court systems and investment in tax bureaucracies, often for the purpose of fighting wars. Possessing one creates a reason to develop the other, and vice versa. We would expect, therefore, higher fiscal capacity to lead to subsequent improvements in the rule of law.

We took the V-Dem data on both variables and tested the hypothesis using standard statistical methods. We obtained a result that looked plausible, with fiscal capacity predicting the rule of law. It was, more or less, directionally in support of our hypothesis. But I do not trust it. If the pattern is “sometimes statistically significant, sometimes not, with no obvious theoretical reason for the difference,” the correct conclusion is that nothing is there. I would even accept treating this as evidence against the theory. We were motivated to find a result and failed to do so, even with our hands on the wheel. All of this can be explained away, but we were not off to a great start. We still published the paper.

Subsequently, I began a separate paper with the idea for the following exercise: what if we removed every component of the Economic Freedom of the World index directly related to the government doing something competently (that is, state capacity), and rebuilt the index? In practice, I removed all variables concerning the legal system and property rights, most the Sound Money variables (because nearly every country has government-controlled money), and a small set of regulation variables that related to the bad government performance (rather than just the government sitting it out).

The countries that rose sharply were not the United States and other countries similar to it. They were failing states that had attempted economic liberalization. Some might have been doing reasonably well, but the new ranking put Haiti ninth and Cambodia eleventh. Meanwhile, clear market economies with social democratic governments, such as France and Norway, fell into the third quartile of the rankings.

The correct middle ground is neither that the classical liberal view that effective government harms freedom nor the Jan Ott view that effective government defines freedom, but that it is an input into economic freedom that we pick up in our existing indicators

Finally, and most recently, O’Reilly and I analyzed the data using vector autoregression. This is not a method on which one should stake a professional view with respect to causality. It was developed in the context of variables like inflation and interest rates and unemployment, and upon imposing various assumptions, how does one variable move in response to the others over time? Because you narrow the scope to your question to only a set of variables that you assume to be the important ones, it is limited in how much you can say about whether one variable really causes another. But in a complex system of variables, it allows you to at least say how they interrelate.

One hypothesis in political science holds that a country needs some amount of economic development, education, and income before it can democratize successfully. Elsewhere, major figures in the state capacity literature argue that a country must develop an effective state before it can do much of anything. Obviously, the mainstream academic view is that democracy is central and economic freedom or state capacity are secondary of secondary importance, at best.

The traditional limiting factor for vector autoregression was the number of years of available data. The data Colin and I had developed loosened that constraint for state capacity. Democracy was not a problem because of V-Dem. We could also treat “economic modernization” as roughly equivalent to economic freedom because it sets the stage for economic development. Economic freedom data extended back to at least 1970. We could therefore ask which impacts which and came first: state capacity, economic freedom, or democracy?

There are several ways to perform this analysis, as with any analysis. But among the assumptions that are specific to vector autoregression, an essential one is that you need to first order the variables, meaning that you must assume which variables can impact each other on a contemporaneous basis. Three variables is not a lot, so we tried all possible orderings of the three. There was only one reasonably robust result, that increases in state capacity were followed by subsequent increases in economic freedom.

I do not regard this as definitive, and I suspect part of the result is a residue of how quickly the effects of state capacity appear in the data compared with the delay before they appear in economic freedom data. But the result is something, and it complicates certain narratives about institutions and economic development.

I should also note that the vector autoregression simultaneously also functioned as a test of the Modernization Hypothesis, which states that a society must first accomplish other steps before democracy can be durably established. We find no evidence in favor of this hypothesis.

Further reading:
Ryan H. Murphy and Colin O’Reilly. 2023. “Freedom through Taxation: The Effect of Fiscal Capacity on the Rule of Law.” European Journal of Law and Economics 56: 69-90.
Ryan H. Murphy. 2022. “Economic Freedom without Quality of Government.” Journal of Private Enterprise 37, no. 4: 21-42.
Ryan H. Murphy and Colin O’Reilly. “Sequencing Institutional Development: The Modernization Hypothesis Reconsidered.” Journal of Comparative Economics, forthcoming.

Share

If the ruler has a short time horizon, he may pillage his own country, (hopefully the other elites restrain him). Everyone benefits if the ruler feels secure that he and his descendants will remain in power. Hence, “long live the king.” Other things equal, therefore, the ruler hopes to tax his people without ruining the economy.

Subscribe now

Given a sufficiently long time horizon, the ruler has several ways to raise more tax revenue, without simply increasing tax rates:

-One option for a desperate sovereign is to sell privileges or monopolies. Selling a monopoly brings in immediate revenue, and perhaps an additional revenue stream, but at the expense of the broader economy.

-The state can spend resources determining who and what it rules, where they are located, and how it can track them. This concept is known as “social legibility.” The state might, for instance, conduct a census. Instead of relying on other elites as intermediaries, or hiring “tax farmers,” it might build a professional bureaucracy to collect taxes. If taxation is then administered more rationally, the ruler can in principle raise more revenue without doing as much harm to the economy.

-The ruler can seek political legitimacy. One way to do this is by actually protecting the people and securing justice. Alternatively, he can portray his rule as religiously ordained. Today, democracy supports political legitimacy, while stable autocracies often derive legitimacy from high economic performance, as in China and Singapore. States perceived as legitimate are more likely to secure voluntary tax compliance, raising revenue or making it less costly to raise the revenue.

-The ruler can take actions to improve the economy, whatever exactly that means, in anticipation of greater future tax revenue.

-The state can build its coercive capacity to extract more revenue at gunpoint or through the threat of imprisonment. But coercion perceived as excessive may reduce political legitimacy, and in extreme circumstances a state that takes too much may cause a famine or other disasters.

-If worst comes to worst, there is always good old fashioned currency debasement.

Greater economic freedom will improve economic performance and therefore increase tax revenue over time, so why not invest in it? Because liberalization either weakens the political coalition sustaining the ruler’s power or causes revenue to fall for other reasons, at least in the short run.

The recurring pattern is that rulers pursue these alternatives to liberalization (that is, more economic freedom) when they lack the state capacity to do the “smarter” thing. Modern countries with open access orders purchase political legitimacy largely by doing things that economists approve of. They achieve strong economic performance and then purchase legitimacy by providing public goods such as education or simply handing out money. They finance this spending with relatively efficient taxes that require substantial state capacity and sophistication to administer, particularly income and consumption taxes (such as sales taxes and the value-added tax).

For each variable, then, we should compare how it operates in an open access order with how it operates when the ruler cannot maintain political power and legitimacy through these more efficient channels.

Begin with Economic Freedom of the World Area 2, the Quality of the Legal System and Property Rights. States with little capacity cannot enforce the rule of law, and their courts, where they exist, are corrupt. But the corruption is often by design. Positions in the legal system are desirable sinecures for elites or people adjacent to them. The ruler could improve the economy by strengthening the legal system and protecting the lives and property of ordinary people, but that often conflicts with the interests of elites who use the courts to get their way and appointees who treat them as opportunities for corruption.

Area 3 of the Economic Freedom of the World index is Sound Money. Hyperinflation is an awful way to raise revenue, and it generally occurs when a ruler is desperate and lacks other means (capacity) to raise it. One Sound Money variable illustrates the point. Some countries prohibit citizens from holding bank accounts denominated in foreign currency, while others prohibit them from holding bank accounts abroad. Some prohibit both. The transparent purpose of these laws is to preserve the ruler’s option to debase domestic money or confiscate bank accounts. A high-capacity state that can raise taxes in some reasonable way has little need to preserve that option.

Area 4, the Freedom to Trade Internationally, is concerned chiefly with tariffs and other restraints on trade. Historically, including in the United States, tariffs were principally “revenue tariffs;” they historically functioned as a main source of revenue for states with middling or low capacity. It is administratively easy to establish offices at a border or port and tax a percentage of the goods entering the country, rather than administer an income tax. The other rationale for tariffs is simple cronyism, whereby leaders purchase the support of politically important people by taxing their foreign competitors. This is less politically necessary in high capacity states, which have less economically costly ways to purchase political support.

The same story can be told about regulation. When countries are compared with one another, much of the variation in the harmfulness of their regulatory environments is driven by cronyism, and political leaders rely on cronyism most in countries with weak state capacity. That leaves one final area of economic freedom, the area Jan Ott regarded as the outlier among them, the Size of Government.

The Size of Government has five components:^[1] government consumption, transfers and subsidies, government investment, the top marginal tax rate, and government ownership of the economy. Government consumption can be understood as paying teachers’ salaries or buying bombs for the military: resources used during that year, or more or less that year. Transfers and subsidies involve the government handing money to another entity. The largest items in the United States are Social Security, Medicare, and Medicaid, though the category also includes spending such as agricultural subsidies. Government investment means building a bridge or school or, in more socialist countries, a factory. The top marginal tax rate is what it sounds like. Government ownership of the economy corresponds to the conventional definition of socialism, “the collective ownership of the means of production.”

The irony of Ott’s argument is that, even as he claimed Economic Freedom of the World was incoherent because Size of Government was not positively correlated with the rest of the index, Size of Government itself is not “internally consistent” in the sense that Ott wanted constructs to be. It appears to measure two or perhaps two and a half concepts simultaneously, and those concepts run more against each other than the Size of Government itself does against the other areas of economic freedom. If Ott’s view were correct, it would imply that the concept of the Size of Government is incoherent, which is silly.

If we simply follow the correlations, the Size of Government variables divide into two groups: (1) government investment and government ownership of the economy, and (2) government consumption, transfers and subsidies, and (maybe) the top marginal tax rate. Government investment and ownership move together because, in actual countries, they reflect how extensively a country attempts socialism. There is something of a paradox in the data. Countries where governments dominate investment are universally poorly governed countries. But when a Denmark (or even a United States) decides to do a bit more government investment and build a bridge, it appears to often benefit the economy. Government investment on average appears to be better than government consumption, because in the end you usually still end up with a bridge.

But when you look at the countries that do a lot of public sector investment, it appears they are often following the earlier model. The investment is just another form of cronyism. And often government ownership also frequently indicates that a state is too incompetent to tax income and must nationalize its source to capture the revenue. Again, less state capacity means less economic freedom.

Everything changes with the final set of variables. Public finance economics repeatedly tells us that if we want to make the poor better off, we should provide public goods and simply give people money, financed through sophisticated tax systems. And what do high capacity states do? They purchase political support by spending enormous sums on public goods such as education or simply transferring money, while otherwise operating essentially as market economies.

What I have given you is a narrative, not a set of established facts. The paper that develops this argument cites other studies that support much of the narrative, if only indirectly. In the next post, I describe three data-driven exercises designed to provide evidence for these claims.

Still, I think the basic correlations are meaningful and align almost perfectly with North, Wallis, and Weingast. Once the essentials of a state are in place and the ruling class has an incentive to move the economy forward, countries shift toward markets, except where rulers need to purchase popular support in the most economically efficient way available. Is this state capacity libertarianism? I do not know. But I think it is correct.

Further reading: Ryan H. Murphy. 2023. “State Capacity, Economic Freedom, and Classical Liberalism.” Constitutional Political Economy 34: 165-187.

Share

^[1] There are really seven variables underlying the Size of Government, but that only overcomplicates matters.

If the groups know that bandits are numerous, no on in the group will have incentive to invest. Any scarce resource in your possession should be consumed immediately because a bandit will simply take whatever you produce. If you chop down a tree, you might as well use all of it for firewood rather than use some of the wood to build a tool that a bandit will later seize.

Subscribe now

Now suppose one bandit stops roving and finds a small group from which he intends to steal continually. He becomes a stationary bandit. Suddenly, the bandit’s incentives change. First, he wants to protect the group from other bandits. Anything they take will no longer be available to him.

Second, the bandit realizes that if he takes everything from the people he now protects, they will stop producing. In the long run, he can steal more by taking only a percentage of their output. He therefore makes an implicit deal with them: he will demand only a share.

The bandit then realizes that he has an incentive to provide certain things that no one else does. Perhaps the group lives near a river that floods periodically, destroying its crops. The people know that constructing a levee would prevent the flooding. But the project requires so much work that no individual or family is motivated to undertake it, and coordinating the entire group is difficult.

Who has the incentive to get the levee built? The bandit. He cares about the group’s overall economic output in a way that its individual members and families do not. The bandit therefore has an incentive to provide public goods.

What do we call an entity that protects borders from outside invaders, taxes a proportion of its economic output, and provides public goods? A state.

Whoever becomes a bandit is likely to be well suited to violence. As a stationary bandit remains in place, he will probably start a family and put down roots. Eventually, the group will be controlled not by a single bandit but by a coterie of family members, cronies, and deputies. These people become a ruling elite in political coalition with one another. Such early polities may be fragile: a family member or deputy may challenge the leader in hopes of becoming ruler.

To quell conflict at the top, the elites establish rules that they must follow. They need more than a line of succession. and the ruler needs the loyalty of those beneath him, he makes credible promises that they can enforce, including a promise not to seize their stuff arbitrarily. Property rights are defined for the elite, so everyone agrees about which elite controls each economic asset and the ruler cannot simply take it away.

As the ruler expands his country, he will probably absorb other stationary bandits’ groups, placing each local bandit in a tributary position or installing one of his own auxiliaries. Eventually, the ruler will want more taxes from these auxiliaries, who in turn tax the locals, probably to finance a war. The ruler may be justified in asking for more taxes. To manage the resulting conflict, the auxiliaries begin meeting and voting on whether to grant them. In other words, they form a parliament.

As economic activity becomes more sophisticated, it becomes useful to form organizations such as businesses or religious organizations which give rulers more legitimacy over their rule. At first, these organizations are highly personalized: they either die with the elite who runs them or remain within a family. But it is useful for organizations to persist indefinitely without being tied to any one name. The state therefore develops structures that allow entities resembling corporations to serve the purposes of the elite.

Late in the game, it is recognized that it no longer necessarily makes sense for the ruler to personally lead the military into war. But a professional army separate from the direct command of a now-civilian ruler is dangerous. To this day, one of the greatest threats to a civilian dictator is losing control of the military. Institutional structures must therefore separate civilian and military roles while placing the military firmly under civilian control.

Throughout this whole process of state building, the elite act solely for their own benefit, even when their actions also help ordinary people. They have secured property rights and the rule of law for themselves, formalized voting rules and procedures, created modern institutional forms such as corporations, and built a military that will not kill the ruler merely because its leaders think they can gain the upper hand.

When a society’s institutions serve a narrow elite, it is a “limited access order.” Once the preceding conditions are met, however, the elites may realize that they themselves would also benefit from allowing more ordinary people (perhaps beginning with the wealthier among them) to use those institutions as well. Democracy gives the elite important feedback and lends the state greater legitimacy for its actions. Extending the rule of law, property rights, and the ability for anyone to form corporations and other organizations improves economic performance, whose fruits the elite can enjoy. If the elites choose to do so (and they may always say no_ they can create an “open access order.”

That is the story of political and economic development from the beginning of time to the late nineteenth century, which is where I date the first appearance of open access orders.

This story also explains the positive relationship between state capacity and nearly every dimension of economic freedom.

Further reading:
The story of roving and stationary bandits originates with Mancur Olson. See also his Power and Prosperity.
The story of limited and open access orders originates with Douglass North, John Joseph Wallis, and Barry Weingast.
One may also consider reading the early 2010s work of Francis Fukuyama.

Share

In an earlier post, I described the Economic Freedom of the World index only in general terms so as not to be boring. Here we are going to get more explicit. Economic freedom has five areas: size of government, legal system and property rights, sound money, freedom to trade internationally, and regulation. For each of these “areas,” we have numerous pieces of data that underlie them. We convert all of them to a scale from zero to ten, with higher scores always corresponding to greater economic freedom. For size of government, higher scores correspond to smaller government; for regulation, higher scores mean less regulation.

Subscribe now

Market economies are generally wealthy, and wealthy countries generally have large welfare states. Look at Europe. One interpretation is that a country gets rich by opening itself to the global economy and, if it is a democracy, eventually votes itself a welfare state. The welfare state is not the only cause of government size. Still, although the other four areas of economic freedom are highly correlated with one another, the correlation between size of government and the rest of the index is basically zero: overall, there is no relationship.

Ott’s criticism had two main points: the size of government area harms the predictive validity and convergent validity of Economic Freedom of the World. Ott claims that if we remove size of government and recalculate the index, the raw correlation between economic freedom and happiness increases. If we believe economic freedom benefits society, he treats this as a problem. We could improve the predictive validity of Economic Freedom of the World, on account of this, by removing size of government.

The second point is that a statistical test can tell us whether the components of an index meant to measure the same thing are internally consistent. When you apply that test to a version of Economic Freedom of the World that omits the size of government, it scores well. The score of that test falls substantially when you then include the size of government. Ott claims that omitting size of government would therefore also improve the convergent validity of Economic Freedom of the World.

One of the biggest things social scientists have learned over the past two decades is how easy it is to obtain “statistically significant” results by manipulating the analysis, often in subtle ways to which no one step would draw an objection. “Statistical significance” just means that, under our own conventions, an observed result is unlikely to be due to chance alone (there could still be many other problems with the analysis, of course). Those conventions are not nearly as strict as they appear if researchers can alter how they perform an analysis until they get the result they want.

In the previous section, I discussed how my co-author and I used predictive validity not to build or adjust our measure of state capacity, but to assess its performance conditional on choices we had already made. The idea that one should use predictive validity to select how one measures a concept gets the logic of social science backward. Adjustments post-hoc based on how well a measure correlates also risks other issues, such as overfitting.

Suppose we laid out all possible approaches to conducting a scientific test on one continuum. At one extreme of the continuum, we simply compare two groups while accounting for nothing else. At the other, we conduct the equivalent of a carefully run medical trial in which the only difference between two groups is whether one receives a new drug. In economics, the quality of our work generally falls somewhere between the two extremes. Ott’s claims about happiness, size of government, and economic freedom were at the first extreme: they relied on the raw relationships among the variables. That is to say, it is weak evidence of much of anything from the standpoint of causality, even leaving aside the question of whether it is appropriate to use predictive validity in this manner.

The problem with Ott’s use of convergent validity is more conceptual. Convergent validity can be important for what is known as a “reflective construct,” something we cannot observe directly and must infer from indirect indicators. This is the situation in my previous post, where test questions are needed to determine whether students understand supply and demand. It is a problem if one question produces a different answer from all the others.

Not all indexes measure reflective constructs. Some measure “formative constructs,” for which the included components follow from the definition of what is being measured. Economic Freedom of the World is such an index. There is no single latent variable that we are trying to detect indirectly. Five distinct concepts constitute economic freedom, and one of them is limited size of government.

There is another aspect of construct validity that we have not mentioned: content validity. This asks whether an index captures everything that it claims to measure. For a formative construct, content validity is especially important when one component is weakly or negatively correlated with the others, because that component supplies a whole lot more information. It is because size of government is not positively correlated with the rest of economic freedom that including it is essential.

In earlier times, academic journals welcomed comments on work they had previously published. That practice has generally faded, but a handful of specialty outlets have taken its place by providing a platform for such scholarly dialogue. One such platform is Econ Journal Watch. I published a paper critically assessing Ott’s arguments here and here.

In his responses, Ott makes clear that his real issue was with the classically liberal definition of freedom and with whether governments can facilitate freedom of choice. I will not get bogged down in the philosophical conflation of positive and negative freedom, or Ott’s attempt to collapse them into one, but economic freedom has a long intellectual history. It constitutes one specific version of the New Institutional Economics hypothesis that institutions matter and are central to economic performance. If that hypothesis is false, then fine, but the data as currently presented permit us to test it, which is one of the index’s fundamental purposes. Ott sought to ruin the index’s content validity while inserting a separate, competing hypothesis (that quality of government or state capacity is what matters) into the index.

As I would later argue, and as we will see next, quality of government and state capacity do not constitute economic freedom, but they may be inputs into it. A country can have high state capacity with low economic freedom, or low state capacity with high economic freedom (though neither may be a very stable equilibrium), and some countries do look like this. More generally, state capacity tends to facilitate economic freedom even as it also facilitates the growth of the welfare state.

As an aside, this did not make it into the final version of the responses, but I want to point out that Ott’s latter papers employed essentialist approaches to the word “freedom.” They presume that this is a “true” definition of the arbitrary combination of noises that we emit when we say a word. My view is that Ott can have a word and I don’t care about terminology, but the index did not invent the definition of economic freedom that it employs. A figure such as Murray Rothbard would spend a lot of time defending the definitions of “freedom” of “liberty” in exclusively negative terms. I honestly don’t care. If you want “true” freedom to involve some version of a positive aspect of choice, I can call negative freedom “freedom_1” and absolutely nothing would change about the empirical arguments. And the empirical arguments are the ones that matter, mostly.

The primary papers the proceeding discussion was based upon were:
Ryan H. Murphy. 2022. “On Whether the Size of Government Belongs in Economic Freedom Indices.” Econ Journal Watch 19, no. 1: 47-57.
Ryan H. Murphy. 2022. “Freedom Stands: A Rejoinder to Ott.” Econ Journal Watch 19, no. 2: 242-246.

Share

A colleague told me about the project at a conference, and I saw how we could use it for Economic Freedom of the World, especially for measuring legal systems and property rights. (It also contains useful material beyond that, because it is V-Dem.) We did end up incorporating it. Later, when I was racking my brain for a way to test state capacity libertarianism, I recalled that V-Dem contained various bits and pieces that fit.

Subscribe now

After the fact, my co-author, Colin O’Reilly, and I realized that we might have something, provided we did it correctly rather than rely on what we had assembled somewhat hastily for the earlier paper. I had not really been part of the state capacity literature, but it seemed important to measure the variable as its proponents intended. We therefore worked through several literature reviews in which those proponents articulated the concepts central to state capacity.

Other classical liberals who were challenging Tyler Cowen’s arguments, as well as the state capacity literature more generally, focused on what was undeniably one dimension of state capacity: coercive capacity. A very weak state cannot control its borders, collect tribute from the locals before they make it into the hills, or enforce its laws. But that is not all that state capacity proponents mean. State capacity is not a single dimension with Somalia at one end and Russia at the other. If we consider only coercive capacity, states around the 40th percentile (think China^[1]) already appear to max out the scale. Coercive capacity alone is simply not the hypothesis in question.

Rather, in addition to coercive capacity, state capacity comprises several overlapping concepts:

1. Fiscal capacity: the ability of the state to administer complex tax systems that are less damaging than alternative tax systems for any given level of revenue.

V-Dem has a variable that categorizes how a state raises most of its revenue along these lines. We used this.

2. Legal capacity: the ability of the state to create the conditions necessary for the rule of law.

V-Dem has several variables measuring the rule of law, which it combines into its own rule-of-law index. We used that index.

3. Bureaucratic quality: the ability of the state to gather information and perform its functions cheaply and without corruption.

V-Dem has three relevant variables: a measure of the quality of public administration, the extent to which public spending goes to genuinely public goods rather than favored interest groups, and a measure of educational equality. On the last measure, note that it captures equality, whether all children receive some basic level of education rather than overall educational quality. Since all governments, on some level, are trying to provide basic education, whether they succeed is a measure of bureaucratic competence.

Finally, V-Dem includes a measure of state control over its own territory, which serves as our measure of coercive capacity. Combining these variables yields our index. From the paper, in the year 2000. Jere are top and bottom of the rankings:

Most countries at the top and bottom of the list align with our intuitions. There are a few surprises: Spain in the top ten; Slovenia, Portugal, Costa Rica, and Botswana in the top twenty; and the United States not quite making the top twenty. (In the current version of the data, the United States ranked 25th in 2000.)

There are several ways to assess whether a measure captures what it purports to measure. The formal study of how to do that is called “construct validity.” Simply inspecting the results to see whether they look right is called “face validity.” But you do not want to change an academic paper merely to give every result perfect face validity. Going down that road leads you to p-hacking, or something roughly equivalent. There is also what is known in sabermetrics as the Bill James 80/20 rule: a metric should confirm what you believe only about 80 percent of the time. If it confirms much more than that, you do not learn anything from it. If it confirms much less than that, it is failing face validity.

Although we did not use those terms, much of the paper performs several additional checks of construct validity. One is predictive validity: does the measure predict what theory says it should predict? Prior scholarship gives us reason to believe that state capacity positively predicts prosperity. Historical threats of war may also produce state capacity by forcing states to get their act together, though civil war apparently just destroys institutions. All three relationships appeared in our data.

We can pose other questions as well. Convergent validity means that a measure is correlated, ideally closely, with other measures of the same concept. Suppose an education researcher is trying to figure out whether a student understands supply and demand. If there are five kinds of questions that could be posed and four identify the same students as understanding the concept, the fifth question’s failure to identify those students counts against it. In the same way, we would hope that our measure and other measures of state capacity identify broadly the same countries as having low or high capacity.

Another form of construct validity is discriminant validity: the ability to distinguish our new measure from closely related but conceptually distinct ideas like democracy. This is an especially important concern because we derive our measure from a data set centered on democracy and, even apart from that, we would already expect state capacity and democracy to be positively related. In the paper, we showed that our measure had a tighter relationship with other measures of state strength (from datasets by the International Country Risk Guide, Worldwide Governance Indicators, and the Fragile States Index) than with three measures of democracy (from V-Dem itself, Polity IV, and Freedom House). We take this as evidence for both convergent and discriminant validity.

Our method may raise some objections concerning face validity, but overall, it appears to perform well in terms of construct validity. And importantly, it provides data for a large number of countries over a long period.

Ideas like predictive and convergent validity may seem arcane, but one scholar used them in an attempt to discredit the Economic Freedom of the World project. The resulting academic exchange is the subject of my next post.

Further reading:
Colin O’Reilly and Ryan H. Murphy. 2022. “An Index Measuring State Capacity, 1789-2018,” Economica 89, no. 355: 713-7145713–745.

Share

^[1] A lot of China’s apparent successes here are not as they appear, including its high speed rail.

Classical liberals tend not to like states very much. Considering modern history, this makes sense. After weakening the shackles of feudalism and unleashing the forces of markets and liberalism, modern states began to grow more capable of regulating every remaining aspect of private life while squeezing out every last dollar. Making states more effective seems like an awful idea if your wish is to achieve liberty.

A state limited to protecting property rights and providing the minimal structures necessary for markets to function can be said to have classically liberal economic institutions. “Institutions,” in this context, means the rules that govern a society. A guild is a feudal economic institution. Free trade is a classically liberal economic institution. Democracy is a political institution. The family is a social institution.

Subscribe now

New Institutional Economics studies the effects of these institutions, including how many of them are essential to the functioning of modern societies and the well-being of their members. One project with which I am closely affiliated is the Economic Freedom of the World index. I do not want to get into the technical details here, but the project assigns a number to the extent of classically liberal economic institutions in every country for which we can obtain data. We refer to classically liberal economic institutions as “economic freedom.”

In the terms above, classical liberals see a fundamental conflict between strong state institutions and economic freedom. A powerful state crushes freedom. If we want a great deal of economic freedom, we should keep states weak and unable to regulate us. “Become ungovernable,” as they say online.

Tyler Cowen argued the contrary. In 2006, he called the libertarian celebration of failures of the state and stories of its ineptitude the libertarian vice. In subsequent years, development economists, economic historians, and political scientists increasingly embraced the idea that strong state institutions are essential for societies to flourish, calling the concept “state capacity.” In 2020, Cowen combined these ideas with classical liberalism to formulate “State Capacity Libertarianism.”

Cowen’s own description of the phrase is effective enough, but the idea is that libertarianism must reckon with the fact that many of society’s biggest problems involve getting the state to perform effectively at what it does, or is supposed to be doing. We should not greet the apparent inability of much of the United States to build infrastructure with glee and argue that it means we should weaken the state further. Rather, we should focus on improving the state’s ability to support markets and make the kinds of investments that promote economic growth and prosperity.

Early in my career, I was always on the lookout for testable hypotheses that had not yet been examined. I did not have a dog in the fight, so when I saw Cowen’s post go up, I saw a scholarly opportunity. I called up a co-author, got him onboard with it, and some months later we published “Assessing State Capacity Libertarianism” in Cato Journal.

A stylized fact of the world is that nearly everything positive is rather correlated with everything that is good. Where you have high standards of material living, you also have high levels of education, good health outcomes, democratic governments, and high levels of economic freedom. Strong states, as conceived in the state-capacity literature, are positively correlated with these outcomes as well.

But what if we looked at how quickly countries were getting richer (that is, economic growth), rather than how rich they were at a single moment? The wealthier a country becomes, the more difficult it is to maintain high growth rates. We can relate both economic freedom and state capacity to the growth rate and have them fight it out in the data. The pure classical liberal view might be that strong states are only superficially associated with wealthy countries; once you account for economic freedom and a few other factors, stronger states may lead to lower growth rates.

In contrast, economic freedom may not “work” without a strong state supporting it. Moreover, the state capacity libertarianism view seems to imply that even in a country such as the United States, where the state is problematic but above average globally, improving state capacity also strengthens the effect of economic freedom on growth. In statistical terms, this is an “interaction” between the two variables. If the strong version of state capacity libertarianism holds, we should observe a positive interaction between economic freedom and state capacity.

As for the data, I have my own measure of economic freedom, of course. But how should we measure state capacity? The most popular measures of institutions come from the Worldwide Governance Indicators, so we could start there. I had previously created another related measure, but it was not a perfect conceptual fit. I also knew that buried in the Varieties of Democracy data set were some other measures we could use, so I gathered them up. What emerged will be the topic of my next post.

The short answer is that, the general pattern is that both economic freedom and state capacity are associated with higher growth in what we believe are the highest quality statistical tests. Their interaction, however, is negative rather than positive. The more economic freedom a country has, the less additional state capacity matters for growth, and vice versa. Read too credulously, some of the statistical analysis implies that further improvements in one eventually make the other harmful to growth, though my co-author and I did not interpret it that way.

Our result therefore supported the ideas that economic freedom is good for growth and that state capacity is good for growth, while simultaneously suggesting that State Capacity Libertarianism, as Cowen described it, was wrong.

And then I realized, did we just invent a way to measure state capacity?

That is the topic of the next post.

Further reading:

Ryan H. Murphy and Colin O’Reilly. 2020. “Assessing State Capacity Libertarianism.” Cato Journal 40, no. 3: 735-767.