Liberty as Amenity: Freedom, Migration, and Growth

Is liberty an “amenity” that people find attractive? We know that people do not necessarily tend to vote for liberty, in part because they are politically ignorant or even irrational, but when it comes to where they choose to live, people can be expected to pay close attention to how the laws in different places affect their quality of life. Economists model migration rates across jurisdictions as a function of economic opportunities (real income differentials) and “amenities” (example). Thus, it is standard practice in the literature to use inter-state migration rates in the U.S. (adjusted for the component predicted by economic growth) as a proxy for the desirability of different states as places to live. The question I address here is whether liberty is an amenity; in other words, do states with more freedoms attract more people?*

My study with William Ruger, Freedom in the 50 States, addressed this issue briefly. We find that both economic and personal freedom are associated with net inter-state migration over the 2000-2009 period. In other words, freer states attract people from less free states. The relationship holds when we control for climate, measured as average January temperature in a state’s largest city. We also find that real personal income growth (total, not per capita) over 2000-2007 is positively associated with economic but not personal freedom. Thus, it remains an open question whether economic freedom attracts people because people find it desirable for its own sake, or whether it attracts people by promoting economic growth. However, it does appear that people are attracted to personal freedom for its own sake.

This blog post offers a first look at a much more sophisticated analysis of the issue, bringing in more control variables and more advanced, appropriate estimation methods. The scope of this analysis is limited to the years 2000-2007 because those are the years for which we have data on consumer prices at the state level, allowing us to test the effects of freedom on both migration and real personal income growth (per capita this time), as well as the reciprocal effects of growth and migration upon each other. I also change the Ruger-Sorens measurements of economic freedom by substituting fiscal data from Fiscal Year 2000 (the 2007 Ruger-Sorens data are based on Fiscal Year 2006 data). I do this so that the measure of economic freedom takes into account the situation as of the start of the period of analysis. There is probably a lag between policy change and change in people’s perceptions of policies. It would therefore be ideal to measure all policies as of 2000, but that’s impossible because most of the freedom data are based on in-depth statutory analysis.

Here is a table showing each state’s total net migration from other states as a percentage of 2000 population (2000-7), real personal income per capita growth on an annualized basis (2000-7), economic freedom, and personal freedom:

State Migration Growth Economic Personal
Alabama 1.23% 1.34% 0.223 -0.073
Alaska -0.43% 0.15% -0.469 0.153
Arizona 12.49% -0.16% 0.162 0.062
Arkansas 2.34% 1.77% 0.009 0.048
California -3.73% -1.28% -0.206 -0.028
Colorado 3.20% -1.20% 0.296 0.093
Connecticut -2.31% 0.50% -0.092 -0.039
Delaware 5.05% 0.49% -0.033 -0.123
Florida 7.67% 1.16% 0.229 0.070
Georgia 5.84% -0.32% 0.223 -0.081
Hawaii -1.52% -0.27% -0.390 -0.120
Idaho 7.68% 0.25% 0.182 0.091
Illinois -4.30% 0.78% 0.062 -0.164
Indiana -0.26% 0.31% 0.277 0.164
Iowa -1.69% 1.33% 0.222 0.025
Kansas -2.59% 1.40% 0.207 0.072
Kentucky 1.61% 0.25% -0.045 0.020
Louisiana -7.78% 2.79% -0.100 -0.080
Maine 2.79% 0.63% -0.230 0.119
Maryland -0.92% 0.99% -0.018 -0.223
Massachusetts -4.27% -0.16% -0.116 -0.144
Michigan -3.53% -1.85% 0.039 0.003
Minnesota -0.65% 0.83% -0.233 0.033
Mississippi -0.96% 1.52% -0.013 0.081
Missouri 0.76% 0.69% 0.246 0.073
Montana 3.53% 1.47% -0.007 -0.039
Nebraska -2.09% 1.48% 0.188 0.023
Nevada 17.84% -0.35% 0.089 0.111
New_Hampshire 3.11% -0.26% 0.420 0.019
New_Jersey -4.47% -0.01% -0.124 -0.106
New_Mexico 1.29% 0.96% -0.328 0.142
New_York -7.69% 0.66% -0.468 -0.167
North_Carolina 6.25% 0.08% 0.181 0.084
North_Dakota -3.01% 2.62% 0.225 -0.039
Ohio -2.52% 0.23% -0.005 -0.115
Oklahoma 0.49% 1.86% 0.182 -0.020
Oregon 3.97% -1.10% 0.036 0.168
Pennsylvania -0.28% 0.37% 0.077 -0.039
Rhode_Island -2.99% 0.95% -0.152 -0.158
South_Carolina 5.65% 0.47% 0.140 -0.075
South_Dakota 0.44% 2.08% 0.382 -0.037
Tennessee 3.75% 0.49% 0.320 -0.019
Texas 2.73% 0.70% 0.197 0.074
Utah 1.23% -0.45% 0.042 0.036
Vermont 0.13% 0.76% -0.156 0.208
Virginia 2.49% 1.29% 0.231 0.027
Washington 2.67% -0.66% -0.161 -0.133
West_Virginia 0.44% 1.27% -0.210 -0.002
Wisconsin 0.02% 0.66% -0.098 0.049
Wyoming 1.97% 2.79% 0.162 0.046

My first step is to run simple, linear regressions of migration and growth on economic and personal freedom and a set of controls. Then I am going to see whether I can parse the channels by which economic freedom might affect migration, and whether the results hold up to a more sophisticated analysis.

The controls in the migration equation are the natural log of annual heating degree days, averaged over the entire state using local population weights (more heating degree days mean the state is colder in winter), and annual precipitation in inches (averaged over the entire state in the same way). I have also tried using violent crime rates, cost of living, and state and national park acreage, but none of these variables contributed any explanatory power to the model. What we expect to see here is that the coefficient estimates on “Economic” and “Personal” (freedom) are positive and statistically significant. Here are the results:


Estimate Std. Error t value Pr(>|t|)
(Intercept) 16.01410 6.44046 2.486 0.01668 *
Economic 7.14774 2.64294 2.704 0.00962 **
Personal 14.05028 6.02897 2.330 0.02432 *
Log HDD -1.46825 0.66415 -2.211 0.03218 *
Precip. -0.07885 0.04465 -1.766 0.08416 .
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3.882 on 45 degrees of freedom
Multiple R-squared: 0.3425, Adjusted R-squared: 0.284
F-statistic: 5.859 on 4 and 45 DF, p-value: 0.0006965

Sunnier, warmer, and freer states attract more people. So far so good. Now for economic growth. To predict growth per capita, I use the following controls: change in population from 2000 to 2007 as a percentage of 2000 population excluding interstate migration (so just natural increase and immigration from abroad), log of capital per worker in 2000, and mining GDP as a percentage of total state GDP. I also tried a wide variety of education and human capital controls (high school dropouts, workforce with college degree, patents, NSF funding, various average scores on national examinations, etc.), but none of them added anything to the model. The coefficient on “Economic” should be positive and statistically significant. Here are the results:


Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.82984 2.66836 1.435 0.158122
Economic 0.92552 0.56861 1.628 0.110573
Pop. Growth -0.10561 0.04212 -2.507 0.015851 *
Cap./worker -0.33696 0.32367 -1.041 0.303411
Mining 0.11544 0.03123 3.696 0.000592 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.8314 on 45 degrees of freedom
Multiple R-squared: 0.3637, Adjusted R-squared: 0.3071
F-statistic: 6.43 on 4 and 45 DF, p-value: 0.0003512

So mining states did much better in per capita income growth over this period, undoubtedly due to rising energy prices. Also, population growth inhibited per capita income growth – no surprise there, since babies don’t add to GDP (and recent international migrants probably earn less on average than the pre-existing population). Economic freedom is positively correlated with per capita income growth, but it’s not quite statistically significant. Also, richer states at the start of the period might have grown more slowly, all else equal, a phenomenon economists call “conditional convergence.”

So at this point the evidence that freedom attracts people looks much stronger than the evidence that economic freedom in particular boosts the efficiency and productivity of a state economy. Does that mean that people are largely attracted to economic freedom for its direct household benefits (for instance, in lower taxes), not because it creates business opportunities and boosts productivity? Not so fast. After all, in-migration should also decrease per capita income. Why? Even if the people moving in are talented, they are increasing the labor supply, holding the supply of capital constant. With more labor working on the same amount of capital, each unit of labor will be less productive, all else equal, and should earn less, since in a competitive economy wages will tend to equal marginal productivity. So migration and income growth should have effects on each other. The story should look something like this:

How can we estimate all those arrows? There’s an econometric model called “three-stage least squares” that allows us to do just this. I’m going to estimate two “simultaneous” equations:

1. Migration = a1 + b1*Economic + c1*Personal + d1*Growth + E1*CONTROLS
2. Growth = a2 + b2*Economic + c2*Migration + D2*CONTROLS

We should see economic and personal freedom boosting migration, growth increasing migration, migration decreasing growth, and economic freedom boosting growth, all else equal. Here are the results, starting with the migration equation and then giving the growth equation:


Estimate Std. Error t value Pr(>|t|)
(Intercept) 13.8992494 6.2524457 2.22301 0.0313994 *
Economic 7.4339327 2.6748958 2.77915 0.0079859 **
Personal 15.0268798 5.8677766 2.56092 0.0139459 *
Log of HDD -1.2332131 0.6479793 -1.90317 0.0635739 .
Precip. -0.0651143 0.0436954 -1.49019 0.1433096
Growth -0.6242469 0.8945412 -0.69784 0.4889486
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3.82333 on 44 degrees of freedom
Number of observations: 50 Degrees of Freedom: 44
SSR: 643.185587 MSE: 14.617854 Root MSE: 3.82333
Multiple R-Squared: 0.376346 Adjusted R-Squared: 0.305476

Estimate Std. Error t value Pr(>|t|)
(Intercept) 4.5372398 2.6363641 1.72102 0.09227255 .
Economic 1.6686013 0.7350826 2.26995 0.02816251 *
Pop. Growth -0.0798077 0.0438334 -1.82070 0.07545675 .
Cap./Worker -0.4295270 0.3205457 -1.33999 0.18712899
Mining 0.1206070 0.0300684 4.01108 0.00023094 ***
Migration -0.0885952 0.0587031 -1.50921 0.13839404
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.808835 on 44 degrees of freedom
Number of observations: 50 Degrees of Freedom: 44
SSR: 28.78542 MSE: 0.654214 Root MSE: 0.808835
Multiple R-Squared: 0.4111 Adjusted R-Squared: 0.34418

It’s almost perfect. 🙂 Economic and personal freedom are still seen to boost migration directly. Migration reduces personal income growth, although the coefficient estimate is not quite statistically significant. Economic freedom increases real per capita personal income growth, and the effect is statistically significant this time. The only weirdness is that growth is not at all associated with migration (indeed, the coefficient is negative). I have a theory about why this might be, but it’s a bit irrelevant here.

What is relevant is that we now have pretty sophisticated evidence showing that economic freedom both increases economic growth in the states (implying that freer economies are more efficient) and attracts people, and that personal freedom also attracts people. Economic freedom does not appear to attract people by generating business growth, since income growth is not associated with more in-migration (surprisingly). It attracts people for its own sake (they want to pay less in taxes, most likely). Opponents of freedom as we define it are going to have to grapple with the consequences of the policies they advocate.

*Note: I have previously addressed the issue of freedom and migration in brief here.

21 thoughts on “Liberty as Amenity: Freedom, Migration, and Growth

  1. Great analysis!

    What’s your irrelevant theory?

    You propose that internal migration would lower per capita income because of the increase in labor, but it’s also possible that per capita income would increase if workers are moving from low productivity areas to high productivity areas. I know this is true in some countries internationally like China where workers are shifting from low productivity activities (rural agriculture) to high productivity activities (urban manufacturing). Other countries, Latin America and Africa, are seeing workers actually shift into low productivity sectors of the economy (service, black market/informal) as their economy focuses on profitable but not labor intense activities (natural resources) and their labor intense manufacturing sectors either decline from global competition or downsize with new, more efficient methods.

    So if we’re seeing per capita income decline due to migration, our internal migrations in the US are not involving themselves in any sort of shift to higher productivity areas. In a way we’re seeing a shift similar to Latin America and Africa, where our labor intense manufacturing sectors are shrinking in terms of employment (globalization, technology), natural resources (like mining) are up but they are not labor intense, and so everyone’s shifting into low productivity sectors (service, health care, and education). And at the same time moving to these more free states. But I wonder if there’s more of a connection.

    1. My irrelevant theory is that in part income compensates for disamenities. So if there were some omitted disamenity factor over this 7-year period that were driving residual states’ differences in income growth, that could account for why income growth is appearing not to attract people.

      It’s true that migration from low productivity to high productivity areas should increase the wages of those workers who move, but consider what happens to wages in the place where they are moving – they should go down. Wages should go up in the place that’s losing people. So take Shenzen, China – wages start out much higher there than in Yunnan or Xinjiang or somewhere. But as rural workers move from Yunnan and Xinjiang to Shenzen, wages will rise in the places they leave and fall in Shenzen. Now, that’s all else equal. Shenzen wages might continue to rise if capital (such as foreign direct investment) continues to flow into Shenzen. So I don’t think the US is necessarily experiencing some kind of inefficiency here. Wages in NY and NJ might rise, because fewer people want to live there so the companies that stay there have to offer workers more to keep them from moving. Meanwhile, wages fall in AZ and FL because lots of people want to be there and are willing to take jobs that pay less.

      1. I don’t know if I agree that labor markets in the United States are that localized, so that more people moving into, say, Atlanta is going to depress wages because there are more Atlanta workers competing for jobs. We have a pretty free and open “immigration” system between the states. Don’t we have just one big national labor market?

        If we assume a national labor market, we’re essentially seeing a migration of the low productivity workers and a leaving behind of the high productivity workers. If you’re ruling out cost of living driving low productivity workers away, and you’re in fact noting that it’s not specifically the economic growth attracting the internal migrants, what you have left is that low productivity workers are drawn to economic freedom. Which I just think is a curious tale to tell.

        I could certainly see an alternate United States, let’s call it Galt-land, in which higher productivity workers are the ones leaving and moving to economically free areas. We seem to have the opposite.

        I’m curious if you could test your own hypothesis that it’s lower taxes. Your economic freedom measurement, if I recall correctly, is 25% spending, 25% taxation, and 50% other. Take a look and see which of those three, if not all or only two of the three, are explaining the trends.

        I’d also be curious if you could do the same with personal freedom.

      2. We do have a very integrated labor market, but that’s precisely why we should expect these supply and demand effects. If Atlanta is an inherently desirable place to live, we should expect supply of labor to be higher there because people are drawn from elsewhere, and therefore wages should be lower. I think that’s the main reason why wages are low in places like Montana, western North Carolina, and Key West – people move there for the lifestyle and then have to make do with less money.

        I don’t think these aggregate results tell us much about which workers are moving. Certainly, a very strong negative correlation between migration and wages might imply that low productivity workers are the ones who are moving – but it really depends on what we assume about the elasticity of labor demand & complementarities between skilled and unskilled labor. Even if high productivity workers are moving in, in-migration could on average depress wages in the receiving jurisdiction and increase them in the sending jurisdiction. (Average wages will still fall if the high-productivity workers’ wages fall and everyone else’s wage rate stays the same.) Again, all of this is ceteris paribus – if capital moves in as well, then wages of everybody should rise ceteris paribus.

        I have run some stuff breaking out fiscal & regulatory indices, and I find that it’s mostly fiscal driving the migration. But I haven’t done the most sophisticated analyses of that part yet (or looked at growth). And I’m willing to bet that some personal freedoms (marriage, education, guns, marijuana) drive migration much more than others (asset forfeiture, campaign finance, tobacco, gambling).

  2. Sure would be interesting to see what factors you considered for “personal freedom”. The bottom ten states, yes BOTTOM ten, constitute pretty much every state where same-sex couples can get married or have a legally recognized relationship.

    I’m guessing you gave very little consideration to that freedom and quite a bit to subsidized private schools, i.e. vouchers.

    1. Check out the study at mercatus.org for all the info you could desire. 🙂 Same-sex partnerships play a huge role in our calculation of personal freedom, although less than educational regulations in toto (more than vouchers alone). Our data are as of January 1, 2009, so some of those states had not passed reforms as of that date. Note also that left-wing states tend to have many other restrictions on personal freedoms, such as restrictive laws against home schooling, gun ownership/carrying, allowing smoking on your property, not wearing a seat belt or helmet, using trans fat in the food at your restaurant, giving money to candidates, parties, or PACs, & so on. They’re also not necessarily better than conservative states on prohibiting police from taking innocent people’s property, allowing gambling, or not arresting people for victimless crimes. Leftist states are superior to rightist states, on average, on just three personal freedoms: same-sex relationships, marijuana, and alcohol.

  3. It would be interesting to see some additional controls included in your model, such as Medicare beneficiaries (to control for retirement-based decisions where economic freedom matters little) or something to account for the spatial determinants of migration.

    This study might be of interest to you (“Patterns of Interstate Migration in the United States from the Survey of Income and Program Participation”): http://research.stlouisfed.org/publications/review/article/8763

    1. Thanks for the link! I’ll certainly try including population over 65 as this moves toward publication. At the moment, I have no strong expectations about the variable’s effect. If you mean by spatial determinants of migration taking into account spatial correlations in the errors, I have tried running spatial lag models with similar results, although I have not yet figured out how to integrate three-stage least squares with spatial lag models.

      1. Thanks! Yeah, Glaeser’s stuff is interesting. The approach is sound economics – the lower is housing elasticity, the more amenities will be priced into real estate rather than reflected in migration – but I am not sure that the housing constraint is as important as he thinks it is, even in places like California. If it were, we should see clear heteroskedasticity in the migration equations (not to get too technical), which I’m not seeing. But I’ll have to respond to the argument more fully moving forward.

  4. Jason,

    I’m no economist (that’s why I read your blog), but at one time I had hopes to become one. Politics unfortunately sucked me away. But in my brief attempt to look at the literature, at least the literature that’s not beyond a paywall, it seems that I can’t find anything firm one way or another. But this study argues that internal migrations during the Great Depression didn’t reduce wages: http://www.econ.ucla.edu/lboustan/research_pdfs/research04_internalmigration.pdf

    Look forward to further research.

Leave a Reply

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

WordPress.com Logo

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

Google photo

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

Twitter picture

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

Facebook photo

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

Connecting to %s