1 Introduction

A large literature has examined the relationship between economic freedom and economic outcomes (Hall and Lawson 2014). Most of this literature examines aggregate economic outcomes. Hall and Lawson (2014) list “gross domestic product (GDP) growth, income levels, poverty, productivity, investment spending” as examples. They also note that “(m)any papers, especially those in finance and other business disciplines, look at firm-level performance measures by country”. Such measures are also generally aggregated. Few papers in this literature examine the influence of economic freedom on the performance of individual firms. Among those that do are (Gropper et al. 2015; Azizi et al. 2016; Bykova and Coates 2020b; Chen et al. 2015; Chen and Lin 2006; Roychoudhury and Lawson 2008). Each of these papers focuses on financial outcomes like return on assets, profitability, equity or stock market returns. While the aggregate literature examines, for example, growth in real output of goods and services, the literature looking at firms has largely addressed financial outcomes.

This paper contributes to the literature by examining the connection between economic freedom and production of real output. The real output in question is wins by the professional football (soccer) team in the jurisdiction covered by a given value of regional economic freedom. Hembre (2017) and Kleven et al. (2013) take the issue a step farther than simply asking whether personal taxes influence location decisions, looking instead to whether those tax differences affect the performance of the firms for which the mobile individuals work. The firms in both cases happen to be professional sports clubs. Both studies also find evidence that clubs from low personal income tax jurisdictions have an advantage over clubs from high tax jurisdictions; the tax system provides low tax jurisdictions a competitive advantage. Neither Hembre (2017) nor Kleven et al. (2013) address the role of differences in economic freedom across jurisdictions. The mechanism by which tax differences in Hembre (2017) and Kleven et al. (2013) affect teams is through movement of players. Clubs with more attractive tax environments are better able to attract the best players and, therefore, perform better in the competitions. However, researchers studying the relationship of economic freedom and economic growth do not generally highlight mobility as the mechanism connecting improving economic freedom with faster economic growth. Hall and Lawson (2014) mention migration and immigration only once each in their review of the economic freedom literature; both instances are in titles of journal articles listed in the appendix. In probably the first paper to utilize the economic freedom index to explain economic growth, Gwartney et al. (1999) refer to Adam Smith’s “invisible hand” and Douglas North’s institutions (North 1990), in each case reflecting decentralized, disaggregated behavior. In other words, economic freedom is a measure of the stringency of the constraints imposed by government on the choices of individual buyers and sellers. This focus on the restrictions faced by individuals in the conception of economic freedom only rarely appears in the literature testing the influence of economic freedom on the economy.

The analysis here follows a similar approach to much of the research on the influence of freedom by first hypothesizing that personal and economic freedom influence or constrain some firm or individual behavior and then using a regression equation to test that hypothesis. Here, we hypothesize that these freedoms affect the outcomes of professional sporting contests. The mechanism by which this occurs is simple but testing it is not. Players negotiate with teams for their services. As Rottenberg (1956) argued, clubs from more attractive cities will be at an advantage in negotiating with players. Clearly, a feature that would make one location more attractive than another is a greater extent of personal and economic freedom in the former than in the latter. This precise question has not been addressed in the literature, though numerous papers have examined the influence of differences in tax regimes between locations (Alm et al. 2012; Driessen and Sheffrin 2017; Hembre 2017; Kleven et al. 2013; Kopkin 2012; Veliotis 2013).

To our knowledge, the question of state and local taxes affecting team success only rarely has been addressed in the US context. The analysis focuses on the influence of state and local tax differences on player mobility and the competition among clubs free agent players, those with the greatest ability to switch teams. Veliotis (2013) examines the influence of state income taxes on clubs’ abilities to produce equivalent after-tax monetary offers for players. For example, several states do not have a state income tax; Texas is one example, which also has numerous professional sport franchises. When considering salary offers between a team from Texas and a team from a state that does impose an income tax, the offer from the Texas team can be smaller than that from the other team while still producing an equal after-tax salary for the player. This differential gives the Texas team a clear advantage in attracting players over the other teams, especially in a league with a salary cap, enabling Texas teams to field a better team than their competitors at the same cost.

The remainder of this paper is structured as follows. Section 2 gives an overview of the literature and introduces the testable hypotheses. Section 3 presents the research design along with the context of the study and data that we used for the analysis. In Sect. 4, we report the results of the estimations. Section 5 provides the robustness check. Finally, Sect. 6 concludes and discusses the results with some suggestions for the future work.

An active literature in recent years has addressed the migration of labor in response to personal tax differentials between states or countries (Akcigit et al. 2016; Kleven et al. 2020; Kleven et al. 2013; Moretti and Wilson 2014, 2017). Hembre (2017) and Kleven et al. (2013) take the issue a step farther than simply asking whether personal taxes influence location decisions, looking instead to whether those tax differences affect the performance of the firms for which the mobile individuals work. The firms in both cases happen to be professional sports clubs. Both studies also find evidence that clubs from low personal income tax jurisdictions have an advantage over clubs from high tax jurisdictions; the tax system provides low tax jurisdictions a competitive advantage.

2 Literature background

There are at least two likely mechanisms by which greater economic freedom affects production by sports clubs. The first, and perhaps most obvious, is through enhancing the club’s ability to attract better players at lower cost than can clubs from less economically free jurisdictions. The extent to which economic freedom influences migration is the subject of a small literature (Arif et al. 2020; Cebula 2014) The second mechanism is through allowing greater flexibility for clubs to respond to the market circumstances, particularly with respect to the labor market for players. Of course, these two mechanisms are closely intertwined, so much so that separating out their influences may be impossible. We try to separate these influences by incorporating indicators of personal freedom.

Migration decisions, players opting for one team over another from a different state, are likely affected by the extent of personal freedom in each state. On the other hand, personal freedoms have little to do with business decisions of sports clubs.

An active literature in recent years has addressed the migration of labor in response to personal tax differentials between states or countries (Akcigit et al. 2016; Kleven et al. 2020; Kleven et al., 2013; Moretti & Wilson, 2014, 2017). Of course, personal income taxes are not the only taxes that individuals may face nor are taxes of whatever type the only location specific influence on the location decision. Tiebout (1956) emphasized that jurisdictions offer a package of taxes and public services and that mobile voters choose among the jurisdictions that which offers the greatest utility. In the literature on locational choices, public services are implicitly or explicitly things like public schools, police and fire protection, and roads and public transportation. Even this definition of public services may be too narrow. People may choose locations based on climate and geography as well, and Hembre (2017) recognizes this by using Albouy's (2016) measure of locational amenities. But people may also select locations based in part on the nature of government policies that have relatively little financial footprint (Dustmann and Preston 2001; Knapp and White 1992; Nair-Reichert 2014; Sinha et al. 2018). For example, some people will find (un)attractive a state that has generous social welfare programs, imposes strict regulations on workplace health and safety or worker licensure, bans same-sex marriages, requires a waiting period before purchasing a firearm, or requires private schools to teach a state-approved curriculum. Ruger and Sorens (2018) and Sorens et al. (2008) produce an index of personal freedom which measures the extent to which states vary in policies related to concerns like those just listed. Both these personal freedoms and economic freedoms may affect location decisions of individuals, including professional athletes.

Kleven et al. (2013) is unique among these papers for three reasons. The first is that unlike the others listed previously, it focuses on international mobility. Secondly, it is about soccer players’ choices of clubs that the other papers above do not consider. Third, and most importantly for our purposes, one part of the discussion shows the average success of clubs in a country declines as the tax rates in the country rise. Moreover, the pattern appears in the time since the Bosman Ruling opened up player mobility across countries in the European Union. Supporting evidence shows that post-Bosman, the proportion of players in a country’s top league who are foreign is smaller in countries with higher tax rates while the proportion of players of a given nationality playing in their home country is smaller the higher are the home country tax rates. The clear implication is that when players are able to change teams, they will take into account tax considerations. Moreover, the evidence suggests that clubs suffer in international competitions because of high tax rates in their country.

Alm et al. (2012) estimate the size of the salary differential for free agent players who have just signed a new contract. Along with standard variables in baseball salary equations, the authors introduce the top state and local tax rate on personal income. Their empirical results indicate that a free agent will get a salary about $24,000 per year higher for each percentage point of the state and local top personal income tax rate. Alm et al. (2012) note that this differential is beneficial to clubs from low tax jurisdictions because players with lower salaries are easier to trade than players with larger salaries and trading away players with low salaries may also bring better terms for the team. The benefit of the lower salaries may have an additional financial reward as clubs from low tax rate jurisdictions can acquire equal talent as those from high tax jurisdictions with less chance of becoming subject to the luxury tax. Altogether, the Alm et al. (2012) discussion suggests that clubs from low tax jurisdictions may have a competitive advantage over clubs from high tax jurisdiction, all other things held constant.

The same result is derived formally by Kopkin (2012) who models the decision of free agent players of which club with which to sign a contract. Clubs from low tax states can offer a lower salary to a player of given quality than can the club from the high tax jurisdiction. Consequently, high tax jurisdiction clubs will, for a given budget, be forced to make offers to lower average quality players than will clubs from low tax jurisdictions. Of course, hiring lower average quality players consigns clubs from high tax jurisdictions to being less successful on the court, pitch or field. Using data on NBA free agent signings, Kopkin (2012) finds that as the income tax rate rises relative to the average income tax rate across the other NBA host cities, the average skill level of the free agent signees declines. The last issue that Kopkin (2012) addresses is the success of teams related to the tax rates. The finding is that a one-percentage point decrease in the relative income tax rate translates into an additional 4 wins in the regular season and nearly 2.5 additional playoff wins, though the former result is statistically significant only at the 10% level and the latter not even at the level.

Hembre (2017) is the most thorough study of the role of taxes and team success of which we are aware. He relates top state marginal income tax rates to winning percentage, finding that a one-percentage point increase in the tax rate reduces the winning percentage by 6-tenths of a percentage point. The upshot is that a one standard deviation increase in the tax rate costs an NBA or NHL club, each of which has an 82 game season, about 1.7 wins. An important aspect of the analysis is a comparison of the relationship between taxes and team success before and after the advent of free agency.

An additional novelty of the Hembre (2017) analysis is an interest in the role of amenities on the performance of teams. Citing the potential inverse relationship between amenities and labor costs and hypothesizing that higher amenity values might give teams from desirable locations bargaining advantages [See also, Rottenberg (1956)], the expectation is that greater amenities will lead to greater success of the club. Hembre (2017) uses the index of amenity values produced by Albouy (2016), which are predominantly influenced by size and education of the population, weather and being on the coast. Interestingly, this amenity value is largely irrelevant for team performance as it is only statistically significant in one equation, that of Major League Baseball, and then only at the 10% level.

One might think of amenities more broadly than is implicit in Hembre (2017). Amenities may be local public goods or institutions, for example. In the sports context, Kufenko and Geloso (2021) use the Economic Freedom of the World Index (EFW) as a measure of institutional quality, though they do not refer to it as an amenity, to study the relationship between economic freedom and country medal counts at the Olympic Games. Their focus is on capturing the incentive effects inherent in the protection of property rights, which is a component of the EFW. Also addressed in the EFW are measures of the size of government, the regulation of labor and other markets, and the severity of taxation. In the EFW, freedom from taxation incorporates all types of taxation, personal, business, property, sales and more. So, while Kufenko and Geloso (2021) do not emphasize the role of taxation in their analysis, implicitly they have addressed a similar issue as in Hembre (2017) and Kleven et al. (2013). Kufenko and Geloso (2021) hypothesized that income inequality would affect Olympic medal counts differently where economic freedom was low than where it was high. Kufenko and Geloso (2021) found that countries’ income inequality negatively affected medal counts in countries with low economic freedom but that inequality had no impact on winning medals in countries with high economic freedom.

Of course, the economic institutions may be no more sufficient to reflect desirable local public goods than are climate and geography. Legal, social and personal freedom may all be important aspects of local “amenities” that make a community more or less attractive. The Cato Institute produced an index of freedom in the states of the United States, which includes an index of personal freedom. A hypothesis of this research is that personal freedom, like economic freedom generally and low personal income taxes specifically, influence free agent athletes’ decisions on the labor market. To the extent that personal and economic freedom capture different aspects of local amenities, more of either will give a club in that jurisdiction an advantage in the labor market for players over clubs from cities with less of either, and therefore a competitive advantage in the sporting contests. Consequently, teams from jurisdictions with high economic and personal freedom will win more often than teams from communities with lower economic and personal freedom, all else held constant.

3 Research design

3.1 Research model

The equation to be estimated has the same structure in each case. Let yit be the outcome of interest for club i in time-period t, efit−1 be the economic freedom score from the previous year in the jurisdiction where the club i plays, and pfit−1 be the personal freedom score from the previous year in the jurisdiction. If player mobility is the source of the effects of freedoms on team performance, then previous levels of freedom are most likely those on which players make their decision to stay or to move. The talent on the team is measured by the team’s payroll relative to the average payroll in the league during that season, rwit and other explanatory covariates are represented by xitk. The basic model is as follows:

$$y_{it} = \alpha_{i} + \beta_{1} ef_{it - 1} + \beta_{2} pf_{it - 1} + \beta_{3} rw_{it} + \mathop \sum \limits_{k = 1}^{K} \gamma_{k} x_{itk} + \in_{it}$$
(1)

where αi are club fixed effects, βi are parameters, and ϵit is a random error term with mean zero and variance σi2 allowing for team specific variance. The basic model in Eq. (1) is expanded to include squares of efit−1 and pfit−1 and interaction terms to allow for the possibilities that economic and personal freedom have nonlinear effects and complement or substitute for one another or influence the effect of the relative salary. Moreover, the construction of the index, described in more detail below, presumes that the jurisdiction with, for example, the least government spending is the most “free”. In the extreme, a jurisdiction with no government spending would be freest while providing no courts for the protection of property rights, indeed, no public goods or public services of any kind. The quadratic specification allows for there to be a point beyond which more freedom, devolving into anarchy in the extreme, say, is harmful to the economy and to performance of the soccer clubs in our analysis.

The most expansive specifications include interaction terms between the personal and economic freedom variables and each of those variables with the relative wage. It is plausible that the degree of personal freedom in a jurisdiction affects the influence of economic freedom on club’s performance and vice versa. The interaction of the economic and personal freedom variables and the team’s relative payroll are intended to proxy for the link discussed in the literature (Hembre 2017; Kleven et al. 2020; Kleven et al. 2013; Veliotis 2013) between taxes and a club’s ability to attract players for lower pay which is particularly advantageous under a salary cap.

The estimating equation is perhaps too simple. One could reasonably argue that explanations of team performance on the pitch will also depend upon other factors beyond those described above; for example, relative wage does not sufficiently capture the quality of the squad. The x variables in the basic model include additional variables that have been found to affect team performance (Bykova and Coates 2020a; Coates et al. 2016) the number of players with international experience for their national team as another indicator of roster quality, and the coefficient of variation, or the Gini coefficient, of the salaries of the players, and the squares of those latter variables.

The inclusion of these measures of the distribution of player salaries reflects the argument that wide salary dispersion may motivate lower paid players to exert more effort (Lazear and Rosen 1981; Rosen 1986) or that dispersion reflects the cohesion within the team, with wide dispersion harming cohesion and, therefore, reducing team performance (Akerlof and Yellen 1990; Levine 1991). Evidence on this relationship is mixed: Breunig et al. (2014), Debrock et al. (2004) and Simmons and Berri (2011), found that increasing expected wage inequality increased team performance; Avrutin and Sommers (2007), Berri and Jewell (2004) and Katayama and Nuch (2011), concluded that increased wage disparity does not result in decreased team performance for the NBA; Breunig et al. (2014) found, for MLB, that greater salary inequality may lead to lower productivity. Franck and Nuesch (2011) argued that the relationship is quadratic, allowing for the possibility, say, that increasing disparity improves team performance when disparity is low but harms team success when disparity is high. Using data from Major League Soccer, Coates et al. (2016) found support for this relationship, though in practice all clubs were on the down-ward sloping side of the quadratic function. In other words, for all clubs the effect of further increases in dispersion in pay reduced the team performance. Bykova and Coates (2020a, b) found that coaches with experience as players worsened the negative effect of dispersion slightly while experience coaching had no effect.

Using the aggregated indices of economic and personal freedom may mask the true relationship between freedoms and firm performance because the personal and economic freedom indices are the result of combining measures of freedom across various categories of economic and personal life. A possibility is that the results for the full indices are driven by only one or a small number of the components or that the individual components contribute differently to the outcome. To assess this possibility, we replace the overall measures of economic and personal freedom with their constituent parts. If the results are linked to components of the index in different ways, then team executives seeking to improve economic outcomes can focus pitches to free agents on those personal and economic freedom issues that affect the team most. More generally, policy makers can target improving freedom metrics to get the biggest response for their efforts.

Two different methods are taken to address the potential differential impact of the components of the freedom measures. In the first, each component is included as a regressor in a linear equation, like Eq. (1) above but with the economic freedom and personal freedom variables vectors of the components. This approach lets each component have its own linear effect. The test of the joint hypothesis that each component of economic (personal) freedom has the same coefficient indicates whether simply summing the components, as the indices do, is valid. The second method builds on the quadratic specification by expanding it and developing implications from the expansion.

The economic freedom index has three components, all of which have equal weight in forming the overall index. The personal freedom index is comprised of 12 separate components, each with its own weight in forming the personal freedom index. Testing for equality of influence of the individual components in the two different indices requires different approaches. Using economic freedom as the explanatory variable implicitly imposes the condition that each of its components has the same influence on the outcome. Treating this as restrictions on the parameter values, we develop and test these restrictions on the parameters. In the case of the personal freedom variable, each component must have its own individual coefficient that is the product of some unknown constant and that variable’s unknown weight in the formation of the index.

Let \(ef_{it} = \left( \frac{1}{3} \right)\left( {c_{1it} + c_{2it} + c_{3it} } \right),\) which is precisely how the three components of the economic freedom index are aggregated into the value used in the regressions above. Dropping the relative wage, international player, and coefficient of variation variables from the models above to simplify exposition, the regression equation is:

$$y_{it} = a_{i} + b_{1} ef_{it} + b_{2} ef_{it}^{2} + d_{1} pf_{it} + d_{2} pf_{it}^{2} + \varepsilon_{it}$$

If it is true that each of the components of the economic freedom variable have the same influence on the outcome, then substitution and collection of terms produces

$$y_{it} = a_{i} + \beta_{1} c_{1it} + \beta_{2} c_{2it} + \beta_{3} c_{3it} + \gamma_{1} efsumsq_{it} + \gamma_{2} efinter_{it} + d_{1} pf_{it} + d_{2} pf_{it}^{2} + \varepsilon_{it}$$
(2)

where \(efsumsq_{it} = \left( {c_{1it}^{2} + c_{2it}^{2} + c_{3it}^{2} } \right)\), \(efinter_{it} = \mathop \sum \limits_{j = 1} \mathop \sum \limits_{k < j} c_{jit} c_{kit}\), \(\frac{{b_{1} }}{3} = \beta_{1} = \beta_{2} = \beta_{3}\), \(\gamma_{1} = \frac{{b_{2} }}{9}\), and \(\gamma_{2} = \frac{{2b_{2} }}{9}\). To test the restriction that the components of economic freedom have the same influence, we test the joint hypothesis that \(\beta_{1} = \beta_{2} = \beta_{3}\) and \(\gamma_{2} = 2\gamma_{1}\). We follow this same procedure for personal freedom and its components.

3.2 Context of the study: MLS background

Major League Soccer is a single-entity organization unlike other professional sports leagues in the United States, which are a group of independent organizations, the clubs or franchises, cooperating to put on sporting events. The different structure of MLS is important because it grants the league substantial and byzantine control over player allocation to teams. In the other leagues, rules like the reverse-order draft, restrictions on free agency based on accumulated service time, rules on player waivers, and salary caps all influence the ability of players to move between clubs. The MLS has similar rules, but MLS also has exceptions to its salary cap, programs for acquiring young players outside the draft system and is involved in more signings of international players, as well as assigning rights to players it has under contract to non-American clubs via the transfer market. All of these rules are, according to analysis by Steven A. Bank (Bank 2018), a means of keeping all the investor-operators striving toward the same goal, profits of Major League Soccer LLC in which all investors share, while also remaining able to respond to new circumstances. The key point about Major League Soccer’s rules regarding player acquisition is that they limit the ability of players to choose the team for which they play. If players cannot select among teams vying for their services, then state and local taxes, public policies, and economic and personal freedoms should have no influence on the roster composition of clubs nor, by extension, on the success of the club on the pitch. MLS players are not so constrained as to have no say in the matter of where they play, but they are arguably more constrained than players in any other of the North American leagues.

For our purposes, the relative immobility of players in MLS is an advantage for two reasons. First, these circumstances are, arguably, the least likely in which to find an influence of freedoms precisely because players cannot so easily switch teams to acquire greater economic and personal freedom. Consequently, finding effects of freedom on team performance in MLS suggests that there will be influences on performance in sports where players are more mobile. Second, limiting player mobility weakens the case for player migration as the mechanism by which freedoms effect team performance and, as a result, strengthens the case that it is better institutions that drive better team performance.

In the next section, we present the data for our analysis.

3.3 Data

Data for this project covers clubs of Major League Soccer (MLS) over the period 2006 through 2016. After accounting for availability of all the variables of interest, the data includes 156 observations from 18 separate clubs. Our data is an unbalanced panel because we include expansion franchises and one franchise, Chivas USA that ceased to exist near the end of the period.

There are four main sources of the data. The first of these is the Economic Freedom of North America data available from the Fraser Institute at https://www.fraserinstitute.org/economic-freedom/dataset. The second source of data is the Cato Institution’s Freedom in the 50 States available at https://www.freedominthe50states.org. This index builds upon and extends the work of Sorens et al. (2008). We calculated the payroll and relative wage using the data on players’ salaries from the MLS website, where we also collected the data on team outcomes and characteristics. We omit from the analysis the teams from Canada. Table 1 provides descriptive statistics for the variables used in the analysis.

Table 1 Descriptive statistics
Full size table

The Economic Freedom of North America (Stansel et al. 2019; Stansel and McMahon 2000) computes for the states of the United States and Mexico and the provinces of Canada indices of the degree of economic freedom in those jurisdictions, though here we use only the data on the 50 United States. The construction of the index begins by identifying broad areas of the economy over which government exercises some level of influence. That influence could be through imposition of regulations, like occupational safety and health or occupational licensure requirements. The influence could be government spending decisions or the setting of tax rates and minimum wage laws. The EFNA defines three areas, Spending, Taxation and Labor Market Regulation, which we utilize individually as well as an Overall index. Maximal freedom is implicitly defined as the least government involvement, and each state’s value on some variable is compared to the value of the state with the greatest government involvement and to the range of values of the variable. In equation form, the index on government activity G for state i is

$$I_{i} = 10*\left( {\frac{{G_{max} - G_{i} }}{{G_{max} - G_{min} }}} \right)$$
(3)

where Gmax is the largest value of the variable, Gmin is the smallest value, and Gi is the value of state i. If the state has the maximum value of G, the index has value zero for that state, indicating the least possible freedom; if the state’s value of G is the minimum, then the index value is 10 for that state, the maximum economic freedom.

A number of stadiums and arenas utilized by professional sports franchises, including soccer-specific stadiums for MLS franchises, have been supported with substantial public support. According to an analysis done for the Washington, D.C. city government (Cost–Benefit Analysis of the Soccer Stadium Development Act of 2014, 2014), the average public share of the total project cost of stadiums built or planned for MLS clubs between 1999 and 2016 was 44.6%. The share of the cost is misleading though because three of the 16 stadiums in the analysis had no public spending contribution while the next smallest public share is 32.2%. Ignoring those three cases, the average public share is 55.75%. Detailed discussion of lease agreements is beyond the scope of this analysis. The point here is that public support for stadium construction contributes to a low or lowered economic freedom score in those jurisdictions that provide it. The subsidy is government spending, an increase in which lowers the spending component of the economic freedom score; taxes must be raised, all other things constant, to finance the bonds used to raise financing for the facility which also lowers the taxation portion of the index. All this means that MLS clubs directly financially benefit from policies that reflect lower economic freedom. The question is whether that translates into better performance on the pitch.

Existing research, predominantly focused on Major League Baseball, has examined the issue of whether opening a new stadium resulted in better team performance. (Clapp and Hakes 2005; Hakes and Clapp 2006; Quinn et al. 2003; Rockerbie and Easton 2019) That research does not address the influence of the economic and social environment on team performance but rather on whether capital, in the form of a stadium, and labor, in the form of playing talent, are substitutes or complements in the production of profit for the team owner. Our approach generalizes these studies by incorporating the influence of the broad economic and social environment which includes the willingness to provide stadium subsidies.

We introduce the index of personal freedom to address the social environment. The index of personal freedom is constructed by Ruger and Sorens (2018) and Sorens et al. (2008) which should be consulted for the details of its construction. In general terms, the index is the weighted average of assessments of 12 policy areas, Campaign Finance, Marriage, Incarceration and arrests, Asset forfeiture, Tobacco, Education, “Mala Prohibita”, Gaming, Travel, Cannabis, Alcohol, and Gun rights. Weights vary from component to component and across individual subcomponents. Mostly, the assessments reflect the presence, absence or strictness of government mandates regarding some activity. For example, Mala Prohibita is a grab bag of issues with little relation to one another: a state-wide ban on transfats in restaurants, whether mixed martial arts are legal, maintenance of a DNA database of arrestees without a probable cause hearing, a ban on affirmative action and a ban on discrimination on the basis of sex, for example. Travel incorporates provisions on licensure of drivers and vehicles; education is about the extent to which state and local governments impose licensing and curricular constraints on private schools. In general, one can argue about whether such issues are what jumps to mind when thinking about the extent of personal freedom one has. That debate is well-outside the scope of this research. Nonetheless, we must acknowledge that one large personal freedom issue is unaddressed; personal freedom does not incorporate abortion and reproductive health freedoms. Oddly, abortion freedom is split into pro-life, moderate pro-choice and extreme pro-choice, so one can add whichever flavor of freedom suits one’s preferences to construct a more complete personal freedom index. Rather than choose among three different measures of abortion freedom, we omit abortion freedoms from our analysis.

The total player salary bill of a club relative to the average player’s salary bill across all clubs captures relative quality of the playing talent on the team. (Frick and Lee 2011; Szymanski 2000; Szymanski and Smith 1997) A club with more talent, holding other things constant, pay out more in player salary compare with a club with less talent. Another variable used to capture player quality is the number of players on the team who have played for their country’s national team. Finally, the model includes a measure of the concentration, or distribution, of salary on the team captured by either the coefficient of variation or the Gini coefficient. The role of salary distribution in company or team performance has been widely studied since the theoretical advances of Lazear and Rosen (1981), who argued greater salary inequality would increase worker productivity, and Akerlof and Yellen (1990) and Levine (1991) who contended that cohesion among the workers would improve productivity. Numerous studies have addressed this issue empirically using sports data from various professional leagues (Debrock et al. 2004; Depken 2000; Franck and Nuesch 2011; Sommers 1998). Specifically focused on Major League Soccer, Bykova and Coates (2020a, b) and Coates et al. (2016) report evidence that team performance is influenced by salary distribution, with the impact different at very low and very high dispersion of salaries on the team.

The outcome or dependent variable in our analysis varies as a type of robustness check. For example, economic and personal freedoms may influence playing success while not affecting attendance or vice versa. It is also possible that the influence will differ between measures of playing success, so rather than limit the analysis to points per game, which is estimated using panel methods, we also use total points and estimate the model using count regression methods. The descriptive statistics are presented in Table 1.

4 Empirical results

Table 2 provides two estimates of Eq. (1), one using the coefficient of variation and one using the Gini coefficient to control for the salary distribution. To be clear, these models include no quadratic terms for any variable and no interactions between any variables. Each model does include year dummies and club fixed effects and is estimated with errors clustered by team. The Adjusted R2 is substantially smaller than the R2 for each model indicating there are right hand side variables that add nothing to the explanatory power of the model that, admittedly, is quite low. One can reject the null hypothesis that the year effects are all zero for the CV model but cannot reject the null for the Gini specification. Looking at the individual coefficients, as the relative wage paid by a club rises, the more points per game the club earns. Team performance falls as the coefficient of variation of the salary distribution rises; that is, more dispersed pay produces worse results than a more equal pay structure. Salary dispersion measured by the Gini coefficient has no effect on team performance.

Table 2 MLS points per game-panel team and year fixed effects regressions
Full size table

The variables of focus for this study, economic and personal freedom, produce unexpected results. Economic freedom has no statistically significant impact on team performance while personal freedom has a negative and statistically significant effect.

Table 3 presents the regression model without personal freedom, in column 1, with personal freedom in column 2, and with interactions between personal and economic freedom and relative wage in column 3. The same models are estimated replacing the coefficient of variation with the Gini coefficient and are reported in Table 4. Because each specification in nonlinear, the best way to understand the relationship between the explanatory variables and points per game is looking at marginal effects, the derivative of the function, at different points on the function. To do this, marginal effects of the relative wage, economic freedom, personal freedom and coefficient of variation, or Gini coefficient of team salary are presented below in graphical form.

Table 3 MLS Points per game-CV
Full size table
Table 4 MLS Points per game-Gini
Full size table

Because the models involve quadratic terms and interactions among the freedom variables and the relative wage variable, joint tests of significance of all the coefficients involving each of the variables are reported in Table 5. For example, in the model in which the coefficient of variation is the measure of salary dispersion, and which does not include the personal freedom variable, one can easily reject the null hypothesis that the coefficients on the relative wage and relative wage squared are both zero. The F-statistic is 5.96 and the p = value is 0.01. On the other hand, one cannot reject the null hypothesis that the economic freedom variable and its square both have zero coefficients. Interestingly, the coefficient of variation and its square are always jointly significant while the Gini coefficient and its square never are jointly significant. Also of note, the year dummy variables are jointly significant in two of the three CV models but in none of the Gini models.

Table 5 F-tests of nonlinearities and interactions
Full size table

Between Tables 3 and 4, the number of players with experience on their country’s national team that are on the roster of an MLS team has a positive and statistically significant coefficient in three of the six regressions. The point estimates of the six coefficients range from 0.0101 to 0.0154. The statistically significant coefficients are the three largest of the six, but clearly, the range of the effect is quite small. Moreover, in practical terms, an additional national team player who raises points per game by 0.015, makes an improvement of about one percentage point relative to mean points per game of 1.39.

When thinking about the remaining marginal effects, it is important to keep in mind that average points per game of 1.39. Figure 1 presents the marginal effects of economic freedom calculated with the coefficients from the model using the coefficient of variation with and without inclusion of personal freedom. The figure shows that the effect of increasing a team’s payroll relative to the average payroll in the league has a positive but declining effect on additional points per game. Raising team payroll increases points per game until payroll reaches a bit less than 2.5 times the league average. At the smallest value of the relative wage variable in the figure, 0.5, the marginal effect is slightly more than half a point per game, or slightly more than one-third of the mean points per game. Even a team whose payroll is about twice the league average would improve points per game by about two-tenths or roughly 14%. The increase in relative pay for this team at twice the league average implies an increase in season points of more than six, easily enough to put a team into the playoffs or even into the top seed in some cases.

Fig. 1
figure 1

Marginal Effects of One-year lagged Economic Freedom with 95% confidence intervals

Full size image

Figure 2 shows how the marginal effect of the salary distribution varies across levels of the coefficient of variation. At values of the CV between about 1 and 2, where most clubs fall, slight changes in the CV reduce points per game by between about 0.35 and 0.2. A club that rasied its CV from about 1 to about 2 would see its points per game improve by about 0.15. In the current 34 game season, the extra 0.15 points per game implies nearly five additional points on the season, easily enough to boost a team into the playoffs.

Fig. 2
figure 2

Marginal Effects of One-year lagged Personal Freedom with 95% confidence intervals

Full size image

Figure 3 shows the marginal effects at different values of the economic freedom variable. The point estimate of the marginal effect is negative at low values of economic freedom but positive at high values. The effect is never statistically different from zero.

Fig. 3
figure 3

Marginal Effects of Relative Wage with 95% confidence intervals

Full size image

Figure 4 shows the marginal effects at different values of the personal freedom variable. The marginal effect is always negative and is statistically significantly different from zero at all but the three highest discreet values of the variable for which the effect was computed. In fact, about two-thirds of the observations have personal freedom scores that indicate negative values of the marginal effect of personal freedom. Very low personal freedom in a jurisdiction translates into less success on the football pitch. Interestingly, point estimates at the statistically significant values of personal freedom indicate a reduction of between 2 and 3 points per game. These are perhaps implausibly large values given that the maximum points per game is three and the mean is 1.39.

Fig. 4
figure 4

Marginal effects of Coefficient of variation with 95% Confidence Intervals

Full size image

The results above suggest the personal but not economic freedom influences the success of MLS teams. These results also indicate that the team’s relative payroll may determine performance on the pitch, but this role may depend upon how the salary distribution of the club is measured. Finally, whether salary distribution is measured by the coefficient of variation or the Gini coefficient matters for the answer to the impact of income inequality among teammates on MLS club performance.

5 Robustness checks

This section addresses the possibility that subcomponents of the economic and personal freedom indices are driving the results. As a baseline for the influence of the individual components, we estimate a completely linear model, analogous to Eq. (1) but with each of the components of economic and personal freedom as regressors. In other words, the baseline model is Eq. (1), but β1 and β2 are 1 × 3 and 1 × 12 vectors of parameters and eft−1 and pft−1 are 3 × 1 and 12 × 1 vectors of variables, respectively. We then estimate models that reflect the inclusion of the quadratic terms in economic and personal freedom as described above and test the coefficient restrictions implied by the aggregated indices. The section also examines how results are affected by alternative specifications, including the use of contemporaneous economic and personal freedom variables rather than the lagged values, and the influence of including or excluding the year effects.

Table 6 present results of estimating Eq. 2 and the models imposing the coefficient restrictions described above.

Table 6 MLS Points per game
Full size table

The first two columns use the coefficient of variation to measure salary concentration, the third and fourth columns use the Gini coefficient. As is clear from the tables, apart from the intercept, the only individually statistically significant variable in the Linear Baseline model, for both CV and Gini, is marriage freedom. Perhaps oddly, greater marriage freedom is associated with lower team performance. The elasticity of points per game with respect to increased marriage freedom is, at the means of marriage freedom and points per game, about 0.07; a one percent increase in marriage freedom raises points per game by 7-hundreths of a percent. In addition, educational freedom is positive and statistically significant, at the 10% level, in one equation. No other components of personal freedom are individually statistically significant and, to be clear, none of the components of economic freedom is individually statistically significant.

Considering joint hypothesis tests of various parameter combinations, the null hypothesis that personal freedom variables all have zero coefficients, in the Linear Baseline models, can be rejected, F = 6.02, p = 0.000 for CV and F = 29.43, p = 0.000 for Gini. The nulls that the relative wage and its square, that the coefficient of variation and its square or Gini and its square, that the economic freedom variables and the year dummy variables are all zero cannot be rejected.

The second and fourth columns of Table 6 report coefficient estimates when the models include the sum of squares and the sum of the interaction terms of the components of economic and personal freedom. In estimating Eq. (2) to test restrictions on the economic and personal freedom coefficients we include all the other explanatory variables, relative wage, its square, coefficient of variation (Gini coefficient), its square, national team players, team fixed effects, and year fixed effects. Considering the Restrictions model using the coefficient of variation, the results are the null hypothesis that \(\beta_{1} = \beta_{2} = \beta_{3}\) and \(\gamma_{2} = 2\gamma_{1}\), and the equivalent hypothesis for the personal freedom variables, are easily rejected; the F = 3.25, p-value = 0.048 for economic freedom and F = 17.14, p = 0.000, for personal freedom. The null that the economic (personal) freedom coefficients are all zero is also rejected. The implication of these results is that some of the coefficients of squared economic or personal freedom components are non-zero and that not all coefficients on squared terms are equal. Likewise, some of the coefficients on the interacted component variables are not zero, and those coefficients are not all equal. Joint hypothesis tests on the Gini Restrictions model result in the same inferences as from the coefficient of variation model.

Finally, note that it is possible to estimate models with all the squares and interactions variables. In such a specification, there are over 90 parameters to estimate just for the freedom variables against our data set of 156 observations. No cross-sectional unit appears more than 11 times and several clubs are in the data 6 or fewer times. Credibly precise parameter estimates of such specifications are improbable until more years of data are available.

In the foregoing analysis, one-year prior freedom variables are hypothesized to affect present team performance. Theory relating player mobility to team performance motivates that hypothesis; we expect players to move based on information available prior to the current season. As a robustness check, the models have been re-estimated replacing lagged values of the freedom variables with the contemporaneous values. These results are shown in Tables 7, 8 and 9. Table 7 is directly analogous to Table 6, with the exception that all the freedom variables are contemporary with the outcome, points per game. The results are rather different, however.

Table 7 MLS Points per game
Full size table
Table 8 MLS Points per game CV
Full size table
Table 9 MLS Points per game
Full size table

As seen above, one can reject the null hypothesis that the coefficient restrictions are valid for economic freedom; in the case of contemporary economic freedom variables, this is not true. However, it remains true for the personal freedom variables. These findings hold whether the salary distribution is addressed by the coefficient of variation or the Gini coefficient. The rejection of the parameter constraints for the personal freedom variables means that these results should not be used to compute marginal effects of the personal freedom variables because coefficients on the relevant square and interaction terms are unknown. Nonetheless, several of the explicit personal freedom variables are individually statistically significant. Like in the case of lagged freedoms, marriage freedom remains negative and statistically significant. Also negative and statistically significant are freedoms related to guns and to assets. Alcohol and educational freedom are positive and statistically significant. Of course, as already mentioned, these variables appear in the sum of squares and the interactions variables as well, so their marginal effects may not be the same sign as their linear coefficient after accounting for these additional influences.

Comparing results from Tables 3 and 8, and Table 4 with Table 9, the similarities are clear. Statistical significance (insignificance) of a variable in one table pairs with its statistical significance (insignificance) in the other; signs though not sizes of the coefficients are quite alike as well. The conclusions for these models in which freedoms are represented by the aggregate index is, therefore, that the implications of the model are not materially changed by whether the freedom variables are current or from one year before the current period. Consequently, based on the theory, we prefer the models with the lagged values (Table 10).

Table 10 MLS Points per game
Full size table

Finally, there is the issue of including or not including year fixed effects in the models. The models reported above all include year effects unless explicitly stated otherwise. From an econometric perspective, including them might impair efficiency of the estimates, but does not result in biased estimates, while excluding them might bias the coefficient estimates of interest. Intuitively, the points per game across the league may be below average in a year, say, due to a spate of injuries to players on otherwise strong teams, resulting in more draws than in a typical year. Many injuries to players on weaker than average teams might reduce the number of draws in a year. In either of these two cases, omission of a year effect could result in biased coefficient estimates. To assess the importance of year effects, we re-estimated the models without year dummy variables.

Tables 11 and 12 provide coefficient estimates for models that are identical to those in Tables 3 and 4 except the former do not include year effects. Comparing Table 3 and Table 11, the results from the model without year effects (Table 11) indicate both lagged economic freedom and its square are statistically significantly different from zero at the 10% level or better.

Table 11 MLS Points per game
Full size table
Table 12 MLS Points per game
Full size table

In Table 4, lagged economic freedom is not statistically significant in any model and its square is only significant at the 10% level if the personal freedom variables are omitted. On the other hand, in Table 12, lagged economic freedom and its square are each individually significant at the 5% level in every model. The lagged personal freedom coefficient is both much smaller and statistically significant at only the 10% level in Table 11 whereas in Table 3 lagged personal freedom is significant at the 1% level. The same holds true comparing Tables 4 and 12. Clearly, inclusion of the year effects makes a difference to estimation results with respect to economic and personal freedom.

This point is made graphically in Fig. 5 through 8 which plot the estimated marginal effects for personal and economic freedom for the models with and without year effects both using the coefficient of variation and the Gini coefficient as the measure of salary distribution. Figure 5 is the marginal effects of economic freedom when CV measures salary dispersion, Fig. 6 is the marginal effects when Gini captures salary inequality. Figures 7 and 8 are analogously defined though the marginal effects are those of the personal freedom variable. The figures show that the impact of the year effects is to reduce the size of the marginal effect of economic or personal freedom at the extreme values of those variables, and to widen the confidence intervals of the estimates, relative to what they would be in models without the year variables. In the observed range of economic and personal freedom, however, the marginal effects are often not different from zero.

Fig. 5
figure 5

Marginal effects of Gini coefficient with 95% Confidence Intervals

Full size image
Fig. 6
figure 6

Marginal Effect of Economic Freedom. Negative Binomial Regression

Full size image
Fig. 7
figure 7

Marginal Effect of Personal Freedom. Negative Binomial Regression

Full size image
Fig. 8
figure 8

Marginal Effect of Relative Wage. Negative Binomial Regression

Full size image

Summing up this section, the evidence is consistent with including quadratic specifications of the economic and personal freedom variables though perhaps it is better, wherever possible, to use the components of the index rather than the index itself. The data clearly rejects a purely linear model in which freedom variables have the same marginal effect regardless of their starting value. In addition, the evidence indicates that inclusion of year effects alters point estimates and statistical significance of key variables, but that those changes do not translate into dramatically different implications (Figs. 9 and 10).

Fig. 9
figure 9

Marginal Effect of Coefficient of Variation. Negative Binomial Regression

Full size image
Fig. 10
figure 10

Marginal Effect of Gini Coefficient. Negative Binomial Regression

Full size image

6 Discussion and conclusion

How political, social and economic policy affects businesses is widely studied. Little of this has focused on the sports business, however. Those papers that have addressed these issues primarily concern the influence of taxes on player mobility and compensation. Here we have examined government influence more broadly by using indicators of economic and personal freedom. Moreover, the outcome measure of interest is team performance. Specifically, we examine the number of points per match earned by clubs in Major League Soccer. Because MLS has more influence over player allocation to teams than do other leagues, we argue that finding an influence of economic and personal freedom on team success should be more difficult in MLS than in those leagues. Finding effects in this case suggests effects may be stronger where leagues have less control.

Our results indicate that economic freedom does influence team performance, though the effects are not large and are concentrated at the extremes of the economic freedom distribution. Specifically, at low levels of economic freedom, economic (un)freedom has small harmful effects on team performance, and at high levels of economic freedom, economic freedom has small beneficial effects on team performance. Most clubs fall in the broad range of economic freedom values where the marginal impact is not different from zero. Nonetheless, the evidence here is that clubs from states with very high economic freedom have a competitive advantage over clubs from states with mid or low levels of economic freedom.

These findings are of interest to the discussion of whether teams improve after opening a new stadium (Clapp and Hakes 2005; Hakes and Clapp 2006; Quinn et al. 2003; Rockerbie and Easton 2019). The consensus in this literature is that there is little evidence that teams improve on the field after opening a new facility. Construction of many of the facilities was financed all or in substantial part by state and local governments which, all other things equal, tends to reduce their economic freedom score. From the results above, a reduction in the economic freedom score reduces the size of any positive marginal effect or increases the size of any negative marginal effect of economic freedom on team performance. The upshot is that the existing evidence of little team improvement from opening a subsidized stadium may result from a combination of some small improvement from the new revenue streams the new facility provides and the harmful effects of reduced economic freedom. Moreover, the results here suggest there are circumstances where the complementarity of capital and labor in sports is sufficient to offset the harmful effects of economic (un)freedom on team performance; and there are circumstances in which very low economic freedom swamps the influence of capital and labor complementarity.

Importantly, the analysis here finds evidence for an influence of personal freedom, at least as it is measured by Ruger and Sorens (2018) and Sorens et al. (2008), on team performance. Low levels of personal freedom harm performance while high levels generally have no effect that is statistically significantly different from zero. This may be a consequence of the policy issues used to define personal freedom or because the aggregation of the individual components is a poor proxy for the influences of the individual components. A topic of future research is to assess the influence of other measures of personal freedom starting with abortion and reproductive rights freedom variables available from Ruger and Sorens (2018) and Sorens et al. (2008) but not utilized here.

Further analysis can be done using the other major professional team sports in North America and expanding the time-frame of the analysis. Existing research on those leagues has focused entirely on the impact of personal income taxation differences across states. The evidence here suggests that economic freedom more broadly, and personal freedom, may be as important as the narrow tax issues. Indeed, to the extent that the influence of the taxes or economic freedom works through player choices about which team to play for, our results suggest the influence may be stronger in other leagues where players face fewer restrictions on mobility. To directly test the hypothesis that club improvement is enhanced when the economic and political environment attract players, tracking free agent signings is necessary. That linkage is the subject of ongoing research.

Finally, the motivation of this research was to evaluate whether economic and personal freedoms influence firm performance as a means of connecting micro-performance to the findings of the studies at the macro-level. Our consistent findings that both economic freedom and personal freedom variables are statistically significant influences on the performance of Major League Soccer teams confirm this connection. The issue of whether this connection is entirely or mostly through the labor market or influences team performance in other ways remains to be established.