Does Economic Freedom Influence Economic Growth? Evidence from Latin America

Abstract

This paper investigates the relationship between economic freedom and economic growth in Latin America and the Caribbean over the period 1997–2023, using data from 14 countries. To capture the multidimensional nature of economic freedom, two widely recognized indices—Heritage and Fraser—are incorporated into an extended Solow-type growth model. The empirical strategy relies on a dynamic panel data approach using the Arellano–Bond estimator, which allows for the control of unobserved heterogeneity, autocorrelation, and potential reverse causality. Robustness is assessed through alternative model specifications and in-sample forecasting using rolling-window techniques and Theil’s U-statistic. The results reveal a negative and statistically significant relationship between economic growth and the Heritage Index, while the Fraser Index shows a positive but generally non-significant effect. These findings highlight the methodological sensitivity of the economic freedom–growth nexus and suggest that context-specific institutional factors may shape how liberalization policies translate into development outcomes. The study contributes to the literature by jointly evaluating the impact of both indices in a unified dynamic framework, providing new evidence for a region marked by institutional heterogeneity and growth volatility.

1. Introduction

Latin America and the Caribbean (LAC) is one of the regions with the slowest economic growth in the world, accentuating structural issues such as limited formal employment, high labor informality, persistent poverty, low educational quality, and significant gender disparities. These economic difficulties have intensified due to external shocks like the COVID-19 pandemic and the ongoing global economic uncertainty. Therefore, it is critical for regional governments and institutions to enact policies that stimulate economic growth to mitigate internal vulnerabilities and external adverse factors, promoting inclusive, resilient, and sustainable development (CEPAL, 2023).

Economic literature suggests economic freedom as a key determinant of growth. Economic freedom, as understood by the classical Anglo-British liberal tradition, refers to individuals’ ability to work, transact, own property, and fully utilize it, with the state ensuring civil rights to produce, exchange, and distribute goods and services. Economies with higher freedom typically experience greater prosperity and stability (Villalta, 2024). Under this perspective, market competition supersedes centralized state planning, promoting privatization, deregulation of labor and markets, reduced taxation, and openness to international trade. Friedman (2002) similarly argues for minimal state intervention focused exclusively on maximizing economic freedom (Giorgio, 2024).

However, is this theoretical relationship empirically valid? Various studies confirm a positive correlation between economic freedom and growth (Jordaan, 2023; Kwablah & Amoah, 2022; Ahmed et al., 2023; Zhu et al., 2024; de Haan & Sturm, 2000), yet some researchers caution that results heavily depend on methodological choices. Misati and Nyamongo (2012) and Carlsson and Lundström (2002) highlight sensitivity in outcomes depending on how economic freedom variables are constructed. Santhirase (2007) finds institutional instability might negatively impact this relationship. Similarly, Hall and Lawson (2014) identify primarily positive associations but acknowledge some negative effects, emphasizing the need for deeper contextual analyses.

In Latin America specifically, results are nuanced. Köppe and Santos (2021) find corruption can distort growth in economically freer countries, suggesting a complex institutional interplay. Acevedo and Lorca-Susino (2023) confirm these findings, indicating economic freedom’s effectiveness depends significantly on institutional conditions. Bengoa and Sanchez-Robles (2003) emphasize economic freedom’s positive role in attracting foreign direct investment (FDI) and fostering growth, while Santiago et al. (2020) present conflicting evidence highlighting methodological sensitivities and uncertainties in the relationship.

Moreover, the literature also reports mixed results regarding economic freedom’s influence on inequality. Ahmad (2017) and Callais and Young (2023) observe slight increases in inequality with greater economic freedom. Conversely, Bennett and Vedder (2013) identify an initially negative correlation before a certain threshold, and Sturm and De Haan (2015) find no robust relationship, attributing this diversity to methodological and institutional variations. In Latin America, the reduction in inequality since the 2000s coincided with stable or moderately increasing economic freedom, suggesting contextual factors and complementary policies significantly influence outcomes.

Additionally, research generally links greater economic freedom to enhanced welfare outcomes, including higher per capita income, life expectancy, and subjective well-being (Graafland, 2023). Bjørnskov (2016) and Alimi (2016) demonstrate economic freedom’s role in stabilizing economies, reducing volatility, and improving resilience to shocks, crucial in historically volatile regions like LAC.

Given this background, the primary aim of this article is to examine the relationship between economic freedom and economic growth in LAC (1997–2023), using Heritage and Fraser Indexes within a robust methodological framework, including random effects and Arellano–Bond dynamic estimators, validated by rolling-window forecasts. This study uniquely contributes to understanding how conceptualizations and measurements of economic freedom influence regional growth policies.

This paper contributes significantly to the existing literature by rigorously examining the complex relationship between economic freedom and economic growth in Latin America and the Caribbean (LAC), integrating methodological innovations and providing nuanced insights into policy implications. By concurrently evaluating two internationally recognized indices—the Heritage Foundation and Fraser Institute—within a robust econometric framework that combines random-effects panel estimation and dynamic Arellano–Bond methods (Arellano & Bond, 1991), this research addresses critical methodological sensitivities previously highlighted by scholars such as Misati and Nyamongo (2012) and Carlsson and Lundström (2002). Moreover, the incorporation of rolling-window forecasts and Theil’s U-statistic as robustness checks further strengthens the analytical validity of the findings. Hence, the paper advances a more comprehensive understanding of how diverse conceptualizations and measurements of economic freedom can shape distinct policy trajectories in regions characterized by institutional volatility, economic fragility, and heterogeneous growth dynamics.

The structure is as follows: Section 2 presents methodology; Section 3 discusses results and robustness analyses; Section 4 critically interprets findings; Section 5 concludes the study.

Literature Review

The relationship between economic freedom and economic growth has been the subject of extensive empirical studies, characterized by a plurality of theoretical and methodological approaches. From the classical Anglo-British tradition, authors such as Friedman (2002) argue that economic freedom—defined as the ability of individuals to undertake, own private property and freely engage in economic transactions—is essential to maximize economic prosperity and social stability (Villalta, 2024; Giorgio, 2024).

Recent studies have consistently evidenced a positive relationship between economic freedom and growth. Jordaan (2023), Kwablah and Amoah (2022), Ahmed et al. (2023), Zhu et al. (2024) and de Haan and Sturm (2000) conclude that economies with higher levels of economic freedom tend to exhibit sustained increases in economic growth, partially confirming the hypothesis put forward by classical liberal theories. However, the literature emphasizes a considerable sensitivity in these results depending on the index and the methodology used. Misati and Nyamongo (2012) and Carlsson and Lundström (2002) find that when decomposing composite indices into individual components, some key factors may even have inverse relationships with economic growth, underscoring the need for a more rigorous and context-specific methodological exploration.

In this regard, Hall and Lawson (2014), after an exhaustive bibliometric analysis of 402 empirical studies on economic freedom, found that although most report positive effects on growth, a significant minority show negative effects, indicating that certain institutional and methodological conditions may reverse or attenuate the theoretically expected relationship. Indeed, Santhirase (2007) shows that institutional instability, especially in developing countries, can impair the effectiveness of the economic freedom model, adversely affecting economic performance and the quality of political leadership.

In the particular context of Latin America and the Caribbean (LAC), previous results highlight a complex and nuanced dynamic. Köppe and Santos (2021) establish that although in general terms, there is a positive correlation between economic freedom and growth, specific institutional factors such as corruption may act as key mediating variables. These authors argue that in Latin American countries with high rates of economic freedom, corruption tends to reduce the expected benefits, while paradoxically, in institutionally more restrictive contexts, corruption can generate conditions that are circumstantially favorable to economic growth. This result is empirically corroborated by Acevedo and Lorca-Susino (2023), who highlight how the effect of economic freedom does not operate in isolation, but is conditioned by the institutional and political environment in which economic policies are implemented.

On the other hand, Bengoa and Sanchez-Robles (2003) provide clear evidence on the positive role played by economic freedom in attracting foreign direct investment (FDI), with subsequent favorable effects on regional growth. However, Santiago et al. (2020) present contrary evidence, finding a negative relationship between economic freedom and growth in LAC, strongly underlining the methodological sensitivity and the importance of strengthening the analysis with advanced econometric techniques that allow controlling for endogeneity and problems inherent to panel data.

Regarding the effects of economic freedom on inequality, the evidence is also heterogeneous. Ahmad (2017) and Callais and Young (2023) identify a slight increase in inequality associated with increases in economic freedom, while Bennett and Vedder (2013) and Sturm and De Haan (2015) find nonlinear effects or absence of robust relationship, respectively. Particularly in Latin America, the significant reduction in inequality since the beginning of the 21st century has coexisted with moderate or stable levels of economic freedom, suggesting that the relationship is strongly conditioned by complementary policies in key areas such as education, labor formalization and social transfers.

Finally, recent studies highlight that greater economic freedom is consistently associated with better outcomes in terms of overall economic well-being and macroeconomic stability. Graafland (2023) shows a significant positive correlation between economic freedom and life satisfaction, operating through increased individual autonomy. Likewise, Bjørnskov (2016) and Alimi (2016) document how freer economies show lower GDP volatility and greater resilience to economic shocks, especially benefiting regions historically prone to recurrent crises, such as LAC.

In summary, the existing literature presents results that, although predominantly positive, are sensitive to institutional contexts and methodological decisions. This paper seeks precisely to bring empirical clarity to this controversy through the simultaneous application of two indices of economic freedom (Heritage and Fraser), using robust econometric techniques (random effects and dynamic Arellano–-Bond estimators), together with a rigorous assessment of stability through advanced predictive techniques.

Finally, although there is broad consensus in the literature on the relevance of economic freedom as a determinant of growth, the available empirical results show considerable heterogeneity depending on both the institutional context and the methodological strategies adopted. In particular, for Latin America and the Caribbean—a region characterized by high economic volatility and institutional weaknesses—these aspects are especially relevant, given the complexity introduced by factors such as corruption, informality and differences in the quality of governance. Given this methodological and contextual diversity, there is a need for robust and comparative empirical analyses to clarify the specific effects of different indices of economic freedom and advanced econometric techniques. Therefore, this article seeks to contribute to the existing debate by means of a rigorous empirical analysis, which answers the following hypothesis:

The relationship between economic freedom and economic growth in Latin America and the Caribbean depends significantly on the institutional context and the methodological specifications employed.

2. Methodology

2.1. Panel Data Estimation

The usability of panel data provides a greater amount of information, since the same observation is studied over time, i.e., it presents a dimension of time and space (Gujarati & Porter, 2010). W. H. Greene (2012) proposes the following representation:

$$w_{it}=\gamma +x_{it}^{\prime }\beta +z_{it}^{\prime }\alpha +u_{it}\hspace{1em}i=1,\dots ,N;t=1,\dots ,T$$

(1)

In this expression, $w_{it}$ represents the explained variable, $i$ is the i-th individual and $t$ is the period, $\gamma$ is the intercept, $x_{it}^{\prime }$ is the vector of explanatory variables and $z_{it}^{\prime }$ contains a constant term and individual or group variables that can be observed or unobserved.

If $z_{it}^{\prime }$ contains only a constant term, it is under a Pooled (POO) model, while if this vector is not observed but is correlated with $x_{it}^{\prime }$, it should be considered a “Fixed Effects (FE)” model. On the contrary, if it is unobserved and uncorrelated, it would be under “Random Effects” (RE).

The estimation in the first model can be performed through Ordinary Least Squares (OLS). In the presence of FE, it should be estimated through Least Squares with Dummy Variables (LSVF), while for RE, Generalized Least Squares (GLS) is used.

2.2. Statistical Tests

For a correct estimation, first of all, the shape of $z_{it}^{\prime }$ must be considered. Additionally, the behavior of the errors must be considered, since they may present heterogeneity, autocorrelation and contemporaneous correlations. In addition, for the choice of the model to be analyzed, informational criteria were used, and some other tests were considered for the choice of the best model. Details of the tests used can be found in Appendix A.

2.3. Econometric Model

Based on Solow (1956), Rincón (1998) proposes the following model for economic growth in LAC:

$$y_{it}={\beta }_0+{\beta }_1\ln \left(y_{oit}\right)+{\beta }_2\ln \left(i_{it}\right)+{\beta }_3\ln \left(n_{it}+g+\delta \right)+{\beta }_4\ln \left({\pi }_{it}\right)+u_{it}$$

(2)

where $y_{it}$ represents GDP per capita growth, $y_{oit}$ is GDP per capita growth from year 0, $i_{it}$ real gross domestic investment (public and private), $n_{it}$ labor participation rate, $g$ rate of technological progress, $\delta$ rate of capital depreciation, ${\pi }_{it}$ is the inflation rate and $u_{it}$ is the error term.

From (2), the index of economic freedom will be added to evaluate the relationship between it and economic growth. Therefore, the model to be estimated is as follows:

$$y_{it}={\beta }_0+{\beta }_1\ln \left(y_{oit}\right)+{\beta }_2\ln \left(i_{it}\right)+{\beta }_3\ln \left(n_{it}+g+\delta \right)+{\beta }_4\ln \left({\pi }_{it}\right)+{\beta }_5{L}_{it}+u_{it}$$

(3)

In this research, it is of interest to establish the relationship between economic growth and economic freedom using different indexes, for which three models to be calculated are established.

Heritage

$$y_{it}={\theta }_0+{\theta }_1\ln \left(y_{oit}\right)+{\theta }_2\ln \left(i_{it}\right)+{\theta }_3\ln \left(n_{it}+g+\delta \right)+{\theta }_4\ln \left({\pi }_{it}\right)+{\theta }_5{H}_{it}+u_{it}$$

(4)

Fraser

$$y_{it}={\alpha }_0+{\alpha }_1\ln \left(y_{oit}\right)+{\alpha }_2\ln \left(i_{it}\right)+{\alpha }_3\ln \left(n_{it}+g+\delta \right)+{\alpha }_4\ln \left({\pi }_{it}\right)+{\alpha }_5{F}_{it}+u_{it}$$

(5)

Heritage and Fraser

$$y_{it}={\gamma }_0+{\gamma }_1\ln \left(y_{oit}\right)+{\gamma }_2\ln \left(i_{it}\right)+{\gamma }_3\ln \left(n_{it}+g+\delta \right)+{\gamma }_4\ln \left({\pi }_{it}\right)+{\gamma }_5{H}_{it}+{\gamma }_6{F}_{it}+u_{it}$$

(6)

To address potential endogeneity problems arising from omitted variables and simultaneity bias, this study employs the Arellano–Bond dynamic panel estimator, which is particularly suitable for macroeconomic panels with a small number of time periods and many cross-sectional units. The inclusion of a lagged dependent variable helps control for unobserved country-specific effects and persistence in economic growth. However, beyond these sources of endogeneity, it is crucial to consider the possibility of reverse causality: economic growth itself may influence the degree of economic freedom. For instance, sustained growth could enable reforms that improve regulatory quality, reduce government intervention, or increase market openness—all components of economic freedom.

The Arellano–Bond method helps mitigate this issue by using deeper lags of the explanatory variables (such as economic freedom) as instruments. These lagged levels are assumed to be uncorrelated with the current error term but correlated with the endogenous regressors. Therefore, the estimator provides a consistent framework to control for both dynamic feedback effects and reverse causality, enhancing the reliability of the coefficient estimates. In addition, the validity of the instruments is tested using the Hansen J-test, and the absence of second-order autocorrelation is confirmed with the AR(2) statistic, strengthening the robustness of the specification.

2.4. Data

The objective of this research is to determine the relationship between economic freedom and economic growth in Latin America and the Caribbean during the period 1997–2023. For this purpose, information was obtained from the World Bank (WB) for the variables corresponding to economic growth.

It should be considered that gross fixed capital formation (% of GDP) was used as an approximation of $i_{it}$. For the construction of $\ln \left(n_{it}+g+\delta \right),n_{it}$ corresponds to the total labor participation rate (% of the total population over 15 years of age); in addition, Mankiw et al. (1992) suggested that δ should be equal to 0.03, while for g, we took the average per capita growth of the countries studied in the period of interest, which was 2.18%, so that $g+\delta$ was 0.052 $\forall i$. Regarding inflation, the annual % based on consumer prices (CPI) was considered.

For the $L_{it}$ Index of Economic Freedom, we worked with the index published by the Wall Street Journal together with the Heritage Foundation, named after the latter ($H_{it}$). It is constructed based on 12 quantitative and qualitative factors, grouped into four categories: (a) Rule of Law, (b) Size of Government, (c) Regulatory Efficiency and (d) Market Openness; the index is presented on a scale from 0 to 100, where a higher value implies greater freedom (The Heritage Foundation, 2023). In addition, the index constructed by the Fraser Institute ($F_{it}$) was also taken, which is based on 44 variables and 26 components, and is divided into five categories: (a) size of government, (b) legal system and property rights, (c) access to sound money (monetary), (d) freedom of international trade and (e) regulations; it is presented on a scale of 0 to 10, where like the previous one, a higher score implies more freedom (Fraser Institute, 2024).

3. Results

For the presentation of the results, we considered the models presented in Equation (4), which established the relationship between growth and the Heritage $H_{it}$ Index, the models in (5), which demonstrated the same evaluation as the Fraser $F_{it}$ index, and the models in (6), which were presented as a whole.

It should be noted that, in panel data analysis, it is essential to determine whether there are fixed and random effects. To identify fixed effects, an F test was applied to each of the models, concluding that (4) incorporates them. On the other hand, the Lagrange Multiplier test of Breusch and Pagan (1980) was used to evaluate the existence of random effects, confirming their presence. Finally, the Hausman (1978) test was applied to determine the choice between fixed and random effects, concluding that in all three models, it is appropriate to consider random effects1. In addition, the behavior of the errors was analyzed with the aim of detecting possible problems of heteroscedasticity, autocorrelation and correlations with time. For this purpose, the WaldTest, Wooldridge (2010) and Pesaran (2021) tests were performed, which showed the presence of heteroscedasticity and contemporaneous correlations in the three models, while autocorrelation affects only (4)2.

Several methodologies were considered to correct the problems mentioned above.

Table 1. Results considering random effects, autocorrelation, and robust errors.

	Autocorrelations	Robust Errors
${\mathit{y}}_{\mathit{i}\mathit{t}}$	(1)	(2)	(3)	(4)
$\mathrm{l}\mathrm{n}\left(y_{oit}\right)$	0.8956	0.919	0.3412	0.4216
	(0.4912) *	(0.4332) **	(0.4920)	(0.5683)
$\mathrm{l}\mathrm{n}\left(i_{it}\right)$	5.3090	5.0277	5.0925	4.5076
	(1.0578) ***	(0.8566) ***	(0.7874) ***	(0.9667) ***
$\mathrm{l}\mathrm{n}(n_{it}+g+\delta )$	4.7591	4.1063	6.5989	5.8992
	(2.8316) *	(3.1589)	(2.9259) **	(3.2786) *
$\mathrm{l}\mathrm{n}\left({\pi }_{it}\right)$	0.0055	0.02471	0.2519	0.4142
	(0.2125)	(0.2009)	(0.2881)	(0.2978)
$H_{it}$	−0.0462	−0.0516		−0.0999
	(0.0011)	(0.0279) *		(0.0554) *
$F_{it}$			0.3200	1.3258
			(0.5489)	(0.8713)
${\theta }_0$	−31.6681	−27.8323
	(13.0590) **	(14.349) *
${\alpha }_0$			−43.4108
			(15.2166) ***
${\gamma }_0$				−39.79
				(15.9667) **
Wald	35.17 ***	53.48 ***	53.58 ***	48.30 ***
AIC	1887.73	1955.3	1676.7	1674.3
BIC	1922.63	1986.7	1706.8	1708.1

The standard error is denoted in parentheses. $p<0.1$ *, $p<0.05$ **, $p<0.01$ ***.

shows the estimations made for all the models, considering random effects and robust errors. In addition, the autocorrelation present in (4) was considered.

From the estimates obtained, it should be noted that the model established in (6) is the most adequate according to Akaike (AIC). It is observed that there is a contradiction between the results, since it establishes that the relationship between the variables of interest is negative when the Heritage is considered, while with the other index, it is direct, although the latter is statistically non-significant. If the indices are analyzed individually, the conclusion remains the same.

In addition, the Arellano–Bond methodology was considered in order to avoid possible endogeneity problems in the results and

Table 2. Results for Arellano–Bond.

${\mathit{y}}_{\mathit{i}\mathit{t}}$	(d)	(e)	(f)
$\mathrm{l}\mathrm{n}\left(y_{oit}\right)$	0.7541	0.2954	0.4166
	(0.5264)	(0.5581)	(0.6732)
$\mathrm{l}\mathrm{n}\left(i_{it}\right)$	7.2443	7.3203	6.5913
	(0.9113) ***	(1.1294) ***	(1.1975) ***
$\mathrm{l}\mathrm{n}(n_{it}+g+\delta )$	7.3724	10.9566	9.9734
	(3.0624) **	(3.1757) ***	(3.5305) ***
$\mathrm{l}\mathrm{n}\left({\pi }_{it}\right)$	−0.0118	0.2792	0.4277
	(0.1627)	(0.2314)	(0.2322) *
$H_{it}$	−0.0508		−0.1247
	(0.0328)		(0.0584) **
$F_{it}$		0.4637	1.6714
		(0.63836)	(0.9129) *
$y_{oit}$	−0.0370	−0.0555	−0.0636
	(0.0425)	(0.0468)	(0.0460)
${\theta }_0$	−47.8930
	(14.3165) ***
${\alpha }_0$		−69.1470
		(15.8502) ***
${\gamma }_0$			−63.7821
			(16.7952) ***
Wald	1074.48 ***	640.15 ***	570.97 ***
AR(2)	−2.21 **	−1.68 *	−1.84 *
Hansen	6.52 **	0.93	0.99

The standard error is denoted in parentheses. $p<0.1$ *, $p<0.05$ **, $p<0.01$ ***.

shows the results obtained.

Table 3. Results for Model 3.

${\mathit{y}}_{\mathit{i}\mathit{t}}$	(1)	(2)	(3)
$\mathrm{l}\mathrm{n}\left(y_{oit}\right)$	0.4283	0.4209	−48.8225
	(0.4192)	(0.5822)	(19.1443) **
$\mathrm{l}\mathrm{n}\left(i_{it}\right)$	4.2982	4.5561	5.7818
	(1.009) ***	(1.0794) ***	(2.4750) **
$\mathrm{l}\mathrm{n}(n_{it}+g+\delta )$	5.6240	5.9802	11.3288
	(2.4693) **	(3.3184) *	(6.8536) *
$\mathrm{l}\mathrm{n}\left({\pi }_{it}\right)$	0.2716	0.4489	1.1403
	(0.2531)	(0.2725) *	(0.3653) ***
$H_{it}$	−0.0816	−0.1052	−0.2765
	(0.0413) **	(0.5444) *	(0.1419) *
$F_{it}$	1.1022	1.3881	1.9237
	(0.5342) **	(0.8174) *	(1.31584)
${\beta }_0$	−37.4300	−40.4335	(.) 1
	(12.0987) ***	(15.8830) **	(.)
$y_{it-1}$			−0.1370
			(0.0501) ***
F	6.09 ***
Wald		48.30 ***	46.44 ***

¹ (.) denotes that the parameter was omitted from the estimation. $p<0.1$ *, $p<0.05$ **, $p<0.01$ ***.

shows the results of model 3. Column (1) corresponds to the estimation of a pooled model with robust errors and it is evident that the two indexes used are statistically significant; however, it is evident that the signs are different, since with Heritage, the relationship is negative, in contrast to Fraser, which shows a direct relationship. Similar findings are obtained when the model is estimated considering random effects and robust errors (2), while when applying Arellano–Bond, the Fraser Index is not statistically significant, while Heritage is statistically significant and maintains the inverse relationship.

From the results obtained, model (6) is taken into consideration, since according to the tests applied, it shows the best behavior. Similar conclusions to those previously presented are obtained, i.e., the relationship between the variables of interest is negative when Heritage is considered, and positive for Fraser. This also demonstrates the robustness of the estimates.

In turn, in order to corroborate the stability of the results, the predictive power of the model was evaluated within the sample, i.e., the data were separated into 80% (1997–2018) for training and 20% (2019–2022) for testing. Using the rolling-windows methodology, the coefficients of a pooled model for equation (f) were estimated; from these results, Theil’s U-statistic (Pyndick & Rubinfeld, 2001)3 was calculated in order to evaluate the predictive quality, which was 0.101, which implies a good estimation.

4. Discussion

The results obtained show a clear disjuncture in the relationship between economic freedom and economic growth according to the index used. In particular, when the Heritage Index is used exclusively, a negative and statistically significant relationship is observed between both variables. This finding is consistent with Santhirase (2007), who argues that, in developing countries, the difficulty in consolidating stable central political leadership can deteriorate the institutional environment, thus negatively affecting economic growth. Added to this is the factor of corruption, which plays a key distorting role in the region. Along these lines, Köppe and Santos (2021) showed that, in Latin America, countries with higher levels of corruption and lower levels of economic freedom presented, paradoxically, higher growth rates, suggesting that in institutionally fragile contexts, the theoretical benefits of economic liberalization may not materialize and may even be reversed. Complementarily, the study by Santiago et al. (2020) also finds a negative relationship between economic freedom and growth for the region, and highlights that the treatment of the data, in particular the omission of outliers or the absence of adequate controls for heterogeneity, can generate significant distortions that alter the empirical results.

On the other hand, the results obtained using only the Fraser Index indicate a positive relationship between economic freedom and economic growth; however, this relationship does not reach statistical significance. This result suggests that this index may not adequately capture the relevant mechanisms in the Latin American context or that its effect is more diffuse due to the institutional heterogeneity of the countries analyzed.

When both indices are considered jointly in the same model, controlling for robust errors and endogeneity problems using the Arellano–Bond methodology, it is found that only the Heritage Index maintains statistical significance, and retains its negative direction. This reaffirms that the relationship between economic freedom and economic growth is highly sensitive to the way in which economic freedom is defined and measured. In this sense, Carlsson and Lundström (2002) warn that the use of aggregate indices can hide relevant relationships between the individual components, since different dimensions can have effects in opposite directions. Similarly, Misati and Nyamongo (2012) argue that the empirical results depend considerably on how the variables are operationalized, which introduces a critical methodological component in this type of analysis. This reasoning is complemented by Hall and Lawson (2014), who, after reviewing 402 studies on economic freedom, conclude that, although most show positive effects, about 4% find negative relationships. Although this proportion is a minority, it is significant from a scientific point of view, as it may reflect specific contexts—such as those observed in Latin America—in which economic freedom generates undesirable effects due to factors such as corruption, informality, institutional weakness or structural inequality.

These observations have important implications for public policy formulation. First, the results suggest that the promotion of economic freedom does not automatically guarantee higher growth rates if it is not accompanied by substantive improvements in institutional quality, transparency and democratic governance. Second, the evidence that different indices produce different conclusions implies that policy makers should be cautious about basing decisions on composite indicators without carefully examining their components. Moreover, the lack of significance of the Fraser Index calls into question its validity as a relevant predictor for the region, which should encourage researchers to construct or adapt indexes that better reflect the structural reality of Latin America. Finally, the results invite a critical reconsideration of the paradigm of exclusively market-oriented reforms, suggesting that their effectiveness is conditioned by the political–institutional context in which they are implemented.

5. Conclusions

The objective of this research was to analyze the relationship between economic freedom and economic growth in 14 Latin American countries, employing a neoclassical growth model adapted specifically to regional conditions. Two widely recognized indices—Heritage and Fraser—were simultaneously used to capture different dimensions of economic freedom and rigorously test the robustness of empirical findings.

The results clearly highlight the methodological sensitivity involved in evaluating this relationship. Specifically, estimations based on the Heritage Index reveal a negative and statistically significant association with economic growth, reflecting structural regional characteristics such as persistent corruption, institutional instability, and weak political leadership. Such factors can significantly distort the theoretical expectations of liberalization policies, potentially producing counterproductive outcomes. Conversely, the Fraser Index demonstrates a positive relationship with economic growth, although this association does not achieve statistical significance. This discrepancy underscores the importance of methodological choices and conceptual definitions in empirical analyses of economic freedom.

Robustness checks further validate these findings. By employing diverse econometric techniques—including random effects models adjusted for heteroscedasticity, dynamic Arellano–Bond estimations to address endogeneity concerns, and predictive stability analyses through rolling-window forecasting—the results proved consistent across different methodological specifications. However, the statistical significance consistently depended on the specific economic freedom index used.

These findings confirm the central hypothesis of this study: the relationship between economic freedom and economic growth in Latin America and the Caribbean significantly depends upon institutional contexts and methodological approaches. Therefore, the effectiveness of market-oriented reforms in fostering sustainable economic growth critically hinges on parallel institutional strengthening, improved governance, and enhanced transparency.

Finally, from a methodological standpoint, this research emphasizes that precise construction of economic freedom measures, alongside careful data management—including the identification of outliers and appropriate handling of endogeneity—is essential to derive reliable empirical conclusions. Consequently, policymakers should approach liberalization reforms with caution, ensuring that they are accompanied by robust institutional frameworks. Future research could fruitfully explore how specific components of economic freedom interact with other variables such as democracy, inequality, and political stability, thereby deepening the understanding of the mechanisms at play within Latin American economies.