How Economic Freedom Mitigates Crisis

When economic crisis strikes, free market is often deemed guilty. Freedom falls during economic crisis, especially sound money and banking, owing to government's reaction. And more regulatory laws are easily passed. Historical records (e.g., free banking) as well as econometric studies (Baier et al., 2012; Chortareas et al., 2013; Lee et al., 2016; Wang & Luo, 2019) show that economic freedom (EF henceforth) is associated with banking stability. The market "hate" seems unjustified. The governmental solution to crisis must also be seriously revised. In fact, more economic freedom causes faster recovery.

Last update: January 31st 2025, added Connors & Norton (2012)

Intro: theory before econometrics

Sound theory is required to avoid misinterpreting econometric models. Consider the following: Assuming all else is held constant, we expect every bit of financial liberalization to improve stability. Or so they say. This however assumes linearity in the relationship and no interaction with the current regulatory measures. For instance, dropping the lender of last resort causes the federal deposit insurance to affect GDP negatively only because this insurance makes the lender of last resort necessary, while in reality the economy would have been much better off after removing both. In this case, the relationship is not linear (initial deregulation causes instability but further deregulation improves stability) and there is an induced interaction effect (insurance x lender).

Regulatory measures indeed have unequal importance. Historical episodes of free banking showed us that the common regulatory laws were not binding, such as capital requirement for opening a bank or restrictions on very small denomination notes, therefore justifying their labeling as free banks, as opposed to some others such as the restriction on branch banking or stringent note backing requirements, which proved being detrimental to the U.S. supposedly free banks, as these laws did not let them function properly as intermediaries.

When reading economics papers, it is no less important to examine closely the variables used in the regression. For instance, economic freedom (EF) is almost always measured using the Fraser Institute's Economic Freedom of the World, a 10-point scale variable, and is composed of 5 pillars: government size (fiscal freedom and spending), rule of law (protection of property rights and from corruption), sound money (inflation rates), trade openness (tariffs, trade taxes, quotas, barriers), regulatory efficiency (business operation, hiring). But not all components are equally important. With respect to financial crisis, austrian economists emphasize, with reason, the role of money. Yet sound money only measures inflation, whereas austrian economists emphasize the distortion of relative prices or rates of profit as the cause of crises.

Financial crisis

Bjørnskov (2016) analyzed the impact of freedom on the risk of crises, severity of crises (peak-trough ratio) as well as the duration and recovery time, separately, from 1993 to 2010 and across 121 countries. He uses regressions with random effects while including regional and annual fixed effects. Other controls (all lagged 1 period) include population size, initial real GDP per capita, trade volume, share of neighbour countries that are in crisis, and a dummy for post-communist countries. EF score is taken prior to the crisis in order to mitigate potential endogeneity since freedom is often negatively affected during crises by governments. Figure 2 displays the growth in GDP per capita during the crisis onset across 80 (non-overlapping) crises. One can observe that countries with low freedom score experience deeper crises and slower recovery. The peak-trough ratio is 10.3% for low EF as opposed to 3.4% for high EF. Both depth and recovery time are robustly associated with EF even after excluding the tails of EF score distribution. An increase in only 10 points of EF is associated with a decline of peak-trough ratio of 4% and a reduced recovery time of 10 months.

Callais & Pavlik (2022) examine the association between EF and adverse outcomes across 382 metropolitan areas in the U.S. as a result of the Great Recession. They use a two-way fixed effects regression, which is a method adjusting for unobserved unit-specific (e.g., firms, countries, states) and time-specific confounders at the same time. Because the effects of the housing market crash likely weren't uniform, the panel model (2002-2012) includes a state-specific time-trend in addition to year and metropolitan area fixed effects. This controls for general state level characteristics that change through time, including the potential for changes in the cost of living. Control variables include income inequality and share of industry employment from 18 industries, such as, construction, education, farming, finance, healthcare, information, mining, real estate, other services, etc. Finally, crisis severity was assessed using 3 dependent variables: unemployment rate, employment per 100 persons, income per capita. The second analysis compare the recovery of "treated" areas (i.e., those that experienced increases in freedom prior to the crisis) with "untreated" areas (those which are very similar to treated ones) in the post-crisis period. It utilizes matching methods (using Propensity Score or Mahalanobis) where changes in EF relate to subsequent changes in outcome measures from 2007 to 2012.

For the first analysis, panel regression showed that EF was robustly related to lower unemployment, higher levels of employment and increased net (i.e., less transfers) income, with and without controls. The strong effect is evident. 1 SD in EF corresponds to 0.61 point reduction in unemployment rate, an increase in 1.36 jobs per 100 persons and 3.8% in income per capita. Even after transfer payments are included in the income variable, the coefficient of EF is somewhat lower but still strong. Income is still higher in areas with more EF despite lack of social safety net.

The second part of the analysis focuses on the changes in the dependent variables between 2007 and 2012. Given the general increase in unemployment, a consistently lower change in unemployment among high EF areas indicates that they had better outcomes. On the other hand, employment estimates were mostly negative for high EF areas but the signs switch directions across matching methods, so the result is ambiguous. Finally, high EF areas grow consistently faster in net income per capita although the association is weaker when considering total income (i.e., after transfers).

Eslamloueyan & Jafari (2019) examined the effect of EF on investment during a financial crisis among East Asian countries between 1984 and 2014. EF is measured using a composite score of institutional quality indices from International Country Risk Guide (ICRG) data. Their first composite score, denoted INS5, is composed of government stability, investment profile, corruption, law and order, bureaucracy quality. A second composite score, denoted INS7, adds democratic accountability and socioeconomic conditions. These two composite scores are obtained from principal component analysis (PCA). The regression model, with investment as outcome variable, includes these predictor variables: lagged investment, real interest rates, trade openness, GDP, real exchange rate, real government consumption expenditure, institutional quality, dummy variables that capture the effects of the East Asia financial crisis of 1997 and the global financial crisis of 2008, and finally the interaction between institutional quality measures and these dummy variables. The analysis involves 9 regressions, each time using a different measure of institutional quality: INS5, INS7, and then each of the 7 variables of INS7.

Several assumptions must be met however. First, the hypothesis of error cross-sectional dependence was tested and rejected. Second, a unit root test reveals that the variables were integrated of order one, meaning their first differences are stationary. Third, check for spurious regression was done through cointegration test which showed there was cointegration in all cases, hence confirming the presence of a long-run equilibrium among variables. Fourth, to deal with endogeneity, instrumental variables were used and they involved all variables in their lagged values.

The regression results (Tables 2-5) show that the dummy variables of the 1997 and 2008 financial crises (denoted D97 and D2008) have negative coefficients whereas their interaction term with the institutional quality measures is positive for 1997 but weaker and not always significant for 2008. This indicates freedom reduces the adverse impact of crises on investment.

Pandemic

Although influenza pandemics are less impactful than financial crises in terms of GDP growth and not caused by (un)sound money, they involve reallocation of resources owing to changed constraint. Regulation is expected to increase the costliness of the adapting process while freedom should reduce frictions in the reallocation.

Candela & Geloso (2021) examine the relationship between EF and six major influenza pandemics from 1957 to 2009. The EF measure was the Historical Index of Economic Liberty (HIEL) which covers 21 OECD countries annually between 1850 and 2007. HIEL is composed of regulation, property rights, international trade and sound money but not government size. Dependent variable was the average GDP growth rate during pandemic. Independent variables include excess death rate, initial income per capita (log), HIEL, urban, and country- and pandemic-fixed effects (as dummies). EDR is calculated as the pandemic crude death rate divided by the last pre-pandemic year's rate in order to account for countries with higher mortality levels regardless of the pandemic. Initial level of income has to be controlled because poorer countries exhibit faster growth by virtue of standard growth models. Two potential issues. The first is that of endogeneity. It might be that governments respond by stimulating the economy, however in none of these events did they enact a measure of stimulus. The second is that the data is restricted to a small number of countries, and those which are generally the freest and richest countries. Because of such restriction in variances in EF and income, the unbiased estimates are likely to be stronger. Generally, studies using EF measures tend to find larger effects on least developed countries.

Results from panel regression show that one extra point in the 10-point scale HIEL increases GDP between 1.53 and 1.79 % points across models. These are models with EDR, with and without fixed effects, and even after discarding the 2009 pandemic which was a minor one. The main limitation of the study is the use of crude rather than cause-specific death rates, which unfortunately capture some of the adjusting behavior of individuals who minimize infection risk.

Geloso & Bologna-Pavlik (2021) analyze the impact of EF on the 1918 flu pandemic across 43 countries. This episode, which compounded the adverse effect of the Great War of 1914-1918, was marked by a sharp contraction followed by a quick recovery, yet is considered as one of the top 5 economic shocks of the 20th century. The regression model includes (lagged) death rates, freedom estimated with HIEL, HIEL*death rate interaction, democracy score, as well as initial levels of GDP per capita and year- and country-fixed effects. Three separate regressions are estimated, each time using a different death rates measure: death from the flu, the Great War and both, lagged in one year. Controlling for initial levels of GDP allows one to estimate GDP growth instead of GDP per capita. Consolidated democracy was measured with "Polity" variable.

The coefficient of whichever death rates measure on either real GDP per capita or GDP growth is strongly negative, evidencing its negative impact. The interaction of death with HIEL is however positive, indicating that HIEL moderates the impact of death on GDP and GDP growth. This holds true regardless of the death rates measures being used. When looking separately at HIEL components for the GDP growth equation, the interaction term is only significant when international trade or regulations are involved, which indicate they are important factors of the crisis recovery. In order to test for the competing role of democracy, the regression was rerun using death rates, HIEL, HIEL*death, Polity and Polity*death. The coefficient of HIEL*death rates was substantially larger than Polity*death.

Unsurprising

Government spending comes at a price. Indeed, Connors & Norton (2012) found that the negative relationship between government spending and economic growth holds within countries as well, and not just across countries, during the 1960-2010 period. Interestingly, the end of that period is marked by lower growth due to recession, yet they notice (see their Table 3) that “the countries with smaller governments had a much smaller reduction in growth rates”. Reducing public spending helps the recovery. This is likely because public spending creates misallocation of resources.

That freedom promotes efficient reallocation of resources is within the prediction of the ABCT. Allowing individual preferences to quickly shift towards more urgent needs, e.g., shorter time preferences, promotes growth that is not based on time mismatching. The kind of which happens after the Keynesian stimulus, assuming it works as intended. As explained before, the cure of a recession is flexibility in labor market while allowing for the availability of idle resources as having higher expected return in the future.