Financial Liberalization, Crises, and Economic Growth

Financial Liberalization and Economic GrowthFinancial Liberalization and Economic Growth

Financial Liberalization, Crises, andEconomic Growth

Inkoo LeeDepartment of InternationalMacroeconomics and FinanceKorea Institute for InternationalEconomic Policy300-4 Yomgok-dongSeocho-gu, Seoul [email protected]

Jong-Hyup ShinDepartment of InternationalMacroeconomics and FinanceKorea Institute for InternationalEconomic Policy300-4 Yomgok-dongSeocho-gu, Seoul [email protected]

Asian Economic Papers 7:1 © 2008 The Earth Institute at Columbia University and the Massachusetts

Institute of Technology

AbstractThe paper computes the effect of financial liberalization on eco-nomic growth by combining the results of a panel model withthose of a probit model. It finds a positive net effect from financialliberalization to growth. Surprisingly, we find that the net effect ongrowth is larger in the crisis-experienced country group than inthe overall sample group. Our guess is that the crisis-experiencedcountries are mostly developing countries that usually enjoyhigher growth rates than the developed countries because of thecatching-up phenomenon. The paper also studies the link be-tween financial liberalization and nominal interest rates, and finds,contrary to expectations, that the direct liberalization effect ispositive. Our guess is that this reflected the overshooting of inter-est rates after crises.

1. The impact of ªnancial liberalization

This paper investigates the impact of ªnancial liberaliza-tion on economic growth and the nominal interest rate. Fi-nancial liberalization affects economic growth throughtwo channels: the direct liberalization effect and the indi-rect crisis effect. The direct liberalization effect is positivebecause ªnancial liberalization removes many frictions inªnancial markets and reduces borrowing costs. The indi-rect crisis effect is negative because ªnancial liberalizationmay also cause ªnancial crises. Our deªnition of a crisis isthe simultaneous occurrence of a banking crisis and a cur-rency crisis, that is, a “twin crisis.” The sum of these twoeffects yields the net effect of ªnancial liberalization oneconomic growth.

The direct liberalization effect is usually estimated by re-gressing economic growth on explanatory variables that

include a ªnancial liberalization dummy (or a ªnancial liberalization index), and acrisis dummy (Rodrik 1998; Bekaert, Harvey, and Lundblad 2005; Henry 2000, 2006).The indirect crisis effect is the product of (1) the marginal effect of ªnancial liberal-ization on the probability of a crisis (Kaminsky and Reinhart 1999; Glick and Hutch-inson 2001) and (2) the coefªcient of the crisis dummy in the previous panel regres-sion.

The impact of ªnancial liberalization on the interest is studied by replacing the out-put growth rate with the interest in the panel equation and probit equation in sec-tion 2.

2. The empirical relationship between ªnancial liberalization and growth

We use a two-step approach to estimate the effect of ªnancial liberalization on eco-nomic growth. The ªrst step makes use of a probit model to measure the marginaleffect of ªnancial liberalization on the probability of a crisis.

Crisisit � c � aFLit � ZitA � �it , (1)

where subscripts i and t denote an individual country and a time period, respec-tively. Crisisit represents a twin crisis dummy variable that is equal to one if the twincrisis occurs in country i at time t, and zero if the crisis does not occur. FLit is theªnancial liberalization dummy variable, Zit is a set of explanatory variables, and �it

is an error term such that �it _ N(0,1). The set of explanatory variables Zit consists ofthe bank liquid reserve ratio and the inºation rate. A, c, and a are constant parame-ters.

The marginal effect of ªnancial liberalization on the probability of a crisis is givenby the change in the predicted probability of crises occurrence that is conditional onthe (discrete) explanatory dummy variable:1

Marginal Effect � Pr(Crisisit � 1|Zit , FLit � 1) � E(Crisisit � 1|Zit , FLit � 0). (2)

In the second step, we run a panel regression model after testing the signiªcance ofthe group effect.

yit � � � �FLit � �Crisisit � Xit � it, (3)

107 Asian Economic Papers

Financial Liberalization and Economic Growth

1 For computing marginal effects, one can evaluate the expressions at the sample means of thedata.

where yit is a dependent variable that is the real per capita GDP growth rate, and Xit

is a set of explanatory variables that includes log(initial real GDP), inºation, tradeopenness,2 and the investment/GDP ratio. �, �, �, and are coefªcients on depend-ent variables, and it is an innovation. As the F-test cannot reject the null hypothesisthat the constant terms are all equal, we choose a random effects model as thebenchmark regression model.

The coefªcient of the ªnancial liberalization dummy (�b) in the panel regression rep-resents the direct liberalization effect. The indirect crisis effect is computed by com-bining the results of the probit regression with those of the panel regression: thecoefªcient on the crisis dummy in the panel regression is multiplied by the marginaleffect of ªnancial liberalization from the probit regression, ( �g � Marginal Effect). Thenet effect of ªnancial liberalization is given by:

E(yit | FLit � 1) � E(yit | FLit � 0) � �b � �g � Marginal Effect. (4)

2.1 DataWe use the annual data of 58 countries between 1980 and 1999, with the numbervarying with the availability of data for each regression. Because most banking andcurrency crises have occurred during 1980s and 1990s, we believe that the sampleperiod is not too short to catch the effects of ªnancial liberalization and the crisis oneconomic variables. We borrow the twin crisis dummy from Ranciere, Tornell, andWestermann (2006), who calculate this dummy using data from Caprio andKlingebiel (2003) and Glick and Hutchison (2001). The investment/GDP ratio comesfrom the Penn World Table.

We have two proxies for the ªnancial liberalization variable. The ªrst proxy is theªnancial liberalization dummy from Beckaert, Harvey, and Lundblad (2005), whichis based on the dates of ofªcial equity-market liberalization in each country, and wecall this variable FL(1). The second proxy is the ªnancial liberalization index fromWyplosz (2001) who made this index by combining the results of Demirguc-Kuntand Detragiache (1998), Mehrez and Kaufmann (2000), and Wyplosz (2001)—andwe call this variable FL(2). We call the sub-sample of countries that have values forboth FL(1) and FL(2) the FL-sample. The rest of the variables are from InternationalFinancial Statistics.

In addition to estimating the models over the full sample, the 58 countries that con-tain values for FL(1), we also conducted the estimations on three sub-samples: the



2 Openness is measured by (export � import)/GDP.

21 countries that experienced a twin crisis (the “crisis-experienced” countries), the11 Asian countries, and the FL sub-sample. The Appendix provides the list of coun-tries in the full sample and in the three sub-samples.

3. The results

3.1 The direct liberalization effect on economic growthThe upper panel in Table 1 reports the coefªcient estimates of the ªnal speciªcationsof equation (3), after dropping some of the statistically insigniªcant variables (e.g.,the government size variable). The ªnal list of regressors is the ªnancial liberaliza-tion variable, the twin crisis dummy, the initial real GDP, inºation, trade openness,and the investment/GDP ratio. Financial liberalization, trade openness, and the in-vestment/GDP ratio are positively associated with economic growth, whereas thecrisis dummy, the initial GDP, and inºation are negatively related to economicgrowth. The trade openness variable is generally insigniªcant, and is dropped in theestimation of the FL-sample.

The point estimates of the FL dummy, FL(1), and the FL index, FL(2), suggest a sub-stantial impact of ªnancial liberalization on output growth through the deepeningof a country’s ªnancial system. For example, ªnancial liberalization, FL(1), appearsto increase the annual GDP growth rate by 1.0 percentage point for the whole sam-ple, 1.3 percentage points for crisis-related countries, and 0.6 percentage points forAsian countries. The coefªcients are signiªcant at the 1-percent level except for theAsian countries sub-sample. The size of the coefªcient for FL(2) is similar that forFL(1).

The occurrence of the twin crisis decreases annual GDP growth rate by 4.2 percent-age points for the whole sample, 4.3 percentage points for the crisis-experiencedcountries, and 6.1 percentage points for Asian countries. The use of FL(2) as the al-ternative measure of ªnancial liberalization does not change the results much, themarginal effect of the twin crisis is 4.4 percentage points. The coefªcients for the cri-sis dummy are signiªcant at the 1-percent level in all cases.

3.2 Financial liberalization and crisesThe lower panel in Table 1 presents the results of probit regressions of equation (1).Only the FL dummy variable is consistently statistically signiªcant across samplesand across speciªcations. The bank liquid reserve ratio and inºation turn out to beinsigniªcant in the incidence of crises. We drop the investment/GDP ratio and theinitial real GDP as independent variables because they are not statisticallysigniªcant.





Table 1. Financial liberalization and economic growth

Panel Regression

Dependent Variable: Per Capita Real GDP Growth

Allcountries

Crisis-experiencedcountries

Asiancountries FL-samplee FL-samplef

Financial Liberalization Dummy, FL(1)a 0.010(3.67)***

0.013(2.92)***

0.006(1.36)

0.011(3.30)***

Financial Liberalization Index, FL(2)b 0.008(1.91)*

Twin Crisis Dummyc �0.042(�7.68)***

�0.043(�6.82)***

�0.061(�6.63)***

�0.045(�6.72)***

�0.044(�6.51)***

Initial Real GDP �0.055(�5.23)***

�0.040(�2.18)**

�0.150(�7.05)***

�0.078(�5.61)***

�0.070(�5.06)***

Inflation �0.002(�4.74)***

�0.002(�4.21)***

�0.208(�7.33)***

�0.002(�4.60)***

�0.002(�4.44)***

Trade opennessd 0.008(1.77)*

�0.001(�0.07)

�0.000(�0.03)

Investment/GDP 0.176(8.36)***

0.211(6.47)***

0.311(8.95)***

0.214(7.87)***

0.222(8.06)***

R2 0.145 0.263 0.479 0.234 0.222

F-statistic �1.324 — — — —

Probit Regression

Dependent Variable: Twin Crisis Dummy

All Crisis Asia FL sample FL sample

Financial Liberalization Dummy, FL(1) 0.272(1.70)*

0.435(2.37)***

0.779(2.25)**

0.595(1.99)**

Financial Liberalization Index, FL(2) 0.362(1.73)*

Bank Liquid Reserve Ratio �0.303(�0.44)

�0.111(�0.12)

�5.628(�1.79)*

1.395(1.46)

1.271(1.35)

Inflation 0.017(1.03)

�0.001(�0.04)

2.405(1.46)

0.014(0.81)

0.010(0.59)

Constant �1.947(�12.40)***

�1.527(�8.30)***

�1.831(�4.62)***

�2.186(�7.59)***

�1.992(�9.04)***

R2 0.005 0.016 0.063 0.010 0.007

Number of Observations 1,160 420 220 540 540

Number of countries 58 21 11 27 27

Marginal Effect of Financial Liberalization 0.020 0.073 0.062 0.054 0.036

Note: t-statistics are in parentheses.

***, **, and * indicate the significance at the 1%, 5%, and 10% level, respectively.

a. FL(1) represents the financial liberalization dummy which is based on the dates of official equity-market liberalization in each coun-

try.

b. FL(2) is the financial liberalization index from Wyplosz (2001).

c. Dummy variable indicating a currency crisis and a banking crisis.

d. Ratio of exports and imports to GDP.

e. This column represents results when FL(1) is used as the regressor.

f. This column represents results when FL(2) is used as the regressor.

The last row of Table 1 reports our computations of the change in the predictedprobability that is conditional on the (discrete) explanatory dummy variable,namely, the marginal effects of ªnancial liberalization on the probability of a crisis.Financial liberalization is expected to increase the probability of crises by 2.0 per-centage points in the full sample, 7.3 percentage points for the crisis-experiencedcountries, and by 6.2 percentage points for the Asian countries. For the FL-sample,the probability of crisis is increased by 5.4 percentage points when FL(1) is used,and by 3.6 percentage points when FL(2) is used.

3.3 The net effect of ªnancial liberalization on economic growthTable 2 reports the net effect of ªnancial liberalization calculated according to equa-tion (4). In every case, ªnancial liberalization leads to a positive net effect on eco-nomic growth. The net effect ranges from a low of 0.22 percentage points in Asia toa high of 0.99 percentage points in crisis-related countries. Surprisingly, the net ef-fect in the “crisis-experienced countries” sub-sample is greater than that in thewhole sample. We guess this is because the crisis-experienced countries are mostlydeveloping countries, and the developing countries usually enjoy higher growthrates than developed countries because of the catching-up phenomenon. Or, the lib-eralization effect is simply just greater in developing countries.

Overall, the results in this section conªrm our view that ªnancial liberalization con-tributes to economic growth through an increase in ªnancial deepness, even aftercontrolling for the probability of crisis occurrence. Our results are similar to those ofRanciere, Tornell, and Westermann (2006) in that our estimate of the net effect ofªnancial liberalization on economic growth is 0.92 percentage points in the full sam-ple and their estimate is 0.86–0.91 percentage points.

3.4 Financial liberalization, crises, and interest ratesThere are two links between ªnancial liberalization and interest rates. On the onehand, one would expect that interest rates will increase due to the borrowing restric-tion in ªnancial markets after crises have occurred. On the other hand, the effect ofªnancial liberalization on the interest rates is likely to be negative because ªnancialliberalization makes it easier to borrow money from abroad.



Table 2. Liberalization and crisis effects on economic growth

All (%) Crisis (%) Asia (%) FL(1) (%) FL(2) (%)

Liberalization effect �1.00 �1.30 �0.60 �1.10 �0.80Crisis effect �0.08 �0.31 �0.38 �0.24 �0.16Net effect �0.92 �0.99 �0.22 �0.86 �0.64

The results in Table 3 show that the crisis dummy has a signiªcantly positive rela-tionship with nominal interest rates as expected, and the FL dummy is also posi-tively related to nominal interest rates, contrary to our expectation. It turns out thatªnancial liberalization directly increases the interest rate by 1.9 percentage points,and the resulting net effect of ªnancial liberalization amounts to 2.05 percentagepoints (Table 4).

The unexpected sign of the liberalization effect can be explained by some combina-tion of the following four considerations. First, it is possible that the interest rateovershoots after crises, which may reverse the true (negative) liberalization effect byscaling up the average interest rate over the entire period. Second, the removal of



Table 3. Financial liberalization and interest rates

Panel Regression

Dependent Variable: Nominal Interest Rate

Financial Liberalization Dummy, FL(1)a 0.019(3.76)***

Twin Crisis Dummyb 0.056(6.93)***

Real GDP �0.009(�3.90)***

Inflation 0.609(34.45)***

Trade opennessc �0.042(�4.40)***

R2 0.682

F-statistic �7.513

Probit Regression

Dependent Variable: Twin Crisis Dummy

FL Dummy 0.379(1.59)

Bank Liquid Reserve Ratio 2.277(1.86)*

Inflation 0.033(1.29)

Constant �2.207(�8.84)***

R2 0.009

Number of Observations 680

Number of Countries 34

Marginal Effect of Financial Liberalization 0.026

Note: t-statistics are in parentheses.

a. FL(1) represents the financial liberalization dummy which is based on the dates of official equity-market liberalization in each coun-

try.

b. Dummy variable indicating a currency crisis and a banking crisis.

c. Ratio of exports and imports to GDP.

***, **, and * indicate the significance at the 1%, 5%, and 10% level, respectively.

ªnancial restraints may not improve allocation efªciency in credit markets becauseof imperfect information. Third, external factors such as the high and volatile worldinterest rate may have been transmitted to the domestic country that has undergoneªnancial liberalization. Last, but not least, ªnancial repression usually includes aninterest rate ceiling that is set below the market-clearing interest rate; and soªnancial liberalization will cause the interest rate to rise.

4. Conclusions

It has been widely recognized that ªnancial liberalization plays an important role ineconomic development. Although an expanding body of literature has documentedthis effect across space and time, the channel through which ªnancial liberalizationaffects the economic growth remains unclear. This paper employs probit and panelregressions to show that ªnancial liberalization is positively associated with eco-nomic growth, with the (positive) direct liberalization effect dominating the (nega-tive) indirect crisis effect. Financial liberalization is expected to increase GDPgrowth by 0.92 percentage points in the whole sample and by 0.99 percentage pointsin crisis-experienced countries, respectively. Financial liberalization contributes toeconomic growth even when the sample is restricted to countries that experiencedªnancial crises.

We also test the effect of ªnancial liberalization on the interest rates. The result iscontrary to our expectation: ªnancial liberalization increases the interest rates. Weconjecture that the overshooting in interest rates after a crisis and the removal of in-terest rate ceiling after liberalization are the main reasons for this phenomenon.

ReferencesBekaert, Geert, Campbell Harvey, and Christian Lundblad. 2005. Does Financial LiberalizationSpur Growth? Journal of Financial Economics 77:3–56.

Caprio, G., and D. Klingebiel. 2003. Episodes of Systemic and Borderline Banking Crises.World Bank Working Paper. Washington, DC.

Demirguc-Kunt, A., and E. Detragiache. 1998. Financial Liberalization and Financial Fragility.IMF Working Paper. Washington, DC.



Table 4. Liberalization and crisis effects on interest rates, %

Liberalization effect 1.90Crisis effect 0.15Net effect 2.05

Glick, Reuven, and Michael Hutchinson. 2001. Banking Crises and Currency Crises: How CommonAre Twins? New York: Cambridge University Press.

Henry, Peter. 2000. Stock Market Liberalization, Economic Reform, and Emerging Market Eq-uity Prices. Journal of Finance 55:529–564.

Henry, Peter. 2006. Capital Account Liberalization: Theory, Evidence, and Speculation. NBERWorking Paper No. 12698. Cambridge, MA: National Bureau of Economic Research.

Kaminsky, Graciela, and Carmen Reinhart. 1999. The Twin Crises: The Causes of Banking andBalance of Payment Problems. American Economic Review 89 (3):473–500.

Mehrez, G., and D. Kaufmann. 2000. Transparency, Liberalization, and Bank Crises. WorldBank Working Paper. Washington, DC.

Ranciere, Romain, Aaron Tornell, and Frank Westermann. 2006. Decomposing the Effects of Fi-nancial Liberalization: Crises vs. Growth. NBER Working Paper No. 12806. Cambridge, MA:National Bureau of Economic Research.

Rodrik, Dani. 1998. Who Needs Capital Account Convertibility? In: Should the IMF Pursue Capi-tal Account Convertibility?, Essays in International Finance, edited by Stanley Fischer, p. 207.Princeton, NJ: Princeton University.

Wyplosz, Charles. 2001. How Risky Is Financial Liberalization in the Developing Countries?G-24 Discussion Paper Series.



Appendix



Table A.1 The countries in each sample used in the estimations

CountryAllcountries

Crisis-experiencedcountries

Asiancountries FL-sample

Sample forstudying effectson interest rates

Algeria — —Argentina — — —Australia — — —Austria — — —Bangladesh — —Belgium — — —Brazil — — —Canada — —Chile — — —Colombia — —Costa Rica —Cote d’Ivoire — —Denmark — —Dominican Republic —Ecuador — — —Egypt — —El Salvador —Finland — — —France — — —Germany — —Ghana —Greece —Guatemala —Honduras —India — — — —Indonesia — — — — —Ireland — —Israel — — —Italy — — —Jamaica —Japan — — — —Jordan — —Kenya — —Korea — — — — —Malaysia — — — — —Mexico — — — —Morocco —Netherlands — —New Zealand — — —Nigeria —Norway — — —Pakistan — — —Paraguay —Peru — — —Philippines — — — — —Portugal — —South Africa — — —Spain — — —Sri Lanka — — — —Sweden — — —Switzerland — —Thailand — — — — —Tunisia — —Turkey — — —United Kingdom — — —United States — —Uruguay — —Venezuela — — —

Note: The regression in each column includes countries with “—” mark.

Documents

Financial Liberalization, Crises, and Economic Growth