Financial Liberalization and Growth: Empirical Evidence

Transcription

1 Financial Liberalization and Growth: Empirical Evidence Arturo Alejandro Guillermo Galindo Micco Ordoñez 1 Arturog@iadb.org Alejandromi@iadb.org Guillermoo@iadb.org Inter-American Development Bank May, 2002 Abstract This paper examines whether financial liberalization promotes economic growth by analyzing its effects on the costs of external financing to firms. Following Rajan and Zingales (1998) methodology, we explore whether economic sectors that rely more on external finance grow faster than others after financial liberalization. Using this methodology and Kaminsky and Schmukler s (2001) dataset on financial reforms, the paper shows that financial liberalization, mainly in the domestic financial sector, increases growth rates of economic sectors intensive in external funding relative to other sectors. Our results are robust to the fact that financial reform usually comes with other structural changes (e.g., trade liberalization). 1 We thank Graciela Kaminsky and Sergio Schmukler for sharing their data with us. We also thank Arturo Bris, Florencio Lopez de Silanes, Andrea Repetto and participants at the Economia Panel in Cambridge, MA, as well as participants at Banco de la República de Colombia s brown bag seminar for useful comments and suggestions. The opinions in this paper reflect those of the authors and not necessarily those of the IADB.

2 1. - Introduction Since the 1970s several countries have liberalized their financial systems. Liberalization has been characterized by opening the capital account and removing financial repression policies and restriction to foreign ownership. A crucial question that deserves attention is whether these liberalizations have somehow contributed to growth. Based on Rajan and Zingales (1998), this paper examines whether financial liberalization reduces the cost of external finance to firms and promotes growth. In particular, we ask whether firms that are relatively more in need of external finance develop disproportionately faster in more liberalized countries and during periods of high financial liberalization. We find this to be true in our sample of 28 countries during the 1970s, 1980s and 1990s. By the early 1970s many developing countries appeared to have slipped into financial repression. Restrictions took the form of deposit and loan interest rate controls, directed credit policies, and exchange and capital controls, among others. These policies distorted the functioning of financial systems. According to Díaz- Alejandro (1985, p. 7) financial repression leads to: segmented domestic financial markets in which some obtained credit (rationing) at very negative real interest rates, while non-favored borrowers had to obtain funds in expensive and unstable informal credit markets. Public controls over the banking system typically led to negative real interest rates for depositors. Financial repression became an obstacle to domestic savings and their efficient allocation, and financial intermediation languished. 2

3 In response to adverse effects of financial restrictions, many countries have engaged in liberalization strategies since the mid-1970s. Liberalization was characterized by opening the capital account and by lifting restrictions on the domestic financial system. With respect to the capital account, liberalization policies included allowing corporations to borrow abroad and removing multiple exchange rate mechanisms and other sorts of capital controls. Regarding the domestic financial system, policies were aimed at removing interest rate controls (lending and deposits), lifting other restrictions such as directed credit policies or limitations on foreign currency deposits, allowing foreigners to own domestic equity, and removing restrictions on repatriation of capital, dividends, and interest. Recent data on financial liberalization collected by Kaminsky and Schmukler (2001) shows how liberalization policies were adopted in several regions of the World (Graph 1). Higher values indicate more liberalization. Graph 1: Financial Liberalization G7 Average Liberalization Index Europe Notes: Financial Liberalization is based on the indicators developed in Kaminsky and Schmukler (2001). We take the simple average of liberalization in the capital account, the domestic financial system, and the stock market. This measures ranges from 1 to 3, where 3 is full liberalization. Average Liberalization Index is calculated as the simple average between each Region s countries in each year. LAC Asia 3

4 By the early 1970s the financial sector of most developed countries was already significantly liberalized, while in most developing countries it was highly repressed. The liberalization process was continuous and gradual in most regions, except in Latin America, where an important reversal was observed in the 1980s. At the beginning of the 1970s liberalization was rapid in Latin American countries. This was mainly driven by the southern cone countries, which had engaged in laissez- faire financial policies that mainly supported unrestricted private participation in financial markets without direct government regulation. As noted by Díaz-Alejandro (1985) this led to massive bankruptcies and a generalized financial crisis throughout the region. Countries then abandoned laissez-faire practices, introducing tighter regulations and restrictions to their financial systems. This came with a de facto nationalization of the banking sector. Later, at the beginning of the 1990s, Latin America engaged once again in the liberalization strategy. The main difference with respect to the previous liberalization wave was the implementation of regulatory and supervision mechanisms to avoid the previous type of crisis. A crucial question that deserves attention is whether liberalization has somehow contributed to growth and through which channels. The literature has no clear answer to this question, either from a theoretical or an empirical point of view. Models of perfect markets suggest that financial liberalization should benefit both borrowers and lenders. Starting with McKinnon (1973) and Shaw (1973), some studies have pointed out that lifting financial restrictions can exert a positive effect on growth rates as interest rates rise toward their competitive market equilibrium and resources are allocated efficiently. In addition, capital account liberalization allows firms to access cheaper foreign funds and puts pressure on the domestic 4

5 financial system to improve its efficiency. However, some authors claim that this efficient-markets paradigm is misleading when applied to financial sector, in particular to capital flows. Removing one distortion need not be welfare enhancing when other distortions are present. Financial markets are characterized by serious problems of asymmetric information and moral hazard issues that may undermine the case for domestic financial liberalization (see next section for a complete discussion of these arguments). While theory does not give us a clear answer on the effect of financial liberalization on growth, empirical studies are not conclusive, either. This is particularly true in the case of capital account liberalization, which remains one of the most controversial and least understood economic policies. 2 The first reason for this disappointing result is the lack of good and homogenous measures of financial liberalization policies across countries and over time. A second serious issue in virtually all studies of financial liberalization is the omitted variable problem, as the financial liberalization process tends to be imposed and removed as part of large a package of policies. Trade liberalization and the privatization of state enterprises, for example, have coincided with the liberalization of the financial sector in many countries. As shown in Graph 2, Latin America is a clear example of this. Therefore, it is crucial to control for all the other policy measures before to be able to conclude about the effect of financial liberalization on growth. For the case of capital account liberalization, Eichengreen (2001) pointed out the main challenges of this empirical literature: 2 See Eichengreen (2001) for a survey. 5

6 Developing adequate measures of capital-account restriction is a particular problem for the literature on the causes and effects of capital controls, but the more general problem of adequately capturing the economic, financial and political characteristics of economies should not be overlooked. (Eichengreen(2001), p.11). Graph 2: Reforms in Latin America Tariffs on Trade and privatizations Tariffs on Trade Privatizations Financial Liberalization Financial Liberalization Notes: Financial Liberalization is based on the indicators developed in Kaminsky and Schmukler (2001). We take the simple average of liberalization in the capital account, the domestic financial system, and the stock market. This measures ranges from 1 to 3, where 3 is full liberalization. Privatizations are measured as the cumulative value of sales and transfers of public companies as a proportion of GDP for each year. Source: Lora (2001). Tariffs on Trade are referred to average tariff. Source: World Bank, WDI (2001). Our paper is able to deal with the previous two issues. Kaminsky and Schmukler (2001) recently produced a homogeneous dataset including several dimensions of financial liberalization for 28 developed and developing countries during 1972 and This is the ideal data not only for analyzing whether liberalization is valuable for financial development and growth, but also for providing specific policy recommendations. Using Rajan and Zingales (1998) 3 For each index, they analyze up to two dimensions of liberalization and allow for different degrees of liberalization. Unlike other dataset, theirs allows for potential reversal of liberalization policies. 6

7 methodology, the panel structure of the data allows us to identify the effect of financial liberalization on growth controlling for most of the other reforms implemented at the same time. The theoretical underpinning of this relationship is rather simple and follows the discussion of Rajan and Zingales (1998) on how financial development can impact growth. They show that more developed financial systems can mitigate the extent to which firms are affected by problems of moral hazard and adverse selection, and hence can reduce the cost of loanable funds. This reduction in the cost of external funds allows firms that depend more on them to grow faster. This differential in growth across firms is the identification mechanism that allows us to isolate the effect of financial liberalization on growth from the other reforms. In this setup, we assume that the effect of these other reforms on firm growth is uncorrelated with firm external financial dependence. The rest of the paper is organized as follows. Section 2 presents a literature review. Section 3 presents the data and the econometric framework. Section 4 shows the evidence on the impact of financial liberalization on growth. Section 5 explores the relationship between liberalization and financial sector development. Section 6 concludes Literature Review Since the pioneering contributions of Goldsmith (1969), McKinnon (1973) and Shaw (1973), the literature has focus on the relationship between financial liberalization and economic growth. Financial liberalization may affect growth through three mechanisms. First it may affect the development of the domestic 7

8 financial system in terms of size and efficiency, second it may affect the access of domestic firms to foreign funds and finally it may reduce agency problems improving corporate governance. Several authors, starting with McKinnon (1973) and Shaw (1973), have pointed out advantages of reducing financial repression in the domestic financial system. They conclude that lifting financial restrictions can exert a positive effect on growth rates as interest rates rise toward their competitive market equilibrium. According to this tradition, ceilings on interest rates reduce savings, capital accumulation, and discourage efficient allocation of resources. Additionally, McKinnon points out that repression can lead to dualism in which firms that have access to subsidized funding will tend to choose relatively capital-intensive technologies, whereas those not favored by policy will only be able to implement high-yield projects with short maturity. Since then, many other authors have noted additional benefits of financial liberalization. Foreign competition, in the form of entry of foreign institutions to the local market (foreign banks) seems to develop the domestic financial sector and reduces the costs of capital to domestic firms. Competition increases efficiency in the domestic banking industry (see Claesens et al. (2001)) and allows countries to benefit from frontier financial technologies as in Klenow and Rodríguez-Clare (1997), and leads to increased growth. Moreover, foreign competition in the form of access to foreign financial markets seems to reduce the costs of capital to domestic firms (cross-listing see Bekaert et al (2001)). This leads to higher investment and growth (Bekaert and Harvey (2000) and Henry (2000a and 2000b)). 8

9 In addition, financial liberalization puts pressure to improve corporate governance. Foreign competition may force the adoption of international accounting and regulatory standards. These improvement may reduce the agency costs that make it harder and more expensive for firms to raise funds in both the banking sector and the securities market (see Claesens et al. (2001), Galindo et al (2001), Stulz (1999) and Moel (2001)). Laeven (2000) shows that financial liberalization can ease financial constraints of the type described by Hubbard (1997) and Gilchrist and Himmelberg (1998). This happens through two mechanisms. A direct one, by allowing more capital -namely foreign- to become available, and an indirect one through an improvement in the allocation of resources towards the sectors that were previously more constrained. Tapping international capital markets can have significant impacts derived from accessing a larger stock holder base, increased liquidity of stocks, and the signaling effects of adopting international accounting standards, among other possibilities (Sarkissian and Schill (2001)). In summary the benefits of financial liberalization can be grouped into increased access to domestic and international capital markets, and increased efficiency of capital allocation. Critics of financial liberalization policies have argued that the efficientmarkets paradigm is fundamentally misleading when applied to capital flows. In the theory of the second best, removing one distortion need not be welfare enhancing when other distortions are present. If capital account is liberalized while importcompeting industries are still protected for example, or if there is a downwardlyinflexible real wage, capital may flow into sectors in which the country has a 9

10 comparative disadvantage, implying a reduction in welfare (Brecher and Diaz- Alejandro (1977) and Brecher (1983). In addition, if information asymmetries are endemic to financial markets and transactions, in particular in countries with poor corporate governance and low legal protections, there is no reason to think that financial liberalization, either domestic or international, will be welfare improving (Stiglitz (2000)). Moreover, in countries where the capacity to honor contracts and to assemble information relevant to financial transactions is least advanced, there can be no presumption that capital will flow into uses where its marginal product exceeds its opportunity cost. Stiglitz (1994) argues in favor of certain forms of financial repression. He claims that repression can have several positive effects such as: improving the average quality of the pool of loan applicants by lowering interest rates; increasing firm equity by lowering the price of capital; and accelerating the rate of growth if credit is targeted towards profitable sectors such as exporters or sectors with high technological spillovers. However, these claims can be doubtful given that they increase the power of bureaucrats, who can be less capable than imperfect markets, to allocate financial resources. This approach also opens doors to rent-seeking activities. Focusing on the development of the domestic financial sector, capital account liberalization that allows firms to list abroad has been attributed to lead to market fragmentation, and reduce the liquidity in the domestic market and inhibiting its development. The inverse correlation between the amount of aggregate capital raised by LAC corporations in the ADR market and the number of Initial Public Offerings in Latin America seems to corroborate this argument (Moel, 2001). 10

11 Finally, liberalization has also been linked to macroeconomic instability. The financial reforms carried out in several Latin American countries during the 1970s, aimed at ending financial repression, yet led to financial crises characterized by widespread bankruptcies, massive government interventions, nationalization of private institutions and low domestic saving (Diaz-Alejandro (1985)). It is worth noting that, as shown by Demirguc-Kunt and Detriagiache (1998), the likelihood of a crisis following liberalization is decreasing in the level of institutional development. In this sense, the arguments that Stiglitz (1994) makes in favor of government intervention in financial markets in the form of prudential regulation and supervision are convincing. The main argument is that the government is, de facto, the insurer of the financial systems, and hence a financial collapse can have significant fiscal repercussions. To the extent that regulation accompanies liberalization, negative impacts such as those noted by Díaz-Alejandro can be contained. The extent to which financial liberalization affects growth remains an empirical question. Several empirical pieces have tried to address it. Researchers have followed two distinct empirical approaches. One proxies financial liberalization with outcome variables, and the other focuses on explicit policy measures. Regarding outcome variables, several measures have been posed to proxy financial repression. Early empirical literature focused on the value of real interest rates as an indicator of repression. The presumption was that countries with negative real interest rates were financially repressed, while those with positive ones were 11

12 liberalized. 4 In short they find that countries with negative real interest rates exhibit lower growth rates than those with positive ones. However, De Gregorio and Guidotti (1993) claim that real interest rates are not a good indicator of financial repression, and that a better indicator of repression, or lack thereof, is the ratio of credit to the private sector to GDP or other measures of financial development. Several papers have analyzed the impact of this and other measures of financial development on growth. There is consensus that financial development has had a significant positive impact on the growth rates of countries. 5 The extent to which these results can be interpreted as being influenced by financial liberalization is dubious. On the one hand, and as noted by Rajan and Zingales (1998) it is unlikely that such empirical strategies are truly identifying the impact of financial development on growth, due to the fact that financial development occurs at the same time that economies go through significant structural transformations. In the case of financial development, Rajan and Zingales (1998, p. 560) note financial development may simply be a leading indicator rather than a causal factor. On the other hand, it is not clear either that financial liberalization has a one to one impact on financial development. Note that it can be the case that financial liberalization has no impact on financial development. Latin America is a clear example of this. Table 1 shows different measures of financial development and liberalization during the past three decades for several regions of the world. As can 4 Lanyi and Saracoglu (1983), Fry (1978), Roubini and Sala-i-Martin (1991), and World Bank (1989) are examples of this approach. 5 Ghani (1992) includes financial development in a Barro type growth equations for 50 developing countries and find significant impacts. De Gregorio and Guidotti find the same for middle and lowincome countries. King and Levine (1993) expand the sample to 77 countries and add several proxies of financial development and find the same results. Levine, Loayza and Beck (1999) and Beck, 12

13 be seen, all regions experience an important impact on their financial systems as they liberalize, with the exception of Latin America, where there is even a decline in the size of the financial system during the liberalization phase. Table 1: Financial Liberalization and Financial Development LIBERALIZATION DEVELOPMENT (% of GDP) Region Decad Domestic Stock Capital Total Credit to Market Fin. M k t A t Private Capitalizatio Asia 70 s S 1.07 t S t31 80 s s Latin America 70 s s s Europe 70 s s s G-7 70 s s s Notes: Data on liberalization is calculated as the simple average between countries of each Region in each decade. For each country the data on liberalization or financial development was calculated as the simple average along each decade s years. Data on Market Capitalization is available only for the 1990s. For the 1970s the sample is from 1973 to 1979, for the 1980s, from 1980 to 1989, and for the 1990s from 1990 to Data on Liberalization are from Kaminsky and Schmukler (2001). Data on Financial Development are from World Development Indicators (WB 2002). Fewer papers have focused on policy measures of financial liberalization. The main reason of this has been the lack of readily available liberalization data. As noted by Eichengreen et al. (1999) gathering adequate policy data has been a constraint. Regarding the direct impact of liberalization on growth, to our knowledge, from a cross-country perspective only Bekaert et al. (2001) have done extensive research on this area. In the context of a dynamic panel of developed and developing countries they find that stock market liberalization has had permanent impacts on growth. Other papers have focused on indirect transmission channels of liberalization to growth using policy measures. Laeven (2000) constructs liberalization indexes for 13 developing countries since the late 1980s and finds that Levine, and Loayza (2000) follow this approach in the context of a dynamic panel and confirm the previous results. 13

14 the liberalization process in general has eased financial constraints faced by large firms in these countries. Galindo et al. (2001) use Laeven s data and find that financial liberalization increases the allocative efficiency of investment. As noted above, these papers are also subject to the identification critique. According to Fry (1995, p. 179), the simultaneity of reforms appears binding for researchers: in practice, however, most clear cut cases of financial liberalization were accompanied by other economic reforms (such as fiscal, international trade, and foreign exchange reforms). In such cases it is virtually impossible to isolate the effects of financial components of the reform package. Bekaert et al., as well as Galindo et al., try to isolate the impact of financial liberalization by controlling for other reforms and macroeconomic events taking place simultaneously. Even if this allows some identification it remains unclear if the set of controls is large enough to isolate the effect of financial reform. This paper improves substantially on these issues since it adopts a very clean methodology that allows the isolation of impact of financial liberalization on growth. As noted, there are several cross-country studies that find evidence of a positive effect of various measures of financial development on growth. Rajan and Zingales (1998) survey this literature. To our knowledge, however, no paper addresses the issue of financial liberalization and growth directly from a crossindustry-country panel data perspective. This is an additional contribution of this work. 14

15 3. - Empirical Method and Data Rajan and Zingales (1998) faced problems similar to those mentioned in the previous section when addressing the impact of development on growth. Namely, identifying the true causal impact of financial development in a context of multiple structural reforms rose as a challenge. In order to overcome these problems they propose a test to identify the true impact of financial development on growth. The basis of their test is the presumption that the development of financial markets and institutions reduces the cost of external funds faced by firms by reducing the impact of problems associated with moral hazard and adverse selection. From this perspective the impact of financial development must differ according to the needs of particular firms for external funds. Firms that rely more on external funds will be more impacted by financial development than those that require little capital. Using firm-level data, Rajan and Zingales identify the need for external finance (the difference between investment and operating income) for several industries at the ISIC three and four digit level. Given that the US market is relatively frictionless, they assume that such technological demands for external capital would apply in other countries once market distortions are removed. Using this identification technique, they estimate whether development affects industry growth by interacting several measures of financial development with the industry demand for external funds proxy in a cross-country-industry growth regression in which they control for country and industry-specific characteristics. By controlling for these specific factors the omitted variable problem is significantly reduced. Moreover, by estimating the interactive term they can fully 15

16 identify the impact of financial development as opposed to other events that, as shown in the introduction, can occur simultaneously. Since our question is subject to the same criticisms as those of the relationship between financial liberalization and growth, our empirical methodology follows Rajan and Zingales in a panel data context. We estimate the following empirical model using time series of cross-industry-country data: growth ijt = α * α1shareijt 1 + α 2 fin _ lib jt req j + µ i + λ jt ε ijt (1) where the dependent variable is the growth of real value added of sector i in country j at time t, share ijt-1 is the share of industry i in country j of total value added in manufacturing at the beginning of the period, fin_lib jt is the measure of financial liberalization of country j at time t (in some regressions we use Financial development to replicate Rajan and Zingales), and req i is the requirement of industry i for external funds. Additionally we include µ i,, an industry-specific fixed effect, and λ jt, a country-year effect. Finally, ε ijt is the error term. Our test of how financial liberalization affects growth is mainly on the size and significance of α 2. Note that this methodology is very clean, in the sense that the inclusion of country-year effects corrects for other types of events possible correlated with financial liberalization, such as the general liberalization trend for Latin America during the 1990s. This, combined with the interaction of financial liberalization with the external dependency measure, allows for a full identification of the financial liberalization contribution to sector growth. Our sector value-added data comes from United Nations Statistical Division and covers the 28 countries for which Kaminsky and Schmukler (2001) have 16

17 information about financial liberalization during the period Our measure of sector dependence on external financing is taken from Rajan and Zingales (1998). Their dataset covers 37 industries; 28 of them correspond to three-digit ISIC codes and 9 correspond to four-digit ISIC code (see Data Appendix). The following section shows our empirical results using this methodology and data. 4.- Financial Liberalization and Growth The first two columns in Table 2 replicate Rajan and Zingales (1998) results using our sample in a panel data setup. As usual in this literature, our measure of financial development is either the ratio of private credit to GDP or the ratio of private credit plus stock market capitalization to GDP. The first column shows that sectors with higher external dependency grow faster in countries that have welldeveloped financial sectors as measured by the ratio of the sum of private credit and stock market capitalization to GDP. The second column redoes the same exercise using just the ratio private credit to GDP. The signs of the coefficients are positive and significant. The line labeled differential in growth at the bottom of the table shows the impact, according to our estimation, on growth differentials across sectors and countries. For example, in column 1 this differential is This should be interpreted as follows: the relative growth rate of an industry in the 75 th percentile of external requirements relative to an industry in the 25 th percentile, in a country with high financial development (in the 75 th percentile of financial development), is 0.82 percentage points higher than that in a country with a weak financial sector (25 th 17

18 percentile). These are large numbers if we consider that the average real rate of growth in the sample is around 5.1% and that the median is 3.8%. These results are similar to those obtained by Rajan and Zingales. 6 Table 2: Financial Liberalization and Industry Growth Dependent variable: Annual value added growth (1) (2) (3) (4) Industry's share of total VA in manufacturing in (t-1) (5.36)*** (7.76)*** (7.80)*** (7.35)*** Financial Development (2.62)*** Credit to Private Sector (2.89)*** Total liberalization (2.92)*** Total Liberalization (t-1) (2.42)** Diferential in Growth Observations Number of countries Number of country-years Country Year Dummies Yes Yes Yes Yes Industry Dummies Yes Yes Yes Yes Notes: Dependent variable is the annual real value added growth for each ISIC industry, in each country and in each year. Financial Development is the sum of credit to private sector and stock market capitalization. Each financial development variable is interacted with external dependence and is expressed as percentage of GDP. MA 3 years is referred as the moving average in t, t-1 and t-2. Total Liberalization is measured as the simple average between domestic financial system, stock market and capital account liberalization. Differential in real growth rate measures (in percentage terms) how much faster an industry at the 75 th percentile level of external dependence grows with respect to an industry at the 25 th percentile level when it is located in a country at the 75 th percentile of financial development rather than in one at the 25 th percentile. In the financial development impact we only have 27 countries because we do not have data on credit over GDP for Taiwan, as is discussed in Data Appendix. Errors are measured considering clusters by Country Industry. This was done following Bertrand et al. (2002) for correcting the bias in the estimated standard errors that serial correlation introduces, allowing for an arbitrary covariance structure between time periods. Absolute value of t-statistics in parentheses. (* significant at 10%; ** significant at 5%; *** significant at 1%). The basic conclusion of the paper are presented in columns 3 and 4 that reports results similar to those above but using current and lagged financial 6 The equivalent differential in growth in the case of Rajan and Zingales (1998) is 1.1. For robustness, we use an alternative measure of financial development that mainly recognizes the fact that the impact of development on growth is not instantaneous. Using a moving average of the past three years of the financial development measure, the results are qualitatively the same. 18

19 liberalization instead of financial development as a regressor, respectively. These results suggest that financial liberalization boost growth in sectors with higher external dependency. Column 3 suggests that industries that depend on external financing (75 th percentile) grow 0.96% faster, relative to industries with low external financing dependence, in periods of full liberalization compared to those of partial liberalization. Once again this is a very significant figure given the average values of sector growth in our sample. 7 To obtain the long run effect of financial liberalization we have to divide its coefficient by the lag sector-share coefficient multiplied by (-1). The results in column 3 suggest that, in the long run, industries that depend more on external financing (at the 75 th percentile level) and are located in a fully liberalized economy will have a share over total value added 1.6% higher than the same type of industries located in a country that is only partially liberalized. This is an important effect if we consider that the average sector-share in our sample is 3.3 %. To identify the mechanism through which financial liberalization boosts growth we open our measure in two components. On the one hand we study liberalization on domestic financial system and on stock market, and on the other we analyze the impact of capital account liberalization. As shown in Table 3, what matters in our estimations is the liberalization of domestic financial markets. Unreported results shows that it is a very robust result: domestic financial system liberalization is the key component regarding the impact of liberalization on growth; 7 As robustness checks, we allow for different lag structures of the financial liberalization measures and also different constructions of the liberalization variable (principal components). Results are qualitatively the same. 19

20 capital account liberalization appears to have very little impact on the growth rate of external finance intensive sectors. Note that the growth effect is very similar to that in Table 2, which confirms the claim that liberalization affects growth through its impact on the domestic financial sector that is the only significant component. Table 3 : Institutional Development and Financial Liberalization Dependent variable: Annual value added growth (1) (2) Industry's share of total VA in manufacturing in (t-1) (7.80)*** (7.35)*** Dom. Financial System Lib (2.18)** Dom. Financial System Lib. (t-1) (1.99)** Capital Account Liberalization (1.16) Capital Account Liberalization (t-1) (0.57) Differential in Growth Observations Number of countries Number of yearcty Country Year Dummies Yes Yes Industry Dummies Yes Yes Notes: Dependent variable is the annual value added growth for each ISIC industry, in each country and in each year. Liberalization measures represent the interaction between them and each industry s external financial dependence. Domestic Financial Liberalization is the simple mean between liberalization in domestic financial system and stock market. Differential in real growth rate measures (in percentage terms) how much faster an industry at the 75 th percentile level of external dependence grows with respect to an industry at the 25 th percentile level when it is located in a country at the 75 th percentile of financial development rather than in one at the 25 th percentile. Errors are measured considering clusters by Country Industry. Absolute value of t-statistics in parentheses (* significant at 10%; ** significant at 5%; *** significant at 1%). As we mention in the literature review section, there are reasons to think that the effects of financial liberalization vary with financial and institutional developments. Removing financial controls may be efficiency enhancing only when serious imperfections in the information and contracting environment are absent. Following La Porta et al. (1997, 1998), we explore whether specific institutional variables that have been proven to promote financial development are also 20

21 associated with the impact of liberalization on growth. There are several usual suspects in this literature. The legal origin of law codes, rule of law, creditor protection and effective creditor rights protection 8 are usual candidates for explaining cross-country differences in financial sector development. Presumably, economies that afford weak legal protections to creditors are less likely to benefit from liberalization. The literature has shown that these are dominant features explaining the development of financial markets. If these protections are not in place the liberalization of restrictions on intermediation are dampened by the adverse effects of institutional disarrays and do not promote financial sector development. The absence of legal protections that guarantee the ability of creditors to minimize their financial loss in case of borrower default can counteract the potential efficiency effects that financial liberalization can induce. To test this hypothesis we re-estimate equation [1] including interactive terms between the liberalization measure, and the different proxies for legal protections. Specifically we estimate: growth ijt = α 0 + α1shareijt 1 + α req 4 j * Leg j + α fin _ lib + µ + λ + ε i 2 jt jt ijt * req j + α fin _ lib 3 jt * req j * Leg j (2) where Leg is any of the legal protection proxies mentioned above. α 3 captures the differential effect, if any, of countries with different legal protections. Higher values of Leg indicate greater legal protection. Hence, one would expect α 3 to be positive if our claims are accurate. 8 Effective creditor rights protection is the product of the rule of law index and creditor rights. It intends to capture the extent to which creditor rights regulations can be enforced or not. See Galindo and Micco (2002) for a discussion. 21

22 Table 4 reports our results. Column 1 reports the regression without any interaction with the legal protections. The results are hence equivalent to those above but now focusing only in domestic financial liberalization. Column 2 reports the result including the interaction of the liberalization measure and effective creditor rights. Note that the significance of both terms drops dramatically. However this must not be interpreted as meaning that economic significance of this interaction is lost. The loss of significance is attributed to the fact that both regressors are importantly collinear (0.76). Hence the relevant test is one on joint significance. We report this at the bottom of Table 4. Note that independent of the legal protectionproxy used joint significance is always above standard level. Column 2 thus suggests that liberalization has a greater impact in countries where creditors are more protected. As above, we present the differential growth impact. Now we compare this impact in countries in the 75 th percentile of legal protection with those in the 25 th. The results suggest that the differential impact is higher where creditor protections are in place. Columns 3 and 4 show the same exercises using creditor rights and rule of law instead of effective creditor rights, respectively. The interpretation of results follows those of effective creditor rights. The previous three measures of legal protection have the caveat that they could be endogenous; the need for financial development may create pressure to improve legal protection. 9 To avoid this problem, in column 5 we use legal origins as a proxy for legal protection. 10 As shown by previous research (see La Porta et al. (1998)), common law origin countries tend to have higher protections. This result 9 In addition these measure are only available for Our proxy for legal protection is a dummy variable that is one for countries with common law legal origin and zero otherwise. 22

23 again suggests that higher legal protections magnify the impact of financial liberalization on growth. Table 4: Financial Liberalization and Growth: Interactions with Legal Protections Dependent Variable: Annual Value Added Growth (1) (3) (4) (5) (6) Industry's share of total VA in manufacturing in (t-1) (7.83)*** (7.89)*** (7.86)*** (7.85)*** (7.96)*** Domestic Financial System Lib (3.00)** (1.69)* (1.98)** (0.18) (2.09)** Domestic Financial System Lib (Int. with Efective Cred. Rights) (1.07) Domestic Financial System Lib (Int. with Creditor Rights) (0.92) Domestic Financial System Lib (Int. with Rule of Law) (0.66) Domestic Financial System Lib (Int. with English Legal Origin) (2.08)** Differential in growth 0.9** (Institutional Measure in average) Differential in growth 0.7** 0.7*** 0.5 (Institutional Measure in percentil 25) Differential in growth 1.0*** 1.1** 0.8*** (Institutional Measure in percentil 75) Differential in growth 0.5** (No English Legal Origin) Differential in growth 2.6** (English Legal Origin) Observations Number of countries Number of yearcty Country Year Dummies Yes Yes Yes Yes Yes Industry Dummies Yes Yes Yes Yes Yes F test /2 3.85** 4.88*** 5.03*** 3.78** Prob > F Notes: Dependent variable is the credit to private sector as percentage of GDP. DFSL is Domestic Financial System Liberalization measured as the simple average between domestic system and stock market liberalization. The impact of financial liberalization on financial development is measured as a percentage of GDP, both for countries at the 25 th and at the 75 th percentile level in each considered institutional variable. Significance level for each impact is calculated by the following linear combination test: Domestic Financial Liberalization + Domestic Financial Liberalization (Interacted with Institutional Measure) * Value of Institutional Measure (at the considered percentile) = 0 All institutional variables are normalized between 0 and 1, 1 being the better possible situation. Errors are measured considering clusters by Country Industry. Absolute value of z-statistics and t-statistics in parentheses. (* significant at 10%; ** significant at 5%; *** significant at 1%). 2 / F test: Domestic Financial Liberalization = Domestic Financial Liberalization (Interacted with Institutional Measure) = 0 Liberalization only seems to boost growth if structural legal conditions, such as the protection of property and creditor rights, are imbedded in the law codes and are effectively enforced. Without these requirements, liberalization does not seem to 23

24 facilitate firm's access to external funds. To check the robustness of these results Table 5 splits the sample between countries with high and low legal protection. Results hold but they are not totally conclusive. Table 5: Financial Liberalization and Growth: Sampling by Legal Protections Dependent Variable: (1) (2) (3) (4) (5) (6) (7) (8) Annual Value Added Growth Sample Effective Cred. Rights Creditor Rights Rule of Law No English English < Mean >Mean < Mean >Mean < Mean >Mean Origin Origin Industry's share of total VA in manufacturing in (t-1) (7.04)*** (5.18)*** (5.13)*** (6.13)*** (6.12)*** (4.81)*** (9.11)*** (3.90)*** Domestic Fin. System Liberalization (1.69)* (2.37)** (1.62) (2.52)** (1.86)* (2.25)** (1.81)* (1.95)* Differential in Growth 0.56* 1.22** ** 1.00* 0.41** 0.44* 3.18* Observations Number of Countries Number of Yearcty Country Year Dummies Yes Yes Yes Yes Yes Yes Yes Yes Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes Notes: Dependent variable is the annual value added growth for each ISIC industry, in each country and in each year. Liberalization measures represent the interaction between them and each industry s external financial dependence. Domestic Financial Liberalization is the simple mean between liberalization in domestic financial system and stock market. Differential in real growth rate measures (in percentage terms) how much faster an industry at the 75 th percentile level of external dependence grows with respect to an industry at the 25 th percentile level when it is located in a country at the 75 th percentile of financial development rather than in one at the 25 th percentile. Errors are measured considering clusters by Country Industry. Absolute value of t-statistics in parentheses (* significant at 10%; ** significant at 5%; *** significant at 1%) Financial Liberalization and Financial Development. In the previous section we show that financial liberalization, in particular domestic financial liberalization, boost growth. These reforms may boost growth through either increasing the size of the financial market, for example credit to the private sector, and/or they can improve its efficiency through reducing agency problems or improving the banking sector efficiency. To test this hypothesis we do two exercises. On the one hand, we see if there is a correlation between the size of the financial sector (measured as credit to the private sector over GDP) and financial liberalization. On the other hand, we test whether financial liberalization has an effect on growth beyond its effect on the size of the financial system. 24

25 Financial Liberalization and the Size of the Financial System Table 6 reports the correlation between financial liberalization and the size of the financial system. We include interactive terms between the liberalization measure, and the different proxies for legal protections of creditors. Beside these variables, we include the lag of log GDP per capita and country fixed effects as controls. Our measure for the financial system size is the current ratio of credit to the private system over GDP. Results suggest that financial liberalization increases the size of the financial system in countries with more developed institutions. Specifically, we find that countries with low creditor protections do not take complete advantage of possible effects of liberalization. On the bottom of the Table 6 we include estimates of the impact of financial liberalization on the financial system size in the extreme countries. For example, in column 2 we show that in a country with low effective creditor rights (in the 25 th percentile) liberalization has a minimum correlation with credit market development. An increase in liberalization leads to a 4.1 percentage points of GDP increase of credit markets. On the other hand, if creditor rights are highly protected (in the 75 th percentile) an increase in liberalization leads to a 23.2 percentage points of GDP increase in the size of credit markets. We obtain similar results for the case of creditor right and rule of law. Finally, column 5 reports the same results using a completely exogenous proxy for legal protections: the origin of the legal code. Not surprisingly, the results suggest that common law countries tend to have advantages from liberalizing due to the fact that they are more oriented toward creditor protection. 25

26 Table 6: The Size of the Financial System and Financial Liberalization with Institutional Interactions Dependent Variable: Private Credit/GDP (1) (2) (3) (4) (5) Log of Real GDPpc (t-1) (3.17)*** (2.24)** (2.61)*** (3.14)*** (2.45)** Domestic Financial System Liberalization (DFSL) (3.29)*** (0.88) (0.47) (0.12) (2.70)*** DFSL * Effective Creditor Rigths (4.76)*** DFSL * Creditor Rigths (3.53)*** DFSL * Rule of Law (2.12)** DFSL * English Legal Origin (4.97)*** Impact of Financial Lib. on Development (in% of GDP) 11.7*** Impact of Financial Lib. on Development (in% of GDP) 4.1** 7.0*** 8.3** (Institutional Var. in percentil 25) Impact of Financial Lib. on Development(in % of GDP) 23.2*** 23.7*** 17.1*** (Institutional Var. in percentil 75 ) Impact of Financial Lib. on Development(in % of GDP) 6.6*** (No English Legal Origin) Impact of Financial Lib. on Development(in % of GDP) 39.7*** (English Legal Origin) Observations Number of countries Country Dummies Yes Yes Yes Yes Yes Notes: Dependent variable is the credit to private sector as percentage of GDP. DFSL is Domestic Financial System Liberalization measure as the principal component between domestic system and stock market liberalization. The impact of financial liberalization on financial development is measure in percentage of GDP, both for countries at the 25 th and at the 75 th percentile level in each considered institutional variable. Significance level for each impact is calculated by the following linear combination test: Domestic Financial Liberalization + Domestic Financial Liberalization (Interacted with Institutional Measure) * Value of Institutional Measure (at the considered percentile) = 0. All institutional variables are normalized between 0 and 1, 1 being the best possible situation. Reported absolute t Statistics of Liberalization at the 25 th and at the 75 th percentile of each institutional variable determines liberalization significance according to the institutional situation. Errors are measured considering clusters by Country. Absolute value t-statistics in parentheses. (* significant at 10%; ** significant at 5%; *** significant at 1%). Financial Liberalization and Domestic Financial System Efficiency The response of credit markets in terms of increasing their size after liberalization is not the only way that financial activity can be affected by liberalization. In addition, as pointed by Galindo, Schiantarelli and Weiss (2001), liberalization can also increase the allocative efficiency of credit as well as it can reduce agency cost in the security market (Doidge et al. 2001). An indirect way to test this hypothesis is to return to our initial growth estimations and test if liberalization has any impact on growth beyond that of developing the size of the 26