Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Economic democracy wikipedia , lookup
Balance of payments wikipedia , lookup
Economic calculation problem wikipedia , lookup
Uneven and combined development wikipedia , lookup
International monetary systems wikipedia , lookup
Rostow's stages of growth wikipedia , lookup
Globalization and Its Discontents wikipedia , lookup
Transformation in economics wikipedia , lookup
Financial Development and Collateral Effect of Capital Account Liberalization Daili Wanga The paper employs an 81-country datasets to address the collateral effect of capital account liberalization on financial development. Besides directly providing foreign capital and transferring technology spillovers, capital account liberalization might foster the development of other sectors, enhancing their importance to economic growth. This is named collateral effect. Previous researches in delving the collateral effect have focused on studying the transmitting mechanism from capital account liberalization to financial development. In this paper we circumvent the difficulty of specifying such mechanism. We test whether different status of liberalization results in varying importance of financial development to economic growth. We find that the less open economies gain more than their more liberalized neighbors, once the financial development takes place. Thus the paper provides a new perspective on investigating the collateral effect. Field of Research: International Finance, Economic Development 1. Introduction Since the building-up of Washington Consensus, debates on pros and cons of capital liberalization have never been as fierce as in present days. Market liberalists suggest capital account controls might impede the optimization of social welfares, thus freeing the constraint is indispensable. On the other hand, countries and regions suffering in the financial turmoil, such as Mexico, Southeastern Asia and Chile, provide straightforward caveats to an inappropriate liberalization of capital account. For politicians, the best response might be a conditional acknowledgement to capital controls, just as reflected by Axel Weber in the Per Jacobsson Foundation Lecture, that “In circumstances of high and a Daili Wang, Ph.D candidate in China Center for Economic Research (CCER), National School of Development, Peking University. The author appreciates comments from Professor Yiping Huang, Assistant Professor Ping Yan, discussion panel in National School of Development and Young Scholar Forum in Xiamen University. The author takes sole responsibility to his opinions and views. Corresponding address: China Center for Economic Research, Peking University, Beijing. P.R China. 100871. Cellphone: +86-134-6671-6699. Email: [email protected]. volatile capital flows … countries need room for maneuver to formulate their own policy … in order to enhance financial stability”.1 Turning from the political perspective to the economists‟ view, neither the neoclassical “laissez-faire”, that liberalization benefits, nor the Keynesian “interference”, that control works better, is strong enough to overwhelm the other. Theoretically, formal discussions are far less than those in trade affairs. Empirically, studies on the impact of capital account liberalization on economic growth at best provide mixed findings. It is no shame to say that, “[Capital account liberalization] remains one of the most controversial and least understood policies of our day.” (Eichengreen, 2003: p.49). In this paper, we plan to elucidate the collateral effect from capital account liberalization on economic growth. The term “collateral” is first proposed by Kose, et al. (2006), referring to the indirect benefit resulting from financial integration. In addition to directly influencing the economic growth by providing necessary capital as well as professional know-hows, freeing international capital flows could nurture the development of financial sector and further benefit the domestic economy. Unlike previous work mainly researching on the transmitting mechanism or relations between capital flows and financial development explicitly, we implicitly address the issue by studying the change in the importance of financial development to economic growth after liberalizing the capital account. To our best knowledge, only Braun and Raddatz (2007) had researched similar questions. While their work was based on the de jure index of capital account liberalization, we presume the de facto index functions better 2 . Quinn (2010) denoted that no consensus has been reached on which measurement of capital account liberalization (nor on the intensity of capital account controls) is superior. We defend our choice of de facto index for following reasons. First, noise embedded in the de facto index could be averaged out in the long run. Besides, by adopting the de facto index, we could distinguish different impact brought by different types of capital flows (e.g. FDI, debt flows and aggregate cross-country capital flows). Last, compared with the de jure index, the de facto one provides more objective measurement on the status of capital account. We expect to detect either a positive effect, implying that opening capital account allows for more investment opportunity which levels up the importance of financial sector, or a negative and neutral effect, indicating that economic agents suffer less from finance constraint after liberalizing cross country capital flows, and premature financial system no long exists as a burden to economic growth. We refer to Levine (2002) to construct the financial development index. Despite the difficulty to distinguish whether or not the index is made under researchers‟ subjective judgment, we focus mainly on the change of status of these indices, rather than delving the transmitting mechanism. We adopt a two-stage regression so as to eliminate the endogeneity occurs with financial development and economic growth. Origins of the country‟s legal system are adopted as an instrument variable3. To preview the main findings, we categorize the research into three types. We first run the OLS and two stage regression so as to confirm the relation between financial development and economic growth. Then we distinguish the crisis and non-crisis period, finding that more liberalized countries suffer in the crisis regardless of their level of financial development. We further find that less open economies would benefit more than more liberalized economies from financial development. This varying importance of financial development to economic growth under different status of liberalization is the reflection of the collateral effect brought by capital account liberalization. 2. Literature Review In this section we connect our research with previous studies. We first summarize the measurement of capital account liberalization, so as to recall the caveat proposed by Quinn (2010) about the measurement error. We then brief the work employing country-wide or industry-wide datasets in the field of capital account liberalization and economic growth. The third subsection reviews literatures on the role of financial development. 2.1. Measurement of Capital Account Liberalization4 Generally speaking, two types of methods are employed in assessing the extent to which capital account has been liberalized. The de jure approach is based on legislative or policy restrictions. It often draws on data from the IMF‟s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER) and the OECD‟s Code of Liberalization of Capital Movement. The simplest method is to use a 0–1 dummy for each category, with 1 representing full control and 0 representing no control, or the proportion of categories without restriction (see, for instance, Epstein and Schor 1992; Klein and Olivei 1999). Some studies try to distinguish different degrees of the controls, such as 0.25 implying light control and 0.75 indicating heavy control (for instance, Quinn, 1997; Montiel and Reinhart, 1999; Gou, forthcoming). The de jure method provides good information about the policy framework or policy intension. However, it may not be a good indicator of the actual controls, as the observed capital flows often exceed the legal limit (Edwards, 2005). Therefore, some researchers prefer the de facto approach, which is literally „outcome based‟. The most popular-utilized de facto indicator is called „quantity index‟, which measures the ratio of gross cross-border capital flows to GDP (Lane and Milesi-Ferretti, 2007). One potential problem of this indicator is that actual flows of international capital are often results of not only capital controls but also several other factors, such as expectations of investment returns and perceptions of sovereign risks. Since the paper is designed to delve the long term impact from capital account liberalization on the importance of financial development to economic growth, the noise brought with implementing de facto index might be averaged out. Besides, it would risk losing generality if we rely on the statute-based de jure index. Thus, we adopt the de facto index as the proxy to capital account liberalization, in contrast to the earlier work done by Braun and Raddatz (2007). 2.2. Capital Account Liberalization and Economic Growth Early studies employed aggregate datasets to track the impact from capital account liberalization. As one of the most cited works, Rodrik (1998) reported no significant relation between capital account liberalization and economic growth. On the other hand, Quinn (1997) employed a larger sample to indicate a significant positive correlation. Klein and Olivei (1999) further proved that in developed countries, removing capital control might facilitate growth. One explanation might be that developed countries are equipped with appropriate institutional environment to capture the positive effects. Henry (2007), however, criticized the previous findings. He claimed that sample selection and different definition of capital control might result in the heterogeneity of findings. He further indicated it might be invalid to conclude from a cross-sectional regression, which simply test the long term relationship. Considering the marginal capital diminishing, it is the dynamic of capital account liberalization that needs emphasis. The problems addressed by Henry (2007) are not easy to settle down by merely connecting capital account policies with economic growth. Consequently, disaggregated analyses have emerged. The first strand of literatures turns to investigate capital account management and domestic investment. Kraay (1998) found no direct influence from capital account liberalization on investment. Since interest rate is a straightforward factor influencing investment, McKinnon (1991) reported that low real interest rate is associated with countries implementing capital controls, which he referred to as financial repression. The second strand of disaggregated researches is based on the notion that capital type matters. As a common belief, FDI is more stable than other crossborder capital flows5. Kose, et al. (2006) provided a comprehensive survey on FDI and growth, concluding that by taking a more nuanced approach, particularly after accounting for the initial conditions, FDI and economic growth exhibits a positive linkage. Researches on FPI reach similar conclusion, which might be explained as that liberalization to portfolio flows would be adopted only if nations have been moderately developed. A more interesting question might be delving the impact on growth volatility from liberalizing different types of capital flows, since lower volatility might facilitate the growth (Ramey and Ramey, 1995). 2.3. Role of Financial development Many researchers have argued that financial development played an important role in facilitating economic growth. Levine (1997) mentioned in his sage work that financial system might influence the economic growth via producing information to allocate capital, monitoring firms and exerting corporate governance, providing risk amelioration and pooling the savings.His claim suggested that a better-functioned financial sector would reap more benefit from capital account liberalization. Following the ideology, Allen and Gale (2000) argued that different financial system, either bank-based or market-based, exerted different influences on the economic growth. At the other end of the line, researchers have shed light on the nexus between capital account liberalization and financial development. Levine and Zervos (1998) discovered that even in low-income nations, capital account liberalization might develop the stock market development, although the systemic impact on financial system is ambiguous. Rajan and Zingales (2003) provided an “interestgroup” theory to illustrate the positive link between trade openness, financial openness and development of financial sector. Balgati, et al. (2009) further elaborated on the issue, claiming that a nation only needed to accomplish one type of openness so as to attain gains in financial development. The third branch of work offered a more synthetic view on the relation between capital account openness, financial development and economic growth. Chinn and Ito (2002) extended the previous findings by providing new datasets of financial development, measurement of openness (the famous KAOPEN) and legal institutions. Eichengreen, et al. (2011) analyzed the effects of capital account liberalization on industry growth while controlling for financial crises, domestic financial development and the strength of institutions. They found no significant difference between liberalized and closed nations when trapped in financial turmoil. To summarize, previous researches have drawn disproportionate attention on uncovering the transmitting mechanism of either capital account liberalization on financial development or the former on economic growth. However, short of acknowledged measurement of liberalization and available datasets hinder the research progress. In this paper we take another view to circumvent the dilemma. 3. Data and Methodology 3.1. Theoretical Reflection With the liberalization of capital account, international capital is expected to flood into the domestic country, promoting economic growth by either increasing the domestic investment, leading to investments with positive spillovers, and/or increasing financial intermediation. We would first illustrate the relationship between capital account liberalization with economic growth in the context of a simple AK endogenous growth model. The model presented here is based on Bailliu (2000). In the closed economy, the aggregate production of the economy is given by: Yt AKt (1) Assume no population growth and only one good exists to be consumed or invested. The capital depreciation rate is given as per period. The gross investment equals to: It Kt 1 (1 ) Kt (2) Financial system could transform the savings into investment with efficiency of (0,1] , which implies that a unit of saving would leave 1 absorbed by the financial system. We could regard the loss of efficient saving allocation as the interest rate spread charged by the financial intermediaries. In the closed economy, the equilibrium must satisfy: St It (3) Combining equation (1) and (3), we could easily derive the steady state output growth rate as: I g A( ) A s Y (4) Here we have s as the saving rate. Equation (4) suggests two main channels through which financial development impact on the economic growth. The first one involves saving-investment transformation efficiency, which is measured as , or the residual of saving after financial intermediaries‟ claim. With the development of financial sector, as well as structural changes taking place in the economy (say, interest rate liberalization), the transformation efficiency would change. The second channel involves the investment allocation efficiency. We implicitly assume the investment allocation efficiency would allocate capital into the sector with highest marginal productivity, thus increasing overall productivity of capital A .6 Once the capital account has been liberalized, international capital could move freely between local economy and their domestic nations. Assume capital flows into the local economy. We might simply have new market equilibrium: * (St NCFt ) It* (3‟) Where NCFt denotes the scale of net capital inflows. The steady state is simply adjusted to: g * A* ( I* ( s NCF ) ) A* * A* * s* Y Y (4‟) Comparing the new equilibrium (4‟) with the one in closed-economy (4), we find that capital inflow might increase the domestic investment by providing more capital ( s* s ). This is also the most popular arguments held by the advocates of capital account liberalization. Surely more restrictions are needed in order to achieve the higher investment, such as not all the foreign capital inflows are transformed into consumption instead of saving, and the investment financed by the foreign capital might not crowd out the domestic investment. The second benefit carried by capital account liberalization is to lead investments with positive spillovers. Since the FDI literature has provided abundant findings to understand the vertical externality as well as horizontal externality brought by capital inflows (see Alfaro, et al. 2010), we would not elaborate too much on this aspect. In the aforementioned framework, the spillover effect might be reflected as a larger productivity A* , compared with the productivity A in closed economy. The third way in which capital flows could influence economic growth is to increase the domestic financial intermediation. In other words, the extent that capital flows are intermediated by domestic financial institutions would lead the financial intermediations themselves work more efficiently (that * ). To sketch the mechanism, Yang and Li (2010) had constructed a more detailed threesector growth model. Before we end the description of the simple endogenous growth model 7, one thing we need to point out. The first and second channels described above are ordinary financing and spillover effect of international capital flow. The third channel, despite its simplicity, is the very effect we called “collateral”. In following sections, we would focus on this third aspect. 3.2. Model Setup and Datasets We have constructed two types of models: one OLS and one two-stage (2SLS) panel regression. The subsection would brief the two approaches respectively. For the OLS model, we would estimate the following equation: gi y0i FDi X i i (5) Where g i is the average GDP per capita growth rate of country i ; y0i is the initial level of real GDP per capita in logarithm; FDi is the corresponding level of financial development. Referring to Levine (2002), we define FDi as ratio of bank credit to private sector to GDP. Alternatives might be: ratio of total value traded in markets to GDP; market capitalization ratio; ratio of bank deposit to GDP). X i in equation (5) depicts a conditioning sets including CPI inflation, government expenditure, gross domestic investment, volume of trade to GDP, level of democracy and primary education enrollment 8 . The government spending variable and gross investment variable might capture the net capital formation effect on economic growth. We expect the government spending would negatively correlate with the economic growth, by crowding out the more productive private investment. The gross domestic investment is positively related to the growth, as shown in equation (4) in section 3. According to Alesina, et al. (1994), higher trade openness would lead to higher economic growth. Institutions such as democracy and autocracy might also influence the effect from capital account liberalization and economic growth. For the two-stage regression model, we would estimate: gi ,t ,t i vt y i ,t Vi ,t X i ,t ,t 1 FDi ,t 2 LIBi ,t i ,t (6) FDi ,t i t yi ,t Vi ,t X i ,t ,t LIBi ,t Zi ,t i ,t (7) First look at equation (6). gi ,t ,t is the averaged growth rate of real GDP per capita in country i and time period (t , t ) . This is a standard panel regression setup, with Vi ,t denotes variables measured at the beginning of period (t , t ) , and X i ,t denotes the growth determinants measured as averages over the period (t , t ) . Considering the span of data is not sufficiently large, we define 5 , which is long enough to eliminate short term business cycle noise and also short enough to capture the changes taking place in the economy. It needs to be notified that the selection of explanatory variables is based on neoclassical growth theory (see Barro, 1991) instead of ad hoc decision. yi ,t is the initial condition of per-capital GDP and LIBi ,t denotes the measurement of capital account liberalization. As discussed in Lane and Milesi-Ferretti (2007), we employ the stock of capital flows including aggregated data and disaggregated data (total scale, FDI, and debt) 9 . i , vt in the model are set to control the country-specific and time-specific effect respectively. Table 1, Data Description Variable Obs. Mean Std. Dev. Min Max Real GDP per Capita Growth initial GDP level 2542 2542 2.08% 11675.8 5.42% 9988.2 -33.78% 982.5 57.53% 97643.2 Total Value Traded (% of GDP) Bank Credit to Private (% of GDP) Market Capitalization Ratio Bank Deposit Ratio 1333 2224 1269 2237 18.4% 43.0% 35.0% 43.1% 36.2% 33.5% 43.2% 29.9% 0.0% 1.4% 0.0% 1.7% 326.3% 200.6% 303.4% 230.1% Gross Capital Stock to GDP Gross Debt Stock to GDP Gross FDI Stock to GDP 2722 2722 2722 114.5% 93.1% 21.4% 152.7% 132.5% 29.0% 0.0% 0.0% 0.0% 1873.0% 1687.0% 270.0% Inflation Government Spending (% of GDP) Domestic Investment (% of GDP) Trade Openness Primary Education Democracy Index Autocracy Index 2408 2342 2442 2835 1002 2539 2539 0.37 15.6% 21.8% 56.6% 100.4% 5.93 2.27 3.25 6.6% 5.8% 40.3% 13.4% 4.15 3.27 -0.22 3.0% 3.0% 0.0% 24.0% 0.00 0.00 117.50 76.0% 48.0% 424.0% 154.0% 10.00 10.00 Secondary Education Black Market Premium Private Investment 985 1120 788 74.8% 0.7% 14.5% 30.0% 4.8% 6.1% 5.0% -0.6% 1.0% 162.0% 139.0% 36.0% Notes: financial development aggregate and financial structure aggregate are built upon a principal component analysis on the disaggregated indicators. The reason why we have differentiated variables measured at the beginning of period, Vi ,t , between variables measured as averages over period, X i ,t , is to accommodate two major types of growth model, namely neoclassical model and endogenous-growth model 10 . To enunciate the idea, notice Vi ,t captures the initial condition of the economy, while X i ,t accounts for the steady-state level. By explicitly distinguishing the two effects, we might explain both the transition path to the steady states (as in a neoclassical model), and the steady states of the economies (or difference in stead-state growth rate) as in an endogenous growth model. With respect to the data measured at the beginning of period, we expect the initial per-capita GDP level yi ,t is negatively associated with the following-year growth rate, which is consistent with the classical growth theory. Human capital, measured as primary education enrollment, included in Vi ,t , which we expect to be found a positive sign. Equation (6), which has been detailed explained above, is the second step in the 2SLS regression. At the first stage, we seek for the instrument variable Zi ,t to provide a consistent estimation (as depicted in equation (7)). As argued in Beck and Levine (2005), the origins of the country‟s legal system perform well when taking account of the financial development and economic growth. We employ these datasets to finish our regression model setup. Table 1 offers the summary statistics of the data. Appendix 1 and 2 further discuss the sources and definition of these datasets. 4. Discussions of Findings In this section, we first provide the regression results from the OLS model. Despite its obvious drawbacks, it is still a suitable start point for further analysis. We then turn to the 2SLS model for remedying the endogeneity and offer discussions with respect to the findings. 4.1 Findings from OLS Model The findings of the OLS model are summarized in table 2. We could find that the coefficients on the initial GDP level are all significantly negative, indicating that a country with higher initial income would growth slower than those countries starting with a lower level. Government spending is negatively related to the economic growth, as what we expect. The improvement of investment rate and trade openness both facilitate the economic growth significantly. The coefficient of education is against our expectation, which is assumed to facilitate the growth but turns out to be insignificant. To the author‟s knowledge, the puzzle has long existed in the cross-country growth literature (see Pritchett 1996). The democracy index also behaves against the expectation. We further run the regression according to the classification of income levels, finding that the coefficient is negative only in the low income group. Within high income and emerging economies, democracy actually facilitates growth.11 Interestingly, capital account liberalization, as measured by gross capital stock to GDP, is associated with lower growth rate. We again run the regression across groups, discovering the negative impact exists in both high income and low income group, while an insignificant positive effect is found in emerging economies. In the last two columns we report replacing gross capital stock with debt stock and FDI stock respectively. The pool effects across countries are insignificant, while the debt flows exerts negative impact in low income economy and FDI stock are significantly associated with growth in emerging economies. In all, the simple regression reveals that benefit of capital account liberalization is nonlinear, which might be consistent with Kose, et al. (2006)‟s “threshold effect”, and the composition of capital flows matter. Table 2, OLS Results Dependent Variable Real GDP per Capita growth Initial GDP level CPI Inflation Government spending (in Ln) -0.05*** -0.011* -0.018 -0.054*** -0.038*** -0.035*** -0.042** -0.01 -0.023* -0.054*** -0.04*** -0.036*** -0.059*** -0.037*** -0.039*** -0.057*** -0.04*** -0.034*** Domestic Investment (in Ln) Primary Education 0.069*** 0.073 0.048*** -0.023 0.076*** 0.088** 0.051*** -0.024 0.051*** -0.014 0.053*** -0.029 Trade Openness (in Ln) Democracy index 0.064*** -0.001 0.028*** -0.003** 0.069*** -0.001 0.029*** -0.002** 0.024** -0.002** 0.02* -0.002* Gross Capital Stock to GDP Debt Stock to GDP -0.036*** -0.014*** -0.035*** -0.013** FDI Stock to GDP Total Value Traded -0.01** -0.001 0.008*** Bank Credit to Private 0.003 Market Capitalization Ratio Bank Deposit Ratio 0.001 0.001 0.006* 0.001 Constant 0.522*** 0.531*** 0.416** 0.538*** 0.562*** 0.573*** Observation 607 779 592 781 778 777 Notes: ***, **, * signal respectively the significant level at 1%, 5% and 10% level With respect to financial development, whatever measurement index we adopt, the result is at least insignificantly positive with economic growth. Since we do not care about the transmitting mechanism, we would not go any further in discovering why employing different index results in different consequence. Before closing the subsection, we might emphasize the purpose of the OLS regression. As is well known in development literature, growth happens in long term. Employing annual data to estimate the impact seems inappropriate. However, we do not draw our definite conclusion from the regression above. Indeed we merely expect to see the sign of both control variables and interested variables appear in consistence with the expectation, so as to confirm the datasets and regression techniques are suitable. In the next subsection, we would adopt the five-year average period datasets and 2SLS regression to detect the change of importance of financial development to economic growth. 4.2. Findings from 2SLS Model We perform three regressions in this subsection. First we run the 2SLS to provide the benchmark results after controlling the endogeneity. We then consider the status in crisis period, which is emphasized in Eichengreen, et al. (2011). To better understand the change of importance, we divide the sample in accordance with their extent of capital account openness, testing if the more open economy gains more from the development of financial sector. Table 3, Two-stage Regression Dependent Variable Real GDP per Capita Growth (5-year period) Initial GDP level CPI Inflation Government spending (in Ln) Domestic Investment (in Ln) Primary Education Trade Openness (in Ln) Democracy index -0.086*** -0.014** 0.006 0.064*** -0.102 0.006 0.001 -0.088*** 0.03 -0.039** -0.01 -0.061 0.027 0.002 -0.085*** -0.006 -0.009 0.046*** -0.113 0.006 -0.002 -0.083*** 0.035 -0.051** -0.016 -0.09 0.033 -0.002 Gross Capital Stock to GDP Debt Stock to GDP FDI Stock to GDP Total Value Traded Bank Credit to Private Market Capitalization Ratio Bank Deposit Ratio Constant -0.017 -0.032 -0.025 -0.037 1.072*** 0.908*** 1.027*** Anderson‟s CC Test Hansen‟s J Test Observation 4.77*** 4.928 199 54.18*** 7.461 322 15.7*** 2.054 194 -0.09*** 0.034 -0.039** -0.01 -0.06 0.026 0.002 -0.092*** 0.022 -0.046** -0.004 -0.099 0.009 0.002 -0.032* -0.008 0.016* 0.112** 0.112** 0.102** 0.162** 0.931*** 0.917*** 0.939*** 138.8*** 8.29* 323 52.45*** 7.21 322 69.65*** 7.89* 322 0.03* Note 1: ***, **, * signal respectively the significant level at 1%, 5% and 10% level. IV variable: origin of the country‟s legal system and lag of financial development. Note 2: GMM-IV regression is adopted since the instrument variables‟ dimension is larger than the explanatory variables‟. Heteroskedasticity and autocorrelation in error terms are taken account of by implementing regression with robust standard errors. Note 3: Anderson‟s CC test evaluates the strength of instrument variables, with null hypothesis implying under-identification. Hansen‟s J test reflects the validity of instrument variables, with null hypothesis indicating exact-identification. Table 3 summarizes the 2SLS regression results. For the model specification, we evaluate the strength of IV by Anderson‟s CC test and the over-identification issue is addressed by Hansen‟s J test. 12 Initial GDP level again appears negatively related to the economic growth. CPI inflation behaves insignificant across model setups. Government spending and domestic investment influence the economic growth in a way consistent with the model predicted. At this stage, however, trade openness, education and democracy do not appear significantly across samples. In every case but one, capital account liberalization does not appear significantly associated with the economic growth. When measuring the openness with the ratio of debt to GDP, we obtain negative result, which is consistent with the argument that debt flows are more volatile, thus hinder the economic growth. In all cases, financial development, whatever measures are taken, is found to facilitate economic growth. We then ask what would happen in the crisis period if the nation adopts more aggressive attitude towards freeing capital flows. Since de facto index performs poorly under extreme condition, taking account of crisis also ameliorate the misuse of measurement. In the estimation we revise equation (6) so as to detect the effect of crisis: gi ,t ,t i vt y i ,t Vi ,t X i ,t ,t 1 FDi ,t 2 LIBi ,t 1crisisi ,t 2crisis FDi ,t 3crisis LIBi ,t i ,t (8) Table 4 summarizes the findings when we take account of the crisis period. We have adopted two types of financial development index and three types of capital account openness index to provide the results. Among the six models, we find initial GDP level and domestic investment are both significant and consistent with the expectation. Inflation, government spending and trade openness are not all significant, while the signs of coefficient match the prediction as well. Model 2, which proxies financial development with the ratio of bank credit to private sector to GDP, in general produces more significant results compared with Model 1. With respect to the coefficient of financial development and capital account liberalization, both models yield similar results. Regardless of the choice of capital account openness measurement, it exerts negatively impact on the economic growth. Financial development is significantly positively associated with the economic growth. Crisis does not change the level of economic growth unless we employ FDI to GDP ratio. The interaction terms suggest that in crisis, more liberalized economies suffer from slower economic growth, while the level of financial development does not matter either in or out of crisis period. Table 4, Capital Account Liberalization and Crisis Real GDP per Capita growth (5-year average) Dependent Variable Model 1 Model 2 Initial GDP level CPI Inflation Government spending (in Ln) Domestic Investment (in Ln) Primary Education Trade Openness (in Ln) Democracy index -0.112*** -0.012* 0.003 0.032*** -0.08** -0.009 -0.001 -0.121*** -0.01 0.006 0.042*** -0.096** -0.003 -0.001 Gross Capital Stock to GDP Debt Stock to GDP FDI Stock to GDP Total Value Traded Bank Credit to Private -0.021** Crisis Crisis * FD Crisis * LIB Constant -0.024 -0.002 -0.021* 1.268*** -0.028 -0.002 -0.02* 1.377*** Observation 199 199 -0.086*** -0.023*** 0.001 0.021** -0.054* 0.009 0.001 -0.117*** -0.002 -0.005 0.026*** 0.01 0.04*** 0.002** -0.112*** -0.013** -0.014* 0.027*** -0.036 0.017** 0.003*** -0.038*** -0.015** 0.027*** -0.12*** 0.002 -0.007 0.026*** 0.009 0.038*** 0.002** 0.028*** -0.036*** -0.025* 0.025* -0.011* 0.279*** 0.276*** 0.239*** -0.059** -0.001 -0.021* 0.934*** -0.012 -0.011 -0.025** 0.989*** -0.016 -0.012 -0.024** 1.004*** -0.059** -0.001 -0.022** 0.949*** 199 338 338 335 Notes: ***, **, * signal respectively the significant level at 1%, 5% and 10% level. IV variable: origin of the country‟s legal system and lag of financial development. The forth to last and third to last rows indicate the interaction between crisis and financial development, as well as liberalization respectively. With respect to the coefficient of financial development and capital account liberalization, both models yield similar results. Regardless of the choice of capital account openness measurement, it exerts negatively impact on the economic growth. Financial development is significantly positively associated with the economic growth. Crisis does not change the level of economic growth unless we employ FDI to GDP ratio. The interaction terms suggest that in crisis, more liberalized economies suffer from slower economic growth, while the level of financial development does not matter either in or out of crisis period. The third regression involves testing changes in importance of financial development to economic growth. We first add an interacting term into equation (8) to construct the following model. gi ,t ,t i vt y i ,t Vi ,t X i ,t ,t 1 FDi ,t 2 LIBi ,t 3 LIBi ,t FDi ,t 1crisisi ,t 2 crisis FDi ,t 3crisis LIBi ,t i ,t (9) We are interested in coefficient 3 . If 3 0 holds, it suggests financial development is less important to economic growth in the more liberalized countries, and vice versa if 3 0 . Table 5 summarizes the estimation results. Table 5, Capital Account Liberalization, Crisis and Interaction Effect Real GDP per Capita growth (5-year average) Dependent Variable Model 1 Initial GDP level CPI Inflation Government spending (in Ln) Domestic Investment (in Ln) Primary Education Trade Openness (in Ln) Democracy index Model 2 -0.084*** -0.085*** -0.08*** -0.121*** -0.124*** -0.119*** -0.015** 0.009 0.065*** -0.119** 0.004 0.001 -0.015** 0.009 0.065*** 0.002 0.001 -0.016** 0.007 0.068*** -0.11** -0.001 0.001 -0.002 -0.006 0.027*** 0.008 0.039*** 0.002** 0.002 -0.008 0.027*** 0.009 0.038*** 0.002** -0.013* -0.016* 0.028*** -0.037 0.018** 0.003*** Gross Capital Stock to GDP Debt Stock to GDP FDI Stock to GDP Total Value Traded Bank Credit to Private -0.008 -0.038*** Crisis Crisis * FD Crisis * LIB FD * LIB Constant -0.029 -0.001 -0.02 -0.007*** 1.0619*** -0.039 -0.002 -0.018 -0.0007** 1.075*** Observation 199 199 -0.008 0.009 0.009 -0.036*** 0.001 0.007 -0.012* 0.286*** 0.285*** 0.255*** -0.565*** 0.006 -0.026 -0.0004 0.999*** -0.011 -0.019 -0.023* -0.011 1.019*** -0.016 -0.012 -0.022** -0.006 1.032*** -0.059* -0.012 -0.022** -0.009 1.001*** 199 322 322 319 Notes: ***, **, * signal respectively the significant level at 1%, 5% and 10% level. IV variable: origin of the country‟s legal system and lag of financial development. The fifth to last and forth to last rows indicate the interaction between crisis and financial development, as well as liberalization respectively. By first glance we might conclude no convincingly significant relation between the financial development and capital account liberalization exists. At best, from the first two columns we find the term FD*LIB is significantly negative, which suggests that more liberalized country would suffer from slower economic growth, given the level of financial development. However, the results do not hold for all six regressions. From the third to last rows, we contend that the partial effect of financial development to economic growth does not enhance with the liberalization of capital account. Instead of adding the interaction term, we classify the sample into “highly liberalized” (HH), “relatively liberalized” (RH), “relatively closed” (RL) and “highly closed” (LL) groups referring to the observation‟s level of capital account liberalization13. Quartiles are taken as the threshold points. Notice the sample here is not grouped merely by their country features, but by their country-time features14. Due to the data availability, we estimate the following model rather than the one in equation (9): gi ,t ,t i vt y i ,t Vi ,t X i ,t ,t 1 FDi ,t 2 LIBi ,t 3 LIBi ,t FDi ,t i ,t ( within noncrisis subsample) (10) Table 6, Within Group Regression LL FD LIB FD*LIB Obs Measurement on Financial Development Total Value Traded Bank Credit to Private Sector Total Debt FDI Total Debt FDI 0.402*** 0.353*** 0.594*** -0.041** -0.043** -0.038** 0.038*** 0.039*** 0.001 72 72 72 RL FD LIB FD*LIB Obs 0.005 -0.018 -0.001 60 -0.005 -0.018 -0.001 63 0.028 -0.053 -0.001 72 -0.033 0.001 0.003 68 -0.089 -0.01 0.004 68 0.088 0.007 -0.001 68 FD LIB -0.002 -0.009 0.03** -0.081*** 0.131 -0.031 0.144 -0.06** 0.081 0.008 FD*LIB Obs -0.001** 69 -0.001** 67 0.081** -0.085 0.0007* 55 -0.007 81 -0.005 81 0.008 81 FD LIB FD*LIB Obs 0.025 -0.033 -0.0008* 59 0.0019 0.0046 -0.001* 58 0.01 -0.016 -0.002** 61 0.146 -0.013 -0.003 85 0.151 -0.012 -0.003 85 0.162 -0.008 -0.013** 85 Group RH HH Note 1: ***, **, * signal respectively the significant level at 1%, 5% and 10% level. LL, RL, RH, HH denotes “highly closed”, “relatively closed”, “relatively open” and “highly open” respectively. Total, Debt, FDI defines the logarithm form of gross capital stocks to GDP, debt stocks to GDP and FDI stock to GDP respectively. The model refers to equation (6) and equation (7). We report the coefficients and observations of the financial development (FD), financial liberalization (LIB) and their cross term (FD * LIB) in the regression. Note 2: Due to the data constraints, we do not have enough observations to conduct within group regression when measuring financial development with total value traded to GDP (the first three columns). To make the estimation consistent, we combine the first two groups, i.e. LL and RL, instead of regrouping the datasets. The report in RL rows in the first three columns reflects the estimation from the combined group. Instead of explicitly estimating the effect in crisis periods, we drop the observation in crisis era (which accounts for 5% of the whole sample), focusing on the “normal” period. Table 6 provides brief summary of the regression results.15 In the first three columns, we define the financial development as total value traded to GDP. Since capital markets are not mature in many less liberalized countries, we do not have enough observations to estimate LL group. To make the estimation consistent, we combine LL and RL groups and run a pooled regression. We first check the impact from the first order effect of financial development and capital account liberalization.16 Though not convincing, the regression reveals a positive connection between financial development with economic growth and a negative relation between liberalization and growth. By detecting the second order effect, it is interesting to notice the interaction terms are significantly negative in RH and HH groups regardless of the choice to measure capital account liberalization. The forth to sixth columns reflect measuring financial development as bank credit to private sectors to GDP. If we deem capital account liberalization as openness to gross capital flows or debt capital flows, the regression suggests a significant positive 3 in LL and insignificant results from other groups. However, one different phenomenon occurs once we define the liberalization as openness to FDI flows. 3 is insignificant in LL, while it is significantly negative in HH. What could we conclude from the findings? The results suggest that, financial development weighs less important in more liberalized countries. The impact on economic growth does not only reflect on the first order positive effect from financial development. Moreover, with the liberalization of capital account, the second order effect from financial development on economic growth shrinks in the more integrated countries. 4.3. Discussions In this section we have offered three types of empirical evidence. We first evaluate the effect of financial development and capital account liberalization from an OLS model and two-stage regression model. Both regressions reveal that financial development does facilitate the economic growth as previous literature suggests, while the effect from capital account liberalization is mixed. We further distinguish between the crisis period and non-crisis period, detecting the role played by financial development and capital account openness. We contend that more liberalized countries suffer more during the crisis, while the financial development does not perform significantly different between the two periods. In the third regression, we evaluate if capital account liberalization could bring along more investment opportunity to enhance the role played by financial development, or providing more fund sources so as to de-emphasize its importance. Upon taking account of the interaction of two variables, we find no convincingly significant impact from the interacting term. By dissecting the samples into groups characterized by different level of liberalization, we come with the result that, countries with highly closed capital account could benefit more from further financial development. We summarize three implications from our research. First, during crisis era we observe the more liberalized countries suffer more, while financial development does not matter. The implication is that even a nation is equipped with a fully developed financial sector; it might risk the stability to push forward the capital account liberalization. The finding challenges the opponents for capital account liberalization, who believe a mature financial sector is the precondition of capital account openness. To make it clearer, we do not argue for the liberalization without prudent consideration. What we propose here is to provide a caveat to those who hesitate over reforms. Nothing comes from nothing. The best strategy towards capital account liberalization is learning by doing. Secondly, we find that the status of capital account liberalization is associated with the importance of financial development to economic growth. Why the less liberalized countries appear to gain more with the development of financial sector? Many factors could be named and we hereby refer to one proposed by Braun and Raddatz (2007), who employed de jure index to reach the similar findings. In closed countries, domestic capital misallocation could not get remedied unless financial frictions being removed. On the contrary, in more liberalized countries, entrepreneurs could access the international capital market so as to circumvent the domestic financing restraints. As the result, the status of capital account liberalization influences the importance of financial development to economic growth. Last, we provide a new perspective on the collateral effects from liberalizing capital account. We have not only explicitly confirmed the positive influence from financial development on economic growth and mixed effect from capital account liberalization, as previous studies have done. One step further has been made to detect the second order effect from capital account openness. With liberalization, we ask what occurs on the importance of other sectors (say, financial sector) to economic growth. We conclude that, besides directly financing and bringing technology spillovers to the domestic economy, capital account openness influences further on the financial sectors, i.e. the collateral effect varies across different status of liberalization. 5. Conclusion For quite a long time, capital account liberalization remains a contentious topic. Researchers employing macro- and micro-level datasets find ambiguous effect of liberalization on the economic growth. However, most previous literatures either focus on the transmitting mechanism or directly test the relationship between capital account liberalization and economic growth. The collateral effect from liberalization has long been overlooked. In this paper, we adopt two-stage regression to investigate if different status of liberalization would affect the importance of financial development. To our knowledge, the perspective is innovative. Our findings are twofold. On the one hand, we find more liberalized countries would suffer more in crisis regardless of their level of financial development. On the other hand, the less open countries would gain more than more liberalized neighbors once the financial development takes place. We contend the findings might be enlightening to the policy decision-makers and further research in the field of capital account liberalization. References Alesina, A., Grilli, V. and Milesi-Ferretti, G. 1994, “The Political Economy of Capital Controls,” in Leiderman, L. and Razin, A. (Ed.), Capital Mobility: The Impact on Consumption, Investment and Growth, Cambridge: Cambridge University Press, pp. 289-328. Alfaro, L., Chanda, A., Kalemli-Ozcan, S. and Sayek, S. 2010, “Does Foreign Direct Investment Promote Growth? Exploring the Role of Financial Markets on Linkages,” Journal of Development Economics,vol. 91,no. 2, pp. 242-56. Allen, F. and Gale, D. 2000,Comparing Financial Systems, Cambridge, Mass.: MIT Press Bailliu, J. 2000, “Private Capital Flows, Financial Development, and Economic Growth in Developing Countries,” Bank of Canada Working Paper 2000-15 Barro, R. 1991, “Economic Growth in a Cross Section of Countries.” Quarterly Journal of Economics,vol. 106, no. 2, pp. 407–44. Balgati, B., Demetriades, P. and Law, S. 2009, “Financial Development and Openness: Evidence from Panel Data”, Journal of Development Economics, Vol. 89, no. 2, pp. 285-96. Barro, R. and Sala-i-Martin, X.1995,Economic Growth, New York: McGraw-Hill Beck, T. and Levine, R. 2005, “Legal Institutions and Financial Development,” in Menard, C. and Shirley, M. (Ed.), Handbook of New Institutional Economics, Springer: Section III, pp. 251-78. Beck, T. and Demirgüç-Kunt, A. 2009, "Financial Institutions and Markets Across Countries and over Time: Data and Analysis", World Bank Policy Research Working Paper No. 4943 Braun, M. and Raddatz, C. 2007, “Trade Liberalization, Capital Account Liberalization and the Real Effects of Financial Development”, Journal of International Money and Finance, Vol. 26, pp. 730-61. Chinn, M. and Ito, H. 2002, “Capital Account Liberalization, Institutions and Financial Development: Cross Country Evidence”, NBER Working Paper, No. 8967. Claessens, S., Dooley, M. and Warner, A. 1995, “Portfolio Capital Flows: Hot or Cold?” World Bank Economic Review,vol. 9, no. 1,pp. 153-74 Edwards, S. 2005. “Capital Controls, Sudden Stops, and Current Account Reversals,” NBER Working Paper no.11170 Eichengreen, B. 2003,Capital Flows and Crises, Cambridge, Mass.: MIT Press Eichengreen, B., Gullapalli, R. and Panizza, U. 2011, “Capital Account Liberalization, Financial Development and Industry Growth: A Synthetic View”, Journal of International Money and Finance, Vol. 30, pp. 1090-1106. Epstein, G., and Schor, J. 1992,“Structural determinants and economic effects of capital controls in the OECD”,in Banuri, T. and Schor, J. (Ed.), Financial openness and national autonomy, New York: Oxford University Press, pp. 136-61 Gou, Q., Wang, D., Yan, P. and Huang, Y. Forthcoming, “Effective or not? Issues on China‟s Short-term Capital Controls,” The Journal of World Economy (in Chinese) Henry, P. 2007, “Capital Account Liberalization: Theory, Evidence, and Speculation,” Journal of Economic Literature,Vol. 45, pp. 887-935. Huang, Y., Wang, X., Gou, Q. and Wang, D. 2011, “Achieving capital account convertibility in China,” China Economic Journal,Vol. 4, no. 1, pp. 25-42 Klein, M., and Olivei, G. 1999, “Capital Account Liberalization, Financial Depth, and Economic Growth,” Journal of International Money and Finance,Vol. 27, no. 6, pp. 861–75 Kose, M., Prasad, E., Rogoff, K. and Wei, S. 2006, “Financial Globalization: A Reappraisal,” NBER Working Paper No. 12484 Kraay, A. 1998. “In Search of the Macroeconomic Effects of Capital Account Liberalization,” World Bank unpublished manuscript Laeven, L. and Valencia, F. 2008, “Systemic Banking Crises: A New Database”, IMF Working Paper No. 08/224 Lane, P. and Milesi-Ferretti, G. 2007, “The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 19702004,” Journal of International Economics,Vol. 73, no. 2,pp. 223–50. Levine, R. 1997. “Financial Development and Economic Growth: Views and Agenda,” Journal of Economic Literature,Vol. 35, pp. 688-726. Levine, R. 2002, “Bank-Based or Market-Based Financial Systems: Which is Better?” NBER Working Paper No. 9138 Levine, R. and Zervos, S. 1998, “Capital Control Liberalization and Stock Market Development,” World Development,Vol. 26,pp. 1169-83 McKinnon, R. 1991, “The Order of Economic Liberalization: Financial Control in the Transition to a Market Economy,” Baltimore, Md.: Johns Hopkins University Press Montiel, P. and Reinhart, C. 1999, “Do Capital Controls and Macroeconomic Policies Influence the Volume and Composition of Capital Flows? Evidence from the 1990s,” Journal of International Money and Finance,Vol. 18,no. 4, pp. 519–35. Obstfeld, M. and Rogoff, K. 1996,Foundation of International Macroeconomics, Cambridge, Mass.: MIT Press Pritchett, L. 1996, “Where Has All the Education Gone?” The World Bank, Working Paper No. 1581. Rajan, R. and Zingales, L. 2003, “The Great Reversals: the Politics of Financial Development in the Twentieth Century”, Journal of Financial Economics, Vol. 69, no. 1, pp. 5-50. Ramey, G. and Ramey, V. 1995, “Cross-Country Evidence on the Link between Volatility and Growth,” American Economic Review,Vol. 85, pp. 1138-51. Quinn, D.1997, “The Correlates of Changes in International Financial Regulation,” American Political Science Review,Vol. 91,no. 3,pp. 531–51. Quinn, D. 2010, “Assessing Measures of Financial Openness and Integration,” Unpublished manuscript, Georgetown University Rodrik, D. 1998, “Who Needs Capital-Account Convertibility?” in Kenen, P. (Ed.) Should the IMF Pursue Capital-Account Convertibility? International Finance Section, Department of Economics, Princeton University, pp. 55-65. Sula, O. and Willett, T. 2009, “The Reversibility of Different Types of Capital Flows to Emerging Markets,” Emerging Markets Review,Vol. 10, no. 4, pp. 296-310. Yang, Z. and Li, H. 2010, “Patterns of International Capital Flows and Endogenous Financial Development,” Journal of International Trade, Vol. 9, pp. 106-16 (in Chinese). Appendix 1, Sample Countries and Categories 81 countries, 1970 – 2004 annual data Since there existsbreak in education series between 1996 and 1997, due to change from International Standard Classification of Education (ISCED76) to ISCED97, the sample span for the OLS regression is: 1970 -1995 and 1998 – 2004. The sample span for the panel regression is: 1970-1995 (average on five years) and 1998 – 2002 (average on five years) 33 High income nations (OECD and Non-OECD countries): Australia, Austria, Belgium, Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Korea, Kuwait, Netherlands, Norway, Portugal, Saudi Arabia, Singapore, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Trinidad and Tobago, United Kingdom, United States 23Emerging economies (Upper middle income countries): Argentina, Brazil, Bulgaria, Chile, Costa Rica, Croatia, India, Cote d'Ivoire, Georgia, Indonesia, Kazakhstan, Latvia, Lithuania, Malaysia, Mauritius, Mexico, Panama, Poland, Romania, Russian Federation, South Africa, Turkey, Venezuela 25Developing countries (Low income and lower middle income countries): Bangladesh, Bolivia, Columbia, Côte d'Ivoire, Ecuador, Egypt, El Salvador, Ghana, Guatemala, Iran, Jordan, Kenya, Kyrgyz Republic, Macedonia, Moldova, Morocco, Nepal, Pakistan, Paraguay, Peru, Philippines, Sri Lanka, Thailand, Tunisia, Zambia Appendix 2, Sources and Definition of Variable Black Market Premium: Levine and Renelt; World's Currency Yearbook (for 1985, 1990-93); Adrian Wood, Global trends in real exchange rates: 1960-84, WB Discussion paper no. 35. 1988 (filling in missing observations); Global Development Finance & World Development Indicators (for 1996-1997, calculated as (parallel Xrate/official Xrate-1)*100 ); values for industrial countries are added as 0) Financial Crisis Episode: drawn from Laeven and Valencia (2008) Democracy and autocracy: research in George Mason University Financial development: from Beck and Demirgüç-Kunt (2009) General government final consumption expenditure): from Global Development Network Growth Database (GDNGD), New York University Gross capital formation (% of GDP):from Global Development Network Growth Database (GDNGD), New York University Growth rate of GDP per capita:from Global Development Network Growth Database (GDNGD), New York University Inflation: from Global Development Network Growth Database (GDNGD), New York University capital account liberalization: from Lane and Milesi-Ferretti (2007) ( FAi ,t FLi ,t ) ratio of the sum of the stock of financial assets and the K i ,t GDP stock of financial liability to gross domestic product (GDP) Origins of the Country’s Legal System: from Global Development Network Growth Database (GDNGD), New York University. Classified as: British common law; Germany civil law; French civil law; Scandinavia civil law; Socialist law. Primary Education Enrollment:from Global Development Network Growth Database (GDNGD), New York University Real Growth of GDP per Capita (Laspyre): from Global Development Network Growth Database (GDNGD), New York University Trade share of GDP: from Global Development Network Growth Database (GDNGD), New York University Endnotes 1 http://www.imf.org/external/np/speeches/2011/092511.htm Generally, de jure index is based on the government statutes, which subject to the enforcement of laws. De facto index is based on economic variables, such as scale of capital flows, which might be influenced by the economic fundamental instead of capital account policies. 3 See Beck and Levine (2005) for the discussion of validity of regarding the variable as reliable instrument variable in the financial development literature 4 The subsection is mainly based on Huang, et al. (2011) 5 See Sula and Willett (2009). However, Claessens, et al. (1995) believed the two types of capital flows are in essence identical. 6 We could also understand the allocation efficiency as better risk diversification effect, thus more capital could flow to originally risky but higher return projects. 7 For discussion of the dynamics of a random AK model, see Obstfeld and Rogoff (1996). The model could also take distortions into account. Future research would address the issue of imperfect market. 8 Upon regression, we have substituted the original setting with secondary education enrollment (vis-à-vis primary education enrollment), private investment (vis-à-vis gross domestic investment), black market premium and level of autocracy (vis-à-vis level of democracy). There appears no material change. 9 Taking account of the critiques from Kose, et al. (2006), we exclude samples that might distract the research interests, such as Hong Kong, China and Luxembourger. These regions are in essence international financial centers and the capital flows cannot be regarded as an appropriate proxy to the capital account liberalization. 10 For more detailed discussion on the growth theory, see Barro and Sala-i-Martin (1995) 11 Due to the words limit, we do not provide the detailed report. The results are available upon request. 12 In the last three columns we observe the IVs are strong; and the over-identification does not matter in the regression. We actually adopt the IV-GMM regression since the dimension of IVs is larger than that of explanatory variables‟. The exclusion restriction implies lag of financial development and origin of legal system are independent of current economic growth, which is argued in Beck and Levine (2005) 13 For instance, when we detect the FDI liberalization and change of importance of financial development to economic growth, we classify the sample with respect to their level of FDI to GDP ratio. The same applies to debt and gross capital stock. 14 This means we could find a country with status HH in given time, while in another time span might be defined with status LL. By this means, we could employ the information more completely and not miss the change in the status of liberalization within country. 15 For full reports of the regression results, readers could reach the author by email. 16 First order effect here means how much the 1 percentage point change in financial development would directly influence the economic growth, if we do not take other indirect effects (such as the interaction from crisis and liberalization) into account. Second order effect, which is quoted in the following passage, denotes the indirect effect from financial development to economic growth, once considering other variables. The reader could regard the second order effect as the marginal effect. 2