NBER WORKING PAPER SERIES CAPITAL ACCOUNT LIBERALIZATION, FINANCIAL DEPTH, AND ECONOMIC GROWTH. Michael Klein Giovanni Olivei

Transcription

1 NBER WORKING PAPER SERIES CAPITAL ACCOUNT LIBERALIZATION, FINANCIAL DEPTH, AND ECONOMIC GROWTH Mchael Klen Govann Olve Workng Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambrdge, MA November 2005 We thank Catherne Humblet for excellent research assstance. Opnons expressed n ths paper are those of the authors and do not necessarly reflect the vews of the Federal Reserve Bank of Boston, the NBER, or of the Federal Reserve System. The vews expressed heren are those of the author(s) and do not necessarly reflect the vews of the Natonal Bureau of Economc Research by Mchael Klen and Govann Olve. All rghts reserved. Short sectons of text, not to exceed two paragraphs, may be quoted wthout explct permsson provded that full credt, ncludng notce, s gven to the source.

2 Captal Account Lberalzaton, Fnancal Depth, and Economc Growth Mchael Klen and Govann Olve NBER Workng Paper No October 1999, Revsed June 2006 JEL No. F36, F43 ABSTRACT We show a statstcally sgnfcant and economcally relevant effect of open captal accounts on fnancal deepness and economc growth n a cross-secton of countres over the perod 1986 to Countres wth open captal accounts over some or all of ths perod had a sgnfcantly greater ncrease n fnancal depth than countres wth contnung captal account restrctons, and they also enjoyed greater economc growth. The results, however, are largely drven by the developed countres n the sample. The observed falure of captal account lberalzaton to promote fnancal deepness among developng countres suggests potentally mportant polcy mplcatons concernng the desrablty of lberalzng the captal account. Mchael W. Klen The Fletcher School Tufts Unversty Medford, MA and NBER mklen@tufts.edu Govann Olve Research Department Federal Reserve Bank of Boston 600 Atlantc Avenue Boston, MA govann.olve@bos.frb.org

3 1. Introducton By the end of the twenteth century, governments of ndustral countres had vrtually elmnated all polces hnderng the movement of fnancal captal across ther borders. Basc economc theory suggests ways n whch captal account lberalzaton may beneft a country: Free captal moblty offers the opportunty to realze the hghest return on savng, to borrow at the most favorable rates, and to dversfy away countryspecfc rsk. 1 A more subtle set of benefts, but ones that have ncreasngly been the focus of dscusson, pertan to the mpact of captal account lberalzaton on the effcency and development of a country s fnancal system. The potental relevance of ths channel s hghlghted by recent work on the mportance of fnancal development for economc growth. 2 There are a number of channels through whch captal account lberalzaton can contrbute to the development of a country s fnancal system. Exposure to nternatonal competton may mprove the effcency of the domestc fnancal system va the ntroducton of nternatonal standards as well as through the potental threat of flght to qualty posed by foregn ntermedares. Subsdares or branches of foregn banks may enlarge the absolute sze of the natonal bankng system, serve formerly neglected nches of the market, and ntroduce fnancal nnovaton that drectly broadens the scope of fnancal servces. These gans n the effcency and scope of the fnancal sector may 1 Two surveys of the emprcal lterature on the role of captal account lberalzaton on growth are Edson, Klen, Rcc and Sløk (2004) and Prasad, Rogoff, We and Kose (2003). 2 Emprcal studes documentng the contrbutons of fnancal ntermedares to economc growth nclude McKnnon (1973), Kng and Levne (1993), Rajan and Zngales (1998), and Levne, Loayza, and Beck (2000). See also the revew paper by Levne (1997). The theoretcal foundatons of fnancal ntermedaton s role n economc development are dscussed by, among others, Bagehot (1873), Schumpeter (1912), Hcks (1969), Greenwood and Jovanovc (1990) and Bencvenga and Smth (1991). 1

4 ncrease the pool of avalable savngs both by elctng hgher domestcally generated savngs and by promotng captal nflows. And, n a vrtuous cycle, ncreased savngs may n turn promote further effcency by enablng ntermedares to reap sgnfcant economes of scale and scope. 3 In ths paper we examne whether there s evdence of a lnk from captal account lberalzaton to fnancal depth and, through ths channel, to overall economc growth. 4 Usng a wde cross-secton of countres, we show that countres wth open captal accounts have sgnfcantly greater fnancal depth than countres wth captal account restrctons. Estmates ndcate an economcally mportant and statstcally sgnfcant effect of captal account lberalzaton on economc growth through the deepenng of a country s fnancal market. We also show, however, that the sgnfcance of the lnk between captal account convertblty and fnancal depth seems to be drven largely by the ndustralzed countres ncluded n the cross-secton. Therefore, one possble nterpretaton of our fndngs s that countres requre a constellaton of economc, legal, and socal nsttutons n order to have captal account lberalzaton translate nto greater 3 See de Swaan (1999). Addressng these ponts, Summers (2000) wrtes: to the extent that nternatonal fnancal ntegraton represents an mprovement n fnancal ntermedaton, [perhaps] because nsttutons nvolved n the transfer of captal across jursdctons mprove effcency wth whch captal s allocated, t offers a potentally sgnfcant ncrease n economc effcency. [F]nancal flows fnance real trade, provde captal to local busnesses on what are often the best avalable terms [and] are closely assocated wth the presence of foregn busnesses and foregn fnancal nsttutons, whch themselves brng sgnfcant benefts. (p. 3) 4 Two works related to ours nclude De Gregoro (1999) and Chnn and Ito (2002). De Gregoro analyzes the effect of fnancal ntegraton on fnancal depth and, through ths channel, on overall growth. Our study dffers from hs n that we use a wder set of countres, a dfferent measure of fnancal ntegraton, and a dfferent emprcal specfcaton. Chnn and Ito s emprcal analyss of the effect of captal account lberalzaton on fnancal development s smlar to ours. We dffer by usng a pure cross-sectonal approach, whereas Chnn and Ito rely on panel data technques. 2

5 fnancal depth. 5 These nsttutons tend to be present n ndustral countres but are less common among developng countres. The rest of the paper proceeds as follows. Secton 2 presents the data used n the emprcal analyss. Secton 3 examnes the effect of captal account lberalzaton on fnancal depth n a wde cross-secton of developed and developng countres. Secton 4 assesses the mpact of captal account lberalzaton on overall economc growth through the fnancal deepenng channel by estmatng a smultaneous equaton model n whch fnancal depth and economc growth are jontly determned. Secton 5 offers some concludng remarks. 2. Measures of Fnancal Depth and Captal Account Lberalzaton It s dffcult to construct a sngle quanttatve measure that captures the extent to whch fnancal markets n a country fulfll ther potental roles, a dffculty compounded when studyng a wdely heterogeneous set of countres. Lkewse, there are several ways to gauge the ease wth whch assets are traded across a country s border. Our ndcators of fnancal depth and captal account lberalzaton largely follow those used n prevous research. Descrptve statstcs of these data, presented n ths secton, foreshadow some of the themes rased n the regresson analyss, ncludng dfferences between ndustral and developng countres and patterns of captal account lberalzaton. Fnancal Depth Several measures of fnancal development have been proposed n the emprcal lterature. In our work, we focus on two ndcators of fnancal ntermedary development, 5 The need for the presence of well-functonng economc, socal, and legal nsttutons n order to reap the benefts from openng an economy has been stressed by, among others, Rodrk (1999) and Klen (2005). 3

6 wth each ndcator constructed n such a way that an ncrease reflects greater fnancal depth. 6 The lqud labltes ndcator, LLY, represents the rato of lqud labltes to GDP, where lqud labltes consst of currency held outsde the bankng system plus demand and nterest-bearng labltes of banks and non-bank fnancal ntermedares. Thus, LLY s a typcal measure of fnancal depth snce t reflects the overall sze of the fnancal ntermedary sector. It does not, however, dstngush between the allocaton of captal to the prvate sector and to varous governmental and quas-governmental agences. In an effort to solate credt ssued to the prvate sector from that ssued to governments, government agences, and publc enterprses, we also employ the ndcator PRIVY, whch equals the rato of clams by fnancal ntermedares to the prvate sector to GDP. Typcally, ths s the ndcator of fnancal development preferred n the emprcal lterature. 7 We wll study the effect of captal account lberalzaton over the perod 1986 to 1995, and over the perod 1975 to 1995, on a country s average level of fnancal depth durng the years We focus on ths average level rather than the measured level of fnancal depth n 1995, the last year of our sample, n order to mnmze random varatons n our ndcators of fnancal depth. 8 6 It s plausble, at least n prncple, that fnancal ntegraton has an mpact not only on the development of fnancal ntermedares, but also on the development of a country's stock and bond market (Levne and Zervos 1998). We lmt our analyss to ndcators of fnancal ntermedary development snce ths allows us to consder a wder set of countres. 7 See, for example, Levne, Loayza and Beck (2000). 8 As we explan later n the text, the sample ends n 1995 because the measurement of our ndcator of captal account openness changed after ths date n a way that makes the post

7 Table 1 presents summary statstcs on LLY and PRIVY, both for the full sample of countres and for the separate samples of ndustral and developng countres. 9 The correlaton between these two ndcators s greater than 70 percent for both OECD and non-oecd samples. Both measures of fnancal development ndcate greater fnancal depth among the 21 OECD countres than among the 74 non-oecd countres n our sample. In addton, whle the OECD countres experenced szable fnancal deepenng over the perod 1986 to 1995, there were much smaller ncreases n both measures of fnancal depth among the countres that consttute the non-oecd sample. In contrast, non-oecd countres experenced a faster rate of fnancal deepenng over the prevous decade, 1976 to 1986, than dd the OECD countres. Ths dfference n the tme-seres evoluton of our ndcators of fnancal depth s one reason for assessng the effect of captal account lberalzaton on fnancal deepenng over dfferent sample perods. 10 Captal Account Lberalzaton Gven the dversty n the ntensty and scope of captal controls, t s dffcult to obtan a consstent measure of captal account restrctons across a wde range of countres. We follow the majorty of studes on captal account lberalzaton n usng annual data from the Internatonal Monetary Fund s publcaton Exchange Arrangements values of ths seres ncommensurate wth the earler values. Furthermore, the rse of emergng markets n the early 1990s makes ths perod a focus of nterest. 9 In ths paper, OECD member countres are those that were n the OECD n The appendx ncludes the lst of countres and notes those that are classfed as OECD members as well as those that, for reasons of data avalablty, are not ncluded n the sample but only n the shorter sample. 10 The results dscussed here are robust to usng the same sample for the shorter perod as s avalable for the longer perod. 5

8 and Exchange Restrctons. 11 One advantage of these data s that they provde a consstent measure over a wde range of countres. Another advantage s that they are avalable on an annual bass. Ther major dsadvantage s that, untl 1996, they provded only the broadest ndcaton of a country s stance towards the free movement of captal. The ndcators before 1996 dd not dstngush between restrctons on dfferent types of captal flows or even between restrctons on nflows of captal and restrctons on outflows of captal. The more recent (from 1996 on) IMF data nclude greater detal, but t s stll mpossble to drectly measure the ntensty of controls or ther effcacy. Also, because we cannot relably map the new classfcaton system nto a qualtatve set of ndcators that matches the older classfcaton system, our emprcal analyss focuses on the pre-1996 perod. From the Exchange Arrangements and Exchange Restrctons we generate a smple 0/1 ndcator varable for each country for each year. 12 Ths ndcator takes the value of 1 f the country had no captal controls n place. Then, for each country, we calculate the varable KALIB, whch represents the proporton of years n whch the country had unrestrcted captal moblty. Ths method of takng the share of years over a perod durng whch the captal account was open has been used by Grll and Mles- Ferrett (1995), Rodrk (1998), and Klen (2005), among others. Panel A of Table 2 lsts the value of KALIB for countres ncluded n next secton s regresson analyss that had some experence wth open captal accounts (that s, 11 Edson, Klen, Sløk and Rcc (2004) show that there s a hgh correlaton between the ndcator of captal account openness used n ths paper and the one developed by Denns Qunn (1997), whch s the other ndcator most often used n emprcal analyss. The Qunn ndcator s avalable for a smaller set of countres than the ndcator used n ths paper. 12 The approprate nformaton s n lne E2 n the Summary Table of Exchange Arrangements and Exchange Restrctons. 6

9 KALIB 86-95? 0) over the perod 1986 to Ths panel shows that KALIB equals 1 for 9 OECD countres (out of a possble 21) and 7 non-oecd countres (out of a possble 74). Whle 58 non-oecd countres had restrcted captal account for ths entre perod, the same s true for only 2 OECD countres, Greece and Iceland. Whle the defnton of KALIB does not necessarly reflect the tmng of captal account lberalzaton, the data n the table nonetheless show a correspondence between KALIB and the actual years durng whch the captal account was open. For example, all the countres for whch KALIB = 0.1 had open captal markets n 1995 only; the countres for whch KALIB = 0.2 had open captal markets n 1994 and 1995; the countres for whch KALIB = 0.3 had open captal markets n 1993 to 1995; and so on. Ths relatonshp generally holds across all values of KALIB wth only two cases of on-agan, off-agan captal account lberalzaton, Ecuador and Uruguay. Panel B of Table 2 shows that there were also no nstances of on-agan, off-agan captal account lberalzaton for OECD countres f we extend the sample perod back to However, 10 of the 18 non-oecd countres that had some experence wth lberalzed captal accounts n the perod between 1976 and 1984, remposed restrctons at some tme over the sample perod. 13 Thus, for non-oecd countres, there s a weaker correspondence between the tmng of captal account lberalzaton and the twenty-year ndcator, KALIB than s the case wth the ndcator based on the shorter sample perod, KALIB Because of ths weaker correspondence, n the next secton we present regresson results for the 1986 to 1995 sample perod, as well as for the twentyyear sample. 13 The fve non-oecd countres that had contnuously open captal accounts over the perod 1976 to 1995 are Bahran, Indonesa, Malaysa, Panama, and Saud Araba. 7

10 3. Captal Account Lberalzaton and Fnancal Depth In ths secton we provde evdence of a systematc assocaton between captal account lberalzaton and fnancal depth, and show that such a result s robust to dfferent estmaton technques, to changes n the sample perod, and to the ncluson of varables controllng for other types of polcy changes or cross-country dfferences n nsttutonal factors. However, we also show that the fndngs are largely drven by the responsveness of fnancal depth to captal account lberalzaton among OECD countres. Specfcaton and Basc Results Fnancal depth s typcally used as an exogenous regressor to explan economc growth. The focus of ths secton, however, s to explan the determnants of fnancal depth. In the absence of a theoretcal model that offers a clear explanaton of these determnants, but wth a focus on the possble role of captal account lberalzaton on promotng fnancal depth, we use the followng specfcaton of the determnants of the average level of fnancal depth n a country over the perod 1991 to 1995, FD = β + β KALIB + β X + ε 1990 s 0 1 j 2 (1) where, FD 1990 s s country s measure of average fnancal depth for the perod 1991 to 1995 (that s, LLY 1990 s or PRIVY 1990 s ), KALIB j ndcates country s stance n terms of captal account lberalzaton over the perod j, where j s ether the perod 1976 to 1995, or the perod 1986 to 1995, X represents a vector of other explanatory varables, and s an error term. In our smplest specfcaton, the vector of controls ε X ncludes ndcator 8

11 varables for the regons of Afrca and Latn Amerca and an ndcator varable dentfyng ol producng natons, Ol. 14 The vector X also ncludes the varable Trade k, representng the rato of the sum of a country s exports plus mports to ts natonal ncome for ether k = 1976 (for the regressons that cover the perod 1976 to 1995) or k = 1986 (for the 1986 to 1995 sample). We nclude the ntal level of trade as a control n the regresson because trade may be postvely assocated wth both fnancal depth (hgh openness to trade may requre a more developed fnancal system) and an open captal account (nternatonal fnancal flows may fnance real trade). Moreover, there s a range of other country-specfc factors that lkely nfluences fnancal depth. In an effort to address the possblty of omtted varable bas, we nclude fnancal depth at the begnnng of the perod (that s, ether or LLY 1986 or LLY 1976 or PRIVY 1976 for the 1976 to 1995 sample, PRIVY 1986 for the 1986 to 1995 sample) as an addtonal control n X. Results based on ths core specfcaton are presented n Table 3. Columns 1 and 2 report OLS estmates usng the 1986 to 1995 sample, for the regressons usng LLY and PRIVY, respectvely, whle Columns 3 and 4 present OLS estmates that use the 1976 to 1995 sample for these two measures of fnancal depth. The estmates n the table show a postve relatonshp between a country s degree captal account openness and ts level of fnancal development. In each of the regressons, the coeffcent on KALIB s postve and sgnfcant at better than the 98 percent level of confdence. Ths nfluence of captal account lberalzaton on fnancal depth occurs whle controllng for the ntal level of 14 The eght ol-producng countres n our sample are Algera, Bahran, Republc of Congo, Gabon, Indonesa, Ngera, Saud Araba, and Venezuela. 9

12 fnancal depth, whch s tself a sgnfcant determnant of the average level of fnancal depth n the frst half of the 1990s. One way to gauge the magntude of the effect of captal account lberalzaton on fnancal depth s to consder the estmated dfference n fnancal depth between a country that had a closed captal account throughout the perod and one that had an open captal account, holdng constant the other varables ncluded n the regresson. For example, the estmated dfference n PRIVY 1990 s for a country that had an open captal account, as compared to one that had a closed captal account throughout the perod, ceters parbus, s Ths represents the dfference between a country at the medan level of PRIVY 1990 s and one at the 64 th percentle. A comparable calculaton for LLY 1990 s, usng the estmates from the 1986 to 1995 sample, gves a dfference between a country at the medan level of LLY 1990 s and one at the 61 st percentle. The estmates from the longer sample show an even larger gan for both ndcators of fnancal depth. 15 We have shown n the prevous secton that vrtually all OECD countres had some experence wth captal account lberalzaton, but there s a relatvely small number of non-oecd countres that lberalzed ther captal accounts at any tme before The queston thus arses as to whether the fndng of a sgnfcant and szable effect of captal account lberalzaton on fnancal depth reflects the nfluence of a restrcted set of countres or rather a more general tendency n the entre cross-secton. Therefore, we next nvestgate whether the responsveness of fnancal depth to captal account lberalzaton dffers between the 21 OECD-member countres and the other countres n the sample by 15 The results from the 1976 to 1995 sample show that the estmated dfference attrbutable to havng an open, rather than a closed, captal account moves a country from the medan value of fnancal depth to the 64 th percentle for LLY, and to the 71 st percentle for PRIVY. 10

13 ntroducng an nteracton term between KALIB j and a dummy varable that equals 1 for each of the 21 OECD member countres. Results of ths exercse are reported n Table 4. These results ndcate that, ndeed, t s the experence of the OECD countres that contrbutes to the sgnfcant effect of captal account lberalzaton on fnancal depth. The coeffcent on KALIB j tself s not sgnfcant at the 95 percent confdence level n any of the four regressons, whle the sum of the coeffcent on coeffcent on the nteracton between KALIB j and the KALIB j and the OECD dummy varable s sgnfcant at better than the 99 percent level of confdence n all four regressons reported n ths table. 16 The statstcally sgnfcant results for the OECD countres also are of a magntude that gves them economc relevance. For example, Australa and Sweden had comparable levels of fnancal depth n 1986, as measured by PRIVY 86, whch was for Australa and for Sweden. Australa had an open captal account n every year of the decade 1986 to 1995 whle the captal account of Sweden was only open over the fnal three years of ths perod. The predcted dfference n PRIVY 1990 s across these two countres s (1 0.3) = 0.200, whch s close to the actual dfference of These results rase the queston of what features enable OECD countres to beneft from an open captal account. One possblty s that hgher nsttutonal qualty n the OECD countres contrbutes to the postve responsveness of fnancal depth to captal 16 One potental explanaton for the lack of sgnfcant results n the non-oecd subsample s that some of the countres had an on-agan, off-agan experence wth captal account lberalzaton. Whle ths could account for the fndngs for the 1976 to 1995 sample, we here note that omttng Ecuador and Uruguay (the only two countres wth on-agan, off-agan lberalzaton n the post perod) from the 1986 to 1995 sample does not sgnfcantly alter the results reported n the frst two columns of Table 4. 11

14 account lberalzaton. To nvestgate ths possblty, we modfy our baselne specfcaton by addng an ndcator of a country s nsttutonal qualty and an nteracton term between ths ndcator and KALIB j to the regressors. The ndcator of nsttutonal qualty we use, labeled Govrep, s an ndex of the lkelhood that a country s government wll not repudate contracts. 17 The Govrep ndcator has a range from 1 to 10, wth larger numbers ndcatng a lower lkelhood of contract repudaton (for our sample, the actual range s 2.68 to 9.77). The average value of Govrep s 9.11 for the 21 OECD countres and 5.62 for the 50 non-oecd countres for whch we have data. Regresson results for ths exercse are reported n Table 5. The bottom rows of ths table present the estmated partal dervatve of fnancal depth wth respect to captal account lberalzaton for dfferent percentles n the dstrbuton of Govrep. 18 These results provde support for the hypothess that captal account lberalzaton s more lkely to promote fnancal depth when n the presence of better nsttutonal qualty. In three of the four cases, captal account lberalzaton has a sgnfcant effect on fnancal depth at the 95 percent confdence level or better. Ths occurs at levels of nsttutonal qualty hgher than 5.30 (for PRIVY 1990 s n the 1986 to 1995 sample), 6.30 (for PRIVY 1990 s n the 1976 to 1995 sample), and 7.80 (for LLY 1990 s n the 1976 to 1995 sample) Ths seres s publshed by the PRS group and s based on the work of Knack and Keefer (1995). 18 The value of nsttutonal qualty at the gven percentles dffers between the 1976 to 1995 and 1986 to 1995 samples, and the values used n constructng the lower part of the table reflect the percentles for the respectve regresson samples. 19 In a related ven, Klen (2005) shows that the effect of captal account lberalzaton on growth depends upon nsttutonal qualty, wth a sgnfcant and postve effect at hgher (although not the hghest) levels of nsttutonal qualty. 12

15 In summary, the baselne results presented n Table 3 of ths secton show a sgnfcant effect of captal account lberalzaton on fnancal depth. Further exploraton, as shown by the estmates n Tables 4 and 5, ndcates that ths effect s concentrated among OECD countres, possbly because these countres enjoy better nsttutonal qualty than the other countres n the sample. Thus, these fndngs suggest that the benefts of captal account lberalzaton are not uncondtonal, but are lkely to depend upon the envronment n whch the lberalzaton occurs. Robustness We now examne the robustness of our basc fndngs wth respect to potental endogenety bas and omtted varable bas. Captal account convertblty s often seen as the logcal culmnaton of developng a deep, mature, and effcent fnancal system. If ths were the case, OLS estmates of β 1 usng specfcaton (1) would bas the results toward fndng a postve relatonshp between captal account lberalzaton and fnancal depth f countres experence a deepenng of ther fnancal system for reasons other than an open captal account whle ther governments also undertake captal account lberalzaton. Therefore, t s potentally mportant to nstrument for our measure of captal account lberalzaton. In what follows, we present nstrumental varables estmates of the effect of captal account lberalzaton on fnancal depth. We also present other estmates drawn from a specfcaton that ncludes addtonal varables whch are potentally mportant for the determnaton of fnancal depth and whch mght be correlated wth captal account lberalzaton. 13

16 The estmates n Table 6 report the nstrumental varable analogues to the OLS regressons presented n Table 4. Specfcally, we adopt a two-step procedure that frst computes ft KALIB, j, the ftted value from an OLS regresson of KALIB j on the chosen nstrument and all the other explanatory varables ncluded n equaton (1). Ths ftted value s then used to evaluate the effect of captal account lberalzaton on fnancal depth by means of an OLS regresson. 20 The nstrument for KALIB j s the smple average of an ndcator of current account lberalzaton and an ndcator of the requrement to surrender export proceeds, wth ths average representng the value of the ndcators for the perod 1979 to 1985 when j = 1986 to 1995, and the value of the ndcators for the perod 1972 to 1975 when j = 1976 to Ths nstrument s always hghly sgnfcant n the frst-stage regressons, wth t-statstcs hgher than 4. As mentoned n the prevous paragraph, we would expect to see a smaller responsveness of fnancal depth to captal account lberalzaton f endogenety plagued our results. However, the nstrumental varable estmates reported n the table offer evdence that ths s not the case, snce the estmated coeffcents for the OECD countres tend to be larger than the correspondng OLS estmates. As s typcally the case, the standard errors assocated wth nstrumental varables estmates are larger than those obtaned wth OLS. 20 The standard errors for the estmates reported n Table 6 are too conservatve because they d not take nto account that error. ft KALIB, j s estmated, and consequently measured wth samplng 21 The ndcators for current account lberalzaton and the surrender of export proceeds are constructed as a share of the number of years consdered, n the same fashon as the captal account lberalzaton measure. We generate a smple 0/1 ndcator varable for each country for each year, accordng to whether the restrctons were n place or not. We use lnes E1 and E4 from the Summary Table of the IMF s Exchange Arrangements and Exchange Restrctons. 14

17 Table 7 addresses concerns of potental omtted varable bas by extendng the specfcaton used n Table 4 wth a range of addtonal varables approprate for the sample perod. These varables nclude the logarthm of ntal real per capta ncome (ether lny 76 or lny 86 ), the growth of ncome per capta over the sample perod (ether lny or ln Y ), the average nflaton over the sample perod (ether π or π ), and three dummy varables related to legal orgn, LegalGermany, LegalFrance, and LegalUK (thus, the omtted legal orgn dummy varable s the one for Scandnavan orgn). All of these varables are potentally correlated wth the dependent varable n our regresson. For example, prevous research has ndcated a lnk between nflaton and fnancal development (Boyd, Levne and Smth 2001). Legal orgn, as well, has been shown to be a determnant of fnancal development (LaPorta et al., 1997, 1998). Lkewse, a country s level of fnancal depth may depend on a country s growth experence and development, wth faster-growng or rcher countres requrng a hgher degree of fnancal ntermedaton. 22 Note, however, that omsson of these varables wll bas the estmate for β 1 n equaton (1) only to the extent that these varables exhbt a correlaton wth captal account lberalzaton. The estmates presented n Table 7 show that, even wth the ncluson of these addtonal varables, captal account lberalzaton retans ts sgnfcant contrbuton to fnancal depth. The estmated effect of captal account lberalzaton on fnancal depth among the OECD countres s margnally smaller when the addtonal control varables are ncluded n three of the four cases, but n all four nstances the coeffcents reman 22 Ths pont wll become mportant for the regressons presented n the next secton, whch draw a lnk from captal account lberalzaton to fnancal depth to economc growth. 15

18 sgnfcant at better than the 95 percent level of confdence. We also note that there s no consstent pattern n the statstcal sgnfcance of any of the varables added to the specfcaton n Table 7. For example, ncome growth s sgnfcant only for the regressons that cover the longer sample perod whle nflaton s only sgnfcant for the regressons n whch the dependent varable s LLY. Overall, we fnd that the results of the prevous secton are robust to the modfcatons n the specfcaton presented here. Captal account lberalzaton contnues to have an mportant role n the determnaton of fnancal depth for the OECD countres ncluded n the cross-secton. We next turn to an analyss of the way n whch captal account lberalzaton, by promotng fnancal deepenng, contrbutes to economc growth. 4. Captal Account Lberalzaton and Growth So far, our results have establshed a lnk between captal account lberalzaton and fnancal depth, holdng constant a range of other possble contrbutory factors. Ultmately, an mportant source of the nterest n fnancal depth resdes n ts mpact on economc growth. In ths secton, we quantfy the contrbuton of captal account lberalzaton to economc growth through ts effect on fnancal depth. The specfcaton we use draws on the results presented above concernng the effect of captal account lberalzaton on fnancal depth. The growth regresson takes the form Y lny Z + 95 = + FD s + Y + u Y ln α 0 α α 2 ln 76 α 3 (2) 76 16

19 where Y j represents real per capta ncome for country n year j (j = 1976 or 1995), ncludes Trade 76, the log of secondary school enrollment n country n 1976, School 76, and regonal dummy varables for East Asa and Afrca. Ths growth equaton s estmated jontly wth an equaton descrbng fnancal depth, Z FD 1990s = β 0 + β1kalib j + β 2X + β3 lny ε. (1') Ths s the same as equaton (1) n the prevous secton, except for the fact that we allow for potental feedback from growth to fnancal depth by explctly ncludng ln Y among the regressors. As n the prevous secton, the vector X ncludes ndcator varables for the regons of Afrca, Latn Amerca, and for ol-producng countres, n addton to the ntal levels of trade and fnancal depth. The system of two equatons, (1') and (2), s estmated usng three-stage least squares. The results for the growth regresson (2), wth one set of estmates usng LLY and the other usng PRIVY as ndcators of fnancal depth, are presented n panel A of Table Panel B of the table reports estmaton results for the fnancal depth equaton (1'), agan for LLY and PRIVY respectvely. The results n Table 8 show that there s a sgnfcant and postve effect of fnancal depth on economc growth for the perod 1976 to 1995 n ths cross secton of countres. Ths fndng s consstent wth prevous emprcal lterature on the mportance of the role of fnancal depth for economc growth (e.g. Kng and Levne 1992). But the focus here s on the effect of captal account lberalzaton on growth va fnancal deepenng. From equatons (2) and (1'), t s possble to see that ths effect s gven by 23 Data avalablty constrans us to 70 observatons, rather than the 73 avalable when we only run the fnancal depth regressons over the 1976 to 1995 perod. 17

20 a 1β1 /( 1 α1β3). We then estmate that an ncrease n KALIB from zero to the mean value n the sample of the non-zero observatons of 0.40 would lead to a change n per capta ncome growth over the years 1976 to 1995 of approxmately 3.0 percentage ponts when LLY s used as an ndcator of fnancal depth, and to a change of 5.5 percentage ponts when PRIVY s used. The economc relevance of these estmates s evdent when we compare them to the sample mean value of per capta ncome growth of 22.1 percent over the perod. Overall, the results n ths secton confrm an mportant lnk from fnancal development to economc growth. However, the results of the prevous secton suggest that captal account lberalzaton appears to postvely affect fnancal depth, and therefore economc growth, only n the subsample of hghly ndustralzed economes. Thus, the estmated economcally szable lnk from open captal account to ncreased growth s not as lkely to be present for a less developed country, at least when such a lnk s presumed to work through an ncrease n fnancal depth. 5. Concludng Remarks In ths paper we have shown a statstcally sgnfcant and economcally relevant effect of open captal accounts on fnancal depth and economc growth n a cross-secton of developed and developng countres over the perods 1986 to 1995 and 1976 to Countres wth open captal accounts over some part or all of these perods enjoyed a sgnfcantly greater ncrease n fnancal depth than countres wth contnung captal account restrctons. 18

21 We have also shown, however, that captal account lberalzaton may not provde the same benefts to all. In partcular, the postve relatonshp between captal account lberalzaton and fnancal depth seems to be concentrated among hghly developed countres. There s lttle evdence of captal account lberalzaton promotng fnancal depth outsde of ths group of countres. Ths may suggest that the benefts of captal account lberalzaton are only fully realzed f ths polcy change occurs n the presence of adequate nsttutons and sound macroeconomc polces. A complete understandng of the mportance of the polcy, nsttutonal, and economc envronment that s requred for successful lberalzaton and ntegraton nto the world economy, however, calls for a more complete understandng of the manner n whch openness alters the performance of an economy. 19

22 Bblography Bagehot, Walter, Lombard Street, 1873, Rchard D. Irwn, Homewood, Illnos. Beck, Thorsten, Asl Demrguc-Kunt and Ross Levne, The Fnancal Structure Database, n Asl Demrguc-Kunt and Ross Levne, edtors, Fnancal Structure and Economc Growth: A Cross-Country Comparson of Banks, Markets and Development, MIT Press, Cambrdge, MA, c. 2001, pp Bencvenga, Valere, and Bruce D. Smth, Fnancal Intermedaton and Endogenous Growth, Revew of Economc Studes, vol. 58 (2), 1991, pp Boyd, John H., Ross Levne and Bruce D. Smth, The Impact of Inflaton on Fnancal Sector Performance, Journal of Monetary Economcs, vol. 47, 2001, pp Chnn, Menze D. and Hro Ito, Captal Account Lberalzaton, Insttutons, and Fnancal Development: Cross-Country Evdence, NBER Workng Paper no. 8967, June De Gregoro, José, Fnancal Integraton, Fnancal Development, and Economc Growth, Estudos de Economía, vol. 26 (2), pp , December de Swaan, Tom, Internatonal Captal Flows: Dscusson, n Jane Sneddon Lttle and Govann P. Olve, eds., Rethnkng the Internatonal Monetary System, Federal Reserve Bank of Boston, Conference Seres No. 43, June 1999, pp Edson, Hal, Mchael Klen, Luca Rcc and Torsten Sløk, Captal Account Lberalzaton and Economc Performance: Survey and Synthess, IMF Staff Papers, vol. 51 (2), August 2004, pp Greenwood, Jeremy, and Bojan Jovanovc, Fnancal Development, Growth, and the Dstrbuton of Income, Journal of Poltcal Economy, vol. 98, 1990, pp Grll, Vttoro and Gan Mara Mles-Ferrett, Economc Effects and Structural Determnants of Captal Controls, IMF Staff Papers, vol. 42 (3), September 1995, pp Hcks, John A., A Theory of Economc Hstory, Oxford Unversty Press, Oxford, England, Kng, Robert and Ross Levne, Fnance and Growth: Schumpeter Mght be Rght, Quarterly Journal of Economcs, vol. 108 (3), August 1993, pp Klen, Mchael W., Captal Account Openness, Insttutonal Qualty and Economc Growth, NBER Workng Paper no. 11,112, February

23 Knack, Stephen, and Phlp Keefer, Insttutons and Economc Performance: Cross- Country Tests Usng Alternatve Insttutonal Measures, Economcs and Poltcs, vol. 7, 1995, pp La Porta, Raphael, Lopez-de-Slanes, Florenco, Shlefer, Andre, and Robert Vshny, Legal Determnants of External Fnance, Journal of Fnance, vol. 52 (3), 1997, pp , Law and Fnance, Journal of Poltcal Economy, vol. 106, 1998, pp Levne, Ross, Fnancal Development and Economc Growth: Vews and Agenda, Journal of Economc Lterature, vol. 35, June 1997, pp Levne, Ross, Norman Loayza and Thorsten Beck, Fnancal Intermedaton and Growth: Causalty and Causes, Journal of Monetary Economcs, August 2000, vol. 46, pp Levne, Ross and Sara Zervos, Stock Markets, Banks and Economc Growth, Amercan Economc Revew, vol. 88 (3), June 1998, pp McKnnon, Ronald, Money and Captal n Economc Development, Washngton, D.C., Brookngs Insttuton, Prasad, Eswar S., Kenneth Rogoff, Shang-Jn We and Ayhan Kose, Effects of Fnancal Globalzatonon Developng Countres, IMF Occasonal Paper no. 220, September Qunn, Denns, The Correlates of Change n Internatonal Fnancal Regulaton, Amercan Poltcal Scence Revew, vol. 91 (3), September 1997, pp Rajan, Raghuram and Lug Zngales, Fnancal Dependence and Growth, Amercan Economc Revew, vol. 88 (3), June 1998, pp Rodrk, Dan, 1998, Who Needs Captal-Account Convertblty? n Stanley Fscher, et al., Should the IMF Pursue Captal Account Convertblty? Essays n Internatonal Fnance, No. 207, Internatonal Fnance Secton, Department of Economcs, Prnceton Unversty, Prnceton, N.J., (May)., Makng Openness Work: The New Global Economy and the Developng Countres, Overseas Development Councl, Washngton, DC, Schumpeter, Joseph A., The Theory of Economc Development, 1912, Harvard Unversty Press, Cambrdge, MA. Summers, Lawrence, Internatonal Fnancal Crses: Causes, Preventon, and Cures, Amercan Economc Revew, vol. 90 (2), May 2000, pp

24 Data Appendx The source for the fnancal ndcators LLY and PRIVY s Beck, Demrguc-Kunt, and Levne (2001). The data can be downloaded at The source for KALIB and KALIB s the IMF Exchange Arrangements and Exchange Restrctons (varous ssues). Secton 2 n the text provdes detals on how these varables are constructed. The level and growth rate of per capta GDP, the rato of the sum of mports and exports to GDP, and secondary school enrollment come from the World Bank World Development Indcators (2000). The source for the varable measurng the degree to whch governments do not repudate contracts, GovRep, s Knack and Keefer (1995) from the Internatonal Country Rsk Gude. The followng s a lst of countres ncluded n the emprcal analyss of Sectons 3 and 4 (wth country codes n parentheses). Countres that are not members of the OECD as of 1986 are grouped by regon (Contnental Afrca, Mddle East, East Asa, South Asa, and Western Hemsphere, respectvely), whle the last group lsts OECD member countres: Afrca (sub-saharan): Burkna Faso, Central Afrcan Republc, Cote d'ivore, Cameroon, The Congo, Ethopa, Gabon, Ghana, Kenya, Lesotho, Madagascar, Maurtana, Maurtus, Malaw, Nger, Ngera, Sudan, Senegal, Serra Leone, Swazland, Chad, Togo, Uganda, South Afrca, Democratc Republc of the Congo, Zamba. Mddle East: Algera, Egypt, Israel, Jordan, Syra. South Asa: Bangladesh, Inda, Sr Lanka, Nepal, Pakstan. East Asa: Fj, Indonesa, Republc of Korea, Malaysa, Phlppnes, Thaland, Samoa. Latn Amerca: Argentna, Bahamas, Bolva, Brazl, Barbados, Chle, Costa Rca, Domnca, Domncan Republc, Ecuador, Guatemala, Honduras, Jamaca, Mexco, Ncaragua, Panama, Peru, Paraguay, El Salvador, Trndad and Tobago, Uruguay, St. Vncent and the Grenadnes, Venezuela. OECD members: Australa, Austra, Belgum, Canada, Denmark, Fnland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Span, Sweden, Unted Kngdom, Unted States. 22

25 Table 1 Fnancal Depth: Summary Statstcs A. LLY - Lqud Labltes as a Rato of GDP Full Sample.36 (73) Industral Countres.57 (21) Developng Countres.28 (52) LLY 76 LLY 86 LLY 1990s Percent Change (95).61 (21).40 (74).46 (95).68 (21).40 (74) 31.2 (73) 18.8 (21) 41.4 (52) Percent Change (95) 11.6 (21) 0.2 (74) B. PRIVY - Fnancal Intermedares Prvate Sector Clams as a Rato of GDP Full Sample.25 (73) Industral Countres.43 (21) Developng Countres.17 (52) PRIVY 76 PRIVY 76 PRIVY 1990s.30 (95).52 (21).23 (74).25 (95).43 (21).17 (74) Percent Change (73) 54.2 (21) 44.1 (52) Percent Change (95) 27.7 (21) 7.7 (74) Note: LLY 1990s and PRIVY 1990s denote average values for the perod Samples szes are n parentheses.

26 Table 2 Captal Account Lberalzaton Index A. KALIB and Years of Open Captal Markets (KALIB ) KALIB Years Open OECD Countres Non-OECD Countres Norway Costa Rca, Nger Span Trndad & Tobago Portugal, Sweden Argentna, Honduras, Peru Ireland Fnland, Austra France, Italy , 1995 Ecuador Guatemala Uruguay Denmark Australa, Belgum, Canada, Germany, Japan, Netherlands, New Zealand, U.K., U.S. Bahran, Bolva, Indonesa, Malaysa, Panama, Saud Araba, Sngapore B. The Evoluton of Captal Account Restrctons, COUNTRIES WITH OPEN CAPITAL ACCOUNTS IN 1976 OECD Countres Belgum, Canada, Germany, Netherlands, US Non-OECDCountres Bahran, Bolva, Ecuador, Guatemala, Honduras, Indonesa, Iran, Malaysa, Mexco, Ncaragua, Panama, Venezuela, Saud Araba COUNTRIES CHANGING CAPITAL ACCOUNT RESTRICTIONS Year OECD Countres Non-OECD Countres Openng Cap. Account Openng Cap. Account Closng Cap. Account 1978 Peru, Sngapore, Uruguay Iran, Ncaragua 1979 Japan, UK 1980 Costa Rca Guatemala, Honduras 1981 Bolva 1982 Paraguay Costa Rca, Mexco 1984 Australa, New Zealand Paraguay, Peru, Venezuela Source: IMF Exchange Arrangements and Exchange Restrctons, varous ssues.

27 Table 3 Fnancal Depth and Captal Account Lberalzaton (1) (2) (3) (4) LLY 1990s PRIVY 1990s LLY 1990s PRIVY 1990s KALIB *** (.0300).1258*** (.0379) KALIB ** (.0420).2670*** (.0577) FD *** (.0564).8429*** (.0797) FD *** (.0872).6547*** (.1170) Trade *** (.0002).0001 (.0003) Trade *** (.0003).0003 (.0005) Afrca *** (.0323) ** (.0380) Afrca *** (.0416) *** (.0544) Latn Amerca *** (.0288) (.0349) Latn Amerca *** (.0373) *** (.0485) Ol *** (.0379) *** (.0465) Ol ** (.0450) *** (.0609) Obs Obs Adj. R Adj. R Note: FD denotes the relevant measure of fnancal depth, LLY or PRIVY. A constant s ncluded n all the regressons. Standard error n parentheses. ** ndcates sgnfcance at the 95 percent level of confdence. *** ndcates sgnfcance at the 99 percent level of confdence.

28 Table 4 Fnancal Depth and Captal Account Lberalzaton: Assessng the Experence of OECD Countres (1) (2) (3) (4) KALIB (.0380) FD ** (.0556) Trade ** (.0003) Afrca ** (.0339) Latn Amerca ** (.0316) Ol ** (.0391) OECD (.0498) KALIB *OECD ** (.0708) LLY 1990s PRIVY 1990s LLY 1990s PRIVY 1990s.0182 (.0433).7313*** (.0797).0006* (.0003) ** (.0362) (.0345) *** (.0449) (.0581).2672*** (.0813) KALIB (.0585) FD *** (.1001) Trade *** (.0004) Afrca *** (.0418) Latn Amerca *** (.0390) Ol ** (.0455) OECD ** (.0549) KALIB *OECD * (.0876).1548* (.0807).5054*** (.1380).0007 (.0005) ** (.0558) ** (.0528) ** (.0624).0621 (.0750).1713 (.1208) Obs Obs Adj. R Adj. R Estmates of the Effect of KALIB for OECD countres KALIB KALIB *OECD ***.2854*** KALIB KALIB *OECD *** (.0598) (.0689) (.0654).3261*** (.0899) Note: FD denotes the relevant measure of fnancal depth, LLY or PRIVY. A constant s ncluded n all the regressons. Standard error n parentheses. * ndcates sgnfcance at the 90 percent level of confdence. ** ndcates sgnfcance at the 95 percent level of confdence. *** ndcates sgnfcance at the 99 percent level of confdence.

29 KALIB (.1440) FD *** (.0703) Trade * (.0003) Afrca *** (.0430) Latn Amerca *** (.0373) Ol (.0468) Govrep.0056 (.0098) KALIB *Govrep.0085 (.0187) Table 5 Fnancal Depth and Captal Account Lberalzaton: Assessng the Impact of Governance Estmates of the Impact of KALIB at Dfferent Values of Govrep (1) (2) (3) (4) LLY 1990s PRIVY 1990s LLY 1990s PRIVY 1990s.0583 (.1684).6708*** (.0943) (.0003) (.0464) (.0415) (.0554).0275** (.0121).0129 (.0219) KALIB (.4052) FD *** (.1053) Trade *** (.0004) Afrca *** (.0507) Latn Amerca *** (.0432) Ol (.0563) Govrep.0012 (.0121) KALIB *Govrep.0456* (.0255) (.2678).3644*** (.1347).0001 (.0005) (.0636) ** (.0538) (.0731).0496*** (.0161).0327 (.0333) Obs Obs Adj. R Adj. R th Percentle of Govrep.0492 (.0587) 50 th Percentle of Govrep.0615 (.0403) 75 th Percentle of Govrep.0825* (.0437) 90 th Percentle of Govrep.0866* (.0497).1222* (.0691).1409*** (.0482).1727*** (.0527).1790*** (.0596) 25 th Percentle of Govrep (.0868) 50 th Percentle of Govrep.0289 (.0587) 75 th Percentle of Govrep.1412*** (.0524) 90 th Percentle of Govrep.1636*** (.0596).1139 (.1132).1613** (.0764).2418*** (.0683).2578*** (.0776) Note: FD denotes the relevant measure of fnancal depth, LLY or PRIVY. A constant s ncluded n all the regressons. Standard error n parentheses. * ndcates sgnfcance at the 90 percent level of confdence. ** ndcates sgnfcance at the 95 percent level of confdence. *** ndcates sgnfcance at the 99 percent level of confdence.

30 Table 6 Fnancal Depth and Captal Account Lberalzaton: Instrumental Varables Estmates ft KALIB.1077* (.0647) FD *** (.0647) Trade *** (.0003) Afrca *** (.0391) Latn Amerca *** (.0344) Ol *** (.0410) OECD (.0596) ft KALIB * OECD (.1060) Estmates of the Effect of KALIB for OECD countres ft KALIB (1) (2) (3) (4) LLY 1990s PRIVY 1990s LLY 1990s PRIVY 1990s.1307* (.0725).6449*** (.0861).0004 (.0003) * (.0409) (.0369) *** (.0462) (.0672).2488** (.1211) ft KALIB (.1053) FD *** (.1073) Trade *** (.0004) Afrca *** (.0494) Latn Amerca *** (.0430) Ol * (.0504) OECD (.0749) ft KALIB * OECD (.1613) Note: FD denotes the relevant measure of fnancal depth, LLY or PRIVY. KALIB s the ftted value from an OLS regresson of KALIB on the nstrument (see man text for detal). A constant s ncluded n all the regressons. Standard error n parentheses. * ndcates sgnfcance at the 90 percent level of confdence. ** ndcates sgnfcance at the 95 percent level of confdence. *** ndcates sgnfcance at the 99 percent level of confdence. ft.2531* (.1446).4547*** (.1469).0005 (.0006) (.0663) ** (.0581) ** (.0689).0616 (.1024).1824 (.2214) Obs Obs Adj. R Adj. R KALIB 95 ft 86 *OECD.2166** 86 (.0841).3795*** (.0993) ft KALIB KALIB 95 ft 76 *OECD (.1264).4356** (.1740)

31 Table 7 Fnancal Depth and Captal Account Lberalzaton: Augmented Specfcaton (1) (2) (3) (4) LLY 1990s PRIVY 1990s LLY 1990s PRIVY 1990s KALIB (.0432).0461 (.0487) KALIB (.05639).1383* (.0798) FD *** (.0603).7181*** (.0870) FD *** (.1065).3555** (.1501) Trade ** (.0003).0002 (.0004) Trade (.0004) (.0006) OECD (.0580) (.0649) OECD ** (.0755) (.1031) KALIB *OECD * (.0815).3102*** (.0909) KALIB *OECD (.0933).1462 (.1303) INFL *** (.0003).0002 (.0004) INFL ** (.0009) (.0013) lny (.0089) ** (.0102) lny * (.0283).0792** (.0395) lny (.0563).1295 (.0629) lny ** (.0498).2074*** (.0692) LegalGermany.1383* (.0811) (.0914) LegalGermany.1759** (.0877).1669 (.1243) LegalFrance.0080 (.0607) * (.0664) LegalFrance.0348 (.0674) (.0870) LegalUK (.0638) * (.0700) LegalUK.0404 (.0685).0247 (.0898) Obs Obs Adj. R Adj. R Estmates of the Effect of KALIB for OECD countres KALIB KALIB *OECD.1480**.3563*** KALIB KALIB *OECD.1890** (.0703) (.0786) (.0750).2845*** (.1037) Note: FD denotes the relevant measure of fnancal depth, LLY or PRIVY. A constant, regonal dummes for Afrca and Latn Amerca, and a dummy for ol producng countres are ncluded n all the regressons. Standard error n parentheses. * ndcates sgnfcance at the 90 percent level of confdence. ** ndcates sgnfcance at the 95 percent level of confdence. *** ndcates sgnfcance at the 99 percent level of confdence.