Democratc Insttutons and Equty Market Lberalzaton Bumba Mukherjee Professor, Department of Poltcal Scence Penn State Unversty sxm73@psu.edu Abstract: In the past two to three decades, the fnancal lberalzaton of equty markets has grown rapdly across the developng world. Concurrent wth fnancal lberalzaton, there has been there has been a global movement toward democracy n several developng countres snce the late 970s. Are these two trends democratzaton and lberalzaton of equty markets related? Are new democraces n the developng world more lkely to ntate equty market lberalzng reforms compared to autocraces? Ths paper attempts to address these questons emprcally by testng the mpact of democratzaton on equty market lberalzaton va estmaton of an orgnal Spatal Autoregressve Error Heckman Selecton model (SAE Heckt model) on a sample of 9 developng countres between 980 and 008. The SAE Heckt model and ts man propertes are presented n some detal as t wll be shown that t s a useful tool for emprcal analyss for scholars of nternatonal fnance. Note that rstmates from the SAE Heckt model provde strong statstcal support for the clam that democratzaton has a postve effect on equty market lberalzaton. Ths result remans robust across dfferent specfcatons.
. Introducton In the past 5 years, the fnancal lberalzaton of equty markets has grown rapdly across the developng world. The sharp ncrease n equty market lberalzaton has been followed wth an equally dramatc ncrease n portfolo captal flows to developng countres and partcularly to large emergng markets n the developng world (Prasad et al 003). Yet the turn to greater lberalzaton of domestc stock markets to especally flows of foregn portfolo captal s by no means ubqutous across the developng world. For nstance developng countres n East and Central and Latn Amerca have been relatvely more receptve to equty market lberalzaton compared to natons n Afrca and Asa (Standard and Poors 000; Bekaert et al 00). Why do some developng countres lberalze ther equty markets more rapdly than others? Put dfferently, why are some equty markets n the developng world more open than others? Interestngly, a large volume of research has developed n emprcal fnance n recent years tryng to understand the consequences of equty market lberalzaton on volatlty of growth, nterest rates and nflaton (e.g. Beck et al 000; Prasad et al 003; Bekaert et al 00, 004). However, relatvely less research has been conducted on the determnants of stock market lberalzaton n especally developng countres. The few extant emprcal studes that examne when developng countres are more lkely to lberalze ther equty markets ether focus on the IMF s role n coercng governments n the developng world to lberalze ther fnancal markets or examne how the degree of domestc fnancal development nfluences the lkelhood of lberalzaton of equty markets. Whle useful, these studes potentally gnore or underestmate another key poltcal factor that may have drven not just equty market lberalzaton n some developng countres but could also See, for example, Abad and Mody (005). A farly large lterature examnes how fnancal development, n partcular the degree of stock market captalzaton, affects equty market lberalzaton n the developed and developng world. In ths regard, see, for nstance, Beck et al 000; Km and Sngal 000.
account for varaton n the lberalzaton of fnancal markets lberalze ther equty markets. Ths paper submts that a crtcal poltcal factor that has arguably drven equty market lberalzaton n the developng world and whch has been underapprecated n the emprcal fnance lterature s the ntroducton of democratc nsttutons,.e. democratzaton, n several developng countres around the globe that were prevously autocraces. Specfcally, from the late 970s onwards, there has been a global movement toward democracy that has been recognzed by poltcal scentsts. For nstance, n 975, there were approxmately thrty democraces n the world. By 994 there were about nnety-fve democraces, whch was slghtly more than half the total number of ndependent countres n the world. By 00, ths fgure had rsen to more than 0 countres. On the other sde of the con, data has revealed that equty market lberalzaton has also taken off n the developng world snce the md-to-late 980s. For nstance, Bekaert et al (00) pont out that more than 30 developng countres have undertaken drastc steps toward equty and fnancal markets lberalzaton startng from the 980s. A later study by these authors that uses a contnuous measure of equty market lberalzaton further confrms that several developng countres and emergng markets were ncreasngly jumpng on the bandwagon of stock market lberalzaton polces from the md-980s. Are these two trends that s, democratzaton and lberalzaton of equty markets related? Are democraces more lkely to ntate fnancal markets, specfcally equty market, lberalzng reforms than autocraces? Ths paper attempts to address ths queston emprcally. That s, I test carefully whether or not the ntroducton of democratc nsttutons n the developng world has had a postve mpact on lberalzaton of domestc markets. There are two key nnovatons n the emprcal tests conducted n ths paper. Frst, as descrbed later, I use a relatvely larger sample of developng countres to test the dea that the phenomenon of democratc transtons may have had a postve mpact on equty market lberalzaton wthn developng countres and across the developng world. Second, I both ntroduce and estmate an orgnal statstcal model, the Spatal
Autoregressve Error (SAE) Heckman selecton (.e. SAE Heckt) model on a sample of several developng countres between 980 and 004. The SAE Heckt model has several methodologcal advantages compared to exstng estmators. For one, t not only accounts for the fact that democratzaton s a non-random phenomenon but also for the possblty that unobservable factors that nfluence the lkelhood of democratzaton may affect equty market lberalzaton. Second, and more mportantly, t accounts n the estmaton process for nternatonal dffuson effects more techncally the mpact of spatal dependence on both equty market lberalzaton and the lkelhood of democratzaton. Estmates from the SAE Heckt model provde strong statstcal support for the clam that democratzaton has a postve effect on equty market lberalzaton. Ths result remans robust when I vary the specfcaton and use dfferent measures of the man ndependent varable. Ths paper proceeds as follows. In the next secton, I brefly justfy why the ntroducton of democratc nsttutons and thus democratzaton s lkely to have a postve effect on equty market lberalzaton.. In secton 3, I present the SAE Heckt model n some detal. In sectons, 4 and 5, I present the data and results from the statstcal model. Ths emprcal paper concludes by dscussng the mplcatons of the fndngs presented here for the lterature on emprcal fnance.. Testable Hypothess As mentoned earler, the emprcal research on the causes of fnancal lberalzaton has been not really focused on developng countres, based almost. Some studes of developed countres have found that the deology of the domnant party affects fnancal lberalzaton, but were unable to provde evdence on the role of the government s deology n developng countres. 3 Does poltcs and/or poltcal nsttutons and specfcally a change n poltcal nsttutons foster equty market lberalzaton n developng countres? Ths paper submts that there are four 3 For ths see, Km and Sngal (000). 3
plausble reasons that may account for why the phenomenon of democratzaton n the developng world has a postve mpact on equty market lberalzaton. Frst, n an nterestng and mportant emprcal study, Qunn (000) suggests that democratzaton plays a crucal role n fnancal, n hs case captal account, lberalzaton. He suggests that democraces provde useful legal rghts that avod arbtrary harm and that democratzaton leads to greater respect for property rghts. As a result, democratzng developng countres can attract more foregn nvestors to help develop ther economy when the rght to undertake nternatonal fnancal transactons s lkely to be respected. Note that f leaders n democratzng developng countres recognze ex ante that foregn nvestors are more wllng to nvest portfolo captal n ther stock markets because these countres respect and recognze the valdty of nternatonal fnancal transactons, then these leaders have greater ncentves to lberalze ther equty markets. Put dfferently, lberalzaton of ther equty markets combned wth the ntroducton of democratc nsttutons that may strongly encourage foregn captal nflows ex post wll provde leaders of democratzng developng countres wth powerful ncentves to lberalze ther domestc equty markets especally gven that foregn captal flows can brng real economc benefts n the short and long run. Second, n a recent study Rodrk (000) suggested and found emprcally that democraces, both new and old, can more easly weather external economc shocks than autocraces. Ths s an nterestng and mportant clam that has crtcal applcatons for ths study. Specfcally, t s well known that lberalzng domestc equty markets leads to a rapd nflow of foregn portfolo captal that can make domestc more vulnerable to exogenous economc and prce shocks. If t s ndeed true that democraces are better than autocraces at wthstandng economc shocks, as clamed by Rodrk, then ncumbent governments n democratzng regmes wll be more wllng ex ante to take the rsk of openng ther equty markets to foregn portfolo captal nflows. Ths s because they wll recognze that they have the ex post ablty to handle more competently external shocks stemmng from greater equty market openness. Thus ratonal antcpaton of beng able to handle ex post 4
exogenous shocks from greater equty market openness wll provde the necessary democratzng leaders wth the necessary ex ante confdence to adopt equty market reforms. Thrd, poltcal scentsts have suggested earler that democratzaton promotes the poltcal power of especally the sklled mddle class n developng countres (see, for e.g., Box 003). Note that sklled mddle class ndvduals gan from open stock markets. For one, they are more lkely to have nvestments n stocks and equty whose value typcally ncreases. Consequently, they wll have a strong preference for open equty markets that allows for foregn portfolo captal nflows. Snce the sklled mddle class attans substantal poltcal power n democratzng regmes, governments n such regmes have poltcal ncentves to pander to the preferences of the sklled mddle class. Ths mples that that they have ncentves to adopt polces that lead to greater lberalzaton of the equty market. Fourth, smlar to democraces, governments n democratzng regmes are more poltcally accountable to ther electorate compared to autocratc leaders. Greater accountablty mples that ncumbents n democratzng states have ncentves to optmally provde publc goods and redstrbutve transfers to votes. To fnance provson of publc goods and redstrbutve transfers, governments n democratzng therefore have ncentves to ncrease captal nflows and nvestment nto ther equty markets as taxes on such nvestment or fnancal transactons wll provde them wth revenue needed to fnance publc spendng. In short, poltcal ncentves to rase revenue wll provde governments n democratzng countres to adopt equty market lberalzaton. Gatherng the dscusson n the precedng paragraphs together thus leads to the followng testable hypothess: Hypothess : Democratzaton ncreases the degree of equty market lberalzaton n developng countres. 3. Statstcal Model To test the clam that democratzaton has a postve effect on equty market lberalzaton n developng countres on avalable data, one needs to explctly account for certan methodologcal ssues n the estmaton process. Frst, democratzaton s a non-random phenomenon and 5
moreover, t s possble that unobservable factors that nfluence the lkelhood of democracy may also affect equty market lberalzaton. For nstance, unobservable strategc factors that nfluence the governng elte's decson to adopt democracy whch can nether be explctly measured nor ncorporated n the specfcaton may also affect ther (or ther successor governments) to lberalze the stock market. Snce unobservable factors that account for a non-random phenomenon such as democratzaton may also lberalzaton of fnancal markets n democratzng developng countres and because equty market s operatonalzed as a contnuous varable for the emprcal tests (ths s descrbed below), I use a Heckman selecton model to test ths paper s man clam. But unlke a standard Heckman model, I estmate a sample selecton model wth spatal autoregressve errors (SAE) n the selecton and the outcome equaton. The sample selecton model s estmated wth spatal autoregressve errors (SAE) that specfes spatally autocorrelated dsturbances n both the selecton and the outcome equaton. I estmate the SAE Heckman selecton model, also known as the SAE Heckt model, on our data because of two reasons. Frst, n the poltcal scence lterature, numerous scholars have suggested that geographcal proxmty plays a crtcal role n the nternatonal dffuson of economc reform polces ncludng lbeberalzaton of equty markets. That s, a country s more lkely to ncrease the degree of lberalzaton of ts domestc stock markets when t observes neghborng countres lberalze ther stock markets (see, for e.g., Smmons and Elkns 004). 4 As descrbed below, a key advantage of the spatal Heckt model I employ s that t allows us to control for the nfluence of nternatonal dffuson on equty 4 Smmons and Elkns (004) use spatal econometrc technques to test the effects of dffuson ncludng dffuson resultng from geographc proxmty on captal account lberalzaton. The spatal econometrc technques are smlar to those used by Franzese and Hays (005, 006) and Beck, Gledtsch and Kyle (006). But none of these scholars estmate a spatal Heckman selecton model, as done here. 6
market lberalzaton va geographc proxmty by explctly ncorporatng spatally autocorrelated dsturbances n the outcome equaton of the emprcal model. 5 Second, poltcal scentsts have long recognzed and found that nternatonal dffuson,.e. spatal effects, strongly nfluence the lkelhood of democratzaton (see, for e.g., Gledtsch and Ward 004; Gledtsch 005; Coppedge and Brnks 006). In partcular, autocratc countres usually face greater pressures to adopt democratc nsttutons and practces when other neghborng countres experence democratc transtons. Smple tests conducted on our sample that s descrbed below reveals spatal dependence effects on the lkelhood of democratzaton and the degree of equty market lberalzaton n our sample. Specfcally, Moran-I Statstcs for the resduals from a panel probt equaton where the democratzaton dummy s the dependent varable strongly (at the % level) rejects the null of zero spatal autocorrelaton n data. Lkewse, Moran-I Statstcs for the resduals from a pras-wnsten model estmated where the degree of equty market lberalzaton s the dependent varable also rejects the null of zero spatal autocorrelaton n the data at the % level. To econometrcally account for both sample selecton and spatal dependence n the data wth respect to democratzaton and the extent of equty market lberalzaton one needs to estmate a two-stage emprcal model that accounts for the selecton of countres that experence democratc transtons and spatally autocorrelated dsturbances. Hence, as mentoned, I estmate a Heckman-type sample selecton model wth spatal autoregressve errors (SAE) n both the selecton and outcome equatons, whch s called the SAE Heckt model. Ths model, developed by Kelejan and Prucha (999) and expanded upon by Flores-Lagunes and Schneer (006), specfes spatally 5 If dffuson n terms of geographc proxmty may be playng a crtcal role n determnng ncreasng levels of captal account lberalzaton, then falng to account for such dffusonmechansms.e. spatal dependence n the data wll lead to nconsstent and neffcent parameter estmates. Snce a SAE Heckt model statstcally accounts for dffuson that occurs va geographc proxmty n the data, t prevents bas n the parameter estmates and s thus a vald econometrc tool. 7
autocorrelated dsturbances (after droppng subscrpt t that denotes tme for notatonal convenence): where * y and () * y 0 x u, u cju j j () * y 0 x u, u cju j j * y are latent varables wth the followng relatonshp wth respect to the observed * * * varables y f y 0and y 0otherwse, and y y y. Therefore, () s the selecton equaton that accounts for partcpaton of countres n the IMF s fnancal stablzaton programs whle () s the outcome equaton that estmates the mpact of covarates on the degree of captal account lberalzaton. The two equatons ( and ) exhbt spatal dependence n ther respectve error term, as u and u depend on the otheru j and u j through ther locaton n space, as gven by the spatal weghts cj and the spatal autoregressve parameters and. I brefly dscuss the operatonalzaton of the spatal weghts n () and () below. At ths stage, observe that we assume that the errors and, =, N are d N ( 0, ). 6 Hence, the statstcal model n ()-() can be presented n a reduced form (3) * y 0 x wj j j (4) * y 0 x wj j j where the weghts wj and wj are the (,j) elements of the nverse matrces ( C ) and ( C ), respectvely, wth C the matrx of spatal weghts c j. Note that both sets of weghts wj depend upon the unknown parameters and. wj and 6 8
Many weghtng schemes have been used to operatonalze elements of the geographc dffuson parameter wj n spatal regresson models by poltcal scentsts. For e.g., Smmons and Elkns (004: 78) use drected trade-flow shares of country j n country s total for w based on the hypothess neghborng countres tend to trade more wth each other. Franzese and Hays (006: 74), however, code w for countres and j that share a border and w 0 for countres that j do not. Snce scholars suggest that a key component of dffuson of equty market lberalzaton may operate va geographc proxmty.e. neghbor effects where a country may lberalze ts stock markets because ts neghbors are dong the same we use a geographc measure of spatal contguty. Ths geographc measure of spatal contguty s operatonalzed as the nverse dstance j j between states and j, where w d j j. As the dstance between states and j ncreases (decreases), wj ncreases (decreases), thus gvng less (more) spatal weght to the state par when j. Whle there s no consensus on how dstance between cross-sectonal unts should be measured, we consder the dstance between captal cty of countres; data for ths varable s drawn from EUgene n the COW dataset. The results reported below reman robust when I use other measures of spatal contguty ncludng drected trade-flow shares of country j n country s total and whether or not states share a border. I use GMM estmaton to estmate the SAE Heckt model, as suggested by Kelejan and Prucha (999), Pnske and Slade (006) and Flores-Lagunes and Schneer (006). 7 The GMM estmaton technque for the SAE Heckt model s descrbed n the appendx. One key advantage of the GMM estmator s that t accounts for heteroskedastcty nduced by the SAE process. Addtonally, to correct for seral correlaton, I use Newey-West standard errors. Fxed effects are ncluded n the outcome equaton of the SAE Heckt model to mnmze omtted varable bas. 8 7 The SAE Heckt model s estmated by usng the MATLAB Spatal Statstcs Toolbox, verson.0. 8 The outcome equaton s also estmated wth random effects but the results from the SAE 9
4. Data and dependent varable To test the clam that democratzaton has a postve effect on equty market lberalzaton n developng countres, I together a tme-seres-cross-sectonal (TSCS) dataset of 9 developng states between 980 and 008. The sze of the sample s determned prmarly by the avalablty of data that s used to construct the contnuous measure of equty market lberalzaton that s descrbed below. The sample s restrcted to 9 developng countres observed between 980 and 008 and t conssts of developng countres from dfferent regons ncludng Afrca, Asa, East and Central Europe and Latn Amerca. The 9 developng countres n the sample are lsted n Table. As descrbed n more detal below, 67 countres n the sample experenced a democratc transton between 980 and 008, another 38 countres remaned as dctatorshps durng ths perod and fnally 4 countres were democratc for the entre tme perod under consderaton. Hence, there s suffcent varaton n the man phenomenon of nterest,.e. democratzaton, to test ths paper s man clam. << Insert Table about here>> In the SAE Heckt model, the key dependent varable of nterest, the level of equty market lberalzaton, s conceptualzed as a contnuous varable. Scholars n emprcal fnance have used a varety of measures to operatonalze equty market lberalzaton (see Bekaert and Harvey 00a,b; Bekaert, Harvey, and Lumsdane 005). Earler measures operatonalzed equty market lberalzaton n dchotomous terms that takes a value of one when the equty market s offcally (by regulaton) lberalzed; otherwse, t takes a value of zero (see Matheson and Rojas-Suarez 993; Bekaert, Harvey, and Lumsdane 00a). Offcal lberalzaton dates that are used to construct ths measure are drawn from the chronology presented n Bekaert and Harvey (005). Whle useful, the dchotomous measure of equty market lberalzaton suffers from three man flaws. Frst, the bnary measure s too coarse and as a result lacks suffcent nformaton about Heckt models wth random effects are not reported to save space. 0
the true extent of stock market lberalzaton. Second, there s controversy about the exact dates n whch governments announced lberalzaton of ther domestc stock markets. Thrd, a cursory examnaton of the news reveals that stock market lberalzaton occurs gradually over tme and s rarely (f ever) an ether closed or open decson taken by governments. That s governments n emergng markets almost never choose to completely open or retan fully closed equty markets, theren rasng the queston of the vablty of the dchotomous measure descrbed above. For the tests n ths paper, I use for the dependent varable a contnuous measure that captures the degree of equty market lberalzaton. Ths measure takes nto account that most lberalzatons are not one-tme events, but are gradual snce they may not be comprehensve at frst. The ndcator that operatonalzes the degree of equty market lberalzaton follows from Bekaert (995) and Edson and Warnock (003), who take the rato of the market captalzatons of the consttuent members of the Internatonal Fnancal Corporaton (IFC) nvestable and the IFC global ndces for each country. Ths measure s defned for each country-year as, IFCI, t IFCG, t MC equty mkt lberalz aton (5) MC where the market captalzaton at tme t of country s () IFC nvestable ndex s MC, IFCI t and () IFC global ndex s MC, IFCG t. The rato of the market captalzaton of the consttuent frms comprsng the IFC Investable ndex to those that comprse the IFC Global ndex for each country-year IFCI IFCG ( MC t MC, t, ) measures the extent to whch foregn nvestors can purchase the stock of domestc frms (Bekaert et al 009). Hence, the contnuous measure n (5) s based on the rato of the market captalzaton of the consttuent frms comprsng the IFC Investable ndex to those that comprse the IFC Global ndex for each country. Note that the IFC Global ndex, subject to some excluson restrctons, s desgned to represent the overall market portfolo for each country, whereas the IFC Investable ndex s desgned to represent a portfolo of domestc equtes that are avalable to foregn nvestors. Hence, a rato of one means that all of the stocks are avalable to foregn nvestors. For example, durng the
990s Korea lfted foregn ownershp restrctons n a number of steps leadng to an ndcator value that gradually moved from zero to one. In the contnuous measure, fully closed countres are assumed to have an ndcator value of zero, and fully lberalzed countres are assumed to have an ndcator value of one. As mentoned earler, the contnuous measure of equty market lberalzaton has been operatonalzed for all the developng countres between 980 and 008. A key advantage of ths contnuous measure s that t contans substantal cross-sectonal and temporal varaton. As a result, one s able to explot the nformaton avalable n the panel data set to test ths paper s man clam as effectvely as possble. 4. Independent and control varables One needs to operatonalze the occurrence of democratzaton n the sample, whch wll serve as the key ndependent varable for testng the central predcton posted earler. To carefully dentfy the exact tmng of democratc transtons for the countres n our sample, I followed the two-step procedure suggested by Papaoannou and Sourouns (004: 8-9). Frst, followng recent research on democratc transtons, 9 I dentfy each epsode of democratc transton n the data at the tmng of the frst nternatonal deemed free and far electons and the adopton of a new democratc consttuton after a prolonged perod of autocratc rule. 0 Based on the aforementoned crtera and the sources mentoned above, I fnd that durng the 980 to 008 tme perod, there have been 58 epsodes of democratc transton across the developng countres n the sample for whch data on trade barrers s avalable. Table lsts each of these 58 epsodes of democratc transtons n the developng countres n the data, the year n whch a democratc transton occurred n these countres, and a bref descrpton of each transton event. 9 For ths see, Box 003; Box and Stokes 003; and Przeworsk et al 000. 0 In most cases of democratc transton, the adopton of a new consttuton and electons occur wthn one or two years.
Next, I dentfed sgnfcant changes n the nature of poltcal regmes accordng to the Polty ndex durng the 980 to 004 perod. Specfcally, f the -pont Polty measure -- rangng from -0 (full autocracy) to + 0 (complete democracy) -- for a country ncreases from a negatve to a postve value and f t remans at the hgher democracy score for at least three years, then that country s consdered to have experenced a democratc transton. Changng the three year requrement to four, fve, sx or seven years does not substantvely or sgnfcantly (n the statstcal sense) alter the results I obtan. Hence the man ndependent varable s a dummy varable Democratzaton that takes the value one n the year n whch a democratzaton epsode occurs --based on the crtera descrbed above -- and n all the years followng n whch the country remans democratc accordng to Przeworsk et al's (000) crtera for a democracy. Fnally, to ensure accuracy, I consulted numerous prmary sources to check whether the tmng of democratzaton epsodes as recorded n these sources s consstent wth the crtera descrbed to code Democratzaton. These prmary sources are: () the country studes/area handbook seres of the Federal Research dvson of the US lbrary of Congress, () the CIA's World Fact Book and () George Kuran (998), World Encyclopeda of Parlaments and Legslatures. The tmng of democratzaton epsodes for countres n our sample followng the crtera descrbed above matches exactly the occurrence of democracy n these countres accordng to the aforementoned prmary sources. Further, the Democratzaton dummy for these countres also matches closely wth the year n I do not code countres that experenced a democratc transton n a year and then relapsed nto an autocracy the year (or two years) after as a democratzaton epsode. Ths s because the effect of democratzaton on trade polces, as envsaged n our model, s meanngful only after some democratc nsttutons are establshed properly and at least partally consoldated. And such partal consoldaton takes at least two years (see O' Donnell and Schmtter 986). Fortunately, there only exsts one case n our data where the country that experenced a democratc transton relapsed almost mmedately nto an autocracy; hence the three year crteron that I use s not strngent. 3
whch these countres become a democracy n Przeworsk et al's (000) data set thus ndcatng consstency of our measure wth Przeworsk et al's (000) operatonalzaton of democracy. Indeed, not surprsngly, the results I report below remans robust when I use the Przeworsk et al (000) dummy measure of democracy. Note that the Democratzaton dummy also acts as a dependent varable n the selecton equaton of the SAE Heckt model. <<Insert Table about here>> I now turn to frst brefly lst the economc and poltcal controls n the outcome equaton and then do the same for the selecton equaton. Wth respect to economc controls n the outcome equaton, scholars have suggested that greater trade openness encourages countres to lberalze ther stock market (Bekaert et al 00; Prasad et al 003). Therefore, I control for the well known measure of Trade Openness n the outcome equaton. It has also been clamed that greater openness to foregn drect nvestment (FDI) fosters stock market lberalzaton. I control for FDI, whch s a country s net foregn drect nvestment as a percentage of ts GDP. The data source s the World Bank s CD- ROM of World Development Indcators. Fnancal development n developng countres s measured by the stock market turnover rato (stock turnover t- ), whch equals the rato of the value of stocks traded to stock market captalzaton. Foregn nvestors should be more nterested n the stock market turnover rato than n other measures of fnancal development (e.g., the stock market captalzaton to GDP rato and the stocks traded to stock captalzaton rato) because ths s a better ndcator of the effcency of the stock market. A small but actve stock market wll have a hgh turnover rato, whereas a large but less lqud stock market wll have a low turnover rato. To control for the possblty that equty market lberalzaton may lead to a larger stock turnover rato, I use the one year lagged value of the stock turnover rato n the outcome equaton. I also ncorporate real GDP per captal n the outcome equaton snce ts plausble that countres wth relatvely hgher per capta ncome mght be more receptve to open fnancal markets. Data for ths varable s drawn from the World Bank s World Development Indcators. 4
Numerous scholars have suggested that external pressure from nternatonal nsttutons such as the IMF that provdes condtonal funds to developng countres may have encouraged these countres to reduce ther nvestment barrers on equty markets (Abad and Mody 005; Dalam 000). The clam here s that the IMF has used condtonal ad programs to "force" developng countres to lberalze ther stock markets. To control for ths effect, I nclude a dummy varable IMF program that ndcates whether the country has partcpated n an IMF program and has thus receved condtonal ad from the IMF. The dummy IMF Program s coded as when the IMF provdes funds to countres that voluntarly opt for IMF stablzaton programs n order to () assst them n dealng wth the effects of externally generated and temporary export shortfalls, () to provde fnancal assstance for exceptonal balance-of-payments dffcultes, () to ncreases reserves and (v) fnally to ncrease confdence n fnancal markets. When the IMF provdes condtonal loans to countres under ts fnancal stablzaton program, t almost always requests the recpent country to further lberalze ts captal account. Note that IMF short-run fnancal stablzaton programs do not nclude programs for long-term economc reform and structural adjustment. Followng the crteron descrbed, three types of IMF fundng are provded under ts stablzaton program: () Stand-by (and extended stand-by) Arrangements (SBA), () Supplementary Reserve Faclty (SRF) and () Extended Fund Faclty (EFF). Therefore, the dummy IMF program s coded as when the IMF provdes ether one or some combnaton of these three types of funds The other three types of IMF short-run fnancal stablzaton programs are () Contngency fundng faclty (CFF), () Buffer Stock fundng faclty (BSFF) and (v) Currency Stablzaton funds (CSF). Unfortunately, nformaton on when and where these types of fundng programs were ntated by the IMF as well as the contents of these programs are extremely weak. Gven the poor nformaton we have about these programs, we chose not to nclude them n our codng of IMF program. Fortunately, these programs are rarely offered by the IMF; ndeed they make up only 7% of all IMF fnancal stablzaton programs that have been ntated and mplemented snce 975. 5
mentoned above to fnancally dstressed countres that opt to partcpate n ts stablzaton program. Data for IMF program s drawn from several sources: the IMF s (004) Revew of Fund Facltes, Hutchson (00), Vreeland (003) and Joyce and Noy (005). 3 5. Fndngs and Analyses In ths secton, I prmarly focus on analyzng the estmates from the outcome equaton snce ths equaton ncludes the key ndependent varable that s used to drectly test the man clam posted n ths paper. I then brefly report the results from the selecton equaton before dscussng fndngs from varous robustness and dagnostc tests. 5. Results from Outcome Equaton A cursory analyss of the data, whch s presented n Table 3, ndcates confrmaton for the dea that democratzaton promotes equty market lberalzaton. In partcular, Table 3 reveals that n the 980 to 008 perod the mean level of the dependent varable, Equty Market Lberalzaton, s ndeed statstcally hgher than the mean level of ths varable for autocraces durng the same perod. Ths descrptve result s encouragng but nsuffcent. <<Insert Table 3 about here>> Model n Table 4 reports the estmates from the outcome equaton of the SAE Heckman selecton model where Equty Market Lberalzaton s the dependent varable and the Democratzaton dummy s the man ndependent varable of nterest. Model presents the results from the outcome equaton wth the same dependent varable but where the ndependent varable s Przeworsk et al s (000) bnary measure of Democracy that has been updated tll 004. <<Insert Table 3 about here>> 3 Note that fund facltes such as the Structural Adjustment Fund (SAF) and the Poverty Reducton and Growth Faclty (PRGF) are not ncluded n the IMF program dummy as these funds are used only for long-run structural adjustment. 6
The coeffcent of the dummy varable Democratzaton s postve and hghly sgnfcant at the % level n model where the contnuous measure of Equty Market Lberalzaton s the dependent varable. Ths result statstcally corroborates theoretcal clam posted earler that democratzaton leads to a declne n foregn nvestment barrers n domestc stock markets and thus engenders equty market lberalzaton n developng countres. For robustness checks, I replaced the Democratzaton dummy wth Przeworsk et al s (000) bnary measure of Democracy n the outcome equaton of the SAE Heckt model. Ths dchotomous measure of democracy also served as the dependent varable n the selecton equaton of the same SAE Heckt model. The results from the aforementoned robustness tests are, as mentoned earler, reported n model of Table 3. Specfcally, observe n model 3, the estmate of Democracy s postve and hghly sgnfcant. Ths result from the robustness estmaton exercse thus reveals that the ntroducton of democratc nsttutons (and n ths case the consoldaton of democracy) as well has a statstcally postve effect on the degree of equty market lberalzaton. Wth respect to margnal effects, smple calculaton from the estmate of Democratzaton n model reveals that when ths varable s ncreased from 0 to and other varables n the outcome and selecton equaton are held at ther mean n the sample, the contnuous measure of the dependent varable Equty Market Lberalzaton ncreases by almost 47.6%. Ths s a substantvely mportant result and t demonstrated emprcally the mportant mpact that the ntroducton of democratc nsttutons has on lberalzaton of stock markets n developng countres and thus emergng markets as well. The llustraton n Fgure broald shows the effect that democratzaton has on the autoregressve movng average n the degree of equty market lberalzaton across tme n the sample of countres lsted n Table. Specfcally, note that n the frst four years after whch a democratc transton occurs,.e. between t=0 and t=4 n fgure, the degree of equty market lberalzaton ncreases, on average, by an astoundng 50%.Ths fgure, whch s derved from the data, thus llustrates the dynamc and substantal effect that democratzaton has on stock market lberalzaton n developng natons. To 7
further derve and fully understand the substantve effect of democratzaton on equty market lberalzaton, I used the the estmates from model x, and parametrc bootstraps, 4 to compute the margnal effect of a 0- to- change n the bnary democratzaton measure on the level of eqauty makret lberlzaton. The resultant frst dfferences n expected values are reported va box plots of ther dstrbutons n fgure The llustraton n ths fgure frst ndcates that n developng states that had (and have) not democratzaed (that s, developng states n whch the democratzaton dummy s equal to 0), the effect on captal lberalzaton s statstcally nsgnfcant and substantvely neglgble. By contrast, and n support for hypothess, I fnd that n developng countres that have democratzed (developng states n whch the democratzaton dummy s equal to ), the bnary ndependent varable has a szable postve effect on eqauty makret lberalzaton. Specfcally, a 0 to ncrease of the dummy democtatzaton varable approxmately yelds an almost % ncrease n the level of equty market lberalzaton n developng countres. Ths margnal effect s statstcally sgnfcant at the 95% confdence level. Hence there exsts strong statstcal and substantve support for hypothess. <<Insert Fgures and about here>> Interestngly, the spatal autoregressve error (SAE) term n the outcome equaton of models and s statstcally nsgnfcant. Ths ndcates that nternatonal dffuson pressures for equty market lberalzaton stemmng for reasons related to geography does not sgnfcantly, n the statstcal sense, affect the degree of lberalzaton of fnancal markets. Hence, although for example, t s wdely beleved by scholars alke that each country n Latn Amerca hastened the pace of fnancal lberalzaton n the 990s because ther neghbors were rapdly openng ther stock markets to foregn portfolo nvestment, the statstcal results reveal that clams about such spatal effects leadng to further equty market lberalzaton may be exaggerated. At the same tme, however, one does not want to suggest from the statstcal nsgnfcance of the dffuson parameter n the emprcal models that nternatonal dffuson mechansms have no effect on fnancal lberalzaton. It s possble, for nstance, that the nfluence of dffuson on equty 4 For the bootstraps n ths case, m=000. All control varables were held to ther means or modes. 8
market lberalzaton s condtonal on other poltcal or economc varables. Alternatvely t s plausble that the absence of temporal varaton n the dffuson parameter that s accounted by spatal autoregressve errors may explan why they do not have a sgnfcant effect on the level of equty market lberalzaton that has changed so dramatcally n the last 0-5 years. At a mnmum, the results n outcome equaton n Table 4 suggest that scholars need to theorze more about the nfluence of nternatonal dffuson on equty market and fnancal lberalzaton n general before testng the mpact of such systemc varables on data. Havng dscussed the results obtaned for the man ndependent varable and the autoregressve error, I now turn to brefly report the results from the economc and poltcal control varables n the outcome equatons n models and. Unlke the strong statstcal support we obtan for the predcton n hypothess, the estmated results of the control varables n the outcome equatons of the dfferent emprcal models provde mxed evdence. For example, the estmate of the controls IMF program dummy, Dvded Government and (surprsngly) FDI are largely nsgnfcant n the outcome equatons of the SAE Heckt models estmated for the sample of developng countres. However, the coeffcent of the other economc control varables relatvely fare much better n the outcome equaton. For nstance, the estmates of GDP per capta, Trade Openness and the lag of the Stock Turnover varable s consstently sgnfcant and n the predcted drecton n the outcome equatons. Fnally, the estmate of the adjusted Inverse Mll s rato s postve and sgnfcant n the outcome equatons of all the estmated SAE Heckt models. Ths ndcates albet ndrectly that unobservable factors that nfluence the lkelhood of democracy are also affectng the degree of equty market lberalzaton n the data. 5. Selecton Equaton Results and Robustness tests Turnng to the estmates of the selecton equaton of models and that are reported n Columns A and B n Table 5 respectvely we fnd that the coeffcent of the lag of the democratzaton dummy s postve and sgnfcant, whch s not surprsng. The estmate of Trade 9
Openness s also postve and sgnfcant, whch s along expected lnes. The economc varables GDP per capta and Growth rate are sgnfcantly postve n the selecton equaton, whle the OIL dummy s nsgnfcant. The estmate of the SAE parameter n the selecton equaton of all the SAE Heckt models s postve and sgnfcant. Ths confrms exstng studes whch suggest that nternatonal dffuson of democratc practces and norms helps to promote democratzaton. <<Insert Table 5 about here>> Fnally, to check the econometrc valdty and consstency of the results reported earler, I conducted some robustness tests and a seres of dagnostc checks. Frst, we added some control varables to the outcome equaton of each SAE Heckt model n table 4 and then re-estmated the model, ncludng the selecton and the augmented outcome equaton. These addtonal controls n the outcome equaton are: log of nflaton, a de facto exchange rate regme varable that s drawn from Renhart and Rogoff's (004) data on de facto exchange rates, a dummy for currency crss and balance of payments crss. I do not report the results from the SAE Heckt models wth the addtonal controls owng to space constrants. Includng these addtonal controls dd not substantvely or sgnfcantly alter any of the emprcal results dscussed earler. Last but not the least, some standard post-estmaton dagnostc checks were conducted on the data set. These dagnostc tests revealed that none of the emprcal models estmated for ths paper suffer from severe multcollnearty, seral correlaton and omtted varable bas and that the resduals are normally dstrbuted. 5 5 The largest VIF value n each emprcal model s substantally lesser than 0, thus suggestng that multcollnearty s not a problem. The Breusch-Godfrey LM test faled to reject the null of no seral correlaton n all the outcome equatons, whle Goureroux et al s (98) score test faled to reject the null of no seral correlaton n the selecton equatons. The RESET test ndcates that the models do not suffer from omtted varable bas and the Jarque-Bera ndcates that the resduals approxmate a normal dstrbuton. 0
6. Concluson Ths paper has examned the poltcal determnants of when x developng countres lberalzed ther fnancal equty markets over the last twenty fve years. More mportantly, t has demonstrated emprcally that the ntroducton of democratc nsttutons,.e. democratzaton, has played a crtcal role n ncreasng the level of equty market lberalzaton n developng countres. The fndngs presented here provde at least three man contrbutons to the study of fnancal markets lberalzaton n emprcal fnance. Frst, all but one of the studes of the determnants of fnancal lberalzaton has used no more than a handful of developng countres n ther sample. Ths study s among the frst to emprcally analyze when fnancal lberalzaton occurs n a relatvely bgger sample of developng countres. Second, ths paper s frst to propose the dea that democratzaton may have fostered equty market lberalzaton n the developng world n the last two to three decades. Ths hypothess receves strong emprcal support n the data. Thrd, from an econometrc perspectve, ths study s the frst, to the best my knowledge, to estmate a spatal autoregressve error (SAE) Heckman selecton model to test the mpact of democratzaton on equty market lberalzaton. Estmatng a Heckt model that explctly accounts for spatal dependence n the data n the estmaton process s necessary gven the presence of systematc spatal autocorrelaton n the degree of equty market lberalzaton across developng countres. Despte accountng for spatal autocorrelaton n the data, one stll found that democratzaton consstently has a postve and statstcally sgnfcant effect on equty market lberalzaton. Hence the results presented here are statstcally robust. An mportant consderaton n developng countres s whether the fnancal markets are suffcently developed to warrant openng them to foregn nvestors. The stock market turnover rato s used to assess the vtalty of the stock market. We fnd that countres wth a more developed stock market open ther equty markets earler, but ths varable has a small mpact on the tmng of fnancal lberalzaton.
In addton to the man fndngs summarzed above, ths paper also provdes some nterestng supplementary fndngs that has largely been overlooked n the extant lterature. Frst, the emprcal analyss provdes some evdence whch shows that hgher levels of trade openness encourages developng countres to open up ther fnancal markets at an earler date. Snce the volume of exports from developng to developed countres has grown dramatcally n the last two decades, ths suggests that trade wth developed countres s an mportant determnant of when fnancal markets are lberalzed. One also fnds, perhaps not surprsngly, that fnancal equty market lberalzaton occurs earler n relatvely rcher countres even among the developng world. Second, ths study shows that the effects of nternatonal dffuson mechansms that operate va geographc proxmty on equty market lberalzaton has been exaggerated. Fnally, the emprcal results presented here suggest that the degree of a developng country s fnancal development s necessary for equty market lberalzaton. Numerous polcy recommendatons emerge from ths study. For one, t suggests that ncreasng trade wth developed countres, whch brngs about greater nteracton wth developed countres, should be encouraged as ths may strongly encourage developng countres to further lberalze ther stock markets. Smlarly, efforts to develop a country s stock market and thus promote fnancal development would help emergng markets to adopt polces that foster lberalzaton of ther fnancal markets. Most mportant of all, ths study demonstrates that encouragng greater poltcal transparency and openness that results from the ntroducton of democratc nsttutons s necessary n order to promote equty market lberalzaton. Future research n ths ssue-area can (and should) develop n two man drectons. Frst, the sample of developng countres used for the tests should certanly be expanded n both crosssectonal and temporal terms. Dong so wll ncrease the generalzablty of the statstcal nferences presented here. Second, scholars also need to develop a logcally consstent theory on why and how democratzaton promotes equty market lberalzaton n the developng world. Theoretcal work n
ths topc wll further enhance our understandng of how the ntroducton of democracy may promote lberalzaton of equty markets. References Abad, A., Mody, A., 005. Fnancal reform: what shakes t? what shapes t? Amercan Economc Revew 95, 66-88. Beck, T., Demrgüç-Kunt, A., Levne, R., 000. A new database on the structure and development of the fnancal sector. World Bank Economc Revew 4, 597-605. Bekaert, G., Harvey, C.R., Lundblad, C., 00. Emergng equty markets and economc development. Journal of Development Economcs 66, 465-504 Bekaert, Geert and Campbell R. Harvey, 005, A Chronology of Important Fnancal, Economc and Poltcal Events n Emergng Markets, http://www.duke.edu/ charvey/country rsk/coundex.htm Bekaert, Geert, Campbell R. Harvey and Robn Lumsdane, 00a, Datng the Integraton of World Captal Markets. Journal of Fnancal Economcs65, 03 47. Bekaert, Geert, Campbell R. Harvey and Robn Lumsdane, 00b, The Dynamcs of Emergng Market Equty Flows. Journal of Internatonal Money and Fnance, 3, 95 350. Bekaert, Geert, Campbell R. Harvey and Chrstan Lundblad, 00, Emergng Equty Markets and Economc Development, Journal of Development Economcs 66, 465 504. Bekaert, Geert, Campbell R. Harvey and Chrstan Lundblad, 005, Does Fnancal Lberalzaton Spur Growth?, Journal of Fnancal Economcs 77,, 3 56. Bekaert, Geert, Campbell R. Harvey, Chrstan Lundblad and Stephan Segel, 005, Growth Opportuntes and Market Integraton, Natonal Bureau of Economc Research, Cambrdge, MA. Benen, H., Van de Walle, N., 99. Of Tme and Power: Leadershp Duraton n the Modern World. Stanford Unversty Press, Stanford. 3
Borjas, G.J., 000. Labor Economcs. Irwn/McGraw-Hll, Boston. Box, Carles. 003. Democracy and Redstrbuton. New York, NY: Cambrdge Unversty Press. Brnks, D. and M. Coppedge. 006. "Dffuson s no Illuson: Neghbor Emulaton n the Thrd Wave of Democracy." Comparatve Poltcal Studes, 39(4): 463-489. Dalam, M., 000. Fnancal openness, democracy, and redstrbutve polcy. Polcy Research Workng Paper Seres 37, World Bank. Flores-Lagunes, A. and Schner, K.E. 006. Estmaton of Sample Selecton Models wth Spatal Dependence. Unpublshed Manuscrpt. Gledtsch, K., and Ward, M. D. 003. Dffuson and the Internatonal Context of Democratzaton", Paper delvered n the Conference on the Internatonal Dffuson of Poltcal and Economc Lberalzaton, Harvard, October 3-4. Haggard, S., Maxfeld, S., 996. The poltcal economy of fnancal nternatonalzaton n the developng world. Internatonal Organzaton 50, 35-68. Hellener, E., 994. States and the Reemergence of Global Fnance: From Bretton Woods to the 990s. Harvard Unversty Press, Cambrdge. Henry, P.B., 000. Stock market lberalzaton, economc reform, and emergng market equty prces. Journal of Fnance 55, 59-564 Internatonal Fnance Corporaton, 999. Emergng Stock Markets Factbook 999. Internatonal Fnance Corporaton, Washngton. Internatonal Monetary Fund, 00. World Economc Outlook. Internatonal Monetary Fund, Washngton. Kenny, L.W., Wner, S.L., forthcomng. Tax systems n the world: an emprcal nvestgaton nto the mportance of tax bases, admnstratve costs, scale and poltcal regme. Internatonal Tax and Publc Fnance. Kelejan, H.H. and I.R. Prucha. 999. A Generalzed Moments Estmator for the 4
Autoregressve Parameter n a Spatal Model. Internatonal Economc Revew, 40: 509-533. Km, E.H., Sngal, V., 000. Stock market openngs: experence of emergng economes. Journal of Busness 73, 5-66 L, Q., Smth, D.L., 00. The dlemma of fnancal lberalzaton: polcy autonomy and socetal demands. Journal of Poltcs 64, 764-790 Pnkse, J. and Slade, M.E.. 998. Contractng n Space: An Applcaton of Spatal Statstcs to Dscrete-Choce Models. Journal of Econometrcs 85 (998), 5-54. Pnske, J., Slade, M.E. and Brett, C. 00. Spatal Prce Competton: a Semparametrc Approach. Econometrca 70: -53. Pnkse, J. and Slade, M.E. 006. Dynamc Spatal Probt wth Fxed Effects usng One-Step GMM: An Applcaton to Mne Operatng Decsons, Spatal Economc Analyss, (): 53-90. Prasad, E.S., Rogoff, K., We, S.-J., Kose, M.A., 003. Effects of fnancal globalzaton on developng countres: some emprcal evdence. Occasonal Paper 0, Internatonal Monetary Fund, Washngton. Przeworsk Adam., Mchael E. Alvarez, Antono Chebub, and Fernando Lmong. 000. Democracy and Development: Poltcal Insttutons and Well-Beng n the World, 950-990. Cambrdge: Cambrdge Unversty Press. Qunn, D.P., 000. Democracy and nternatonal fnancal lberalzaton. Workng paper, Georgetown Unversty. Qunn, D.P. Inclan, C., 997. The orgns of fnancal openness: a study of current and captal account lberalzaton. Amercan Journal of Poltcal Scence 4, 77-83. Standard & Poors, 000. Emergng Stock Markets Factbook 000. Standard & Poors, New York. World Bank, 00. World Development Indcators 00. World Bank, Washngton. 5