POLI(CY RESEARCH WORKING PAPER 2320 Does N4ore Intense Empirical evidence indicates a strong correlation between Coripetition Lead long-run growth and effective To FligT'eir Gtowth? enforcement of antitrust and competition policy. Mark A. Dutz Aydin Hlayri The World Bank Development ]Research Group Public Economics U April 2000 POLICY RESEARCH WORKING PAPER 2320 Summary findings The relationship between the intensity of competition in country study. They examine rhe impact on growth of an economy and its long-run growth is an open question various measures having to do with intensity of domestic in economics. Theoretically, there is no clear-cut answer. competition - beyond the effects of trade liberalization. Empirical evidence exists, however, that in some Their results indicate a strong correlation between sectors more competition leads to more innovation and long-run growth and effective enforcement ot antitrust accelerates productivity growth. and competition policy. To complement those findings and capture economywide effects, Dutz and Hayri conduct a cross- An earlier version of this paper - a product of Public Economics, Development Research Group - was presented at a conference, Industrial Reorganization and Development, in Toulouse, France (November 1998). The study was funded by the Bank's Research Support Budget under the research project "Does More Intense Competition Lead to Higher Growth?" (RPO 682-47). Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Hedy Sladovich, room MC2-609, telephone 202-473-7698, fax 202-522-1154, email address hsladovichaworldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/ workingpapers. The authors may be contacted at mdutz Ciworldbank.org or ahayri(@ddttus.com. April 2000. (25 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas a ount development issues. An objective of the series is to get the findings out quickly, even if the preseistations are lets than fully polishedl. The papers carry the namnes of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, i the countries they represent. Produced by the Policy Research Dissemination Center Does NMore Intense Competition Lea(d to Higher Gromwth?* Mark A. Dutz The World Bank and EBRD Aydin Hayri Deloitte & Touche LLP An earlier version of this paper was presented at a conference on "Industrial Organisation and Development" in Toulouse, France, in November 1998. Co-mments by Richard Caves, Ross Levine and conference participants are gratefully acknowledged. Research assistance by Gregorio Impavido is very gratefully acknowledged. This paper was prepared under funding from the World Bank research project 'Does more intense competition lead to higher growth' (RPO-682-47). I. Introduction Whether the intensity of domestic competition beyond trade liberalisation has a positive influence on economic growth is an open question, both theoretically and empirically. The existing theoretical work does not provide a clear-cut answer to whether a monopolist's higher tendency to innovate is outweighled by the productivity gains induced by competition (see Rey 1997 for a survey of this literature, anid Aghion and Howitt 1998 for a recent theoretical treatnent). A number of studies have attempted to settle this issue by using industiy or firm level data: (i) increases in concentration are associated with reductions in technical efficiency (Caves and Barton 1990, Green and Mayes 1991, and Caves and Associates 1992); (ii) fewer competitors and higher average rents are associated with lower productivity growth (Nickell 1996); (i.ii) trade liberalisation and industrial deregulation can have positive effects on iinn-level productivity (for example, Harrison 1994 cn C6te d'Ivoire, LaPorta and Lopez-de-Silanes 1999 oln Mexico, and Graham et all. 1983 on U.S. airline deregulation); and (iv) increases in concentration and other measures of monopoly power dampen innovative activity (Gercski 1990 and Blundell et al. 1995). The ava.ilable empirical studies fail to capture economy-wide elffects. More importantly, they focus on manufacturing industries and do not include any service or network-based industries such as computer software, telecommaunications, and advanced logistics services. Financial services also are increasingly assuming network characteristics as banks market others' products in addition to their own and play the role of market makers. In this paper, we adopt a more direct approach and study whether different availabler measures of intensity of competition at the economy-wide level are positively associated with economic development using data from over 100 countries over thle ten-year period 1986 through 1995. Specifically, we investigate whether higher levels of domestic competition, while controlling for the degree of trade liberalisation, are significantly and robustly correlated with faster cunrent and ]^uture rates of per capita economic growth rates. The existi,ng empirical growth literatute provides the techniques for testing this kind of hypothesis using cross-country regressions (Barro 1997). Although the methodology is straightforward (see Temple 1999 for a recent evaluation of methodology), the major empirical challenge is to define and assemble on a cross-country comparable basis variables that can adequately capture the intensity of economy-wide competition. We compile and construct three types of variables related to policy, structure and mobility. First, we compile qualitative policy measures to capture the quality of the microeconomic incentive regirne and the enabling legal and regulatory framework in areas that directly promote competition. Second, we compile qualitative variables and construct quantitative variab;les to reflect the extent to which market structure is concentrated from 1 an economy-wide perspective. Finally, we construct quantitative mobility variables to capture the ease with which new enterprises can enter and grow in any market. Although each competition indicator has shortcomings, this array of measures provides a richer picture of intensity of competition at the economy-wide level than using only a single measure. We develop as robust a testing methodology as possible, given that we have fewer observations of our competition variables than of the more standard growth variables. The smaller subset of countries that each competition indicator covers also varies. Our analysis proceeds in three steps. First, we build parsimonious growth models using core variables on which there is agreement in the literature. We require that the explanatory variables included in these models pass a test based on an extreme-bounds analysis (EBA) as in Levine and Renelt (1992). The second step is to study the strength of the partial correlation of our competition variables with unexplained growth from the growth models. Finally, the third step is to test the robustness of these partial correlations using a. modified EBA procedure. We find that key measures related to intensity of economy-wide competition are positively associated with unexplained growth. Most importantly, after controlling for the traditional fragility in growth models, we find that one policy measure, namely whether antitrust or antimonopoly policy is perceived as effectively promoting competition, has a reliable, independent and robust statistical relationship with unexplained growth. This variable is particularly appealing because it captures the effectiveness of implementation of competition policy as perceived by key local market participants. II. Indicators of Intensity of Economy-Wide Competition We begin our analysis by compiling and constructing as many as possible relevant quantitative and qualitative variables that may, however imperfectly, capture intensity of competition at the economy-wide level. We classify such measures into three categories. The variables and the data sources are described in Table 1. Competition policy variables. A first set of measures captures the quality of the microeconomic incentive regime and the enabling legal and regulatory framework in areas that directly promote economy-wide competition. These measures are indirect measures of intensity of competition, in the sense that they reflect relevant policy input rather than any directly resulting intensity of competition output. We have identified seven cross-country comparable policy variables that reflect economy-wide competition beyond trade liberalisation. The most promising policy indicator, since it is most directly related to the effectiveness of competition policy, is a qualitative variable that we call ANTITRUST. This variable is based on direct responses from over 3,000 top business 2 executives of large international and domestic finns in 53 countries to the question: "Does anti-trust or anti-monopoly policy in your country effectively promote competition?" This variable became first available in 1996, based on an overhaul of the former IMD and World Economic Forum World Cornpetitiveness Report.I A second related indicator is provided by the qualitative variable UNFAIR. This variable is based on a somewhat different question: "Do anti-trust laws prevent unfair competition in your country?" We use the 1989-1996 averaLge for this indicator. The shortcoming of this question as posed is that 'unfair competilion' might mean predation against cashl-constrained rivals, but also might allude to pressure on a less efficient enterprise applied by a more efficient rival (with possibl[e overtones of deceptive advertising-style consumer protection, employment pirotection or even anti-dumping style protection against lower priced foreign products). To the extent that respondents impute the latter types of rneaning to 'unfair competition', the indicator may fail to reflect the intensity of competition or effectiveness of anti-trust policies. We considered three other policy indicators. PCONTROL refers to the extent to which price controls are used throughout the economy on various goods and services. Variables BUSFREE1 and BUSFREE2 reflect different tabulations of responses to the question: "Are busirnesses and co-operatives free to compete?' Finally, we included two intensity of competition policy variables constructed on the basis of confidlential internal assessments by 'world Bank country economists. DISTRIBUTION WB reflects the extent of pro-competition marketing and public procurement policies; PRODUCT_WB reflects the extent that pro-competition product market policies and anti-competitive behaviour by enterprises is strictly checked by a fully effective competition policy. The appeal of these indicators is that they not only reflect the judgement of economists working closely on specific countkies but also a careful effort to achieve internal consistency in rankings across countries. Structural va!riables. A second set of qualitative and quantitative variables reflects the extent to which market structure is concentrated from an economy-wide perspective. It is problematic to construct cross-country comparalble industry concentration measures in light of cross-co-untry differences in relevant market definition, technology, multi- market contact and diversification of large companies, not to mention the overwhelming data collection requirements. Given these data challenges, the most informative variable may well be a qualitative variable, MRKTDOM, which is based on answers from Under the guidance of Professor Sachs of Harvard University, the traditional annual World Competitiveness Report (WCR) variables were revised in 1996, and resulted in the compilation of a new Global Competitiveness Report series published by World Economic Forum. The ANT]:TRUST variable appears for the first time in this new publication. The original WVCR was renamed World Competitiveness Yearbook, and is nowv published exclusively by IMD. 3 business executives to the question "Do you agree that market dominance by a few companies is not common in your country?" To the extent that this indicator is negatively associated with economy-wide concentration, it is expected to be positively associated with competition and growth. The advantage of such a qualitative variable is that it incorporates the country-specific judgement of high-level executives regarding the relevant size of markets, the actual degree of market power facilitated by cross-company industrial-financial and service sector ownership links, and other difficult to quantifyr local factors.2 The extent of direct state involvement in the economy is generally expected to be positively associated with economy-wide market concentration, as state-owned enterprises usually hold exclusive monopoly franchises or often have advantages such as soft budget constraints to pre-empt entry. We use two different measures of the relative size of the state-owned sector as of 1985-SOE1 and SOE2.3 Finally, in spite of all the conceptual and data limitations involved, we attempted to construct cross-country comparable economy-wide concentration ratios. In order to ensure consistency of data collection across countries, we based the indicators on a recently available international company database produced by Dun & Bradstreel:, Principal International Businesses, which contains data on some 90,000 largest companies ranging all sectors of the economy and spanning all emerging market economies. This database includes companies that make annual sales figures publicly available. We calculated two economy-wide thirty-firm concentration ratios. The first variable, S30, gives the ratio of total domestic sales of the top 30 companies to GDP.4 The second concentration variable, HERF30, is an economy-wide Herfindahl-based concentration index, the sum of squares of the total sales shares in GDP for the largest 30 firms. Unlike an average of industry-level concentration ratios, these measures take multi-market contact into account and avoid cross-country problems with the selection cf representative sectors. 2 Nickell (1996) uses a similar indicator of the intensity of competition for individual firms based on their responses to the question "Do you have more than 5 competitors in the market for your product(s)?" 3 SOE2 is included as an alternate measure of the size of state enterprises in the economy. In contrast to SOEI, a higher rating indicates that government enterprises play a less significant role in the economy. SOE2 is expected therefore to be positively associated with growth. 4 Since the database does not include the value of exports for each firm, we attempted to separately collect this information by contacting local exporters association, business magazines, government statistical offices, chambers of commerce and, in some cases, the individual firms. Based on results for 20 countr:ies, we decided to proxy domestic sales by re-scaling worldwide company sales by national exports since: (i) information on company exports was usually based on surveys conducted on a random sample of firms and not necessarily carried out every year; (ii) in some cases, company exports included both home and foreign plant production; (iii) the correlation between our proxy and the more detailed export accounting was C.79 with a p-value of zero. 4 Mobility variables. A third set of measures attempts to capture enteiprise mobility, that is, the ease wvith which new enterprises canl enter and grow in any market.5 ENTREPRENEUR is a proxy for the relative size of the entrepreneurial pool. It is defined as the sihare of total employers in the labour force, namely owner-managers working on their own account or with one or a few partners, making the operational decisions affecting the enterprise, and engaging one or more persons to work for them in their business. T'o the extent that the relative size of the pool of owner-managers reflects economy-wide ease of entry into markets (among other factors), it would be expected to be positively associated with growth. Our second mobility indicator, AGE25, is a variable that measures the average age of 25 of the largest 30 companies by sales in each country based on the year of incorporation.6 AGE25 should arguably be lower in countries where there exists the potential for entrepreneurs with new ideas to start a new company that if successful could become a major national player over time. Perhaps the ideal example is provided by Microsoft Inc. in the United States. To the extent that lack of competitive pressure limits turnover among the largest firms or that entry and expansion barriers prevent smaller innovative firms from growing into larger firms over time, this measure should be negatively associated with growth. As with other variables, it is important to emphasise that this measure is at best an imperfect proxy, given that incumbent large firms may successfully remain large over time in a sufficieintly competitive environment by practices such as aggressively introducing new products, constanitly adopting technological[ly cost.-efficient practices and modifying their core business in response to changing demands. III. Parsimonious but Robust Growth Models The methodology chosen to conduct our exercise is strictly related to our main objective. We want to determine the impact of indicators of intensity of economy-wide competition on growth and to provide results which can be regarded as general as possible, that is, which abstract from the specific sample size chosen or from industry, sector or country-specific issues. The methodology must also adequately deal with the fact that we have fewer observations of our competition variables than of the more standard growth variables. The smaller subset of countries that each competition 5 Although a range of variables related to turnover exist (capturing processes of entry and exit, variations in sizes and market shares of continuing business units, and changes in the control of enterprises), considerations cf data availability for a sufficiently large number of countries restricted our focus to the reported measures. C)n other mobility measures and links with p:roductivity growth, see Caves (1998). 6 Observations on the year of incorporation were missing in an apparently non-systematic manner for a small number of larger firms, which is the reason for limiting the variable to 25 of the largest 30 companies for each country. 5 indicator covers also varies. To deal with this problem, we developed a parsimonious but robust methodology, combining the 2-stage OLS analysis with the Extreme Bounds Analysis (EBA). Because the intensity of competition variables had not been compiled in the past, we cannot construct panel data. This further restricts our ability to use the framework of Mankiw et al. 1992. Consequently, our only alternative is to use what Temple 1999 calls "informal growth regressions," with innovative applications of the EBA. We believe that our methodology allows us to use all available information in an efficient manner. Several issues have been thoroughly studied in the empirical literature on growth andL for some of them a general agreement has been reached.7 Table 2 describes the list of potentially important growth variables, identified by past studies, that will be used at this stage.8 As dependent variable for all models, we use the average annual growth rate of real GNP per capita (RGNP_G) over the period 1986-95. We focus on four core explanatory variables where there appears to be a reasonable degree of consensus: * Convergence: We include the pre-period log value of real GDP or GNP per capita, in line with the findings that higher initial levels of income constrain growth possibilities, reflecting the catch-up potential by poorer nations. * Openness: We include indices of trade openness or liberalisation, in line with the findings that a country's outward orientation and trade liberalisation enhance its growth potential. * Human Capital: We include several variables reflecting the level of human capital accumulation, both pre- and in-period, which have been found to favour growth. * Investment in Physical Capital: "There is a robust correlation between investment rates and growth...," Temple (1999). Accordingly, we use the share of Gross Domestic Investment in GDP as an explanatory variable. 7 For a survey of studies focusing on the politics of growth, income, consumption distribution and fiscal policy, see Alesina and Perotti (1997); for studies focusing on "catching-up" or mean convergence, human capital and production factor accumulation, see Barro (1991, 1997, 1998), Baumol (1986), De Long (1988), Hansson and Henrekson (1997), Jones (1995), Lucas (1988), Mankiw (1995), Mankiw et al. (1992), Romer (1990), Young (1995); for a study focusing on schooling, see Summer and Heston (1988); for studies focusing on financial development see Bencivenga and Smith (1991), Greenwood and Jovanovic (1990), King and Levine (1993), Levine (1991); for studies focusing on economic openness, see Dollar (1992), Harrison (1996), Sachs and Warner (1995). 8 Note that some variables thought to have an important explanatory power in growth models, such as schooling-related measures, are not considered at this stage. The reason for their absence is related to the fact that the sample of these variables only partially overlaps with the other growth variables. Their use at this stage would have considerably reduced the number of countries taken into consideration, with negative effect on the efficiency of our estimates. These variables will be used later on smaller samples, when the correlation between growth and competition variables is tested. 6 Each one of ourT growth models have two "core" variables belonging to two of these four groups (see Table 2 for the definitions of variables): Model 1: RGNP_G; - a, + LGDP85i P1i + SACWAR9Si, P21 + Uil Model 2: RGNP_G; = cc2 + LGDP85i 112 + GD195i P22 + Ui2 Model 3: RGNP_Gi = a3 + LGDP85i 113 + LLIFEM85i 1323 + Ui3 Model 4: RGNP_Gi = a4 + SACWAR95i 114 + GD195i 124 + Uj4 Model 5: RGNP_Gi = a5 + SACWAR95i 115 + LLIFEM85i 125 + Ui5 Model 6: RGNP_G; = OC6 + GDI95i P16 + LLIFEM85i 326 + Ui6 To ensure parsimony, we augmented these models only with variables that passed a modified version of a robustness test based on EBA.9 The robustness test for including variable I in Model k is conducted as follows. We add I and a rotating set of three other variables (denoted by Z) as regressors to Model k above. RGNP_GCi = (xk - Xi,k + Ii PI + Zi z + Uik We therl run this regression with all possible sets of Z-variables to find the extreme bounds (i.e., maximum and minimum) for the estimate of /3,.10 If the estimate is significant at both extreme bounds, variable I is coinsidered to be a robust variable. We repeat this exercise for all six models and for all variables listed in Table 2.1 l The results are reported in Table 3. Each cell in Table 3 reports the minimum and maximum estimates of the coefficient of the variable of interest and their significance levels for the model indicated at the top of the column. Robust variables are identified by the shaded cells. For the purposes of this exercise, we used the same set of 83 observations that are common to all variables included in our investigation. We also created a seventh model by using the variables that appear in at least two of the six regressions. We report the results from these seven regression models in Table 4. 9 See Edward Leamner (1983) for a general discussion of EBA and Levine and Rernelt (1992) for an application to cross.-country growth regressions. 10 The pool of Z-variables varies according to which model and variable of interest we are considering. For instance: wherL the variable of interest is POPG95 in model 4, Z variables are chosen from the following set: {LGNP85, GDFI95, LLIFET85, LLIFEF85, LFERT, LPOP85, XGDP95, MGDP95, OPEN95, BUDG95, TA'K95, INFL95, GD195, LLIFEM85, FAREAST, OIL, TRANS, LATIN, AFRICA}. 1 l We disregard results from regressions that exhibit significan: multicollinearity, as evidenced by having a variance inflation factor greater than 10. Chatterjee and Price (1991) provide a definition of the variance inflation factor and a discussion for selecting a cut-off level. Although the EBA, with this screen, can potentially include imore variables as being robust, the use of the screen (lid not have a material effect on our results. 7 Note that Model 4 is nested in Model 2. Moreover, it is the only model that fails the Ramsey test for functional mis-specification and the Jarque-Berra test for the normality of the residuals. We therefore disregard Model 4 for the rest of our analysis. According to the applications of the J-test (Davidson and MacKinnon 1981) and the Cox-Pesaran-Deaton test (Pesaran and Deaton 1978) Model 1 appears to be better than Models 6 and 7; and Model 3 than Models 5, and 7. But the tests concerning other binary comparisons were inconclusive. Of these models, all explanatory variables of Model 7 are measured before the period. The only in-period variable in Models 1 and 5 is the openness variable, which reflects the reform efforts rather than ex post growth. The other models include in-period investment variables, measured as the relative intensity of investment (e.g., average investment-to-GDP ratio) rather than the absolute amount of investment. Therefore, they are not so much subject to the usual endogeneity criticism. IV. Intensity of Competition and Unexplained Growth Our primary aim is to study the strength of association between intensity of economy- wide competition and growth. In order to accommodate the fact that we have fewer observations of our competition variables than of the more standard growth variables and that we would like to use all available information as efficiently as possible, we test for the correlation between our competition variables and the residuals from our robust growth models. In our attempt to utilise as much information as possible, we also have extended the growth models 1, 2 and 5 to the maximum number of available observations. This approach is justified because, as reported in Table 5, these models remain stable with respect to the extension whereas the others do not. We also have checked and report in Table 5 whether the statistical properties of all models apply to smaller sets of countries for which competition variables are available. We use these test results in evaluating the reliability of our conclusions. In the second stage of our analysis, we test whether any of our competition variables exhibit a robust correlation with residual growth rates. As some of our qualitative competition variables arguably could reflect institutional factors not directly related to intensity of competition, we have compiled an additional list of variables that could potentially explain growth to control for such factors. These variables are described i.n Table 6. We include them at this stage, instead of the first stage for two reasons: * These variables, unlike the ones used in building our parsimonious growth models, do not have as solid an established link to growth in the existing literature. * They represent alternative hypotheses to our investigation. For example, one could argue that the ANTITRUST variable reflects general institutional quality rather than the more focused government efforts to foster competition. Were this alternative true, 8 we would expect the institutional quality variable, INSTITQUALITY, to exhibit as strong a link to r esidual growth as ANTITRUST. Variables that appear to be correlated with unexplained growth. We report the correlation between the competition variables (and alternatives to compet:ition variables) and residual growth in Table 7. Only correlation coefficients significant at 10 percent are shown (starred coeifficients are significant at 5 percent). ANTITRUST and its earlier version, UN1FAIR, appear as top performers, but they are not alone. Competition policy variables by far exhibit the highest degree of correlation with residual growth tharn any other group. Besides ANTITRUST and UNFAIR., PCONTROL and DISTRIBUT ION_WB are significant at the 5 percent level in at least one model. Among the structural variables, MRKTDOM and SOE2 have significarLt and positive correlations with residual growth, supporting the hypothesis that more competitive economies tend to have higher growth rates. Mobility variables too, despite their tentative nature, show some correlation with residual growth. These results support our belief that thiere should be more serious efforts to collect and compile international data on measures of corporate and entrepreneurial mobility. Two of the alternative-to- competition variables appear to have significant correlation with residual growth. One of them reflects the quality of environmental policies and regulations, and the other that of general policy making. These variables are not correlated with ANT:[TRUST; they probably reflect other factors than competition policy. However, in our second stage EBA analysis reported below, their correlations with residual growth are not robust. Tests of robustiess. In interpreting the correlations in Table 7, we should keep in mind the fragility of cross-country statistical relationships as noted by Levine and Renelt (1992). It is the:refore important to test their robustness. We once again use the EBA, treating each variable in Table 7 as a variable of interest. We thus determine whether controlling for different sets of factors weakens the raw correlation with residual growth. The techniqu;e is similar to the one previously described with the difference that now, there are no core vaxiables and the rotating set of "other" variables is restricted to only two variables due to sample size concerns. For each variable in Table 7, we run all possible regressions with two additional variables chosen froim the pool of variables in Tables 1 and 6. The results of this EBA are reported in Table 8 where each cell shows the maximurn and mninimum coefficient estimate for the variable of interest and their significance levels. Only the shaded cells have both extreme bounds significant at 10 percent. Our analysis identifies a relatively robust relationship with growth for ANTITRUST and to a lesser degree for UNFAIR and AGE25. Only the extreme bounds of these three variables remain significant at 10 percent level thrDughout the rotations of additional explanatory variables for at least one model. ANTITRUST and UNFAIR have robust 9 correlations with the residuals of models 1 and 3 which are, as discussed above, superior to models 5, 6, and 7. Moreover, models 1 and 3 could be reliably restricted to the sample size of these variables. All other associations are eliminated in our test for robustness. In particular, the variables reflecting institutional quality that had significant raw correlations with growth are not robust. Moreover, these three variables complement one another in explaining growth: AGE25 is robust in the only model where ANTITRUST is not. In models where ANTITRUST is the most robust (models 5 and 6), UNFAIR is not robust at all. In fact, UNFAIR is the predecessor of ANTITRUST and emphasises the effectiveness of competition policy in dealing with unfair practices rather than its ability to prevenl explicitly anti-competitive practices. AGE25 is robust in only one model, Model 2, which could be reliably restricted to the sample size of AGE25. Model 2's unique feature is the absence of the Far East dummy. As we argue below, the link between competition and growth appears to be most tenuous in that region, and hence the poor showing of ANTITRUST and UNFAIR for Model 2.. However, AGE25 captures the youthfulness of the leading companies in this region. For all Far East Asian countries except Philippines, AGE25 is below the sample median. Based on these EBA findings reported in Table 8, the correlation between ANTITRUST and growth is robust. The size of the coefficient, varying between 0.28, and 0.47, implies that the link between active promotion of competition policy ancl growth may be economically important. Although causality cannot be inferred from our analysis, a 1-point increase in the perceived effectiveness of antitrust enforcement is associated with an increase of about 0.4 percentage points in the annual growth rate. Causality. Not surprisingly, ANTITRUST appears with a positive and significant sign when included as an additional regressor in any of the cross-country growth regressions that we tested. Although the association between ANTITRUST and long- term growth is irrefutable, the causal link between ANTITRUST and long-term growth cannot be established in a simple regression analysis because ANTITRUST is simultaneously determined with growth. The same simultaneity problem applies tLo SACWAR95. An application of an instrumental variable technique for each one of the three variables provides some support for the hypothesis that each variable has a distinct causal effect on long-term growth. For the instrumental variables approach, it is necessary to create a model with only exogenous variables except for the endogenous variable for which instruments are used. Model 7 augmented by one of the endogenous variables satisfies that requirement:. Instrumental variables for the endogenous variable are selected from the categories identified in Table 2. All instrumental variables measure pre-period values (i.e., 1985) and thus are exogenous. 10 The best instruments for ANTITRUST appear to be terms of trade in 1985, growth in government consumnption (three-year average as of 1985), and population in 1985. With these instrumentls, ANTITRUST retains its significance as a regressor. Using other indicators of fiscal health as instrumental variables, instead of growth in government consumption, retains the same results at somewhat weaker significance levels. A relatively advantageous foreign trade position, a large domestic marLet, and small government appear ito be conducive to meaningful antitrust enforcement. On the other hand, the best instruments for the trade opernmess index, SACWAR95, appear to be exports (as a percentage of the GDP in 1985), population growth (three-year average as of 1985), and inflation in 1985. Rather perversely, inflation appears to have positive correlation with SACWAR95. One plausible explanation would be that high inflation coumtries in 1985 were mostly Latin American countries that subsequently opened up their economies in the 1990s. However, after controlling for inflation and life expectancy (another significant determinant of SACWAR95), Latin American countries have lower SACWAR95 values. Considering that politics in practice appears to dominate a country"s decision to open up its markets, the strange mix of these variables should not be surprising. A closer look at individual countries. We can identify four distrnct groups of relatively comparable countries for which we have ANTITRUST observations. Exploring the competition and growth link among thiem is instructive for understanding both the source and shortcomings of our results. Findings are reported in Table 9. The three Latin American Southern Cone countries have the same rankings with respect to both ANTITRUST and growth residuals (Panel A). During this period, Chile was the leading reformer in Latin America, buildirig a cornpetitive economy through privatisation and deregulation. Other macro and trade policies moved roughly in tandem in these three colmtries, with Chile following a more cautious capital account liberalisation and achieving stabilisation earlier than the others. Yet all these macro factors are controlled for in the models and in the EB A analysis. Similarly, there is almost a perfect correlation between competition and growth among the peripheral members of the European Union (Panel B). Most observers would likely agree that the Irish or the Portuguese business environment has been far more competitive than that of Greece during the period under investigation. It is reasonable to postulate that lack of competition is one of the leading explanations for Greece's sub-par growth performance. For the group of small European economies, toD, there is a very strong correlation between ANTITRUST and growth (Panel C). What is interesting is that for the so-called Asian tigers, this correlation disappears (Panel D). These findings suggest that the 11 effectiveness of competition policy may not be uniform across different groups of countries. One contentious issue is whether there is any role for competition policy beyond trade liberalisation in a small open economy. We have several findings that confirm the plausibility of such a role. First, Models 1, 2, and 5 explicitly control for trade openness and their residuals still show a robust correlation with ANTITRUST. Second, alternate measures of trade openness appear in the EBA procedures and they do not appear to be weakening the correlation between growth and ANTITRUST. Third, the instrumental variables approach, discussed above, shows that ANTITRUST and SACWAR95 impact growth through different channels. Finally, the link between ANTITRUST and growth appears to be more significant for small open economies in Europe. Our findings therefore suggest a strong complementary role for competition policy in stimulating growth beyond trade liberalisation and international openness. V. Conclusion Despite difficulties and data problems, we have developed different sets of variables that measure the intensity of economy-wide competition. We then created traditional and robust cross-country growth models and explored the correlations between competition variables and residual growth rates. Our results indicate that there is a strong correlation between the effectiveness of competition policy and growth. We tested the robustness of this relationship by controlling for other factors that arguably may be proxied by our competition policy variables. The relationship appears to be robust. Our analysis suggests that the effect of competition on growth goes beyond that of trade liberalisation, institutional quality, and ;a generally favourable policy environment. However, this link appears to be more tenuous for Far Eastern economies. This observation cautions us against being overly simplistic in promoting the importance of competition policy as a major and independent determinant of long-run growth. It suggests that there remains ample scope for further empirical work in this area. Given the tentative but promising links between mobility- related variables and growth, there should in particular be more systematic efforts to collect and compile internationally-comparable data on measures of corporate and entrepreneurial mobility. 12 References Aghion, P., and P. Howitt (1998) Endogenous growth theory, MIT Press, Cambridge, Massachusetts. Alesina, A., and R. Perotti (1997) The Politics of growth, in: Government and growth, Bergstr(im V. (Ed.), Clarendon Press, Oxford, U.K. Barro, R. J. (1991) Economic growth in a cross-section of countries. Quarterly Journal of E,conomics, 106, 407-443. Barro, R. J. and J.W. Lee (1996) International Measures of Schooling Years and Schooling Quality, American Economic Review, 86, 218-23. Barro, R. J. (1997) Determinants of economic growth, MIT Press, Cambridge, Massachusetts. Barro, R. .l. (1998) Overview of empirical evidence on the deterninants of economic growth. Processed. Baumol, W. J. (1986) Productivity growth, convergence and welfare. The American Economic Review, 76, 1072-85. Bencivenga, V. R., and B. D. Smith (1991) Financial intermediation and endogenous growth. Review of Economic Studies, 58, 195-209. Blundell, R., R. Griffith and J. Van Reenen (1995) Dynamic Count Data Models of Technological Innovation, Economic Journal, 105, 333-344. Bumnside, C., and D. Dollar (1997) Aid, policies and growth. The World Bank Policy Research WorkingPaperNo 1777. Caves, R. E. (1998) Industrial organization and new findings on the turnover and mobility of firms. Journal ofEconomic Literature, 36, 1947-1982. Caves, R. E_. and D. R. Barton (1990) Efficiency in US. Manufacturing Industries. MIT Press, C'ambridge, Massachussets. Caves, R. E. and Associates (1992) Industrial E)ficiency in Six lations. MIT Press, Carnbridge, Massachussets. Chatterjee, S., and B. Price (1991) Regression analysis by exa iple, John Wiley & Sons, New York. Cox, D. (1.961) Tests of separate families of hypotheses. In: Proceedings of the 4th Berkeley symposium on mathematical statistics and probability, 1, University of Californiia Press. Cox, D. (1962') Further results on tests of separate families of hypotheses. Journal of Royal Statistical Society, B, 24, 406-424. Davidson, R., an(d J. MacKinnon (1981) Several tests for model specification in the presence of alternative hypotheses. Econometrica, 49, 781-793. De Long, J. B. (1988) Productivity growth, convergence and welfare: comment. The American Economic Review, 78(5), 1138-54. Dollar, D. (1992) Outward-oriented developing economies really do grow more rapidly: evidence f-rom 95 LDCs, 1976-1985. Economic Development and Cultural Change, 523-44. 13 Dun & Bradstreet (1997) Principal International Businesses, Disc Series 1997. EBRD (1995, 1996) Transition report. European Bank for Reconstruction and Development, London. Freedom House (1996) Freedom in the world: The annual survey of political rights and civil liberties, 1995-1996. Freedom House, New York, NY. Geroski, P. A. (1990) Innovation, Technological Opportunity, and Market Structure. Oxford Economic Papers, 42, 586-602. Green, A., and D. G. Mayes (1991), Technical Inefficiency in Manufacturing Industries. Economic Journal, 101, 523-538. Greenwood, J., and B. Jovanovic (1990) Financial development growth and the distribution of income. Journal ofPolitical Economy, 98, 1076-1107. Gwartney, J. D., and R. A. Lawson (1997) Economic freedom of the world. annual report. Fraser Institu{w, Vancouver. Hansson, P., and M. Henrekve-n (1997) Catching up, social capability, government size and economic growth. In: Government and growth, Bergstrom V (Ed.). Clarendon Press, Oxfort U.K. Harrison, A. (1994) Productivity, Imperfect Competition and Trade Reform: Theory and Evidence, Journal of International Economics, 36, 53-73. Harrison, A. (1996) Openness and Growth: A Time-Series, Cross-Country Analysis for Developing Countries, Journal of Development Economics, 48, 419-447. IMD, The world competitiveness report (various years, to 1995); The world competitiveness report (1996). International Labour Organisation (1997) Yearbook of Labour Statistics, Geneva. Jones, H .(1975) An introduction to modern theories of economic growth. King, R. G., and R. Levine (1993) Finance and growth: Schumpeter might be right. Quarterly Journal of Economics, 717-37. LaPorta, R., and F. Lopez-de-Silanes (1999) The Benefits of Privatization: Evidence from Mexico, Quarterly Journal of Economics, 114, 1193-1242. Learner, E. E. (1983) Let's Take the Con Out of Econometrics, American Economic Review, 73, 31-43. Levine, R. (1991) Stock markets, growth and tax policy. Journal of Finance, 46, 1445- 1465. Levine, R., and D. Renelt (1992) A sensitivity analysis of cross-country growth regressions. The American Economic Review, 82(4), 942-63. Lucas, R. E. (1988) On the mechanics of economic development. Journal of Monetary' Economics, 22, 33-42. Mankiw, N. G. (1995) The growth of nations. Brookings Papers on Economic Activity, 1, 275-310. Mankiw, N. G., D. Romer, and D. N. Weil (1992) A contribution to the empirics of economic growth. Quarterly Journal of Economics, 107, 407-3 7. Nickell, S., Competition and Corporate Performance, Journal of Political Economy. 1996. 14 Pesaran, H., and A. Deaton (1978) Test in non-nested non-linear regression models. Econometrica, 46, 677-694. Political Risk Services Group (1998) International Country Risk Guide. Rey, P. (1997) Competition policy and developmenit. University of Toulouse Mimeo, Toulouse. Romer, P. M. (1990) Human capital and growth: theory and evidence. Carnegie- Rochester Conference Series on Public Policy, 32, 251-86. Sachs, J. D., anid A. Warner (1995) Economic :reform and the process of global integration. Brookings Papers on Economic Activity, 1, 1-118. Summers, L., and H:eston (1988) A new set of international comparisons of real product price level estimates for 130 countries, 1950-1]985, Review of Income and Wealth. Temple, Jonathan (1999) The new growth evidence, Journal of Economic Literature, XXXVII, 112-156. WIPO (1997') Industrial property statistics. World Intellectual Property Organisation, Geneva. World Bank (1998), World development indicators, NVashington D.C. World Economic Forum (1996) The global competitiveness report. Geneva. Young, A. (1995) The tyranny of numbers: confronting the statistical realities of the East Asiani growth experience. Quarterly Journal of Economics, 641-380. 15 Table 1. Measures of Intensity of Economy-Wide Competition Variables Definition Period Source (1) #obs. Std. Dev. Min Max ANTITRUST Anti-trust or anti-monopoly policy effectively promotes competition (2) 1996 GCR 52 0.7921 2.130 5.470 BUSFREEI Are businesses and co-operatives free to compete? (3) Average EFW 115 2.1252 2.500 10.000 BUSFREE2 Are businesses and co-operatives free to compete? (3) Average FH 115 2.1344 2.500 10.000 94-95 Policypeariables DISTRIBUTION_WB State intervention in marketing and public procurement systems (4) 1997 PREM 130 0.8225 1.000 5.000 PCONTROL Extent of price controls on various goods and services (5) Average EFW 112 2.2475 0.000 9.500 89-95 PRODUCT_WB State intervention in product markets (6) 1996 PREM 130 0.7686 1.000 5.000 UNFAIR Do anti-trust laws prevent unfair competiton in your country? (7) Average WCRIWCY 49 1.0818 2.197 6.902 89-96 MRKTDOM Market dominance by a few companies is not common (2) 1996 GCR 52 0.7802 2.000 5.480 S30 Concentration ratio of top 30 firms ranked by domestic sales over GDP 1996 D&B 53 0.2297 0.009 0.932 Structural HERF30 Herfindahl index of top 30 firms by sales (shares of GDP) 1996 D&B 59 19.2477 0.001 87.486 Variables SOEI SOE value added as % of GDP (8) 1985 WD198 49 0.0774 0.006 0.350 SOE2 Size of govemment enterprises in the economy (9) 1985 EFW 103 2.3569 0.000 10.000 Mobility AGE25 Average age of 25 firms within the top 30 firms ranked by total sales 1996 D&B 42 14.2192 9.000 71.160 Variables ENTREPRENEUR Share of owner-managers in labour force Average ILO 43 0.1433 0.003 0.639 Notes: (1) D&B: Dun & Bradstreet (1997) (with company data on largest companies by employment, based on stock exchanges, employment bureaus, ministries of labor, post offices, manufacturing censuses and surveys); EFW: Economic Freedom of the World, Gwartney et.al. (1997); FH: Freedom House (1996); GCR: World Economic Forum, Global Competitiveness Report (1996); ILO: International Labour Organisation (1996); PREM: Poverty Reduction and Economic Management Network, "Country Policy and Institutional Assessment", The World Bank (confidential intemal assessments by staff economists, various years); WCRIWCY: IMD, World Competitiveness Report(to 1995)/ World Competitiveness Yearbook (1996); WD198: World Development Indicators, The World Bank (1998). (2) 1= strongly disagree to 7 = strongly agree (3) The higher the rating the greater the freedom to compete (10 = countries for which businesses and cooperatives were most free to compete). EFW modified the original FH survey team ranking by reducing the rating for several countries based on EFW's substantial evidence that the FH rating was overly generous. (4) 1 = widespread interventions with state marketing monopolies over agriculture and exports; 3 = some entry/exit barriers with reform program underway; 5 = no marketing monopolies, pro-competition public procurement system in place (5) 0 = widespread use of price controls throughout economy; 10 = no price controls, more than 90 percent of companies can set prices freely (6) 1 = widespread price interventions and reservation policies for selected products; 3 = progress towards price decontrol and full cost recovery for utilities, with effective implementation of competition policy; 5 = no price controls, full cost recovery, anti-cornpetitive behavior by firms strictly checked (7) 0 = strongly disagree to 10 = strongly agree (8) Value added of state enterprises is estimated as their sales revenue minus the cost of their intermediate inputs, or the sum of their operating surplus and wage payments. (9) 0 = economy dominated by SOEs (employment and output in SOEs exceeds 30% of total non-agricultural employment/output); 10 = very few SOEs, less than 1 % of country's output _ Table 2. Variables Used in Parsimorious Growth Models Std. Variable Definition Period Source (1) # obs Dev. Min Max long-run growth rgrp_g Real GNP per capita growth (annual %) Average WD198 161 0.04 -0.14 0.09 86-95 Igdp85 Log of GDP per capita (constant 1987 US$) 1985 WD198 153 1.44 4.31 10.14 convergence Ignp85 Log of GNP per capita (constant 1987 US$) 1985 WD198 148 1.45 4.43 10.23 mgdp95 Imports of goods and services (% of GDP) Average WD198 172 0.26 0.04 1.77 86-95 open95 Import plus export (% GDP) Average WD198 172 0.48 0.06 3.64 openness 86-95 sacwar9!5 Sachs and Wamer openess index Average SW 108 0.44 0.00 1.00 86-93 xgdp95 Exports of goods and services (% of GDP) Average WD198 172 0.25 0.02 1.87 86-95 Ifert85 Fertility rate, total (births per woman) 1985 WD198 187 0.52 0.31 2.19 Ilifef85 Log of life expectancy at birth, female (years) 1985 WD198 187 0.20 3.63 4.39 human capital llifem85 Log of life expectancy at birth, male (years) 1985 WD198 187 0.19 3.55 4.31 llifct85 Log of life expectancy at birth, total (years) 1985 WD198 187 0.20 3.59 4.35 gdfi95 Gross domestic fixed investment (% of GDP) 9verage WD198 170 0.09 0.09 0.69 investment 6- gdi95 Gross domestic investment (% of GDP) Avera WD98 172 0.09 0.09 0.69 86-95 budg95 Overall budget deficit, induding grants (% of GDP) Ave rage WDI98 128 0.05 -0.25 0.32 86-95 9 2 .5-.503 fscal policy infl95 Inflation, consumer prices (annual %) Average WD198 140 2.71 -0.03 27.13 tax95 Tax revenue (% of GDP) Average WD198 130 0.10 0.00 0.47 86-95 Ipop85 Log of population, total 1985 WD198 200 2.11 10.52 20.77 population popg95 Population growth (annual %) Average WD198 200 0.01 -0.01 0.06 africa Sub-saharhan countries Average 209 0.40 0.00 1.00 86-95 fareast East Asia dummy 209 0.19 0.00 1.00 dummies latlin Latin Amrican country dummy dummy 209 0.31 0.00 1.00 oil Oil producing countries dummy 209 0.31 0.00 1.00 trans Transitional economies East Europe dummy 209 0.33 0.00 1.00 Notes: (1) SW: Sachs and Warner (1995); WD198: World Development Indicators, The World Bank (1998). Table 3. Results of the EBA Analysis for Growth-Related Variables Ml M2 M3 M4 MS M6 significance significance significance significance significance significance coefficent level coefficient level coefficient level coefficient level coefficient level coefficient level lgdPmS -O.i 4 a .. .. . .. -0.018 0% 0 0 01 0% nM -0002 a 8% -0.009 1% -0>000 2ax gdn95 mli 0.083 6% 01 8% .. ) .. na 0.006 39 %a na .0 sacwar95 Max 0.028 0% 35 O n . 0.023 0na IlHe5 min 0.071 0.068 13% 0.0300.071 l0% Max 0.231 0% 0.198 0% na .a 0223 0% llifet85 mli 0.037 6% 0.037 6% na ,a -0.012 56% na Max 0.124 0% 0,114 0% .a na 0.096 0% na fgnp8as mini na na nas na na nas -0.016 % O 0,020 0 -0.0 :0 Max na itsa it na na it -0.0o2 27% o -0.00 gdfI95 min 0.08 53% its 93% 0.000 7% 0a 2ts 0.084 1093% ita i Max 0.235 00 O it is 0. 207 0% na it 0.221 0% na it llIfefBs mlin 0.030 14% 0.030 14% na it -0.024 34% its it it i Max 0.120 0% 0.110 0% na it 0.095 O% its its it its lifet:5 min 0.037 O% 0.037 1% its its -0.013 51% its it na its Max 0.124 0% 0.114 0% na ns 0.098 O,% its02 its is 6 fareast rnint * 17 0.014 14% 0>4 0.017 5% O.021 1% 0093 Max .040 0% 0.034 0% 4.047 0% 0.033 0% 0.038 0% -0.04 0% ffert85 m'itt -.5 % -.4 0 002 0 -0.044 0% -0.040 0% -0.039 0% Max 0.02 83% -0.00 7% -0.022 0% 0.014 15% 0.002 87% 0.009 43% lipop85 min 0.001 53% 0.000 93% 0.000 79% 0.002 23% 0.001 53% 0.000 93% Max 0.009 3% 0.006 0% 0.006 0% 0.007 0% 0.008 0% 0.006 0% popg95 mlin -16480 0%A -1.528 0% -1.28 0 -1.412 0% -1.142 0% -1.256 0% Max 0.368 38% 0.047 90% -057 5 -0.364 71% -0.104 82% -0.371 38% olt min -0.020 0% -0.019 1% -0.011 12% -0.019 0% -0.021 0% -0.018 0% Max 0.002 7738 0.001 3% 0.003 67% -0.002 72% 0.002 77% -0.001 88% trans min -0.040 7h -0.047 Wo -0.047 0% -0.038 O% -0.038 0% -0.044 0% Max 0.004 83% -0.005 73% -0.016 35% -0.002 88% 0.011 51% -0.002 88% xqdp9 m5 i -0.022 4/ -0.013 3% -0.019 O,% -0.014 2% -0.020 0% -0.013 2% Max 0.006 39% 0.012 7%. 0.001 90% 0.007 21% 0.007 27% 0.008 20% africa min -0.030 0% -0.023 O. -0.009 23% -0.021 O% -0.014 13% -0.008 34% Max 0.006 38% 0.007 34% 0.008 31% -0.016 12% 0.007 40% 0.008 33% mgdp95 min -0.020 97% -0.021 % -0.009 35% -0.080 1% -0.019 8% -0.019 8% Max 0.039 0% 0.020 11% 0.040 0% 0.017 18% 0.038 0% 0.019 13% xgdp95 min -0.011 32% -0.015 191% -0.004 69% -0.016 12% -0.014 13% -0.015 15% Max 0.043 0% 0.025 31% 0.042 O% 0.083 2% -0.013 21% 0.021 4% open95 min -0.007 19% -0.009 16o -0.003 49% -0.009 9% -0.008 14% -0.009 10% Max 0.021 0% 0.012 6% 0.021 0% 0.010 7% 0.019 37% 0.010 10% budg95 mlin -0.002 97% -0.002 97% 0.039 48% -0.010 87%1 0.016 76%1 0.018 76% Max 0.255 O% 0.208 O,/ 0.275 0% 0.162 1% 0.252 O,/ 0.186 0% tax95 mlin -0.081 1% -0.047 11% -0.085 1% -0.091 01% -0.135 0% -0.089 0% Max 0.063 9%1 0.088 1% 0.060 8% 0.048 14% 0.032 29% 0.044 14% IFi'IBS mlin -0.002 1% -0.002 09% -0.002 1% -0.002 0%/ -0.002 1% -0.002 0% Max -0.001 43% -0.001 7% -0.001 41% -0.001 6% -0.001 37%/ -0.001 6% Table 4. Parsimonious Growth Models Ml M2 M3 M4 M5 M6 M7 significance significance significance significance significance significance significance Regressors coefficient level coefficient level coefficient level coefficient level coefficient level coefficient level coefficient level convergence: lgdpB5 -0.014 0% -0.015 0% -0.013 0% Ignp85 -0.010 0% -0.008 0% -0.013 0% human capital: Ilifem85 0.055 0% 0.046 2% 0.061 0% 0.083 0% 0.087 0% 0.061 0% Ifert85 -0.024 0% -0.028 0% -0.020 8% -0.028 0% openness; sacwar95 0.016 0% 0.016 1% 0.017 0% 0.019 0% investment: gdi95 0.143 0% 0.204 0% 0.092 4% gdfi95 0.113 1% fareast 0.032 0%/ 0.022 1% 0.037 0% 0.029 0% 0.034 0% constant -0.087 25% -0.065 40% -0.133 10% -0.040 0% -0.260 0% -0.302 0% -0.105 18% statistics F-testforjointsignificance: F(5,77) 25.430 F(5,77) 24.150 F(6,76) 20.070 F(2,80) 26.210 F(4,78) 25.650 F(4,78) 22.390 F(4,78) 27.440 R-square: 0.623 0.611 0.613 0.396 0.568 0.535 0.585 Adjusted R-square: 0.598 0.585 0.583 0.381 0.546 0.511 0.563 Standarderroroftheregression: 0.016 0.016 0.016 0.020 0.017 0.018 0.017 Table 5. Goldfeld-Quandt Test for Variance Constancy with Respect to Sample Size Restrictions Ho: Model cannot be extended to the largest possible sample size. Model test statistic distribution probability Ml 1.18 F(13,77) 0.31 M2 1.33 F(12,77) 0.22 M3 2.99 F(45,76) 0.00 M5 1.05 F(14,78) 0.42 M6 2.38 F(48,78) 0.00 M7 2.96 F(51,78) 0.00 Ho: Model cannot be restricted to the set of observations where ANTITRUST, UNFAIR and MRKTDOM are available. Model test statistic distribution probability Ml 1.59 F(52,39) 0.07 M2 1.22 F(51,38) 0.22 M3 1.67 F(34,36) 0.06 M5 1.43 F(53,40) 0.12 M6 1.57 F(36,38) 0.09 M7 2.06 F(36,38) 0.02 HO: Model cannot be restricted to the set of observations where S30, HERF30 and AGE25 are available. Model test statistic distribution probability Ml 1.75 F(59,32) 0.04 M2 1.45 F(57,32) 0.13 M3 1.75 F(42,28) 0.06 M5 1.53 F(60,33) 0.09 M6 1.96 F(44,30) 0.03 M7 1.91 F(44,30) 0.03 Ho: Model cannot be restricted to the set of observations where BD-based (1) alternative-to-competition variables are available. Model test statistic distribution probability Ml 1.66 F(58,32) 0.06 M2 1.43 F(58,31) 0.14 M3 2.11 F(46,23) 0.03 M5 1.22 F(59,33) 0.27 M6 1.39 F(48,25) 0.19 M7 2.19 F(48,25) 0.02 Ho: Model cannot be restricted to the set of observations where ENTREPRENEUR is available. Model test statistic distribution probability Ml 1.38 F(65,25) 0.19 M2 1.32 F(64,25) 0.22 M3 1.05 F(46,23) 0.47 Ms 1.66 F(66,26) 0.08 M6 1.53 F(48,25) 0.13 M7 1.20 F(48,25) 0.32 Note: BD: Burnside and Dollar (1997). These altemative-to-competition variables includes ASSAS, CIVLIB, ETHNFRCT, INSTITQUALITY and POLICY (for definitions, see Table 6). Table E. Definitions of Additional Variables Used in the Second Stage EBA Variable Definition Period Source (1) Obs Std Dev. Min Max ASSAS Number of assasinatons Average 86- BD 71 0.7685 0.000 4.125 93 CMVILIB Civil liberhes 11985 BD 70 1.3545 1.000 7.000 CORRUPT CoIruption inclex (0 to 6, high to low) 1985 ICRG 126 1.2320 1.000 6.000 ENVRNMNT_WB Environmental policies and regulations (2) 11998 PREM 135 0.8750 1.000 5.000 ETHNFRCT Ethnic fractionalisation index Average 86- BD 68 29.895 0.000 93.000 93 INSTITQUALITY Instituional quality Average 86- BD 64 1.4026 2.270 8.560 93 PATENTS Number of patents granted 11995 WIPO 93 1.8E+04 0.000 1.1E+05 POLICY Economic policy index Aver3ge 86- BD 92 1.2834 -3.230 4.030 933 RULE WB Private economic activity is facilitated by a rule-based 1997 PREM 130 0.8155 1.000 4.670 govemance structure (3) corruptavg Coiruption inclex (0 to 6, high to low) Average 86- GRG 126 1.3126 0.450 6.000 95 crpriv95 Credit to private sector (% of GDP) Average 86- WDI98 166 0.3156 0.002 1.943 95 govc95 Genri. government consumption (% of GDP) Average 86- WD198 169 0.0698 0.055 0.493 95 govcgY5 Geri. government consumption growth (% of GDP) Average 86- WD198 167 0.0687 -0.340 0.521 95 land85 Land (SqKm) 1985 WD198 173 1.5E+06 20.000 9.3E+06 m1m295 M1/M2 ratio Aver3ge 86- WD198 165 0.2004 0.086 0.944 95 m295 Money and quasi money (M2) as % of GDP Average 86- WD198 163 0.2735 0.002 1.604 95 m395 Liquid liabilities (M3) as % of GDP Average 86- WD198 164 0.2860 0.003 1.715 m3ml Quasi-liquid liabilities (% of GDP) Average 86- WD198 164 0.2771 0.001 1.562 prim2585 Percentage of primary school attained in male 1385 BL 110 0.1695 0.073 0.740 population oldier than 25 pyrm2585 Average years of primary schooling in the male 1385 BL 110 1.8493 0.610 8.020 population aged 25 and over shpuppB5 Ratio of Cgov. current educaton expend. per primary 1985 BL 90 0.0851 0.029 0.449 school puJpii t) per capita GDP spread95 Interest rate spread (lending rate minus deposit rate) Average 86- WD198 144 1.4944 -0.029 16.939 Notes: (1) BD: Bumside and Dollar (1997); BL: Barro and Lee (1996); WD198: The World Bank (1998); ICRG: Intemational Countty Risk Guide, The Political Risk Services Group (1998); PREM: confidential intemal assessments by staff economists, The World Bank; WIPO: Worid Intellectual Property Organisation (1997). (2) on a 1-6 scale where 2 = no policies or investments for sustainable manage -nent of natural resourcs or pollution control, regulations inadequate or weakly enforced; 5 = comprehensive policies accompanied by credible enforcement capadity to sustainably manage key natu resources, regulations consistent with intemational norms implemented effectivsly (3) on a 1-5 scale where I = laws and regulations lack certainty and applicaton lacks predictability, property rights not well-defined or enfor 3 = credible reform process underway, limited discretion; 5 = well-functioning legal and regulatory system with low transactions costs, confli of interest regulations for pu"blic servants strictly enforced. Table 7. Correlation Matrix for Residuals and Competition Variables residuals from Ml M2 M3 M5 M6 M7 Variables # of obs. 102 101 83 102 83 83 . ^.jfl1 52 0.2826 0.3891* 0.3815* 0.4870* 0.3474* BU$FR~~~~E1 ~115 0.1758 0.1854 BUSFREE2 ~~~~~15 0.1775 0.1903 Competition 130 O.2128* Policy Variables I =. 130 0.2128* POONROL 112 0.2517* 0.2237* 0.2990* 0.2142* PRODUCTMRKT_WB 130 lJFAIR lXs2 49 0.2626 0.5091* O.4379* 0.4692* MgRKW 52 0.2738 0.3390* 830 53 Structural HeRP3O X 00> 559 0.2531 Variables SOEI 49 0E 103 0.2048 0.2348* 0.1944 Mobility Variables A2 42 -0.3031 -0.2920 ; TREPRENEUR 43 0.3555* 0.2793 0.3239 ASSAS 71 CIVILLIB 70 Alternative CORRUPT 124 variables that R U P 130 0.2004 0.3030* 0.3295* O.3794* could potentially ETHNFRCT 68 explain residual INSTITQUALITY 64 growth PATENTS 93 POLICY 94 0.2454 0.2626* 0.2553* 0.2666* RULE WB 130 Note: Table shows correlations that are significant at 10%, star indicates significance at 5%. Table 8. ETA Results: Robustness of Competition and Other Variables Ml M2 M3 M5 M6 M7 signifh- signifi. signifi- signifi- signifi- signifi- coef. cance coef. cance coef. cance coef. cance coef. canoe coef. cance XNTfMUSD I min 0o, 8% 0.005 23% 0.007 ' 0.0 '3% 0.010 2% 0.005 9% max 0.0t0 M% 0.009 5% 0.010 2% 0.d14 0% 0.014. 0% 0.009 2% n BUSFREEI min 0.000 80% -0.001 64% 0.000 84% 0.001 43% 0.001 45% 0.000 90% max 0.003 7% 0.002 19% 0.003 9% 0.003 4% 0.004 2% 0.003 5% t BUSFREE2 min 0.000 80% -0.001 64% 0.000 .34% 0.001 43% 0.001 45% 0.000 90% mrnax 0.003 7% 0.002 19% 0.003 9% 0.003 4% 0.004 2% 0.003 5% o DISTRIBUTION_W13 min -0.010 14% -0.010 17% -0.010 t8% -0.006 41% -0.006 44% -0.010 18% max 0.001 84% 0.000 95% 0.003 367% 0.005 50% 0.008 32% 0.005 50% PCONTROL min 0.000 74% -0.001 54% 0.000 30% 0.000 89% 0.000 85% -0.001 80% *. max 0.002 11% 0.002 13% 0.002 6% 0.002 8% 0.003 3% 0.002 10% 0 min 0.004 55% 0.005 53% 0.007 30% 0.004 59% 0.008 33% 0.006 41% O PRODUCTMIRKT_WBI max 0.014 2% 0.014 3% 0.019 V. 0.013 6% 0.019 1% 0.018 1% UNFAIR,- min 0.07 8% 0.004 36% .0108 3% 0.003 46% 0.005 24% 0.007 5% max 010 .)0% 0.009 4% 0.013 0% 0.009 3% 0.011 1% 0.013 0% MRKTDOM min 0.003 40% 0.005 33% 0.001 75% 0.006 16% 0.003 44% 0.001 84% max 0.009 2% 0.011 2% 0.009 3% 0.014 0% 0.014 0% 0.007 8% x S30min -0.004 78Y -0.018 18% -0.008 55% 0.001 97% -0.002 90% 0.000 97% 0 max 0.009 40% 0.003 85% 0.009 44% 0.010 38% 0.009 43% 0.013 25%A t HERE30 min 0.000 40% 0.000 76% 0.000 73% 0.000 33% 0.000 60% 0.000 42% max 0.000 19% 0.000 67% 0.000 33% 0.000 15% 0.000 31% 0.000 14% min -0.065 10% -0.054 11% -0.088 13% -0.070 10% -0.079 11% -0.073 12% SOEI (I) max 0.058 26% 0.055 23% 0.033 54% 0.053 35% 0.023 71% 0.042 36% SOE2 min 0.001 62% 0.001 46% 0.000 85% 0.000 98% 0.000 73% 0.000 72% max 0.003 2% 0.003 1% 0.003 1% 0.002 15% 0.002 15% 0.003 1% min -0.001 2% -0.001 2% -0.001 4% 0.000 15% 0.000 24% -0.001 4% E0 nmax 0.000 16% 0,000 8% 0.000 42% 0.000 37% 0.000 72% 0.000 28% ENTREPRENEUR min -0.011 62% -0.014 58% -0.007 78% -0.020 35% -0.022 35% -0.007 77% max 0.018 45% 0.017 44% 0.026 26% 0.004 84% 0.008 72% 0.022 35% ASSAS min -0.004 29% -0.005 23% -0.004 37% -0.003 40% -0.004 43% -0.005 29% max 0.001 85% -0.001 83% 0.002 60%1. 0.000 94% 0.002 66% 0.001 71% 2n CrvlLB min -0.004 8% -0.003 33% -0.005 9% -0.004 16% -0.004 17/o -0.005 10% am CIVILUB 0 max 0.000 93% 0.002 44% 0.000 99% 0.000 96% 0.001 84% 0.000 95% min -0.003 10% -0.002 41% -0.004 7%6 -0.002 33% -0.003 24% -0.005 4% 0 CORRUPT max 0.002 31% 0.002 33% 0.002 18% 0.002 22% 0.003 14% 0.002 29% m ENVRNMNT_WB min 0.003 53% 0.004 37% 0.005 24% 0.004 49% 0.006 29% 0.002 59% 0 max 0.010 5% 0.011 5% 0.014 1% 0.011 5% 0.016 1% 0.013 2% 50 ETHINFRCT min 0.000 7% 0.000 12% 0.000 10% 0.000 11% 0.000 17% 0.000 10% E Tmax 0.000 82% 0.000 44% 0.000 52% 0.000 44% 0.000 32% 0.000 59% ; INSTITQUALITY min -0.002 42% 0.000 85% -0.003 12% -0.001 70% -0.003 23% -0.003 12% ,,, max 0.005 10% 0.006 9% 0.005 16% 0.005 9% 0.005 16% 0.005 13% PATENTS mmin 0.000 4% 0.000 12% 0.000 8% 0.000 11% 0.000 16% 0.000 10% max 0.000 92% 0.000 90%6 0.000 50% 0.000 99% 0.000 24% 0.000 53% min 0.001 43% 0.002 18% 0.001 45% 0.001 64% 0.001 77% 0.001 54% POLICY E max 0.007 0% 0.007 0% 0.008 0% 0.006 0% 0.008 0% 0.008 0% a a RULE_WB min -0.004 47% 40.005 37% -0.001 81% -0.003 60% -0.001 90% -0.001 80%/0 max 0.008 14% 0.006 29% 0.011 5% 0.010 8% 0.014 2% 0.011 4% Note: Table reports the maxinum and minimum coefficient estmates from EBA analysis and theirsignrficance levels for each model. Table 9. Anti-Trust and Residual Growth Rates A. LATIN AMERICAN SOUTHERN CONE COUNTRIES AGE25 ANTITRUST RESIDUAL GROWTH Chile 10.52 4.71 3.25% Brazil 50.16 3.91 0.03% Argentina 20.96 3.06 -0.08% 3.89 1.07% B. EMERGING EU COUNTRIES AGE25 ANTITRUST RESIDUAL GROWTH Ireland 29.32 4.87 2.91% Portugal 27.8 4.45 0.77% Spain 50.6 4.08 0.03% Greece 3.92 -1.53% 4.33 0.01 C. SMALL EUROPEAN ECONOMIES AGE25 ANTITRUST RESIDUAL GROWTH Norway 5.27 1.26% Denmark 48.04 4.88 0.66% Austria 33.64 4.76 0.39% Belgium 48.72 4.59 0.45% Iceland 3.67 0.37% Netherlands 37.72 5.42 0.33% Finland 4.23 -0.05% Sweden 29.32 4.71 -0.43% Luxembourg 4 -0.46% 4.53 0.00 D. ASIAN TIGERS AGE25 ANTITRUST RESIDUAL GROWTH Korea, Rep. 23.52 3.79 1.84% China 24.12 4.73 0.95% Thailand 22.32 3.06 0.60% Indonesia 21 3.25 0.17% Malaysia 10.92 3.38 -0.14% Hong Kong, China 10.52 3.5 0.04% Singapore 14 4.48 -0.19% Philippines 31.48 4 -1.36% 3.77 0.00 TRANSITION ECONOMIES(1) AGE25 ANTITRUST RESIDUAL GROWTH Poland 3.42 -1.37% Hungary 4.01 -2.62% Slovak Republic 40.64 3.55 -3.08% Czech Republic 59.72 4.01 -3.35% Russian Federation 2.78 -6.56% 3.55 -0.03 Table 9. Anti-Trust and Residual (Growth Rates ANGLO-AMERICAN ECONOMIES AGE25 ANTITRU';T RESIDUAL GROWTH United States 5.09 0.78% United Kingdom 44.52 5.39 0.47% Australia 25.4 4.58 0.23% Canada 41.28 5.03 -0.43% New Zealand 33.2 5.11 -0.63% 5.04 0.00 LEFTOVERS Israel 18.96 4.83 1.77% India 28.92 3.82 1.68% Turkey 3.3 1.42% France 32.96 4.25 0.59% Egypt, Arab Rep. 3.43 0.55% Italy 32.64 3.86 0.22% Colombia 2.33 -0.23% Switzerland 4.29 -0.41% Zimbabwe 33.84 2.59 -0.48% Venezuela 3.66 -0.53% South Africa 39.84 3.26 -0.59% Japan 39.56 4.52 -1.66% Peru 29.6 3.95 -1.94% Mexico 20.6 4.15 -2.29% Jordan 2.13 -3.44% Notes: (1) For transition econornies, the EBRD publishes annually a 'transition indicator' for competition policy (ANTITRUST). For the years 1995 and 1996, Poland, Hungary, Slovak Republic and Czech Republic were each assessed 3 and Russia 2, on a scale from 1 tc 4+. See Transition Report for 1995 and 1996. Policy Research WDrking Paper Series Contact Title Author Date for paper WPS2302 Why Liberalization Alone Has Not Klaus Deininger March 2000 M. Fernandez improved Agricultural Productivity Pedro Olinto 33766 in Zambia: The Role of Asset Ownership arndc Working Capital Constrainls WPS2303 Malaria and Growth F. Desmond McCarthy March 2000 H. Sladovich Holger Wolf 37698 Yi Wu WPS2304 Disinflation and the Supply Side Pierre-Richard Agenor March 2000 T. Loftus Lodovico Pizzati 36317 WPS2305 Tne Impact cf Banking Crises on Maria Soledad Martinez March 2000 A. Yaptemco Money Demand and Price Stability Peria 31823 WPS2306 Intemational Contagion: Implications Roberto Charg March 2000 E. Mekhova for Policy Giovanni Majnoni 85984 WPS2307 Surveying Surveys and Questioning Francesca Racanatini March 2000 P. Sintim-Aboagye Questions: Learning from World Bank Scott J. Wallsten 37644 Experience Lixin Colin Xu WPS2308 How Small Should an Economy's Paul Beckerman March 2000 H. Vargas Fiscal Deficit Be? A Monetary 38546 Programmring Approach WPS2309 What Drives Private Saving around Norman Loay2:a March 2000 E. Khine the World? Klaus Schmidl-Hebbel 37471 Luis Serven WPS2310 How Politics and Institu,ions Affect Mitchell A. Orenstein March 2000 M. Leenaerts Pension Reform in Three 84264 Postcommunist Countries WPS231 1 The Buenos Aires Water Concession Lorena Alcazar April 2000 P. Sintim-Aboagye Manuel A. Abdala 38526 Mary M. Shirley WPS2312 Measuring Governance. Corruption, Joel S. Hellman April 2000 D. Bouvet and State Cap-ure: How Firms and Geraint Jones 35818 Bureaucrats Shape the Business Daniel Kaufmann Environment in Transition Economies Mark Schankerman WPS2313 How Interest Rates Changed under Patrick Honohan April 2000 A. Yaptenco Financial Liberalization: A Cross- 31823 Counlry Review Policy Research Working Paper Series Contact Title Author Date for paper WPS2314 Technological Leadership and Beata K. Smarzynska April 2000 L. Tabada Foreign Investors' Choice of 36896 Entry Mode WPS2315 Investment in Natural Gas Pipelines Alejandro jadresic April 2000 M. Salehi in the Southern Cone of Latin America 37157 WPS2316 Distrubutional Outcomes of a Emanuela Galasso April 2000 P. Sader Decentralized Welfare Program Martin Ravallion 33902 WPS2317 Trade Negotiations in the Presence of Keiko Kubota April 2000 L. Tabada Network Externalities 36896 WPS2318 Regulatory Reform, Competition, Mark A. Dutz April 2000 H. Sladovich and Innovation: A Case Study of the Aydin Hayri 37698 Mexican Road Freign Industry Pablo Ibarra WPS2319 Externalities and Production Gunnar S. Eskeland April 2000 H. Sladovich Efficiency 37698