Thanks to the Internet explosion and CNN’s images broadcasted from all the corners of the world, everyone has had an encounter with "globalization". Academic as well as coffee-pause discussions on its positive and negative impacts abound. On one extreme, certain commentators are tempted to consider globalization very much like the Internet: an inherently good thing that, with time, will bring beneficial effects to the whole world. Others, observing the widening gap between rich and poor countries, find in the globalization a useful escape goat to be blamed for many of the unresolved problems of the world. Clearly, there are a lot of analysts holding more reasonable moderate positions, but it is undeniable that the debate around globalization is hot and often ideologically charged.
Although some recent media-hyped events, such as the Seattle backlash against the WTO meeting, may make us think differently, this debate is part of a wider and, at least, twenty year long dispute around the evolution of the world economy and its global and local governance arrangements. A 1989 article by Francis Fukuyama entitled "The End of History" may be considered the first widely known document explicitly describing a society that had entered a new and lasting phase. He claimed that the change was so dramatic that it could be accurately depicted as representing the end of history. This new phase represented the worldwide triumph of liberal democracy, the collapse of Communism, and the advent of true global economic progress. Back then, at the time of the fall of the Berlin Wall, and translated in economic terms, that thesis seemed to say that the market had finally triumphed over the state. A "blind" reliance on the market and numerous recommendations for a minimal state combined to form what came to be known as the "Washington Consensus". A few years later, innovative and cheaply available communication possibilities and the ensuing new economy revolution reinforced the view that the market and its globalising forces would bring huge benefits for all.
However, the main problem of this Washington Consensus is that, even after repeated attempts, it has not really delivered a "Moscow Success", nor a "Latin American Miracle". Indeed, even the East Asian one, which superficially looked like a diligent application of the Washington paradigm, had recently to sail through stormy waters. Due to these setbacks and their own scepticism for the standard recommendations, economists have reconsidered the important role of public intervention in fostering the 20-year long East Tigers’ booming phase; they have contrasted the recent mediocre growth performance of regions following orthodox recommendations, such as Latin America and Eastern-Europe, with the success cases of China, India and others that joined the global economy in an unorthodox – gradual, sequential and still partial – manner; and many are persuaded that effective states as well as efficient markets are both crucial ingredients for a successful human society. The preparatory works for the Encyclical Centesimus Annum clearly showed that markets operate in particular environments and that their performances vary greatly depending on how other institutions operate. The stress on the natural and human environments, together with a rejection of central planning as a viable alternative to the market, were seen as a major innovation in the economic and social thinking of the Catholic Church.
Even within this broad consensus, the issue on how state and market, or globalisation and governance, should be combined and whether institutional arrangements should be uniform across countries or wide local variations should be encouraged are still contentiously debated.
At the beginning of the new millennia, the ‘end of history’ seems far away in the future and it may still take time to fully understand the very complex globalization-governance nexus.
By investigating how globalization influences governance, this paper offers some interesting new evidence on a particular but potentially quite valuable dimension of this nexus. More specifically, we attempt to understand why increased openness and international integration should affect the quality of domestic institutions, and try to measure by howmuch. The paper is organised in 4 sections: the next section discusses some stylised facts about the globalization and governance nexus emphasising some of the major difficulties researchers encounter in studying this subject, section 3 briefly reviews theories on the potential channels through which openness may influence governance’ quality, and presents our empirical assessment of the strength of these links. A final section concludes by summarizing main findings and pointing out potential policy implications.
Figure 1 summarizes in graphical form some important links between globalization, governance and economic performance that operate at a level of individual country. A nation’s resource endowments and its productivity determine how fast it can grow and the level of its economic well being in terms of income per capita: arrows 1 and 2 in the figure represent the causality links. It is immediately evident that feed-backs are possible: a richer country growing fast may invest more resources in scientific research and technology development and thus enjoy higher productivity levels than a poorer, slow-growing economy; this explains why arrow 1 is double-sided.
Globalization and governance, in the bottom panel of the figure, together with other important variables not shown here, have also important relationships with economic performance. For instance, through trade, capital flows or migration, globalization can influence the level of endowments available in an economy, or even, through international technology transfers, its productivity. Conversely a country’s endowments of natural resources, labour, and capital, as well as its geographic location and efficiency of its production structures may determine how much it trades with the rest of the world.
Figure 1: A map of Globalization, Governance and Economic Performance’ links
Similarly, a country with a good governance, namely a democratic state with high-quality institutions, effective corruption-free accountable bureaucracies, and a flourishing civil society may likely increase the quality, if not the quantity, of its most important endowment: its own people. Once more, cause and effect can be swapped: well-endowed countries may evolve towards democratic forms of government more easily, or, at least, they may afford investing more resources to build well-functioning institutions.
Arrows 1, 2, and 3 have been at the core of economics since its beginnings, and even arrow 4, if not constantly, has always attracted economists’ interest. In some sense, the ‘end of history’ debate mentioned above has drawn new attention towards arrow 5 and "how globalization and governance interact to affect economic performance" has become a topical question.
Surveying the large literature on this question is beyond the scope of our study, however emphasizing some stylised facts and a few key theoretical findings may help clarifying our own research motivation and may put our contribution in the right perspective.
The following figures offer a clear graphical summary on five of the most important relationships among economic performance, globalisation, and governance. A large cross-country sample covering various statistics for the 1995-98 period has been used to construct these figures.
Three sets of histograms in Figure 2 show the links between government size and the level of development, between government size and trade openness, and between trade openness and level of development.
Figure 2:Income, Openness and Government size
More in detail, the first histograms on the left depict the level of government spending as a percentage of GDP for different country groups. Countries are assigned to a specific group according to their level of development measured as income per capita. The clear pattern that appears here is that, moving from the "very low" income per capita group to the "very high" one, government spending increases; in other words, richer economies on average display larger governments. Honouring a nineteen-century German economist who first noticed it, this pattern took his name: "Wagner’s Law". Based more on historical observations than basic economic principles inference, Wagner (1883) formulated a simple theory in which public expenditure growth was seen as a natural consequence of economic development. A very large literature followed and the validity of Wagner Law has survived recurrent scrutiny.
The second histogram in the centre of Figure 2 measures government spending when countries are grouped according to their level of trade openness. Another remarkable pattern appears here: countries trading more intensively have larger governments. It seems as if countries heavily relying on global markets tend to compensate the ensuing risks they take with a bigger public sector. In fact, this is exactly the theory advanced by Rodrik (1999) who constructs a model where government size provides an indirect insurance against risks originating from global markets. By employing more people or through their social safety nets, large governments partially insulate their citizens from global markets fluctuations. Although contending theories may explain the size of the public sector, the simplicity of this model is appealing and its predictions are validated by empirical tests.
The final histograms display another important relation. Here countries are grouped according to their level of GDP per capita and trade openness of the resulting five categories is measured by the height of the histograms. The figure seems to offer a neat positive answer to the question "do more open countries grow faster?", however the links between trade openness and growth are much more complex and the debate among those who support globalisation as a positive growth factor and those who are more sceptical is not yet settled.
For a more precise approximation of governance instead of using the size of government expenditure, Figure 3 employs an index of the quality of the bureaucracy. Once more, its histograms show that richer and more open countries have higher quality administrations. Alternatively, Figure 4 displays the level of import openness (the ratio of imports on GDP), international capital flows intensity (foreign capital flows on GDP), and a liberal trade policy index for different countries grouped according to their levels of perceived corruption. The histograms suggest that countries perceived to be more corrupt are less opened to imports, have a lower degree of integration in the international financial market, but do not necessarily seem to have a less liberal trade policy, at least according to the particular trade policy index used here.
Figure 3: Quality of the Bureaucracy according to Income and Openness
Observing these simple histograms, one may be tempted to conclude that richer countries, as opposed to poorer, have a tendency to be more open to international trade and to have a larger and better public sector; or, in other words, that successful economies are able to combine the right mixes of market and state, of globalisation and governance. Unfortunately these graphs display just broad connections for each pair of variables with no indication of causality and, more fundamentally, they do not provide any clue on how, or why, variables mixes could be achieved.
As we have already noticed in commenting Figure 1, causality is one of the fundamental problem social scientists have to face when studying the relationship between economic performance, institutions and global markets. On the one hand, many papers document how high-quality institutions foster economic growth; on the other hand, evidence is accumulating on how developed countries may afford better institutions.
Complexity is the other fundamental problem. The histograms shown above are appealing due to their simplicity: they depict patterns linking pairs of variables, however these apparent patterns may be caused by the influence of other variables. Economic development, for instance, may be at the origin of a spurious relationship between government size and openness by simultaneously increasing the levels of both of them.
Explaining how globalization affects governance and how it helps or hinders economic development means to clearly identify causes and effects and to take into account multiple factors. To do that the next section moves from simple graphs to more complex theories and empirical tests.
Figure 4: Imports, capital flows and trade liberalization links with perceived corruption
Note: M is the ratio of imports to GDP; GPKF stand for
Gross Private Capital Flows and this is also a ratio with respect to GDP;
higher values for the trade liberalisation index correspond to more liberal
trade policies.
- Theory
- Empirical evidence: Model Specification and Data
- Results
Why and by how much does openness influence the level of perceived corruption in a given country? This is the central question we attempt to answer here. On the surface, no relationships seem to link openness and corruption directly, and a brief digression on the theoretical determinants of corruption is useful before considering our results.
Increased private gains are corruption’ main objective, however, among its crucial causes, we find economic as well as cultural and social variables. In a recent study, Treisman (2000) tests several hypotheses for the causes of corruption and offers a quite informative ranking on several corruption determinants. Clearly corruption is lower when its costs, including psychological and social, are higher than its benefits, and he finds that, on average, this is the case for those countries with Protestant traditions, those that are more developed and have a higher quality governments. On the contrary, corruption is more pervasive when the state is federal, or its democratic basis has been established only recently (less than 20 years), or, finally, when a country is less open to trade. More succinctly and using Klitgaard’s words, corruption thrives when monopoly power is combined with discretion and low accountability. Incentives to bribery do not arise in a society where all economic activities are carried out in a perfect competition setting and no single agent is able to affect the price or the quantity of the product he sells or buys. By the same token, corruption is reduced when economic rents do not depend on the discretionary power of some public official, or when monopolistic economic activities as well as governments are subject to strict rules of accountability.
Within this general framework, openness to international trade and capital flows may alter the balance between corruption costs and benefits through various mechanisms, which we now consider in turn.
Krueger illustrates the first mechanism in her 1974 article focusing on rent-seeking activities caused by quantitative restrictions to imports. In contrast to tariffs, quotas, and other official permissions to imports, generate considerable economic rents due to the monopolistic power they grant to legal importers. In order to appropriate these rents, agents may legally compete or embark in illegal rents-seeking activities such as bribery, corruption, smuggling and black markets. Kreuger proves that these rent-seeking activities induce an economy to operate at a level below its optimal, generate a divergence between private and social costs, and, thus, entail a welfare cost additional to that due to tariff restrictions. In successive papers, Bahgwati, Srinivasan et al have generalized Kreuger’s original idea to a whole array of Directly Unproductive, Profit-seeking (DUP) activities providing further arguments in favour of trade liberalization. More recently Gatti (1999) presents some empirical evidence of the explicit link between restrictions to trade and capital flows and corruption. In fact, her empirical study aims at disentangling two effects of inward-oriented policies on corruption: the "direct policy distortion" and the "foreign competition effect". High barriers to international transactions directly encourage private agents to bribe public officials in exchange for favouritism, the first distortion, and, through the second effect, reduce competition between domestic and foreign firms so that margins for rent seeking, and corruption, are kept high.
This second competition-reducing mechanism deserves some additional attention. Ades and Di Tella (1999) provide evidence that the level of rents in general and market structure in particular determine corruption intensity in an economy. Interestingly they argue that changes in rents size due to variation in the degree of competition may have ambiguous effects on corruption. On the one hand, larger rents resulting from a low competition environment increase the amounts bureaucrats can extract as bribes; on the other hand, in such a situation, it becomes more valuable to a society to increase the monitoring and accountability of its bureaucracy (more on this below). Determining the correct sign of the net effect on corruption due to these two opposing tendencies may be theoretically important, however, looking at real world situations, one finds many examples of positive connection between rents and corruption. A clear-cut case, cited by the authors, is that of oil-exporting countries: Nigeria, where 1980s oil export generated about 80% of government revenues and spurred a construction and import boom especially favouring the ruling party’s officials, provides a striking example of how rents cause corruption. These observations provide sufficient justification to Ades and Di Tella to build a model that links directly increased product market competition, to lower rents, and to lower corruption levels. In their model three sets of variables determine corruption: wages of the bureaucracy, the level of monitoring, and the level of profits that, in turn, depend on the degree of competition. Bureaucracy wages and monitoring are indirectly captured by a society’s level of economic (GDP per capita, schooling) and political (respect of political rights, Gastil index) development, whereas competition is proxied by the share of imports in GDP, the sector-concentration of exports, and the distance from the world’s major exporters. For the same level of the other variables, countries less exposed to foreign imports, or with a large share of its exports due to natural resources, should suffer higher levels of corruption that those countries more integrated in world markets and with a differentiated export basis.
Wei (2000), by explicitly considering differences in the costs of monitoring public officials due to the degree of international integration, advances a final third mechanism linking globalization to institutional quality. The basic idea is straightforward: improving the quality of institutions and their capacity to fight corruption depend on the amount of resources a society devotes to this end. A society invests more into building good institutions the larger the benefits it receives or the smaller the costs. Given that foreign producers may divert their exports or investments from a national market to another more easily than domestic producers, one would expect corruption and bad governance to discourage more strongly international trade and capital flows than domestic commerce and investment. This differential effect of corruption induces stronger incentives towards good governance investments for those economies that are more open. Other things being equal and because of the resulting larger benefits, an economy more exposed to international markets would allocate more resources to fighting corruption and end up with a lower level of it than a more isolated inward-looking one.
This model main result rests on two crucial assumptions. Firstly, corruption must truly affect more severely international transactions than domestic ones. Wei provides justification for this arguing that, thanks to their better opportunities to do business elsewhere, foreigners enjoy stronger bargaining power vis-à-vis domestic agents. Alternatively, enforcement costs for international contracts, already starting at a higher level than those for domestic ones, increase more steeply with corruption.
The second crucial assumption is concerned with the direction of causality: for Wei ‘being open’ comes before and independently of corruption, it is not a result of economic policy or business choices. In fact it may be useful to examine further this general issue of causality for all the three openness-corruption links we have described.
In Kreuger model, trade policy is exogenous and causality goes from trade barriers to corruption (or other rent-seeking activities) via a reduction of foreign competition and the creation of artificial rents in import regulated activities.
In Ades and Di Tella study, the degree of competition influences corruption, but this, in turn, by reaching certain intolerable thresholds, can provide incentives to alter the rules of competition. To overcome this circularity in their corruption regression, they measure import openness, a crucial proxy variable for competition, as if this was determined only by countries’ population and land sizes. The identifying assumption here is that these variables affect corruption only through their effect on import openness, and that they cannot be altered by corruption. For the other variables – natural resources share in total exports or trade distance – the direction of causality does not present problems.
Finally, Wei’ solution to the causality problem is to consider two types of openness. The first, labelled natural openness, is the potential cause for corruption and the second, residual openness, is the possible consequence of corruption. In his model, natural openness is determined by geographical measures, such as a country’s distance from major trading nations weighted by bilateral trade flows. In this way, corrupt official erecting artificial trade barriers cannot alter this type of openness and will only affect what Wei calls residual openness.
Our discussion on corruption determinants and causality issues is summarised in Figure 5. The three blocks on the left side represents three different sets of variables and the arrows depict the influences they exercise on corruption and among themselves. Apart from cultural andsocial variables and geography, all the other factors considered here can be affected by corruption, and this explains the double direction of the arrows.
Figure 5: Corruption Determinants
A complete model would take into account all this figure’s variables and arrows and provide guidance on how to disentangle causality directions, however, as reported above, economic theories of corruption have not yet reached this ideal stage and we have to adopt several simplifying assumptions. In particular, in our empirical assessment, we take into account potential reverse causality from corruption to relevant economic variables and correct for this by using geographical determinants as instruments; we construct openness measures that may influence corruption but are unaffected by it. Additionally, we always consider economic policies as exogenous (i.e. not influenced by corruption levels or by economic variables). Finally, we introduce additional controls to avoid potential omitted variables bias. If, for instance, we knew that ethnic fractionalisation determines low levels of corruption, is strongly correlated with trade openness, and is omitted from our regressions, then the positive influence we would attribute to openness should in reality be assigned to ethnic fractionalisation.
The theories we briefly reviewed contribute to explain why openness has an effect on corruption; this and the next sub-sections test these theories on real world data offering an empirical assessment of how much openness influences corruption.
All our estimations are based on the following equation or on some of its variations:
CORRUPTit = b0 + b1 Opennessit + b2 log (GDPit)+ b3 PolRit + b4 Otherit + b5 EcPolit +eit (1)
Our focus is on the sign and magnitude of b1, the marginal effect of openness on corruption; however, as suggested by the theory on the causes of corruption, we introduce several additional explanatory variables. The level of development of a country, by influencing cultural attitudes towards corruption and by affecting the amount of resources that may be devoted to monitor public officials, is a key determinant of corruption levels and enters our equation in terms of Gross Domestic Output per capita (GDP). Similarly a country’ score on basic political rights (PolR) may be a good proxy for its degree of accountability, another important factor explaining corruption. As shown above in Figure 5 and to avoid omitted variables bias, a series of other variables taking into account additional social causes for corruption is introduced in our empirical estimation (the Other variables we used are briefly described below). Finally, we test whether economic policy variables (EcPol), such as the degree of trade liberalization or more general state intervention in the economy, have a direct effect on corruption.
Among the variables included in equation (1), Openness and GDP, at least, suffer the problem of reverse causality; a corrupt bureaucracy may induce a lower degree of international integration by erecting discretionary barriers or even slow down development process through excessive regulations and direct waste of resources. If not corrected reverse causality can be a serious drawback altering not only the magnitude of our bsbut also their meaning: instead of verifying whether openness or GDP influence corruption we would be picking up how much corruption affect our regressors. Fortunately, by identifying suitable variables (instruments) that are highly correlated with openness and GDP but that do not directly influence corruption, standard econometric techniques allow us to bypass this problem. Suitable is the crucial attribute here. This means that we need new theoretically sound explanations for how openness and GDP are determined independently of corruption.
Consider openness first. Our solution consists of estimating a new measure of openness using Wei (2000) method, which, in turn, is similar to that of Frankel and Romer (1999). The basic equation is the following:
Opennessit = b0 + b1 Remotenessit + b2 log (Popit)+ b3 Otherit +eit (2)
Remoteness, population (Pop), English speaking and other geographical determinants (Other) are our suitable variables. Remoteness is a weighted average of each country distance from its trading partners in which the weights are given by the share of exports of the country’s partners in global exports. Formally it is constructed in this way:
Remotenesskt = ?i?k wilog (distancek i) , where: wi = exporti / ?i?k exporti (3)
This approach has its theoretical foundations in the well-established gravity equation that links bilateral trade flows to distances from major trading partners. According to equation (2), a country degree of corruption-independent openness increases with its proximity to the largest world traders, or if the country’s official language is English. Conversely, the larger is the size of a country’s domestic market, proxied by its population, the lower its openness.
The resurgence of economic geography in the late 1970s provides valuable instruments for GDP. A series of recent papers, studying the strong links between geography and the level of economic development, present empirical evidence on positive correlation between GDP per capita and the absolute value of latitude, or argue that lower development at the tropics may be caused by poorer human health and inferior agriculture productivity due to tropical climate, or, still, consider that winter frosts in temperate regions may boost agriculture productivity and thereby development. Geographical variables such as these are convincing instruments because their impact on corruption could only result through their influence on GDP. A very simple equation with the absolute value of latitude and a dummy for tropical countries provides us with a valid corruption-independent GDP estimate.
A brief description of the data we used in our equations concludes this section. Quantitative studies of the determinants of corruption are relatively recent given that numerical measures for corruption have not been ready available in the past. In this study we use two subjective indices of corruption as our dependent variable. These indices, produced for the use of international investors, are derived by standards questionnaires subjected to large random polls so that, by construction, they facilitate cross-country comparisons. In addition their commercial value partially guarantees their accuracy. Objective indices would be preferable if they were measured consistently across countries and were independent of corruption itself. Consider, for instance, a measure such as reported fraud cases: its objective value may depend on country-specific definitions and local corruption-fighting systems so that a country with a true low level of corruption and efficient monitoring schemes may report more numerous fraud cases than a more corrupt country. Corruption indices used here are those produced by Transparency International (TI) and by the International Country Risk Guide (ICRG) of the PRS group. It should be noticed that we have rescaled these two indices to vary in a continuous interval between 0 and 10, where 10 reflect the best score, i.e. the lowest level for corruption. Given that these two indices cover different country samples and time periods, we use both indices to test for robustness of our results. TI sample contains yearly corruption data covering 53 countries for the period 1980-85, the same countries for 1988-92, and 75 countries for 1995-98; ICRG sample includes 119 countries for the three periods of 1984-88, 1990-94, and 1995-98. Instead of using yearly data, for both TI and ICRG, we calculated three averages corresponding to the time intervals for which corruption indices were available. Due to the fact that yearly estimates for all our dependent variables and for all the countries covered by TI and ICRG do not exist, we preferred to adopt this averages approach to fill the gaps rather then restrict the sample to the few countries which present all the necessary data. In this way we maximize cross-country variation sacrificing little time variation.
Our initial measure for openness is given by the ratio of imports on GDP, however, following Larrain and Tavares (2000), we extend the concept of openness to include capital flows share on GDP. In fact, legal barriers, foreign competition, and monitoring costs, the three theoretical mechanisms through which openness may influence corruption can be equally applied to trade and capital flows.
The other economic variables, namely GDP per capita, the share of natural resources exports on total exports, government expenditure and consumption, and population were collected from the World Bank’s World Development Indicators. Political rights index – varying between 0, worst score, and 10, best score – was obtained from Freedom House, ethnic fractionalisation and protestant traditions dummies were derived from La Porta et al (1999), colonial past and democracy dummies were taken from P. Hensel website and Treisman (2001), trade liberalisation index was kindly provided by the IMF, and geographical data (distances, latitude, tropics dummy) come from various sources.
Table 8 in appendix presents summary statistics for the main variables used in our regressions. The number of observations for the ICRG and TI groups reflects the largest samples we were able to use in our most complete specification of equation (1), and it does not necessarily equal the sum across periods of all the countries covered by the corruption indices. Besides, the trade liberalization index is only available for the 1995-98 time interval, and that explains the drop in the number of observation. A major difference between the ICRG and TI’ samples consists of the latter’s exclusion of a fair share of developing countries: this is noticeable in its higher mean for GDP per capita (TI’s GDP average is almost 30% higher than that of ICRG). With respect to their means, trade openness and GDP per capita present lower degrees of variability than capital flows openness. In general, the large dimension of our samples, providing significant cross-country and time variation, should result in high quality estimates of the effect of the globalisation on corruption, a main contribution of this paper.
Simple correlations, the most basic statistical measure of quantitative relationship, are a good starting point and are shown in Table 1.
Table 1: Corruption and explanatory variables: simple correlations
Although they do not give any indication on causality, correlations in Table 1 represent a first approximate test for the corruption theories we described and offer an initial indication of the strength of the relationships. Openness, measured as a ratio of imports or capital flows on GDP, has a positive effect on corruption: our data show that countries with a higher degree of openness will, on average, also record lower levels of corruption. The same tendency applies, with stronger intensity, to the level of development. For each corruption index, two correlation values are shown in the top panel of the table where the first is calculated using ‘raw’ data and the second using data adjusted through the instruments approach: with respect to raw data, instrumented import openness displays a rather stronger correlation with corruption, whereas instrumented capital flows openness and GDP register a similar or slightly lower value. A single correlation value is shown for the other exogenous variables. In general, countries showing stronger accountability, proxied by the political rights index, lower sectoral concentration in their exports, lower ethnic fractionalisation, larger government involvement in their economy, and a high degree of trade liberalisation, also register low levels of corruption. Therefore in all cases but for the size of the government (more on this below) the sign of the relationship corresponds to that predicted by theories on the causes of corruption.
Before considering more sophisticated regression analyses, the results of the first-stage instrumenting approach, namely the results of estimating equation (2), are shown in Table 2.
Remoteness and size of the domestic market are very significant variables and together with a language dummy are able to explain almost 50% of total variation in import openness. Adding additional geographical variables, such as island, or landlocked country dummies, or the ratio of coast length to land area, does not change the coefficients neither the significance of the estimators included in Table 2. These are also quite powerful in explaining capital flows intensity on GDP.
Table 2: Constructing Corruption-Independent Openness and Development Variables – First stage regressions
t statistics are shown in italics below the estimates; * for the GDP regressions we use time dummies not reported here.
The two rightmost columns of Table 2 contain the results for our instrumentation of GDP per capita. International disparities in the level of development measured by GDP per capita can, by a large degree, be attributed to the physical geographic variables on the position of a country in the globe.
The fitted values obtained from the regressions shown in Table 2 represent the corruption-independent openness and level of development that we have used in equation (1) to measure the their influence on corruption.
Table 3 and Table 4 show our main results. The first table illustrates the importance of adopting a two stages regression analysis. Using instrumented variables (IV) in a rather parsimonious specification of equation (1) increases the explanatory power of the regressions as well as the magnitude and significance of the estimated coefficients. All the included IV variables are significant at the 1% threshold for both samples. It clearly appears that, because of reduced rent-seeking wasteful activities or due to their larger investment in institution building, countries more exposed to international imports experience a lower level of corruption.
Table 3: Globalisation and Corruption : Instrumental versus Non-Instrumental Variables (IV and No-IV)
t statistics are shown in italics below the estimates; * ICRG sample is restricted to include only countries covered by TI.
Different samples produce different results as it clearly appears by comparing the estimated coefficients for import openness in the second and the fourth columns of Table 4. As shown in the rightmost columns, such discrepancy almost disappears when the larger ICRG sample is restricted to include only countries covered by TI.
Table 4 is organised as follows: the first two panels present the effect of instrumented import and capital openness on perceived corruption. In the last rightmost one policy variables that are likely to influence the degree of trade openness are explicitly introduced. Results for the ICRG sample are reported in the upper part of the table, while those for TI are reported at the bottom.
The parsimonious specifications (1) and (5) indicate a positive impact of openness on the quality of governance, in our case a reduced level of corruption, for both samples.
Our regressions are in lin-log specification, meaning that the dependent variable, corruption, is in linear format and the independent variables are in logarithmic format. In this specification we can interpret the bs as the marginal effect on corruption of a change in the logarithm of the dependent variable, or, as the marginal effect due to a relative (percentage) change in the dependent variable in linear format.
Equation (1) predicts Therefore, forthat the ICRG sample, a 10% increase in imports openness results in 0.07-point change in the corruption score (0.74 x 0.1) in the ICRG sample, and in 0.2-point change (2.06 x 0.1) in the TI case. This is a sizeable effect, especially when compared to These are very significant values comparable to the 0.13 and 0.12-point changes due to a 10% increase in log GDP per capita; in the TI case, a given increase in openness contributes to reduce perceived corruption more than that same increase in development.
Instead of an arbitrary 10% change, it may in fact be more instructive to consider more realistic variations in the dependant variables such as their observed standard deviations. This exercise results in a 0.40 reduction of corruption (0.74 x 0.54) for the ICRG sample and a 1.11 reduction (2.06 x 0.54) for the TI sample.
Compared to import intensity, capital openness has a smaller impact on corruption. A 10% increase in capital openness results in a 0.05 and 0.11 reduction of perceived corruption in the ICRG and TI sample respectively. Considering a one standard deviation increase in the log of GPKF, ICRG is reduced by 0.65 point (0.48 x 1.36), while TI falls by 1.25 points (1.06 x 1.18).
To isolate the direct impact of openness on governance we need to consider other important simultaneous determinants of corruption: columns (2) and (3) in the first panel and (5) and (6) in the second introduce these additional controls to the basic specification.
Controlling for dependence on oil and mineral exports (equations (2) and (5)) and ethnic fractionalisation ((3) and (6)) does not change the overall picture. In these specifications a high explicative power is achieved, even if not all the included variables are significant at conventional levels. The basic results concerning openness and corruption are unchanged: the magnitude of import and capital openness is slightly increased and the coefficients remain statistically significant.
Interestingly enough, while dependence on natural resources turns out
to be a significant determinant of higher level of corruption, ethnic fractionalisation
is never significant, nor has the expected negative sign.
Table 4: Globalization and Corruption: extended results
t statistics are shown in italics below the estimates
These results are not surprising. A vast literature points to higher rent-seeking behaviour in natural resource abundant countries. The effect of ethnic fractionalisation on corruption is not clear. While recent investigations indicate high fractionalisation as a negative determinant of growth, studies focusing on the causes of corruption do not find such a clear-cut result. Gatti (1999), for instance, finds that fractionalisation is significant and reducing corruption. This finding is explained in terms of the increased difficulties bureaucrats encounter in extracting bribes from ethnic groups they do not belong to.
Columns (7) and (8) in the last panel of Table 4 introduce policy variables as potential explanations of corruption. Due to the difficulty of finding exogenous variables proxying policy input, the analysis is restricted to policies affecting import openness.
Column (7) shows how results vary when an index of trade policy liberalisation is introduced. Notice that this variable is available for the last period only, so that results are not strictly comparable to those reported in previous columns. Basic findings are unchanged, but the impact of the new variable on both ICRG and TI turns out to be not significantly different from zero. Furthermore, in the ICRG sample, the liberalization index has the wrong sign. This result should not be too surprising given what we observed in Figure 4 above.
Column (8) takes into account the extent of government intervention, approximated by government consumption as a share of GDP. This additional variable is highly significant in both samples, but its effect on openness differs greatly. Using the TI sample, openness remains significant and sizeable; conversely, with ICRG, it seems that introducing government size among the regressors makes openness’ effect on corruption much smaller and statistically insignificant.
Recent empirical research on the causes of corruption and the quality of governments points out that a further series of social and historical variables should be considered. In particular the role of the colonial past, that of the religious traditions, and that of long-term stable democratic institutions are seen as important explanatory variables for the level of corruption. In Table 5 we add to specification (3) of the previous table these additional historical controls and notice the following. Firstly, the coefficients on "colonial past", "protestant" traditions, "democracy" and "OECD" membership are significant, show the right sign, and increase the R-squared of the regressions. Countries that have never been colonies, where protestant is the largest confession, where democracy has been uninterrupted for the last 50 years, and that belong to the Organization for Economic Cooperation and Development record lower levels of corruption.
Secondly, the estimation of the effect of import openness on corruption is not affected by omitting these historical variables; on the contrary, with their introduction its explanatory power is actually increased. It seems that these variables, by lowering the explanatory power of the GDP and political rights coefficients, are in fact accounting for deep institutional and social cross-country differences.
Table 5: Additional controls – historical variables
t statistics are shown in italics below the estimates
Finally and most importantly, it should be noticed that these variables are all in dummy formats thereby they are really just labels used to describe a, sometimes quite loose, common characteristic of a particular group of countries. In fact the only proper label is the "OECD" one: this group of countries adopted common measures to fight corruption and is trying to enforce them through "peer pressure" mechanisms. Rather than testing serious hypotheses on how, for instance, being a democracy may affect a country’s corruption level, they provide an indication that our corruption theories are still incomplete.
In summary, our main empirical result, that the causal link from openness to corruption is strong and statistically significant, is robust to the introduction of a whole set of additional explanatory variables used in the literature on the causes of corruption, and it is not affected by sample bias. Table 6 confirms that our results are in the broad range of other studies’ estimations, providing further support to the thesis that corruption declines in more open societies.
Table 6: Corruption and Openness – comparative results
In this paper we analysed how openness can affect the quality of domestic institutions, a specific yet rather important dimension of the globalization and governance nexus. Microeconomic theory helped us identify trade policy, competition by foreign producers and international investors, and openness-related differences in institution building costs, as three major transmission mechanisms through which openness affects a country’s corruption levels. Examining a large sample of countries covering a 20-year long period, we found robust empirical support for the fact that increases in openness do indeed cause reductions in corruption. The magnitude of the effect is also quite strong. After controlling for many cross-country differences, openness’ influence on corruption is quite close to that exercised by the level of development.
An important point should be stressed: we were not able to measure a significant direct effect from trade policy, confirming previous authors results, and found that, at least for the ICRG sample, the addition of government size among our explanatory variables decreases the magnitude of the openness effect and its statistical significance.
Although this does not invalidate our findings – openness in the TI sample is unaffected by government size and it may as well be the case that this variable is caused by corruption – some caution should be used when drawing economic policy implications. Firstly, reducing trade barriers may not bring immediate positive corruption reductions. It is true that in the long run, more open economies, enjoying more foreign competition and investing abundantly in institution building, will register lower corruption levels; however, in the short run, domestic policies may be more valuable than pursuing globalization at all costs. This may be especially important for poorer countries that may face serious trade-offs between complying with international agreements and investing in basic development infrastructures such as education, health, and social security.
Secondly, our support for a positive effect of globalization onto governance
is based on a cross-section study, and it is well known that this type
of analysis has several problems. Cross-country differences in the levels
of the dependent variables are the central explanation for the variations
in the dependent variable, and, no matter how many controls are added,
it will always be possible that some additional relevant variable is missed
or wrongly measured and that results are thus distorted. It is possible
to account for many local characteristics, yet comparing China to the USA,
or India to Argentina, will always be a bit stretched. This suggests that
future research should be focussed on in-depth country specific case studies
and, as in the case of Srinivasan and Baghwati who examine the links between
openness and growth, we are confident that the virtues of outward orientation
as quality enhancer for domestic institutions and growth will not be refuted.
