ECONOMIC CONVERGENCE OF ITALIAN REGIONS: THE ROLE OF ORGANIZED CRIME AND OF PUBLIC EXPENDITURES,

ECONOMIC CONVERGENCE OF ITALIAN REGIONS: THE ROLE OF ORGANIZED CRIME AND OF PUBLIC EXPENDITURES, 1960-1993. by Giuseppe Tullio and Stefano Quarella University of Brescia (June 1998) Current address: Università degli Studi di Brescia, Facoltà di Economia, Via San Faustino 74/B, 25122 Brescia, Italy. e-mail address : gtullio@altavista.net 1

ECONOMIC CONVERGENCE OF ITALIAN REGIONS: THE ROLE OF ORGANIZED CRIME AND OF PUBLIC EXPENDITURES, 1960-1993 1. Introduction Giuseppe Tullio 2 and Stefano Quarella 3 From 1960, the first year accurate data are available, to 1996 the differences in real GDP per capita between the regions of the South and of the Center-North have been reduced considerably. Real GDP per capita of the South increased as a ratio of North-Central GDP from 47% in 1960 to 58% in 1996. However, the differences still persisting are large, despite massive public investments and current transfers from the Center-North and a very deliberate policy of all Italian governments of the period to reduce these differences. Worst still, since 1984 the convergence has come to an end: GDP per capita of the South fell from 62% of Central-Northern GDP in 1984 to 58% in 1996. Unemployment in the South remains above 20%, while the Center-North is very close to full employment. It should therefore not be a surprise that the debate on what during the last century was labeled the Southern question has revived again in the last years both from a political and an economic point of view. After the decision of May 2nd 1998 to accept Italy into the European Monetary Union the focus of the Prodi government has shifted from policies aimed at achieving entry into the Monetary Union to policies aimed at increasing public investment in the South to fight unemployment. Increasing public investment in order to fight unemployment is a very sensible policy in a normal socio-political environment. It may achieve nothing and even be counterproductive in an environment in which public investments are decided on the basis of 1 This paper is based on a chapter of the laurea dissertation of Stefano Quarella presented at the University of Brescia on February 4th 1997 and entitled Diseguaglianze economiche fra Nord e Sud e sviluppi del mercato del lavoro. We would like to thank Prof. Francesco Daveri for useful suggestions on earlier drafts. 2 Professor of Economics, University of Brescia, Via San Faustino 74/B, 25122 Brescia. 3 Banca San Paolo di Brescia, Corso Martiri della Libertà 13, 25121 Brescia. 2

political-electoral motivations rather than on the basis of objective economic criteria, in which local administrations are corrupt, in which the informal connections between organized crime and local authorities and sometimes even the central authorities are strong, and in which the property rights are not guaranteed because of the activities of organized crime in the presence of a judicial system which is always extremely slow and sometimes not sufficiently distant from organized crime and corrupt officials. We do not mean to say that all of Southern Italy presents these features, although the reluctance of Northern Italian and foreign capital to move there despite high subsidies, lower labour costs and fiscal incentives seems to suggest that there are significant problems of the type mentioned above 4. The Sicilian judge Giovanni Falcone killed by Mafia in 1992 wrote in his book Cose di Cosa Nostra that he thought Mafia was getting about 20% of his total income from public investment. In addition there may well be a link, so at least some newspapers suggest, between the large number of killings among different criminal organizations in Naples in the first half of 1998 and the fact that public money is about to arrive to the city. If the South of Italy is a sick socio-political environment, as we are firmly convinced 5, the right economic policies to get again the 4 For very large Italian firms like Fiat these problems are less serious because they have a strong bargaining power with the Central government, local admistrations and anyone else who comes along. 5 It may be worth mentioning that one of the writers of this paper has inherited property in the Center-South of Italy and he has incurred in the following three interesting episodes. They are quite revealing of the lack of the rule of law in the South. First in 1981 and 1982 his farm incurred into a VAT credit with the government. When his father died in 1984 the inheritance tax was promptly paid on the VAT credits. By 1988 he still had not received the VAT credits. The Ministry of Finance in Rome said that the case of the 1981 reimbursement was closed because this is our standard practice for reimbusements of less than half a million lire. Therefore they would not pay neither the capital nor the interest despite the fact that inheritance taxes were paid on the 1981 VAT credit. As to the 1982 reimbusement, they would pay the capital, but if he wanted also the interest he would have to go to court. He went to court for the 1981 credit (capital and interest) and for the interest on the 1982 credit. The case was closed with full reimbursement in 1998, after 17 years! This episode makes you wonder how people can trust such a government and may even provide a full moral justification for tax evasion. Second he owns a strip of land by the sea in a Central-Southern beach resort. According to the regulations of the town one can build on it and it is already built all around. He asked for a building permit for the first time in 1983. The request has been rejected two times in the course of the succeeding years on the ground that the town has not yet built a sewage system in the area, although he had planned to construct at his own expenses a link to the closest point of the existing sewage system. In the course of time he was repeatedly told by city officials first that a recommendation to the mayor from Mr. Andreotti would greatly help and secondly that money was still lacking for the next town 3

convergence process of the South in motion consist mainly in fighting corruption more seriously than has been done so far, increasing enormously the resources of the judiciary and the police, give more power to the judges and police officials who are known to be honest and remove all the others, make sure that all pending civil and criminal lawsuits are resolved rapidly, make careers of judges and policemen dependent on honesty and professional abilities rather than on political affiliations or other dubious affiliations, pass special laws as requested by honest judges in the South 6, take measures to increase the confidence of honest citizens in the police and in the judicial system so that they will report crimes with less fear. Public investments should wait until the environment has been sufficiently cleaned. On the other hand if our judgement is wrong, then increasing immediately public investment in the South is the right policy. celebrations. In March 1994 after the second application for a building permit had been turned down with a two year delay, he went to court against the decision of the town administration to turn the project down and to this day he is still waiting (15 years after the first application and almost 5 years after the appeal to the courts!). The third episode is even more disturbing: in March 1994 he sold the farm in Southern Italy (the same on which VAT credits were due) and he agreed with the buyers that the European Union subsidy on hard wheat for the agrarian year 1993/94 would go to him and the receipts from the sale of the product to the buyers. There was no objection and the tranfer of the property took place regularly on March 30 1994. He was accused by the buyers in July 1994 of extorting from them with threats the clause that the EU subsidy would go to him. The buyers have false witnesses to whom they have probably promised a part of the subsidies, should the courts find him guilty of extorsion. He has not heard from the judges in four years. Meanwhile his name is on the police records as a potential criminal. We could tell of a fourth episode of injustice caused by the virtual absence of a legal system, but we do not want to bore the reader further. The point is either that one of us is extremely unlucky or that something must be dead wrong in Southern-Central Italy. We should add that the family of the one of us who incurred into the above three legal problems lives in Rome while its property is/was scattered in several towns of Southern/Central Italy; as a result in the eyes of local criminal elements its social standing is small. In addition we are not members of a party, we do not have close contacts with the higher hierarchy of the Catholic Church and we are not free masons; these organizations are known to frequently solve problems of the type mentioned above through connections in the right administrations and by speeding up processes bringing them to the right conclusion. He won the battle concerning VAT reimbursement and he is confident he shall win the other two. The question is when and at what cost. We have lost all confidence in the Italian government and the Italian legal system and the one of us involved in these matters is at the moment packing all his belonging to leave Italy for good and in disgust. 6 See the clear and strong appeal by judge Morvillo of Palermo, brother-in-law of Giovanni Falcone, on the Maurizio Costanzo show (Channel 5 of Italian Television) in May 1998. He appealed for special laws on the ground that there is a war in the South and that laws designed for normal criminal offenses cannot be sufficient to fight the types of crimes committed there by organized crime. 4

The answer to the question concerning the right policy to fight unemployment in the South and to reverse the economic divergence of the South recorded since 1984 ultimately hinges on how important organized crime and corruption really are and on what has been their role in stopping the process of convergence of the South. This paper intends to analyze by means of regression analysis the factors which influenced the convergence/divergence of Italian regions since 1960 with a particular emphasis on the role of regional crime rates and regional public expenditures. As to the latter we shall consider both total expenditures and their split into investment and current consumption. The above mentioned variables show a significant difference in behaviour both across regions (in particular between the South and the Center-North) and through time. For instance Figure 1 shows the number of voluntary homicides as a fraction of the resident population in the Center-North and in the South. The systematic difference in the crime rates between the two main areas of the country is impressive. Even more so is the sharp increase in the Southern crime rates since the mid 1970s and then again in the second half of the 1980s and early 1990s. Several empirical studies already exist on convergence of Italian regions 7. Many do not consider crime as an explanatory variable and among the few which do no one finds it to be significant. However, existing studies use the number of crimes against property and the economy or of crimes in general instead of voluntary homicides. We feel that in the South there is a large underreporting of crimes against property and of crimes in general because of the obvious dangers involved in reporting them to the authorities. The number of voluntary homicides, although it is certainly still not a good measure of the strength of organized crime, is instead less subject to underreporting except when bodies are dissolved in acid or hidden in cement walls, both of which have occurred. 7 See Piras M.G. (1992), Di Liberto A. (1994), Paci R.-Pigliaru F. (1995), Cosci S.- Mattesini F. (1995), Fabiani S.-Pellegrini G. (1997), Bianchi C.-Menegatti M. (1997), Ferri G.-Mattesini F. (1997), Del Monte A.-Papagni E. (1997). 5

Fig. 1- Italy: voluntary homicides in the South and in the Center- North for every 100.000 inhabitants, 1960-1995. 0,7 0,6 0,5 0,4 South 0,3 0,2 0,1 Center-North 0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 It is important to mention that in Italy there is virtually no killing going on in connection with small robberies in the streets of big cities of the type which is common in US cities or in Rio de Janeiro, so that it is safe to assume that regional voluntary homicides are truly exogenous with respect to regional GDP growth. They are probably much less exogenous with respect to public spending for the reason mentioned above. Concerning public expenditures the hypothesis has been advanced that they may have had a beneficial effect on growth and convergence of Southern regions in the 1960s, when the marginal productivity of public investments was very high, current expenditures still moderate and corruption much lower, while in the 1980s the effect may have become negative as a result of a much lower productivity of public investment, high corruption, a deeper link between politics and business, systematic use of public money for personal enrichment or to finance lavish expenditures by political parties in power and increasing distortions in the labour market caused by current expenditures (Micossi and Tullio, 1992a, 1992b). In the 1980s a widespread system of false invalidity pensions and a 6

generous pension system (not linked to actual contributions), coupled with strong intergenerational solidarity within the Italian family caused a halt in the emigration from the South despite increasing unemployment especially among the young. It cannot be excluded that in the last two decades or so many public investment projects were decided with the main objective of maximizing bribes and votes for politicians and profits for contracting firms without any regard to economic rationality 8 and as a result an increasing fraction of what in the government budget and in the accounts of state industries and agencies is recorded as investment is really something else, namely bribes 9 and sheer waste of money. In this study we intend to test explicitly the hypothesis that the effect of public expenditures on convergence changed through time and across regions. Only one of the existing studies on convergence has tried to do so. Piras (1992) finds a positive and significant relationship between GDP growth and the variable public works for the period 1963-75. The variable becomes insignificant for the period 1963-84. Finally with respect to previous studies on convergence of Italian regions we extend the sample period to the last year for which all the data were available at the time the study was started. Our proxy for organized crime turns out to be of utmost significance and goes a long way in explaining why divergence set in after 1984. We show that the very high significance of our proxy for organized crime in our regressions for the whole sample period is 8 In the early 1980s a unit was created at the Ministry of the Budget in order to evaluate public investment projects on the basis of the rigorous techniques used by the World Bank. Experienced staff from the World bank was hired. Initiallly politicians were favourable. When they understood that they would loose their discretionality in deciding which investments to make the Unit was sacked by inventing criminal offenses against the best and most honest officials and forcing them to leave. Later the personnel of the unit was replaced with safe officials. 9 In the late 1980s in the Southern region where the Earthquake of 1980 caused over 1000 casulaties a small piece of land of one of us was expropriated to build a road linking a town of 3000 inhabitants to a village of about 300 souls. The road was built with bridges and tunnels at about midway between the bottom of the valley and top of the mountain. Planned cost: 32 billion lire. Although we are not civil engeneers we noticed that the road could have been built much further down the valley without so many bridges and galleries. We found out that an alternative project existed. Planned cost: 4 billion. However, local authorites must have made sure that the alternative project disappeared so that it could not be even considered. The presumed reason is that the margins for bribes and unjust profits are in absolute value much higher on a 32 than on a 4 billion project. This reflects a very high esteem for the taxpayer on the part of public administration in some areas of the country. If the courts had been functioning well and fast we may have asked them to investigate the matter more deeply than we could with our limited means and only for the sake of satisfying our intellectual curiosity. 7

explained mainly by the effect it exerted in the South after 1980. The change in sign of public expenditures through time is also confirmed by the data as the non-significant effect in the regressions for the whole sample period hides a very positive and significant effect on growth until the mid 1970s, a much weaker positive effect in the second half on the 1970s and a neutral or negative effect in the 1980s and 1990s with significant differences among the major areas of the country. We conclude that in the South organized crime and the waste of public money have interfered so heavily in the mechanism of allocation of resources and on the factors affecting growth that after 1980 their negative influence has even become visible on (macroeconomic) growth rates of regional GDP per capita. The paper is structured as follows. Section 1 presents a brief historical overview since 1960 of the various phases of the convergence process of Southern GDP per capita towards the level of the Center-North. We shall distinguish two periods of rapid convergence (1960-71 and 1978-84) and two periods during which the two areas of the country diverged (1971-78 and 1984-96). According to preliminary data the phase of divergence continued in 1997. Section 2 presents the regression results. We have divided the period 1960-93 into 7 5-year periods, except for the last one which has only four, and we use 5-year averages of annual data. Since Italy has 20 regions we have 140 observations for each regression (7 times 20) and usually well over 120 degrees of freedom. The use of 5-year averages allows us to study better how the impact of the explanatory variables (in particular organized crime and public expenditures) on convergence changed through time and across the three main groups of regions (South, Center, North). None of these three groups is perfectly homogeneous and thus by comparing the aggregate for the South with the aggregate for the Center-North as we do in Section 1 and Figure 1 for convenience of exposition we hide important information. In Section 3 we analyse the effect of the crime rate and of public expenditures on the ratio of employment to the resident population. Section 4 summarises the main results and draws conclusions for economic policy. 1-A brief historical overview of convergence of the South towards the Center-North, 1960-1996. 8

Figure 2 shows the behaviour of Southern GDP per capita in relation to the one in the Center-North from 1960 to 1996. For the whole of the sample period there was a significant degree of convergence as GDP per capita in the South increased from 47.3% of Central-Northern GDP in 1960 to 58.2% in 1996. Fig. 2- Real GDP per capita in the South in percentage of GDP per capita in the Center-North (1980 prices) 0,64 0,62 0,6 0,58 0,56 0,54 0,52 0,5 0,48 0,46 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 However, the convergence process did not proceed smoothly. A period of very rapid convergence between 1960 and 1971 was followed by a period of divergence in the period 1971-78. Convergence resumed at a more moderate pace in the period 1978-84 which was followed again by divergence in the period 1984-96 during which Southern GDP fell from 62 to 58.2 % of Central-Northern GDP. The 1960-71 period: rapid convergence. Most of the convergence recorded in the period 1960-96 occurred from 1960 to 1971. During this brief 12 year period Southern GDP per capita increased from 47.3% to 61.1% of Central- Northern GDP (a reduction of 13.8 percentage points in the gap; more than 1% per year). The reasons of this strong convergence are first the very favourable behaviour of the whole Italian economy which grew in the period at a rate close to 5% per year 10 helped by increased integration with the European economy, a stable financial environment and falling real prices of raw materials. Second the high migration from the South strongly contributed to the increase in 10 4.6% to be precise. 9

productivity and GDP per capita in the area. Third public expenditures were mostly limited at least in the first half of the 1960s to productive investment in infrastructures the main objective of which was to create the conditions for growth rather than to intervene directly in the production process. On the other hand the investments of state owned industries made in the area during the period were still mostly motivated by economic rationality rather than other objectives and must therefore have exerted a positive impact on Southern GDP per capita. The 1971-78 period: sharp interruption of convergence From 1971 to 1978 Southern GDP per capita fell from 61.1% of Central-Northern GDP to 57.4% or 3.7 percentage points. The Italian economy was weakened during this period by the world recession and the aftermath of the Bretton Woods crisis od 1971, leading to the August devaluation of the US dollar, the more militant attitude of labour unions since 1969, the convergence of wages nationwide, the oil crisis of 1973-74 and increased uncertainty. The lira was repeatedly devalued in 1973 and then again in 1976. Inflation accelerated and became very variable. Government expenditures accelerated in a disorderly fashion in an attempt to fight unemployment originating in the excessive increase in real wages and in the increase in the real price of oil. The level of remittances of wage income on the part of Southern Italian workers abroad was held back by expectations of devaluation of the lira and by negative real interest in Italy for most of the period. In sum, when things turned for the worse, the weakest Italian regions were affected more heavily than the rest of the country and the economic policy of the government which started to be characterized by disorderly deficit spending certainly did not contribute to the long run solution of the economic problems of the South. Public industries which in the previous period had started to direct a large share of their investments to the South suffered severely from the increase in labour costs and the first oil crisis. In addition Italian medium and small enterprises were the fastest to adapt to the new environment and the fact that they were predominantly located in the Center-North left the South in a weak position. Rather than stimulating the development of private small and medium sized enterprises in the South and creating the legal and social environment needed to attract foreign direct investment, the government preferred 10

to fight unemployment by sharply increasing public employment and by making huge investments in very capital intensive cement, steel and chemical mega-factories in the hope, so at least the story goes, that these huge factories would stimulate the birth of a sufficient number of small enterprises around them. Our preferred explanation for the choice of mega-industrial projects is instead that brides and electoral considerations were already becoming important in the period. Far from creating the environment necessary to stimulate direct investment from abroad and from the Center-North in small and medium enterprises these policies of building cathedrals in the desert reduced the role of the market in the South, increased the corruption and arrogance of the local political class and made organised crime stronger, an outcome which may have been quite welcome in the eyes of the then ruling political class given that at the time of elections hundreds of thousands of votes could be mobilised by organized crime 11. The 1978-84 period: resumption of convergence. After the foreign exchange crisis of 1976 the lira was stabilized, the monetary financing of budget deficits was highly reduced, the Trasury bill market was developed and Italy recorded large current account surpluses in 1978-81. In March 1979 the lira joined the European Monetary System (EMS), albeit with a larger band (+/-6%) than other members (+/-2.25%). In July 1981 the bank of Italy was relieved from the obligation to buy Trasury bills and bonds at primary auctions. Despite the second oil shock of 1979-80 the above institutional changes and especially the joinin of the EMS significantly reduced uncertainty. During this period Southern GDP increased from 57.4% of the level in the Center-North to 62.3% (a postwar peak) or 4.8 percentage points. When in Section 2 we use regional 5-year averages rather than annual data aggregated for macro-regions (like in Figs. 1 and 2) we find that in the first half of the 1980s government expenditures and especially the crime rate in the South were already affecting 11 Mr. Andreotti, many times prime minister and virtually always in the government or a leading figures of the Christian Democratic Party since 1947 has been charged of committing such a crime in Sicily. It would be a big mistake to think that these practices have ended after the fall of the Christian Democratic-Socialist regime in 1992. Leading Sicilian figures of Forza Italia, the party created by the TV tycoon Berlusconi in 1993, have been charged and at least one is under arrest at present for maintaining very close connections with Mafia at the time surrounding the 1994 elections. In addition Mr. Dell Utri, head for decades of one of the most important firms of Berlusconi (Publitalia) has been charged of association with the Sicilian Mafia. 11

convergence negatively. Thus the continuation of convergence in the first half of the 1980s may be the result of a sharp fall in the crime rate in the South during the period (see Fig.1) coupled with a more favourable national macroeconomic environment characterized by reduced uncertainty with respect to the 1971-77 period. The 1984-96 period: the South looses ground again. Between 1984 and 1996 Southern GDP per capita fell from 62.3% to 58.2% of Central-Northern GDP per capita or 4.1 percentage points. Most of the deterioration in the South occured between 1991 and 1996 when the gap increases by 3.7 percentage points in the presence of a virtual stop of growth in the South and very moderate growth in the Center-North. In 1996 the gap falls back to the level of sixteen years earlier (1979) or 58.2%. The causes of the 1991-96 developments are two. In the South organized crime reached a strength never seen before and after the large September 1992 devaluation of the lira and its exit from the EMS the Center-North compensated the fall in domestic demand with a strong increase in exports. The very significant increase in voluntary homicides in the South in the period 1990-93 (see Fig. 1) went hand in hand with a remarkable increase in government consumption expenditures there (see the comments to Fig. 4). In many ways the South of Italy can be considered as one of the few remaining truly centralized economies of the world (together with Cuba and North Corea) as the public sector dominates the economy to an incredible extent. Micossi and Tullio (1992a, 1992b) calculate that in 1988 the share of government expenditures to GDP in the South amounted to 76%. In addition public enterprises still play an important role and their outlays are not fully included in the above figures. If one considers that agriculture, which is still very important in the South, is heavily subsidized by Brussels (but the money is actually distributed by Italian agencies) and that virtually no economic activity can be started without official authorizations 12, it becomes clear that local political or criminal groups (or both together) fully direct the allocation of economic resources in the South. Even the large investment of Fiat in Melfi decided in the early 1990s cannot be labelled as truly private as out of an investment of 5000 billion Italian lire (about 4 billion dollars) Fiat received a generous grant (not 12 See footnote 5, the second episode. 12

a loan!) of 3000 billion lire. In addition not enough has been done yet to re-establish law and order at least at the level of the early 1960s and speed up the solution of pending legal cases. 2-Effects of Organized Crime and Public Expenditures on Convergence of Italian regions: empirical evidence from 1960 to 1993. 2.1-Absolute or unconditional convergence. According to the neoclassical theory of economic growth if two countries or regions are characterized by the same structural parameters (production functions) but different initial levels of income per capita they will tend in the very long run towards the same level of income per capita (the same steady state) 13. The region with a lower initial income per capita will grow faster than the richer region and its income per capita will converge towards that of the richer region owing to the neoclassical hypothesis of declining marginal productivity of capital (along the same production function for the two regions). In this case one talks of absolute or unconditional convergence. The absolute convergence is measured in the empirical literature by the β-indicator or by the σ-indicator. An estimate of the β-indicator is obtained by running the following regression: (1/T). ln ( y i,t / y i,t-t ) = a + b. ln y i,t-t + u i,t (1) where y is GDP per capita, ln stands for the natural logarithm, a reflects its equilibrium (steady state) level, b measures the effect of the initial level of GDP per capita on its average growth rate from time t-t to time t, u is a stochastic variable, i represents the i-th region, t the time index and T the time interval considered (in our case 1960 to 1994). Thus the dependent variable is the average geometric growth rate of real GDP per capita of each region from period t-t to period t and b is an approximation of the β-indicator with the sign reversed. It measures absolute convergence or the speed of adjustment of actual GDP per capita to its steady state level 14. 13 See Ramsey (1928), Solow (1956), Cass (1965), Koopmans (1965). 14 See Barro and Sala-i-Martin (1991). 13

In particular there is absolute convergence when b is negative (β positive). In such a case the regions which are poorer in the initial period tend to grow faster in the succeeding periods. The larger is b in absolute value the shorter is the period needed for the poorer regions to catch-up with richer ones. Table 1 presents OLS estimates of equation (1). It includes estimates of the equation for the whole sample periods 1960-1993 and 1960-1994 and for all the subperiods analysed in the historical overview of Section 1. The estimate of β for the period 1960-93 indicates a statistically significant (at the 1% level) but rather slow convergence of 1.4 per cent per year. If we add 1994 to the sample period the convergence falls to 1.3 per cent per year. The first number implies that the gap between the poor and the rich regions tended to be closed at the rate of about 1.4 percentage points per year. Remembering that in 1960 the initial gap of the Southern GDP per capita with respect to Central- Northern GDP was 52.7% this estimate implies that it would take 54 years for the gap to be closed 15. However, as we have already seen in Section 1 the convergence process was not stable through time. A period of strong absolute convergence from 1960 to 1971 at a rate of 3.2% per year (statistically significant at the 1% level) was followed during the 1971-78 subperiod by a trivial and statistically insignificant convergence of 0.5% per year. Then from 1978 to 1984 convergence becomes again high and statistically significant at the 1% confidence level (2.9%). Finally in the subperiod 1984-94 we observe divergence at a rate of 0.73% per year. Comparing the last two rows of Table 1, it is interesting to observe how the addition of only one year (1994) changes the degree of divergence of the last subperiod from 0.46% to 0.73% per year and makes the coefficient significantly different from zero. The finding of a small albeit non significant convergence in the period 1971-78 is in apparent contradiction with the divergence observed in Fig. 2. However, the contradiction is only apparent as Fig. 2 shows average GDP per capita of the South compared with average GDP per capita of the Center-North, while the regressions are based on 20 regional observations. The continuation of convergence of poorer Central and Northern regions towards their richer neighbours has outweighed in the regressions the divergence which 15 The calculation is done as follows: (1-0.527) * (1+0.014) t = 1. Hence, solving for t one obtains t = ln (1/0.473) : ln (1.014) = 53.849 54. 14

set in between the Southern regions and the rest of the country. In effect several regions of the Center-North like the North Eastern regions (Veneto, Friuli, Trentino), Emilia-Romagna, Marche and Abruzzi, characterized by medium-small industrial firms were much more able than the traditional industrial regions like Piemonte, Liguria and Lombardy to adapt to the wage shocks, militant unions and the two oil shocks. The same considerations could explain the fact that in Table 1 the divergence observed from 1984 to 1993 is not statistically significant, while it seems substantial in Fig. 2. Table 2 shows for each year the σ-convergence which is the dispersion of the natural logarithms of real GDP per capita of the regions. It is measured by the standard deviation of the natural logarithms of real GDP per capita of each region. If the σ-indicator falls there is absolute convergence. The σ-indicator falls significantly and without interruption from 1960 to 1971. It oscillates until 1978 at a higher level than in 1971, then it starts to fall again from 1978 to 1984 when it reaches the minimum of the whole sample period. Thus the behaviour of the σ- indicator fully confirms the conclusions based on the β-indicator. 2.2-Conditional convergence: the role of organized crime, government expenditures and education. If we add to the initial level of GDP per capita, the only independent variable contained in equation (1), additional explanatory variables like the regional crime rate, regional government expenditures, and a proxy for education, the estimate of the coefficient b is called conditional β-convergence. The inclusion of additional explanatory variables takes into account the fact that the various regions may be structurally different (i.e. have different steady states) and that regional econonomic policy, organized crime, education and micro or macroeconomic uncertainty may speed up or interfere with the natural convergence process determined by the declining marginal productivity of capital. The additional explanatory variables we shall consider are: 1. crim = the regional crime rate, defined as the number of voluntary homicides per 100.000 inhabitants, as a proxy for the strength of organized crime. 15

2. ip = regional public investment in construction and public works, in constant 1980 lire and in proportion to the population resident in the region. This variable excludes subsidies to private investment and investment by public enterprises; it is thus an imperfect measure of the investment activities of the government in each region. 3. cp = regional public consumption in constant 1980 lire and in proportion to the resident population. The data used for public consumption and investment are national accounts data and not data from the budget. Hence transfers to households, the most dynamic part of Italian government expenditures in the last decades are excluded. Had we been in possession of more comprehensive data, our econometric results may have been better. 4. sp = ip+cp = the sum of public investment and consumption as defined above. 5. e = number of students enrolled in secondary schools in percent of resident population. For a more detailed description of the data used and their sources see Appendix 1. The coefficient of the education variable never turned out to be significantly different from zero, thus confirming what had been found by other studies on Italian convergence, and as a result we dropped it in the initial stages of our research. Another variable which was never significant in other studies was net migration. After some initial attempts on our part which confirmed that it is not significant, we decided to drop also this variable. National accounts data for the Southern regions are systematically biased by the fact that government services are estimated at factor costs and that the government has sharply increased public employment after 1970 as unemployment was going up there. For instance Paci (1997) shows that in the 1980s virtually all the growth recorded in Sardinia (a Southern region) was due to the growth of the non-salable services sector which virtually coincides with the government sector. We therefore performed all the tests presented below also with the GDP per capita redefined to exclude non-salable services as we were expecting to find stronger results than with official GDP per capita growth. However, the results did 16

not differ significantly and we present therefore in this paper only the regressions with the official definition of GDP growth 16. A few comments concerning our proxy for organized crime are in order. First in Italy there is virtually no killing going on in connection with small robberies in the inner cities of the type which is common in US cities or in Rio de Janeiro, so that it is safe to assume that the excess of regional voluntary homicides per capita of the South with respect to Center-North is closely related to the interests and activities of organized crime. The latter systematically kills all people who greatly disturb its activities: judges, entrepreneurs, policemen, members of rival bands, former affiliates who know too much or collaborate with judges. Concerning the latter, if organized crime does not succeed in killing them they sometimes systematically kill close relatives, including children. Also politicians who were no longer able to guarantee to criminals protection from condemnation in courts have been killed 17. Second, as the voluntary homicides of organized crime are well pondered joint decisions based on economic-political considerations, they can safely be assumed to be truly exogenous with respect to economic growth unlike homicides in inner cities in the US which must be very sensitive to the business cycle. Third contrary to other proxies for crime used in other studies on Italy which have been found not to be significant (crimes against property or crimes in general), our proxy is much less subject to the risk of underreporting for reasons which hardly need an explanation. Basically as the government is not capable of reacting in real time (court cases take decades to be solved 18 ) and as organized crime has the virtual monopoly of the use of violence, people are afraid of reporting illegal acts to the police. Worse than that, when it comes to gathering information from public offices about potentially illegal activities by criminals in order to formulate clearly the accusations against them to the police, some officials in some regional branches in the South seem to know very well which information they have to conceal and from whom 19. 16 The interested reader can consult Quarella (1997), Chapter III. 17 I am thinking of the Sicilian member of Parliament Salvatore Lima, killed in March 1992. 18 See footnote 5. 19 Recently one of us managed to get for the moment only tentative and preliminary information from AIMA, the state agency in charge of paying out agricultural subsidies of the European Union, on traces of illegal demands for subsidies presented in 1993 by third parties on his farm (see episode no.3, footnote 5), only after writing to the European 17

At a more theoretical level it is worth spending a few words on why organized crime and corruption should affect economic growth negatively. They do not create wealth, they only redistribute it and by doing so they discourage productive economic activities much like taxation 20. They reduce the efficiency of each lira spent on public investment because of the percentage cashed by organized crime and corrupt officials and/or entrepreneurs and because they distort choices towards bribe-maximizing at the expense of truly productive investments 21. In addition they distort occupational choices in favour of activities of redistribution and rent-seeking rather than of true creation of wealth. Organized crime is a threat to property rights and it discourages therefore the accumulation of capital. True entrepreneurs existing in the territory controlled by organized crime have the tendency to leave; worst still, potential investors from other areas of the country or of the world have no incentive to invest in the area 22. If in addition the presence of organized crime is linked, like in many parts of Southern Italy, with an inefficient and slow judicial system and a non-transparent public administration the officials of which are also often colluded with organized crime and corrupt, it is easy to understand why the conditions for convergence are not present. Also the efficiency and equity of the tax system is influenced negatively by corruption 23. Table 3 shows the estimates of equation (1) in which the explanatory variables mentioned above are added to the natural logarithm of the initial GDP per capita (lnyin). The sample period is 1960-93; 1993 is the last year for which almost all explanatory variables are available 24. In order to obtain statistically more robust results and to be able to study better the changes through time of the coefficients of the crime rate and of government expenditures we Commission in Brussels. The Commission put AIMA under pressure. Before writing to the Commission the answer to his request of information was: there is now a law on privacy and we cannot answer. 20 See Bardhan (1997). 21 See footnote 9. 22 Not surprisingly foreign direct investment in Italy as a percent of GDP is among the lowest in Europe and most of it goes to the North. For instance on January 1st 1996 the percentage of factories with foreign participation was 73% in the North, 16% in the Center and 11% in the South (Quintieri, 1998). 23 For a comprehensive analysis of the channels through which corruption affects negatively the efficiency and equity of the tax system and the power of the state to enact economic policy see V. Tanzi (1995). 24 ip and cp were available only until 1992, so we took the averages over three years. 18

divided the sample period into 7 subperiods of 5 years each 25 and calculated 5-year averages of all the additional explanatory variables. Exceptions are the initial level of income, which refers to the initial year of the 5-year period, and the dependent variable which is a 5- year average geometric growth rate. We thus run regressions with 140 observations (7 periods times 20 regions) which guarantee robust results. The procedure used also has the advantage that excessive yearly volatility especially in the crime rate is not allowed to bias the estimates. We observe first that the coefficient of initial real GDP per capita is negative and significantly different from zero at the 1% level. Comparing with the estimate of absolute β-convergence of Table 1 the coefficient almost doubles in absolute value from 1.4% to 2.1-2.5%. This implies that the influence on convergence of the additional explanatory variables included in Table 3 has been significantly negative i.e. that organized crime and government expenditures have significantly reduced convergence for the average of the sample period below what it would have been otherwise. Second the crime rate has a coefficient which is significantly different from zero at the 1% level. Its effect on convergence has been negative. This finding is in sharp contrast with what found by other studies on Italy. Third total government expenditures and investment are not significant and government consumption has a negative influence on convergence but only at a confidence level of 10%. We shall see below that the crime rate and government expenditures are not independent from each other: in the period 1980-93 the crime rate in the South seems to be particularly influenced in a positive way by public consumption, slightly less by public investment. In the regressions of Table 3 all coefficients are constrained to be the same through time and across regions. This is a very heroic assumption. In the remaining of this section we shall remove this assumption and by doing so learn a lot about where and when the crime rate and government expenditures had a negative effect on convergence. First we test whether the coefficients of the additional exogenous variables changed through time. The periods eventually chosen are three: 1960-74, 1975-79 and 1980-93. They were decided following Micossi and Tullio (1992a, 1992b) on the basis of their findings that regional economic policies were characterized by some 25 The 5-year periods considered are: 1960-64, 1965-69, 1970-74, 1975-79, 1980-84, 1985-89, 1990-93. The last period is an exception as the data are averaged over 4 years instead of 5. 19

efficacy in the 1960s, by a neutral transition period in the 1970s and by distinctly negative effects since the early 1980s associated with the degeneration of the Italian political system. The main doubt was where to place the first half of the 1970s and we let the data tell us by placing this period first with the second half of the 1970s and later with the 1960s. Three dummy variables were constructed d1 which assumes the value of 1 in the first three periods (1960-74) and zero otherwise, the dummy d2 which assumes the value of 1 only in the period 1975-79 and the dummy d3 which is equal to one for the last three periods (1980-93) and zero otherwise. Then we constructed the variables spd1, spd2 and spd3 as the the product of sp and the temporal dummies. The variables ip, cp and crim were split in the same way. The results are presented in Table 4. As before the initial level of income is statistically significant with the right sign at the 1% confidence level, suggesting the presence of conditional convergence. The crime rate is statistically significant at the 1% confidence level only in the third period during which the Italian political system had degenerated (1980-93). Without the variable government expenditures the statistical significance of the coefficient of the crime rate in the period 1980-93 increases sharply and it is noteworthy that the Adjusted R 2 is hardly influenced by the suppression of the former (cfr. regr. 4 with regr. 1-3). This may be an indication that these two variables were closely linked to each other in this period. If our views about the Italian political system and about the Southern society since 1980 is correct, the crime rate was influenced positively by government expenditures. The coefficient of government expenditures is positive and statistically significant at the 1% confidence level in the first period (1960-74), it is still positive but less significant in the second period (at the 2.5% confidence level) and becomes insignificantly different from zero in the third. The same holds for its components ip and cp. This fully confirms the views expressed by Micossi and Tullio (1992a, 1992b) and others about the enourmous waste of public resources in the South after 1980. Of the two components the variable cp seems to maintain a slightly more significant and positive role than ip. From a temporal decomposition of the explanatory variables we move to a geographical decomposition by the three major areas of the country: the North, the Center and the South. In order to verify whether the coefficients of the crime rate and government expenditures are significantly different across macro-regions, we define three dummy variables: dn, dc and dm. The first assumes the 20

value of one for each region of the North and the value of zero otherwise 26, dc assumes the value of one for each region of the Center 27 and dm for each region of the South 28. Then for each exogenous variable we define three new variables; for instance crim is split into crimdn, crimdc, crimdm where crimdn is simply the product of crim and dn. A similar procedure is used for sp, ip and cp. The results are reported in Table 5. The conditional β-convergence ranges from 2.6 to 2.9% per year and is significantly higher in absolute value than the unconditional β-convergence estimated in Table 1. The coefficient of the crime rate is not significantly different from zero in the North and in the Center. It is instead different from zero at the 1% percent confidence level and with the expected coefficient in the South. Government expenditures are nowhere significant. The evidence from Tables 4 and 5 seems to suggest that organized crime and corruption became a drag on growth mainly in the period 1980-93 and mainly in the South and that government expenditures may have stimulated growth only in the first period (1960-74) and somewhat less in the second (1975-79). In order to find out more we decided to decompose the exogenous variables jointly by major period and by macro-regions. We define the appropriate dummies and then split each exogenous variable into 9 variables. The results are reported in Table 6. For instance the variable crimd1dn is the product of crim and the dummies d1 and dn and represents the crime rate in the North during the first period (1960-74). Since we have 140 observations and since the exogenous variables are maximum two in each regression we are still left with at least 120 degrees of freedom. There is again strong multicollinearity between the government expenditures variables and the crime rate, especially between total government expenditures and crim, so we present regressions also without the crime rate. If we introduce government spending and the crime rate simultaneously no variable is significant. The crime rate is a factor significantly retarding economic growth in the South in the period 1980-93 (see the coefficient of crimd3dm). Government expenditures has significantly contributed to convergence in the first 26 The regions of the North are 7: Lombardia, Piemonte, Valle d Aosta, Liguria, Trentino- Alto Adige, Veneto, Friuli-Venezia Giulia. 27 The regions of the Center are 5: Emilia-Romagna, Toscana, Marche, Umbria and Lazio. 28 The regions of the South are 8: Abruzzo, Molise, Campania, Puglia, Calabria, Basilicata, Sicilia and Sardegna. 21