ESTIMATION OF UNEVEN DEVELOPMENT OF LITHUANIAN REGIONS

The researchers have aimed to analyse the processes of divergent development of Lithuanian regions by estimating the basic parameters of the process of inter-regional stratification. We have carried out an estimation of the process of differentiation using three parameters: per capita gross regional product (further GRP), per worker GRP and per capita foreign direct investment (further FDI) in a separate region. Using these indicators and the economic classical con vergence-divergence approaches of analysis, the principal dimensions of the unequal development of territorial units in Lithuania have been assessed.


Introduction
During the process of reform, the socioeconomic situation of separate regions in Lithuania has been influenced by a set of new factors. These include: the rate and scale of economic transformations; the development of market sectors; foreign economic co-operation; mutual relationship of each separate region with the Central Government; and ability to adapt to new conditions and to make good use of them. All these factors have promoted the growth of regional differentiation.
The issue of uneven development of the Lithuanian regions has been actively discussed for the last ten years when it became possible to refer to the county level statistics. A new push for discussions emerged at the end of 1999 after Lithuania had been invited to negotiate its membership in the EU, therefore the necessity to form the policy of regional development as one of the requirements arose. During the last decade, the subject of development of Lithuanian' regions has been actively discussed, however, a more detailed analysis of the tendencies of uneven economic growth in Lithuanian regions, covering a longer-term period, is lacking.
In 2000, the Regulation of Regional Development which declares acceleration of a wellbalanced development of separate regions and a decrease of social and economic disparities among the Lithuanian regions as one of the goals was prepared and launched, with agencies of regional development established. A measured extent of Lithuania's uneven development of administrative territorial entities would be a sufficiently good indicator enabling to ground the effectiveness of a pursued regional policy.
In the article, we will statistically evaluate Lithuania's uneven regional economic growth by using a system of indicators, and the object of the research is the process of convergencedivergence in the economy of regions.
Beyond the theories of regional economic convergence, there are a number of other approaches attempting to explain the reasons why regional growth disparities exist and persist. Theories of regional underdevelopment and polarised growth regard underdevelopment as self-perpetuating, in contrast with convergence theories which claim that the less favoured regions can bypass the problems accruing due to their underdevelopment and enter a path of stable economic growth. The latter theories (for example, neoclassical approach) indirectly state that free market conditions tend to eliminate the regional economic disparities through the price mechanism, while regional growth is mainly the outcome of technical progress and of efficient allocation ofrecourses.
Authors give a qualified approval to conclusions of these theories and maintain that without conceptual regional politics, in conditions of market economy, the territorial differentiation of Lithuania's regions is the bigger the faster is its economic growth. 8 Problem of regional underdevelopment and its estimation Regional development questions have attracted the attention of diverse groups of scholars during the past fifty years. The topics that were initially of interest only to economists and geographers are now being investigated by sociologists, political scientists, and researchers from other social science disciplines. This growing interest to regional development studies is due in part to the recognition that the processes driving innovation and national economic growth are fundamentally spatial in nature.
Regional economies are not uniform. Localspecific factors like sectoral composition, history of development, geography, the degree of their integration to the national and international economy, etc. affect the growth prospects of each region. In general, three types of regions can be identified: prosperous or growing, stagnating or declining, and underdeveloped or developing.
Most of the theories of regional growth focus on one or two specific factors in their task to describe the growth process and lack a holistic analysis. In this sense, it would be more appropriate to refer to them as "models" rather than as "theories".
Most of the early theories of regional economic growth were spatial extensions of neoclassical economic theories of international trade and national economic growth. Together, these early neoclassical theories predict that over time the differences in the price of labour and other factors across regions will diminish and tend toward convergence. This prediction has generated considerable controversy among theorists, particularly in light of the apparent tendency toward international divergence between the per capita incomes of industrialized nations and less developed nations. Early theories of regional economic development emerged out of this controversy and can be distinguished from one another in terms of differences in the theoretical predictions regarding interregional convergence or divergence in per capita incomes and factor prices overtime.
The concept of convergence, even in its weaker formulation as long-run constant per capita income growth rates, or conditional convergence, has come under attack from many sides. One criticism is largely empirical. The field of development economics emerged in the post-WorldWar II period in recognition of the growing economic disparities between industrialized nations and less developed countries. Although empirical studies (Perloff et al., 1960;Williamson, 1965) supported a trend toward economic convergence at the regional scale, at least in the United States, critics pointed to the persistent poverty in most less developed countries as an evidence that some regions of the world were not conforming to the predictions of the neoclassical growth models.
Another criticism focuses on the unrealistic assumptions underlying neoclassical growth theories, particularly those having to do with the assumption of constant returns to scale, zero transportation costs, etc.
One response to the convergence critique has been to directly incorporate the prediction of divergence into extant theories of regional economic growth. Here two such theories are examined: cumulative causation theory and growth pole theory. These theories of polarised growth are proposed to explain the existence of regional economic disparities and to describe the process of regional growth. Unlike other (Marxist, socio-economic and the vicious-cycle) approaches these theories were not initially proposed to explain national economic growth. This is probably one of the reasons why these theories are most popular among the different approaches that have been employed by policymakers in order to assist regional development.
Mydral (1957) argues that increasing returns to scale produces clustering of economic activity within the regions that are first to industrialize. Moreover, the process of growth tends to feed on itself through a process of cumulative causation. Although underdeveloped regions offer the advantage of low-wage labour, these benefits tend to be offset by the agglomeration economies found in the industrialized regions. Kaldor (1970) elaborates on and expands Myrdal's theory of cumulative causation by introducing ideas from export base theory and the concept of an efficiency wage. Like Myrdal, Kaldor assumes that increasing returns to scale gives early industrializing regions an advantage in international trade. Cumulative causation sets in when an exogenous shock increases the worldwide demand for an industrial good. Actual monetary wages may be the same in all regions, but efficiency wages, defined as monetary wages divided by a measure of labour productivity, tend to be lower in industrialized regions due to scale economies. Since regions with lower efficiency wages can produce more output, which in turn leads to further reductions in the efficiency wage (and so on), growth may build on itself without bound.
The growth pole theory of regional economic growth places Myrdal's theory of cumulative causation into a spatial context. Perroux's (1950) "space as force" view of spatial interaction, which defines space as a type of network that is held together by centripetal forces. has formed the basis of most growth pole theories. In Perroux's (1950) original formulation, a growth pole referred to linkages between firms and industries. "Propulsive firms" are those that are large relative to other firms and generate induced growth through interindustry linkages as the industry expands its output. Hirschman (1958) argues similarly in his discussion of backward and forward linkages amongfirms. Boudeville (1966) is credited for placing Perroux's formulation into geographic space. For Boudeville, a growth pole is defined in terms of the presence of propulsive finns and industries that generate sustained regional growth through linkages with other firms in a region.
To conclude, the convergence-divergence debate is no longer simply an academic debate when viewed in light of policy issues related to efficiency and equity. If one accepts the convergence hypothesis, then one can assume that lagging regions will tend to grow faster and approach the standards of living in developed regions over time, and inequities will be resolved in the long run simply by improving the functioning of the market. If, on the other hand, there are substantial market imperfections (researchers advocate this approach) in regional trade and knowledge diffusion, as suggested by Stiglitz (1989), then market inefficiencies will result in interregional inequities. The appropriate strategy for improving interregional efficiency and / or equity depends on the nature of the original source of divergence, its speed and the benefits and costs of diverting the path of growth in the other direction.
The earliest interests in the assessment of the process of convergence are found in the works of Easterlin (1960), Borts and Stein (1964). They focus on the so-called a-convergence that enables to size up whether dispersion of per capita GRP is decreasing in the course of time. In more recent works by Barro and Sala-i-Martin (1991, Mankiw and others (1992), also Vohra (1996), J3-convergence is analysed by measuring whether regions with a smaller initial 10 per capita GRP were growing faster than regions with a larger initial per capita GRP.
To measure the uneven economic growth of regions, classical evaluation methods of income distribution can also be applied. The obtained results to be trustworthy, the selected concentration statistics have to fulfil the conditions that are characterised and formulated into five axioms in Atkinson's (1970Atkinson's ( , 1983), Cowell's (1985), Sen's (1973) works.
Practical evaluation of uneven economic growth and identification of its determinant factors are necessary in the formation of regional national development policy and while planning appropriate means to reduce the uneven economic and social development.
The publication "Territorial Outlook", prepared by the Organization for Economic Cooperation and Development, states that the selection of indicators enabling to evaluate disparities of interregional development depends upon the goals of public policy. In the majority of cases, this goal is economic growth. In this case, typical indicators could be the growth of output, the change in one employee's level of produced output, wage, full employment, investments, etc.
The article aims to evaluate the process and rate of uneven economic development of the Lithuanian regions, applying the classical method of economic convergence-divergence process. For the assessment, the index of GRP per capita and per employed person in a region was applied. Each indicator of the territorial inequality reflects only one dimension. A multidimensional index would enable to reveal the nature of disparity in the interregional development, but then we would have to consider which indicators ought to be taken into account and what is their comparative weight.
Estimation of the comparative weight of regions in the economy of the country The Lorenz curve and the Gini coefficient as concentration standards would enable to evaluate the comparative weight of regions as territorial-administrative entities in the Lithuanian economy as well as their economic power in the economy regardless of the population size in each of them.
The Lorenz curve was developed by O. Lorenz as a graphical representation of income distribution. It portrays observed income distributions and compares this to a state of perfect income equality. It can also be used to measure the distribution of GRP. The Gini coefficient is a measure of inequality developed by the Italian statistician Corrado Gini. It is usually used to measure income inequality, but can be used also to measure any form of uneven 0.9 § o.s .; 0.7 .~ 0.6 distribution. The Gini coefficient is a number between 0 and 1 where 0 corresponds to perfect equality and 1 corresponds to perfect inequality. The Gini coefficient is calculated as a ratio of the areas on the Lorenz curve diagram. If the area between the line of perfect equality and the Lorenz curve is A and the area underneath the Lorenz curve is B, then the Gini coefficient is A/(A+B).
According to Figure 1, which reflects GDP distribution among the Lithuanian regions, it is possible to claim that the input of administrativeterritorial entities into the country's economy is very uneven. The total annual value added of five counties (Alytus, Marijampole, Taurage, TelSiai and Utena) where GDP comparative parts are the smallest amounts to only 20 per cent of the whole country production output, while a comparative part of the single Vilnius county during the period ranges from 28 to 35 per cent. The two most economically productive regions of Vilnius and Kaunas, in their own turn, create about 50 per cent of the total value added. The uneven distribution of GDP among regions is inclined to change. In Figure 1, during the period of 1995-2003 the Lorenz curves are receding from the absolute straight line reflecting an increasing uneven regional input into the country's economy.
For the evaluation of the Gini coefficient, the Lorenz curve was directly used. Firstly, to characterise the empirical Lorenz curve (F(x», a polynomial function was applied. The approximation degree of the obtained cumulative distribution function F(x) was estimated with the determination coefficient R 2 (functions with the determination coefficient over 0.9 were chosen for the calculation). Since the function 12 of absolute linearity can be described by the dependence y = x, the Gini coefficient was estimated with the help of determinate integrals (Dikhanov, 1996). The estimated Gini coefficients enable us to evaluate the rate of the regional divergence process more precisely (see Figure 2).
In Figure  and means of production. Attraction of investments is a relevant indicator of regional economic growth depending on production growth, business infrastructure, political decisions related to taxation, privatization, reduced bureaucracy. Every country should regulate investments so that they are directed towards the most underinvested territories. The regional distribution of investments is one of the most relevant issues of a territorially well-balanced development.
The uneven presence offactors necessary for foreign investments in the territories influenced the uneven regional distribution of investments (see Figure 3a and Table).
In 1995, when FDI data became available by regions, the Gini coefficient which estimated its concentration was 0.63 (FDI per capita was 0.34). The concentration ofFDI reached its peak in 2001 when the coefficients were respectively 0.74 and 0.5, ie. increased by 18 and 44 per cent over a fiveyear period (the average annual growth rate was 2.7 and 6.6 percent respectively). Since 2002, the uneven regional FDI distribution began decreasing, but the reason might have lied in the change of the FD I calculation method. J J Since 2002 FDI and MI data have been registered by investment address. previously -by enterprise registration address.
14 Analysing FDI in smaller territorial administrative units, i.e. in municipalities, a far higher degree of concentration is visible; here the estimated Gini coefficient exceeded 0.9, and the FDI per capita was 0.6 since the year 2000.
Analysing FDI by the type of activity, it becomes obvious that the majority of investments in 1995-2003 went to the sectors of industry and services. According to the statistical yearbook of the Lithuanian counties, industrial investments concentrate in the field of manufacturing where about 40 per cent goes to the food industry and 10 per cent to the textile. In the sector of public services, investments concentrate in trade (comparative part is decreasing) and intermediation services (comparative part is increasing). In the sector of trade, most of investments go to wholesale and in the sector of intermediation to financial intermediation. The sectors of wholesale, financial intermediation and food industry attract around 40 per cent of all FDI. Thus, having the distribution of FDI ~ The inequality of distribution of FDI has been estimated since 1995 as before that year the data had not been collected. The distribution of FDI and MI at the level of municipalities has been analysed since 2000. because the changing number of municipalities and size of territories did not allow to measure the distribution of investments. over activity sectors and the plans for economy development on hand, it is possible to predict the potential geographical locations for investments.
The fact that most of the enterprises are concentrated in the five major cities of Lithuania increases the attraction for investments in the cities. The material facilities and infrastructure created there are equally important in attracting investments. Even though some of the light and food industry enterprises emerged in other Lithuanian cities and towns and this is a premise enabling to distribute the foreign investments more evenly, high concentration nonetheless persists. This tendency is very likely to survive in the future, particularly when the comparable part of investments is increasing in the sector of services, the territorial concentration of which is higher than of the industry. Hence, it is possible to state that foreign direct investments depend upon the concentration of industry and service sectors in the counties. The bigger an industry centre, the greater chances to attract investments. The sector of agriculture, which is large enough in Lithuania according to the generated value added, receives almost no foreign direct investments. The situation is complicated by high production costs which result from the low work efficiency; in the absence of a possibility to make a profit foreign capital bypasses this branch of economy.
The counties' statistical outlook states that in 2003 the biggest portion of FDI (60 per cent) goes to the Vilnius County, and more specifically to the cityofVilnius (98 per cent of it). The rest of the county gets the remaining 2 per cent. Klaipeda County absorbs 11.3 per cent of all investments, out of which 72 per cent go to the city of Klaipeda. Kaunas County receives 13.5 per cent of all FDI, 88 per cent of which goes to the city of Kaunas. The remaining counties share only 15 per cent of the FDI, and investments here are not so strictly concentrated in county centres as in the above-mentioned cases.
Analysing the distribution ofMI (see Figure 3, chart b, and Table), it is possible to state that the degree of concentration of MI here is lower than that of FDI both on the level of counties and municipalities. Up to the year 2002 the concentration was increasing, but recently it has dropped which could be the result of the change in the method of calculation. Seventy per cent of the investments are concentrated in Vilnius, Kaunas and Klaipeda counties. The data on the sources of MI and the directions of investments are presented only on the country leveL therefore it is impossible to perform a more detailed analysis.
According to the Statistical Yearbook of the Lithuanian Counties, in 1995-2003 TelSiai County invested the biggest part of its GRP (18 per cent); then economically strongest regions of Klaipeda, Vilnius and Kaunas follow with 17.8, 16.5 and 12.3 per cent, respectively. The least was allocated by the weakest regions: Siauliai and Marijampole -9 per cent each and Taurage -7 per cent; consequently, most of the generated GRP went to consumption, and the accumulation of investments was not sufficient. For the investments to enable the renovation of economy (where new production facilities exceed the depreciation of the old) they have to reach at least 25-30 per cent of the GRP. In the developing regions this percentage must be even higher.
The calculated correlation between the Gini GDP and the Gini FDI per capita (0.79) gives a clear view that the distribution ofFDI and GDP interact with each other. Similar dependence is visible for GDP and MI distribution (the correlation coefficient equals 0.62).
In the countries that develop rapidly, the overall amount of investment reaches 40 per cent from GDP. Lithuanian regions like Klaipeda and Vilnius in 1998-2001 accumulated 15-20 per cent of total value added for investment and stimulated a 25.5 per cent and 52.8 per cent economic growth, respectively. The regions that invested less than 10 per cent from all income weakly promoted their economic activity.
Analysis of the convergence-divergence process of the Lithuanian regional economy based on per capita GDP After the restoration of independence, there were expectations that the development of the market would level up all greater regional disparities. However, it soon became obvious that the metropolitan areas had more benefit from the market economy -owing to short distances, large sales market potential and the accessibility of capital markets. The insufficient mobility of the capital and labour force in the country predetermined the steady structural unevenness to become the reason of regional disparity. It has become obvious that in rural areas the income per capita is lower, the level of unemployment is higher, the dependence on agriculture is stronger, there are more obsolete technologies and more of slowly developing branches of industry.
The course and rate of the uneven development of Lithuanian regions were evaluated by applying the classical methods of convergence-divergence process analysis. For the estimation, the indicator ofregional per capita GDP (GDP u) and per one employed person GDP (GDP g) was used.
The analysis starts from R. Barro  In the first stage, analysing per capita GDP and annual growth rate, we obtain the following results (see Figure 4).
In the computed regression equation (where the independent variable is a region GDPg's logarithm at a time period Tand the dependent variable is a region G D Pg'S logarithm of annual growth rate), the obtained f3coefficient is negative (0.144). However, in the Lithuanian regions the absolute /3-convergence process that takes place due to the rapid growth of the least developed areas, is not evident for three reasons: (1) the obtained equation of regression is not statistically significant due to the small determination coefficient (R2 = 0.2), though the hypothesis of the equality of the a coefficient to zero while using Student's test is rejected; (2) instead of the annual GDP growth in the . g region, an average GDPggrowth within a certain period may be used. In such a way, it is possible to evaluate whether in neglected regions where the initial GDP g (year 1995) was lower, this indicator was growing more rapidly and was catching up with the regions of a high initial GDP g level. The results of the analysis are shown in Figure 5. Figure 5 gives a clear view of the regions where the initial GDP g was higher; the growth rate of this indicator was also higher of this regression equation was equal to 0.52, so statistically it is more significant than the previous equation; the hypothesis of the coefficient {3 being equal to zero when applying the Student's test could also be dismissed. According to the calculation, it may be asserted that there exists an absolute {3divergence based on GDP g in Lithuania; (3) a higher percentage of the annual GDP g growth in less prosperous regions may have been obtained owing to the low initial indicator from which the increment was calculated (e.g., an increase from 8 to 10 will make 25 per cent and from 20 to 24 only 20 per cent). For this reason, in the more prosperous areas the growth of absolute GDP will result in a relatively smaller g .
percentage than in the less prosperous regIOns with a low initial GDP g indicator.
In the subsequent stage of this research, which seeked to evaluate the uneven development of the Lithuanian regions, the analysis of eT-convergence hypothesis was applied.
The concept of a-convergence, suggested by Easterlin (1960) and Borts and Stein (1964), may be characterised in the following way: a group of regions converge according to 6 if the cross-regional dispersion of per capita GDP declines over time: al+ T <a" where er, is the standard regional log (y;) deviation within the period t (Sala-i-Martin, 1996). The author does not provide the formula of calculation of the standard deviation, i.e. he does not emphasize whether the calculated means have to be simple or weighted (weights here are popUlation sizes across the regions) and whether the indicator of the arithmetic mean has to represent the country. According to Foster and Ok (1999), there are practical cases when the dispersion of general distribution is declining and a, is increasing, which leads to a doubt whether the latter statistics gives a full account of eT-convergence. Instead, the adjusted weighted variant of standard deviation ought to be employed: rage of log (y;), and Pi" are the corresponding weights. See results in Figure 6.
The findings displayed in Figure 6 allow to assert that Lithuania's regions diverge according to 6. The process of divergence is continuing, but its rate is declining. The fastest uneven development occurred in 1996-1998 (with the average annual increment of 28 per cent), a slower development took place in 1998-2002 (with the average annual increment of 8.8 per cent), whereas the remaining period is marked with the lowest divergence rate (the average annual increment is 3 per cent). During the whole research period the indicator reflecting eT-convergence increased almost 2.5 times, implying that over an incomplete decade the inequality of the Lithuanian regions according to GDP g increased to the same degree.
The rest of the indicators of uneven economic development are presented in Table 2.
The three generalised entropy (GE) indicators presented in the table above permit us to identify the uneven distribution of GDP g in separate regional groups. The indicators that belong to the GE class have a common estimation formula: where Il is the number of regions,y;" is the indi-   The results of the calculation enable us to assert unambiguously that GDP g-based inequality is far smaller (2.6 times on the average during the research period) in neglected regions in comparison to the prosperous ones. However, the disparity came down from 3.3 times in 1995 to 2.1 times in 2003. It came as a result of a faster process of divergence in the least favoured regions in comparison to the richest ones (see Figure 7, where the changes of indicators estimated in Table 2 during the research period are presented and where the figures of 1995 are equal to 100).
Estimating the indicator of GE (1), equal weight is given to gaps among the regions under research, and this indicator doubled twice during the time of the research. The estimated Gini coefficient reflects the general extent of uneven 20 GDP g distribution, and its value has the same change direction as the above-analysed crconvergence indicator (see Figure 6). The estimated correlation coefficient between GDP g and the Gini coefficient is equal to 0.964. This shows a strong direct relationship between these indicators with a 99.5 per cent accuracy, i.e. the bigger the increase of GDP g in Lithuania, the more unevenly it distributes across the regions.
The curves in Figure 7 suggest that there is no polarisation in Lithuania, because inequality based on GDP g in separate regional groups is not declining. Notionally, it is possible to make such an equation for all class GE indicators: ~oi"'= =Iwithil/grollPS +Iumol/ggrol.lps. As mentioned earlier, findings of the plot highlight that l\G(O)+l\G(2)= =l\G(I); hence we may suggest that the country's overall tendency of unequal development between regions based on GDP g is determined not by a growing gap between the economically most and least developed regions, but by the growing inequality inside the groups of these regions.
All findings of concentration suggest that the GDP g gap among the regions in the course of time is widening, consequently, the process of divergence in Lithuania is taking place. Interregional disparities measured by the GDP g coefficient of Gini in the majority of cases were growing faster than economy itself: during the period 1995 through 2003, GDP per capita grew by 120 per cent and the Gini coefficient increased by 170 per cent.
All of the above analyses performed with the help of the Gini coefficient based on GDP g have produced the following prognosis of unequal economic development across the Lithuanian regions: the predicted Gini coefficient for 2004 is 0.161 (with a reliable 95 per cent interval and possible ±O,014 change range), for 2005 -0.173 (with a reliable 95 per cent interval and a possible ±O,014 change range). A small deviation of prognosis (MAPE=5.13%) enables to assert quite reasonably that, first, if the established conjuncture and, second, a direct interdependence between the rate of economic growth and the regional distribution of this growth do not change, the gap across the Lithuanian counties will be widening in the future, too.
Analysis of convergence-divergence process based on GDP per employed person in the economy of the Lithuanian regions 5 For the practical evaluation of the uneven economic development of the Lithuanian regions it is important to define the rate of the convergence-divergence process based on GDP per employed person. In the first stage, it is relevant to check the hypothesis of [3-convergence. Figure  8 illustrates the obtained results. S Reliable data for this analysis are available from year 1998. In accordance to the plotted regression equation that reveals statistical dependence between the initial (1998) level of GDP u in the counties and the average annual growth rate of the indicator during the study period, one cannot assert that an absolute {3-divergence based on GDP u does exist, since f3 = 0.016. The determination coefficient of the plotted regression equation equals 0.044, i.e. it does not statistically significantly define a linear dependence between the initial level of GDP u and the subsequent average annual growth rate of the indicator. Table  3 presents the remaining indicators that reveal the process of convergence-divergence based on GDP u . Indicators in Table 3 and their variations in Figure 9, as compared with 1998, disclose the process of divergence in the Lithuanian regions assessed by GDP u' Analysing the index of aconvergence it is obvious that its rate of variation partly correlates with the changes of GDP u on the level ofthe whole country (the coefficient of correlation is 0.74, reliability 95 per cent). Over the period of economic decline, the coefficient of concentration lowered. In the beginning of the economic recovery, it remained constant, but in 2002 it grew by almost 27 per cent. This jump may be explained by the different growth rates of the regions: the leading ones have absorbed FD I and making use of new technologies raised their labour productivity, while the weakest regions only partly took advantage of the economic improvement.
The differences between the processes of 0convergence assessed by GDP g and GDP u (according to the former, the interregional inequality is more pronounced) may be the result of the discrepant unemployment level in the Lithuanian counties as well as of the different structure of the population age.  unemployed persons the difference between the GDP g and GDP u in the region would not be great (such a situation may emerge in metropolitan regions where the unemployment level in urban areas is especially low, with lots of young people coming from other regions); on the other hand, with a high level of unemployment and a great number of the unemployed, the difference between GDPgand GDP u will be greater (this 22 may emerge in the least prosperous regions which are abandoned by young people in favour of the largest cities).
In different groups of regions the rates of development vary: in the group of regions with economic retardation the disparity over the discussed period grew 1.16 times, whereas in the group of the most prosperous regions only 1.06 times; therefore, the difference between the  This phenomenon could be explained by the uneven rates of development of neglected regions, i.e. some of them (TelSiai., PaneveZys, Alytus) managed to achieve fast growth rates over the research period, while other regions (Taurage, Marijampole, Utena) produced only a slight growth. The growth rates in the group of the most favoured regions have been rather steady, therefore the gap that existed in the beginning of the considered period remained practically unchanged.

Conclusions
The research of the quantitative assessment of the degree of uneven economic activity in Lithuanian regions was carried out using the classical methods of convergence-divergence analysis. The results of this research provide an opportunity to identify and assess the evolution and rate of the divergent regional development in Lithuania. For estimation, we have selected two basic indicators which reflect economic performance: gross regional product per capita and per worker. All the used distribution indexes reflect the constantly growing regional disparities. Moreover, the rate of this process is not steady and correlates with the economic cycles.
(1) Results of assessment ofthe J3-convergence process based on per capita GDP In the presented regression equation (where the independent variable is a region GDP g'S logarithm in 1995 and the dependent variable is GDP g'S logarithm of average growth rate during the study period), the obtained {3 coefficient is positive (0.11). The coefficient of determination of this regression equation equals 0.52, hence statistically it is significant and the hypothesis of the coefficient {3 being equal to zero when applying the Student's test has been rejected. On the ground of these calculations it may be asserted that there exists an absolute {3 divergence based on per capita GDP g in Lithuania.
(2) Results of assessment ofthe u-convergence process based on per capita GDP The regions in Lithuania diverge according to 6. The process of divergence is continuous, but its pace is slowing down. The fastest unequal development took place in 1996-1998 (the average annual increase amounted to 28 per cent), a slower one occurred in 1998-2002 (with the average annual increase of 8.8 per cent), and the latter was marked by the slowest rate of divergence (annual increase of 3 per cent). Over the period of analysis, the indicator reflecting 6convergence grew 2.5 times, implying that during a period of incomplete ten years the inequality of the Lithuanian regions based on per capita GDP grew by the same degree. (3) Results of assessment of the tJ-convergence process based on GDP per employed person According to the equation of regression that disclosed a statistical correlation between the initial level of GDP u in counties (yearI998) and the average annual growth rate of the indicator in the study period, it is impossible to assert that an absolute tJ-divergence based on GDP u is evident, because f3 = 0.016. The coefficient of determination of the presented regression equation is equal to 0.044, i.e. it does not statistically significantly characterise a linear dependence between the initial level of GDP u and a subsequent average annual growth of GDP u' ( 4) Results of assessment of the u-convergence process based on GDP per employed person Analysing the indicator of a-convergence it is obvious that its rate of variation partly correlates with the changes of GDP u on the level of the whole country (the coefficient of correlation is 0.74, reliability 95 per cent). Over the period of economic decline, the coefficient of concentration decreased. At the beginning of the economic recovery it remained unchanged, but the rise of the coefficient by 27 per cent in 2002 may be explained by different growth rates of the regions: the leading ones have absorbed FDI and making use of new technologies raised their labour productivity, whereas the weakest regions only partly took advantage of the economic improvement.
All findings of concentration suggest that the GDP g gap between regions in the course of time is widening, consequently, the process of divergence in Lithuania is taking place. Interregional disparities measured by the GDP g coefficient of Gini in the majority of cases were growing faster than the economy itself: during the period 1995 through 2003, GDP per capita grew by 120 per cent and the Gini coefficient increased by 170 per cent. The calculated correlation between the Gini GDPg and the Gini FDI per capita (0.79) gives a clear view that the distribution of FDI and GDP interact with each other. A similar dependence is obvious between GDP and MI distribution (correlation coefficient equals 0.62).