What is the theory of economic convergence?

6The neoclassical model is based on the idea that an economy grows by saving and investing in its capital stock and that once a certain level of capital per person is reached – the steady state – the per capita growth rate will be zero (Acemoglu, 2009; Barro & Sala-i-Martin, 2004).

7As the economies’ capital stock per capita is increasing, returns to capital are diminishing and tend towards zero at the steady state where savings (and thus investments) do exactly offset the diminution of the capital stock due to depreciation and population growth.

8Factors that determine growth and steady state in the neoclassical model are all exogenous. Changes in the savings rate, the population growth rate or the depreciation rate can raise the level of the steady state but, due to the diminishing returns to capital, any economy will sooner or later converge to its own steady state. The further away an economy is from its steady state, the higher its growth rates per capita will be.

9Does it mean that economies will converge?

10Yes, if they have the same steady state. Under the assumption that the economies only differ in terms of their initial capital stock, but have the same savings rate, the same population growth rate and the same rate of capital depreciation, the neo-classical growth theory predicts convergence among economies.

11The absence of convergence among countries worldwide might thus not be a reason to reject the neoclassical growth theory. Convergence occurs only if the countries share the same steady state.

12If economies do not have the same savings rate, the same population growth rate and the same rate of capital depreciation, convergence to their own steady state does not lead to inter-economies convergence. However, controlling for special characteristics of the economies in the sample, conditional convergence can be tested.

13Endogenous growth theory does not predict convergence. Contrary to the neo-classical growth theory, it does not assume diminishing returns to capital and there is no convergence towards a steady state and, as a result, capital accumulation is sufficient to ensure growth (Lucas 1988; Romer 1986; Rebelo 1991). Under the assumption that both countries have the same characteristics, meaning the same savings rate, same technology, same population growth rate and same depreciation rate, both countries will have the same growth rate. If they differ in these parameters, they might have different growth rates. But there is no indication of convergence in the sense of the neoclassical model, as there is no systematic relation between the initial level of capital stock and the growth rate of a country. Catching up can occur or not.

14A rather new field of studies which might add to explaining the differences in income and GDP between spatial units and the possible lack of convergence in particular among regions is that of economic geography.

15The aim of economic geography is to explain why certain economic activities localize in certain geographical zones. The main forces driving the new economic geography are increasing returns to scale and the existence of transport costs. For the analysis of regional disparities the two opposite forces of agglomeration and dispersion are particularly interesting. Depending on which of the two is stronger, models of economic geography could explain regional divergence (if agglomeration forces dominate) or convergence (if dispersion forces dominate). Agglomeration forces are due to the fact that a firm producing under increasing returns to scale profits from centralizing its production activities in one region and serving the market in other regions from there. Dispersion forces mainly come from the existence of transport costs but also from congestion effects (Puga, 2002).

16The most important work in this field of study is the Krugman’s core-periphery model (Krugman, 1991) which explains exactly these dynamics. The main setup is that there are two regions, one big “core” region with a higher population share and one small “peripheral” region. Workers are immobile across regions. Firms will locate in the region that has the higher demand. The big region will attract a bigger share of industrial production.

17With regards to explaining regional inequalities in Europe, transport cost should receive special attention. The Krugman’s model predicts that the lower the transport cost, the stronger agglomeration forces will be. In practise, economic integration like in the European Union decreases transport costs and consequently increases the share of sales each firm located in the core region can make in the peripheral regions. Hence, according to this model, economic integration favours agglomeration in general. However, as the size of industry in the core region increases further, so do factor prices because labour is assumed to be immobile across regions. This tends to drive some firms out of the core region.

18Krugman (1991) further elaborates this model, taking into account the mobility of workers. If workers are totally mobile, they tend to cluster together with firms since wages are higher due to increased labour demand in the core region. More workers with higher income will in turn attract more firms and this leads to an endogenous mechanism which results in total agglomeration in one region. Two regions that are initially similar might diverge substantially, if one of the two regions is able to attract only slightly more companies, which triggers the above described dynamics.

19Puga (1999) summarizes the dynamics between economic integration and spatial concentration as follows: If trade costs are high, firms will split between regions to meet local demand. If trade costs are intermediate, regional disparities might occur but not necessary lead to full agglomeration. If trade costs are very low, firms will locate in regions with lowest wages and agglomeration unravels.

20Another important aspect is the existence of externalities that encourage firms to choose a location that is close to other firms and thus leads to the emergence of clusters. Fujita, Krugman & Venables (1999) and Duranton & Puga (2004) suggest agglomeration externalities in learning, sharing and matching. Productivity should be higher in areas with a high level of economic activity. This effect should therefore further encourage agglomeration and adds to understanding regional disparities.

24The main motivation for redistributing income among individuals is that it is commonly assumed that market forces lead to outcomes that leave some better off and others worse off and that a society should aim at promoting an equal wellbeing for all its citizens. This is also referred to as equity and involves the concept of solidarity among the individuals of a society. Moreover, it is assumed that more equity is better for a society as a whole and that some externalities exist from equity. Rich people tend to be happier when others around them are well-off too (Tesch, 2008).

25The question of distributive justice is a difficult one to answer, because it has to take into account the losses faced by those who pay through taxation and the gains made by those who receive transfers. From welfare theory it follows that social welfare can be increased by transferring income from relatively rich towards relatively poor individuals, because the increase in social welfare from giving the poor person an additional unit of income exceeds the loss incurred to social welfare from taking a unit of income away from a rich person (Tesch, 2008).

26Another concept justifying the redistribution of wealth from richer to poorer individuals is the benefits-received principle (Tesch, 2008). The main idea is that rich individuals might benefit more from the provision of public goods than others. The application of a transfer scheme including progressive income taxation and attribution of benefits and social transfers is then a means of compensating low income individuals for the extra demand for public goods exhibited by high income individuals (Lambert, 2001).

27It is intuitive that regional redistributive policy should focus on poor regions, but it is possible that by doing so, resources are allocated to places where they are less productive. Agglomeration of economic activities can be an important source of efficiency gains, while the spatial redistribution of economic activities might entail efficiency losses (Gérard-Varet & Mougeot, 2001).

28In general, economic theory predicts a negative relationship between the level of taxes and economic growth. The explanation is that taxes reduce the incentive to invest in the taxable activity. Income taxes typically reduce the incentive to work and discourage effort (Poulson & Kaplan, 2008). Barro (1990) and Barro and Salai-Martin (1992) analysed the effect of tax policy and government expenditure in various models of endogenous growth. They consider the growth effects of taxes. They show that the growth rate is decreasing in the rate of distortionary taxes while it is unaffected by non-distortionary taxes.

29Gemmell and Kneller (2002) classify among the distortionary taxes income and profit taxes, social security taxes and payroll and manpower taxes. The redistribution of income is to a very big extent done through income taxation and social security taxes. The theory thus predicts a negative impact of taxes and redistribution on growth.

31This section will try to summarize some of the most important empirical works on convergence in Europe. It will highlight the fact that empirical works find convergence as well as divergence among countries and regions, depending on the dataset, the time period and the methods applied.

32Most empirical papers find no proof of absolute convergence in worldwide datasets. If we consider the framework of the neoclassical growth theory, this is not too surprising, since it is unlikely that countries in a worldwide dataset have similar characteristics and steady states. Contrary to that, absolute convergence is usually found among countries or regions that are similar, such as countries belonging to the European Union.

33Probably, the most important empirical studies go back to Robert Barro and Xavier Sala-i-Martin. In one of their first papers on convergence they analyse the growth and dispersion of income in the United States since 1880 (Barro & Sala-i-Martin, 1991). They find evidence that poor states grow faster than rich states and they conclude that the gap between poor and rich states diminishes at a rate of about 2% per year. Applying the same framework to 73 regions in 7 countries in Western Europe since 1950, they find similar results.

34Sala-i-Martin (1995) and Gaulier et al. (1999) confirm the existence of convergence across European regions.

35Barro and Sala-i-Martin extended their analysis from 1991 to 90 European regions and longer time periods (Barro & Sala-i-Martin, 1992). For the period 1950-1990, they find absolute convergence across European regions. They further estimate the within country convergence for Germany, United Kingdom, France, Italy and Spain and find estimates very close to what has been found for the US.

36Fagerberg and Verspagen (1996) test absolute convergence among European regions over the same period and find similar estimates. However, convergence rates are lower if within-country convergence is considered.

37In a more recent study, Arbia and Piras (2005) use panel data to test for convergence among European regions. They test a cross section model on a sample of 125 regions belonging to 10 European countries during the period 1980-1995 and find convergence also when accounting for the spatial dependence among regions.

38Rather mixed results are found by other authors. Cappelen et al. (2002) analyse 95 European regions from 1980 to 1997 and find that the standard deviation of per capita GDP has declined only very little. When Spain, Portugal and Greece are excluded from the sample, it has even increased. This suggests that the small decline in inequalities is due to a catching up of the Southern European countries during this period, while inequalities within countries persist or even increase.

39Bouvet (2010) finds that generally inequalities have decreased since 1977, but this decrease is attributed to between-country reduction in inequalities. She studies a sample of 197 NUTS2 regions between 1977 and 2003 and finds substantial fluctuations in inequalities until the beginning of the 1990. Only from 1993 onwards, she finds a continuous decrease in regional inequalities across all regions. Within country inequalities have increased since the mid-1990s except in the Southern European countries after their accession to the EU.

40Boldrin and Canova (2001) find no evidence for convergence among 185 European regions between 1980 and 1996. They report that there is strong evidence of a catching up of poorer regions to the EU-average during the period 1950-1975, but this process largely stopped over the next two decades.

41The framework for analysing the effect of fiscal policy on growth can be extended to investigate regional convergence. If fiscal policies are able to enhance growth in poorer regions relative to the growth rate in richer regions, fiscal policies will ultimately contribute to convergence across regions.

42Concerning the impact of fiscal policies on growth, one important contribution comes from Barro (1999) who refers to the well-known argument that transfer payments and the levying of taxes distort economic decisions and the more income is redistributed in an economy, the bigger the distortions are. Investments are reduced and consequently economic growth. He empirically studies a broad panel of countries from 1960 to 1995 and finds that income-equalizing policies are growth enhancing for poor countries, but lead to a reduction in overall growth rates in rich countries. This enhances convergence but at the risk of a lower overall economic growth.

43Furthermore, according to Barro, redistributive fiscal policies aiming at reducing disparities might have a negative impact on labour mobility. This is also the conclusion of Obstfeld and Peri (1999) when comparing the US to some European countries and Canada.

44Some recent studies estimate the impact of redistribution on regional convergence in Europe.

45Meunier et al. (2007) investigate if national interpersonal transfer policies affect regional economic growth rates. Using a sample of 230 NUTS2 region from 1995 to 2002, they proxy the level of redistribution with an index of relative disposable to relative primary income of a region. Their transfer index allows for an analysis of net contributing regions and net benefiting regions separately. Taking into account the spatial dependence among regions, they find that transfers received have a positive impact on a poor region’s growth rate while they acknowledge the presence of a simultaneity bias in the case of net contributing regions that does not allow for any evidence.

46Algoed and Persyn (2009) who analyse convergence across 140 NUTS2 regions find different results. They first estimate the rate of interregional redistribution for each country in the sample, expressing how much, on average, of the relative difference in primary income is eliminated through redistribution. When accounting for the rate of redistribution in a cross-section setting with average growth rates over the period 1995-2007 and independent variables evaluated at their levels in 1995, they find convergence and a positive impact of the level of redistribution on growth. Including country fixed-effects, they find a typically lower rate of within-country convergence and a negative impact of redistribution on growth. Furthermore, by interacting the level of redistribution with the initial level of income, they find that redistribution slows convergence because the negative effect on growth is larger for initially poorer regions.

47Another approach was adopted by Checherita et al. (2009). They analyse the role of net fiscal transfers for income and output convergence across 230 European regions between 1995 and 2005. They use a simultaneous equation system that allows for net transfers to be endogenously determined and also accounts for effects on labour motility. They proxy net fiscal transfers with the ratio of disposable to primary income and estimate the effect of net fiscal transfers in a split sample of what they call “heavily taxed regions” and “receiving regions and marginal payers”. The findings suggest that higher taxes have a negative impact on growth in both sub-samples. Furthermore, they find that fiscal transfers have a stronger growth-reducing impact on richer regions than on poorer regions. Moreover, they estimate a negative impact of net fiscal transfers on labour mobility and a negative impact of labour mobility on growth of the poor regions. As lower labour mobility has a positive impact on output growth, transfers have indirectly a positive impact on the growth of poor regions.

49We will focus on the growth of primary household income instead of GDP, due to the importance of interregional commuting workers which impact the regional income. If interpersonal transfers discourage working and outward commuting, they should have a negative impact on the growth of primary income.

50The dataset used for this part is entirely taken from Eurostat’s regional household statistics [2]. Eurostat measures primary and disposable income of households which reside in a particular region. Primary income includes several components, the most important of which are compensation of employees and property income. Disposable income is based on primary income which is reduced by current taxes on income or wealth, social contributions and other transfers paid by households and increased by social benefits and other transfers received by households.

51The geographical unit of analysis is the NUTS [3] 2 level. [4]

52The dataset includes the following 16 countries: Austria, Belgium, Czech Republic, Finland, France, Germany, Greece, Ireland, Italy, The Netherlands, Poland, Portugal, Slovakia, Spain, Sweden and UK. This amounts to a total of 229 regions [5]. The dataset includes the years 1995 throughout 2008. The remaining EU-countries are excluded from the analysis due to a lack of data available for the entire period or due to the fact that they comprise only one region at NUTS 2 level, which does not allow for interregional transfers analysis.

53Absolute convergence implies that regions with lower initial income should have a higher growth rate, in other words, they are supposed to catch up.

54In order to verify convergence empirically, a standard model is estimated. For each region, the average growth rate of per capita primary income (PINC) between 1995 and 2008 is calculated and is then regressed on the respective levels in 1995. The average growth rate g of the primary income of a region i over the entire period under observation is calculated as the average logarithmic difference.

55Then, the following regression is estimated:

56

57A significant negative coefficient for ß is a sign of absolute convergence. The results are reported in table 1.

Nobs: 229 asterisks indicate the statistical significance level: * = 10%, ** = 5%, *** = 1%; this notation will be kept throughout the remainder of the analysis.

58The coefficient is negative and significant at a 1% level, suggesting absolute convergence among European regions. Regions with initially lower income – as measured by the relative levels in 1995 – have been growing faster. The main assumption made here is that all regions are structurally similar and they only differ in the initial level of income, since no other regional characteristics are controlled for.

59In a second step, the standard model is estimated again across all European regions, but considering that the national “steady state” can vary from country to country. This hypothesis is accounted for by including country dummies in the regression. By doing so, we isolate the within-country convergence from the inter-country convergence.

60Country dummies are jointly significant at less than 1%. This confirms the existence of different national steady states. As shown in table 2, within-country convergence in income – even though significant at a 5% level – Is much lower than the one estimated previously across all European regions.

61Does this mean that, once considering national fixed specifications, poorer European regions are not converging to richer ones?

62To answer this question we estimate the following ad-hoc model:

63

64where RELPINC95i indicates the regional initial primary income relative to the national average.

65Results are reported in tables 3 and 4. Estimations are done with, as well as without country dummies.

66It clearly seems that the growth rate is higher for initially poorer regions but that it is lower if the region is poor with respect to the country it belongs to. This result is only a statistical observation but could be attributed to the fact that the growth of relatively poor regions suffers from the persistence of structural elements that lowers their potential for economic development. One of those elements could potentially be the existence of transfers, which will be tested in the next section.

67To estimate the impact of transfers on growth, our ad-hoc model is augmented with an index of transfers’ intensity. Transfers are considered separately for receiving regions and contributing regions as their impact on regional growth follows different channels. For receiving regions the transfer intensity will be noted TRSFR, for contributing regions TRSFC.

68The transfer index TRSFR for a receiving region i is computed as the following ratio:

69

70where PINC and DISP are the primary and disposable income respectively, measured in region i and in country n. For receivers, this ratio is higher than one and a higher transfers’ intensity corresponds to a higher index. For contributors, TRSFR is null. For simplicity, the transfers’ index for contributors TRSFC is computed with the inverse of the receivers’ formula such that, also in this case, a higher index corresponds to a higher transfers’ intensity. For receivers, TRSFC is zero.

71Estimations are done in a panel setting which allows controlling for regional fixed heterogeneity.

72Growth rates are computed for 4 ten-year periods (1995-2005, 1996-2006, 1997-2007, 1998-2008). Independent variables are considered at their initial level for each period.

73Table 5 presents the results taking into account time dummies and successively no territorial dummies, either only country dummies or only region dummies.

74Globally, panel estimation without territorial dummies confirms the cross section results in tables 3.

75Panel estimation 3 with country and time dummies shows a significant negative impact on growth of transfers’ intensity both for receiving and contributing regions. This result is similar to Algoed and Persyn’s. However, this conclusion is not maintained when taking into account regional fixed effects as in panel estimation 3. Once controlling for regional fixed heterogeneity, transfers show no significant impact on regional growth and hence on regional convergence.